Rotation Held Up While StopLoss Bounces Persisted

Abstract

This report reviews six supplied CoinRobot.AI strategies for the closed week ending 2026-10-04, using only the provided historical snapshot, strategy formulas, KPI windows, execution-model diagnostics and optimization records. The market context was constructive but not uniformly easy: Bitcoin gained strongly over the 90-day lookback, the supplied Binance majors all showed positive 90-day returns, yet the CoinRobot market vector still classified most covered symbols as sideways rather than bullish. That combination is important because it rewards systems that can participate in sustained upside while filtering intraday chop. The most compelling measured result in this packet came from Strategy 934, a dedicated CONVERT relative-momentum rotation system, whose 30-day aggregate showed a 4.9308 profit factor, 6.1074% convert increase and 8.4985% portfolio profit across six active plans. The report does not treat that as a universal conclusion: several other strategies demonstrated high win rates with poor or marginal profit factor, sparse active evidence, inactivity, or execution-model mismatch. A repeated theme across Strategy 932, Strategy 933 and Strategy 918 is StopLoss bounce risk: many stop-loss exits were followed by rapid recovery above the exit price, so the applied changes generally widened or confirmed stop-loss branches rather than removing emergency protection. The research also highlights customizable trading intelligence as an auditable process. Users and community authors express market views through formulas; AI-assisted changes remain visible, scenario-validated and testable rather than hidden optimization claims. The correct reading is therefore conditional: current evidence supports conservative iteration and explicit hypotheses, not guaranteed improvement.

Executive summary

Market backdrop was positive over the longer window, with BTC up about 38.9% on Binance over 90 days, ETH up about 56.3%, SOL up about 55.6% and BNB up about 39.9%, while the CoinRobot symbol regime mix remained dominated by sideways classifications.
Strategy 934 delivered the strongest robust CONVERT profile in the supplied set, with 30-day profit factor 4.9308, convert increase 6.1074% and portfolio profit 8.4985% across six active plans.
Strategy 935 showed why win rate is insufficient: its 30-day aggregate win rate was 68.2594%, but profit factor was only 0.9987 and portfolio profit was slightly negative, prompting a conservative profit-guard SELL adjustment.
Strategy 932 had attractive 30-day CONVERT evidence on one active plan, including 87.3735% convert increase, but its broader 90-day evidence and prior validation-blocked episode require caution.
Strategy 933 retained strong 30-day aggregate profitability but active-revision evidence showed recent StopLoss bounce risk, leading to a StopLoss-only confirmation change.
Strategy 918 and Strategy 930 were largely historical or inactive in current windows, so their applied changes are hypotheses rather than proven live improvements.

Market analysis

The supplied market snapshot describes a market with strong large-cap momentum but mixed breadth quality at the strategy level. CoinGecko global market capitalization was approximately 2.924 trillion dollars with 24-hour volume near 85.165 billion dollars and BTC dominance at 59.319082%. The same snapshot showed a negative 24-hour global market-cap change of about 2.014235%, so the week closed with some immediate pressure even though the 90-day context was strongly positive. CoinGecko’s Bitcoin 90-day record showed a 36.5474% return, a maximum drawdown of only about 6.9827%, and annualized volatility near 38.0261%. Binance data for BTCUSDT was similar but slightly stronger, with 38.9083% 90-day return, 6.9224% maximum drawdown and 37.8759% annualized volatility. This is a favorable backdrop for trend and momentum systems, but it is not a blanket approval for all entry styles because intraday reversals, fee drag and stop placement still determine realized outcomes.

The four supplied Binance majors all advanced over 90 days. ETHUSDT returned 56.3199% with annualized volatility of 51.5858%; SOLUSDT returned 55.6212% with higher volatility and a maximum drawdown of 11.2333%; BNBUSDT returned 39.9202% with lower annualized volatility at 33.5371%; BTCUSDT returned 38.9083%. Breadth across those four instruments was therefore 100% positive. However, the broader CoinRobot market vector covered 153 symbols and classified 89 as sideways, 35 as bull and 29 as bear. The median daily RSI was 55.1, median daily normalized ATR was 3.743902%, median daily Bollinger width was 10.244736 and median daily trend strength was 70. This combination matters: trend strength was high enough to justify momentum participation, but the majority-sideways regime distribution implies that many individual symbols were not in clean directional trends. Formula thresholds that require EMA alignment, positive MACD histograms, positive short-window momentum and volatility caps should therefore be useful filters. Conversely, broad entry formulas that react too quickly to short bounces are exposed to churn.

Daily major-asset sequencing supports that interpretation. BTC moved from weakness early in September into a strong mid-month surge, including 5.8454% on 2026-09-18 and 6.7038% on 2026-09-21, then consolidated around the mid-80,000 area before a 2.096% rise on 2026-10-04. ETH similarly posted strong advances on 2026-09-18 and 2026-09-21, then moved sideways-to-slightly-higher into the end of the week. SOL was more volatile, with a 10.9012% gain on 2026-09-18, later pullbacks, and renewed strength into 2026-10-04. BNB had a strong early September jump, a later consolidation, and positive closes into the final days. These sequences are consistent with strategies needing both trend-following participation and profit-protection patience; exits triggered too close to intraday lows could miss rebounds, while loose entries could buy late extensions.

Event data provides context but not causality. The supplied event calendar includes ongoing incentive programs and upcoming network or governance events for assets such as SOL, ETH, ADA, SUI, APT, LDO, HYPE, NEAR and others. The rules explicitly require dated events to be connected to price only when temporal evidence supports it. No strategy-specific trades are mapped to those events in the packet, and news and macro feeds were disabled. Therefore, events should be treated as possible future volatility sources, not explanatory causes for the measured strategy outcomes.

The strategy results align with the market state in a differentiated way. Strategy 934’s CONVERT design benefited from relative rotation in a rising but uneven market. Strategy 935’s BUY/SELL cohort suffered because higher win rates did not overcome loss size and early-exit behavior. Strategy 932 generated strong one-plan CONVERT evidence but broader historical results remained mixed. Strategy 933 produced good medium-window BUY/SELL results yet active-revision stop-loss exits showed bounce risk. Strategy 918 and Strategy 930 had mostly historical evidence and inactive current windows, which reduces confidence in immediate conclusions. In short, the market was favorable for momentum, but only formulas with suitable selectivity, rotation logic or patient risk handling converted that backdrop into robust measured performance.

Asset	90d return	Max drawdown	Volatility	Taker buy share	Funding
BTCUSDT	38.91%	-6.92%	37.88%	49.49%	0%
ETHUSDT	56.32%	-5.62%	51.59%	49.74%	0%
BNBUSDT	39.92%	-7.54%	33.54%	49.29%	0%
SOLUSDT	55.62%	-11.23%	53.89%	50.68%	0%

Methodology

The analysis is restricted to the supplied historical snapshot as of 2026-10-04T23:59:59+00:00. No external prices, news, macro data or exchange data were added. Where the packet states that current-only global, local-calculator or exchange summary fields may reflect collection time rather than exact historical closes, this report treats those values as contextual rather than precise close marks. News and macro feeds were disabled, so the report does not infer macro causality. The primary evidence layers are: market context, Binance and CoinGecko lookback metrics, CoinRobot market vector, strategy formulas, daily KPI windows, execution-model KPI windows, revision history, active-revision diagnostics, StopLoss cohorts and optimization validation status.

The calculator reference table used throughout is deterministic and comes from the supplied calculator catalog. Momentum fields include 1m-2, 1m-10 and 1m-20, each defined as percent change versus the open price two, ten or twenty minutes ago. RSI fields such as 5m-RSI and 1d-RSI are Wilder RSI values on the named timeframe, with values from zero to one hundred. MACD fields such as 1m-MACD-hist, 5m-MACD-hist and 1d-MACD-hist are MACD line minus signal line on the stated timeframe. EMA fields are price-level trend measures on their stated timeframe. Bollinger fields include upper, lower, middle and width definitions; width is percent where supplied as such. NATR fields are normalized ATR percentages and are compared across assets more safely than raw ATR. MarketRegime-1d is a deterministic bull, sideways or bear classification based on EMA and MACD structure. TrendStrength-1d is an index from zero to one hundred. Position-state fields include TradeCount, TimeSinceLastBuyMin, TimeSinceLastSellMin, LastBuyPrice, LastSellPrice, ProfitSinceBuyPct and DrawdownFromHighPct. Null values are treated as unavailable evidence unless the formula explicitly uses null logic, for example “TimeSinceLastSellMin = null OR …”. Missing values are never coerced to zero.

Performance windows are interpreted separately. The seven-day window captures immediate behavior but can be dominated by sparse trades or temporary noise. The thirty-day window is usually the most useful balance between recency and sample size. The ninety-day window provides background but may include old formulas, inactive plans, backtests or payloads with limited coverage. Coverage metadata is therefore a constraint: a metric with zero win-rate coverage or limited convert-increase coverage is not treated as comparable to a fully covered metric. Plan eligibility is also considered when portfolio-profit percentages are reported.

Execution-model attribution is handled carefully. BUY/SELL plans attempt to enter and exit asset positions into cash or stable quote exposure according to formula triggers. CONVERT plans primarily target growth in asset units by rotating among selected assets; fiat profit remains secondary evidence. A CONVERT switch is not automatically proof that the strategy BUY or SELL formula evaluated true. When the diagnostics provide decision patterns such as sell+neutral or neutral+buy, those are interpreted separately from ordinary BUY/SELL formula triggers. Source and destination legs are assessed only where paired execution confirmation and target-versus-source return evidence are available. Incomplete or unavailable switch evidence is treated as a data limitation.

Adaptive changes are separated from evidence. An applied change is described as a hypothesis until post-change evidence accumulates. A validation-blocked change is treated as a research success in governance because it prevented invalid formula deployment. The report compares supplied base formulas, active formulas and optimization notes without assuming that AI revisions improve results. Formula interpretation follows expression order and explains key predicates, thresholds, AND/OR structure, and position-state dependencies. StopLoss formula logic is distinguished from plan-level stoploss_policy; disabled plans may still have a formula present but not apply that layer in execution.

Observation windows: 7, 30 and 90 days. Formula changes remain attributable by strategy version and adaptive change ID.

BUY/SELL and CONVERT execution models

The supplied strategies use two materially different execution models. BUY/SELL plans are judged primarily by whether formula-confirmed entries and exits create positive realized profit after loss size, churn and any exchange-specific costs. CONVERT plans are judged primarily by whether rotations increase the quantity of target assets or improve relative exposure; their fiat profit and portfolio value remain useful, but a convert increase can be the more direct objective. This distinction is crucial because several strategies share formulas across models, while others are dedicated to only one model.

Strategy 934 is the cleanest CONVERT example. Its BUY/SELL cohort has no active plan count or events; all meaningful evidence is CONVERT. Over thirty days it recorded 52 trades, 72.7273% win rate, 4.9308 profit factor, 6.1074% convert increase and 8.4985% portfolio profit. Active execution diagnostics for the latest packet show four CONVERT events, two plans, one switch, one complete execution pair and one stop-loss event. The one paired switch had a sell+neutral decision pattern and target outperformance over source at 15, 60 and 180 minutes, with mean 60-minute target-minus-source return of 0.89823%. That is constructive but very small sample evidence. It supports the idea that the model can rotate well, not that every future neutral convert will outperform.

Strategy 935 is shared across BUY/SELL and CONVERT, but the cohorts are not comparable in scale. BUY/SELL has eight active plans in the execution-model KPI block and 624 trades over thirty days, with profit factor 0.993 and slightly negative portfolio profit. The single CONVERT plan has only six thirty-day trades but a very high profit factor of 14.6769 and 28.0716% portfolio profit. The CONVERT sample is too small to declare superiority for the shared strategy, especially because its single switch underperformed the source at 15 and 60 minutes but outperformed at 180 minutes. For shared formulas, changes must not overfit the small CONVERT cohort at the expense of the much larger BUY/SELL cohort.

Strategy 932 is also shared in structure but currently has meaningful recent evidence mainly from a single CONVERT plan. Its thirty-day CONVERT cohort shows 18 trades, 85.7143% win rate, 2.3102 profit factor, 87.3735% convert increase and 16.9747% portfolio profit, while its BUY/SELL cohort has no active seven- or thirty-day evidence and mixed ninety-day historical evidence. The CONVERT execution diagnostics include one switch but incomplete switch pairing and unavailable target-versus-source post-switch returns. Therefore, the convert increase is important, but the mechanism cannot be fully attributed to validated switch timing.

Strategy 933 is BUY/SELL only. It has two active plans and no CONVERT cohort. Its thirty-day BUY/SELL evidence is strong in aggregate, with 1521 trades, 78.4024% win rate, 1.3148 profit factor and 14.0474% portfolio profit. However, seven-day performance deteriorated to a 0.906 profit factor and negative portfolio profit. Execution diagnostics show 157 BUY/SELL events in the active analysis, with 89 buys and 56 sells. That means the formula is active enough to evaluate, but the recent edge is weaker than the longer window.

Strategy 918 is BUY/SELL only and largely inactive in current windows. It has zero active seven- and thirty-day trades and only historical ninety-day evidence, including 126 trades and a tiny 0.1953% portfolio profit with no win-rate or profit-factor coverage in the aggregate KPI block. Its revision KPI shows 90-day profit factor 1.0793 from 126 trades, but all plan activity ended weeks before the analysis date. Strategy 930 is mixed BUY/SELL and CONVERT historically, but all plans are inactive by the end of the snapshot. Its ninety-day aggregate portfolio profit is positive at 7.1224%, driven mainly by CONVERT portfolio profit of 11.4806%, but active-revision evidence is historical and the latest change is a hypothesis.

Overall, CONVERT produced the strongest measured result where the strategy was designed for it, especially Strategy 934. BUY/SELL produced strong longer-window results in Strategy 933 but weaker recent results in Strategy 935 and Strategy 930. The report therefore does not conclude that one execution model is universally better. It concludes that formula intent, sample size, execution evidence and rotation diagnostics must be aligned before comparing models.

Strategy	Execution model	Plans	30d trades	30d profit	Profit factor	Win rate	Unit increase
CONVERT - Relative Momentum Rotation	CONVERT	8	52	207.54 $	4.93×	72.73%	6.11%
BUY/SELL - Trusted Asset Recovery Profit Guard	BUY/SELL	9	624	-3.31 $	0.99×	68.4%	—
BUY/SELL - Trusted Asset Recovery Profit Guard	CONVERT	1	6	1.4 $	14.68×	60%	6.43%
Universal - Flexible Momentum Quick Profit Guard	BUY/SELL	3	0	0 $	—	—	—
Universal - Flexible Momentum Quick Profit Guard	CONVERT	6	18	50.34 $	2.31×	85.71%	87.37%
Universal - Selective Trend Momentum Profit Guard	BUY/SELL	2	1,521	57.8 $	1.31×	78.4%	—
Universal - Selective Safe Entry 1pct	BUY/SELL	2	0	0 $	—	—	—
OKX - Trend Pullback Guard Fee Aware	BUY/SELL	1	0	0 $	—	—	—
OKX - Trend Pullback Guard Fee Aware	CONVERT	3	0	0 $	—	—	—

CONVERT decision-pattern outcomes

Strategy	Pattern	Switches	Avg unit increase	Target beat source after 60m	Mean target − source return
CONVERT - Relative Momentum Rotation	sell+neutral	1	1.27%	100%	0.9 pp
BUY/SELL - Trusted Asset Recovery Profit Guard	sell+neutral	1	5.99%	0%	-0.63 pp
Universal - Flexible Momentum Quick Profit Guard	unavailable+unavailable	1	87.37%	—	—

StopLoss analysis

StopLoss behavior is one of the clearest cross-strategy themes. The supplied data repeatedly shows a tension between emergency protection and wick-driven exit churn. The correct interpretation is not “disable StopLoss.” StopLoss is a separate risk-control layer that can protect against sustained adverse movement, and plan-level stoploss_policy determines whether it is applied. The evidence instead supports more confirmation or wider thresholds in selected strategies where exits often recovered quickly after triggering.

Strategy 932 provides a direct example. In its baseline revision history, stop-loss analysis recorded 8 StopLoss exits, 6 early exits, 4 exits near the following 60-minute low, 4 reentries above exit by 3 hours and 4 by 6 hours, with recovery above exit within 5 minutes in 6 of 8 cases. The active analysis still shows enabled StopLoss evidence of 8 exits with 75% early-exit rate, 50% exit-near-60m-low rate, 62.5% to 75% short recovery depending on the recovery horizon, and only 12.5% negative stop-loss exits in the detailed cohort. That directly justifies a StopLoss-only adjustment that preserves hard severe-loss exits while requiring stronger momentum confirmation around intermediate loss thresholds.

Strategy 933 shows similar but more current evidence. In the latest active revision, the enabled cohort recorded 12 StopLoss exits; all were negative, 75% were early, 58.333% were near the following 60-minute low, and 66.667% to 75% recovered above the exit price within one to five minutes. The adaptive change tightened the intermediate 95.2% loss branch by requiring stronger 1m-2, 1m-10 and 1m-20 weakness, while retaining the hard 94.6% severe-loss exit and unchanged bear-regime and volatility-breakdown emergency branches. This is coherent: it does not remove protection, but it asks for clearer local confirmation before accepting a stop that historically bounced quickly.

Strategy 918 also has StopLoss bounce evidence, though it is historical rather than current. The enabled cohort recorded 34 StopLoss exits, early-exit rate 0.764706, exit-near-60m-low rate 0.411765, recovery above exit of 0.676471 within one and two minutes and 0.705882 within five minutes. Its applied change widened price and drawdown triggers and required local downside confirmation when the broad market-composite stop was non-neutral. Because seven- and thirty-day active evidence is absent, this remains a hypothesis.

Strategy 935 is more ambiguous. It has one active-revision StopLoss exit in the latest packet, with early-exit and negative-exit flags, but the sample is too small to justify further StopLoss modification. Prior evidence showed many stop-loss exits recovered quickly, leading to a previous wider loss threshold and hard 5% catastrophe branch. The latest AI appropriately left StopLoss unchanged and adjusted only the SELL profit guard.

Strategy 934 uses StopLoss as a volatility-aware profitable-peak guard rather than a normal loss-cut rule. Its formula requires ProfitSinceBuyPct above 1.00 before checking volatility-tiered drawdown from high, MACD weakness and either RSI-trend sell or sufficient 1m-10 weakness. Because most plans have StopLoss disabled and active StopLoss evidence is sparse, the report treats this as a protective overlay rather than a core exit engine. Strategy 930 shows mixed exit evidence: stop-loss exits had bounce behavior, but longer follow-up sometimes deteriorated, so the applied change appropriately did not loosen StopLoss and instead tightened a BUY branch. Across the packet, the evidence supports a disciplined rule: confirm StopLoss where bounce rates are high, but preserve hard emergency branches for fast adverse moves.

Strategy analyses

Indicator reference: from market concepts to formulas

Formulas combine indicator comparisons and position state into trading rules. AND requires all conditions in its branch; OR allows alternative branches. Read parentheses, comparison operators and units together. A condition describes eligibility, not proof of execution or future profit.

Timeframes below describe calculator inputs; they are distinct from the 7, 30 and 90 day performance windows. Timeframe-prefixed tokens are grouped by indicator; use the listed prefix plus the indicator name in formulas (for example, 1m-RSI). Similar indicators on different timeframes can confirm or contradict each other. Position and market-state fields are included alongside technical indicators.

Formula token / category	Timeframe	Meaning and formula use	Value / unit / range
`BuyHoldBenchmark` Position-state	State / composite; see definition	Plan return minus the return of the run-scoped equal-weight buy-and-hold basket. Positive values mean the Plan outperformed HODL. Calculation: PlanReturnPct - EqualWeightSelectedBasketReturnPct. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	percent or null until all selected symbols have valid prices
`2` Momentum	1m	Relative change versus the open price 2 minutes ago using 1-minute data. Calculation: (current - open_2m) / open_2m * 100 Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	percent change
`10` Momentum	1m	Relative change versus the open price 10 minutes ago using 1-minute data. Calculation: (current - open_10m) / open_10m * 100 Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	percent change
`20` Momentum	1m	Relative change versus the open price 20 minutes ago using 1-minute data. Calculation: (current - open_20m) / open_20m * 100 Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	percent change
`action` Market	State / composite; see definition	Action signal derived from the last 2/10/20 minute changes. Calculation: Rules based on 2m/10m/20m relative change thresholds. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	BUY / SELL / WAIT
`RSI` Momentum	1m, 5m, 1d	20-period Wilder RSI calculated on selected-timeframe closes. Calculation: Wilder RSI(20) using smoothed average gains/losses over selected-timeframe close prices. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	0-100
`RSI-trend` Trend	1m, 5m, 1d	Direction of the latest selected-timeframe RSI trend. Calculation: buy if RSI > 55, sell if RSI < 45, otherwise neutral. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	buy / sell / neutral
`ATR-EMA-14` Volatility	1m, 5m, 1d	1m-ATR-EMA-14, 5m-ATR-EMA-14: 14-period Average True Range (EMA) on selected-timeframe candles. Calculation: EMA(14) of True Range using high/low and previous close on selected-timeframe candles. 1d-ATR-EMA-14: 14-period ATR using EMA smoothing on selected-timeframe candles. Calculation: EMA ATR(14) of True Range on selected-timeframe candles. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price range
`NATR-EMA-14` Volatility	1m, 5m, 1d	1m-NATR-EMA-14: 14-period normalized ATR on selected-timeframe candles (ATR / current price * 100). Calculation: selected timeframe-ATR-EMA-14 / CurrentPrice * 100. 5m-NATR-EMA-14, 1d-NATR-EMA-14: Normalized EMA ATR(14) on selected-timeframe candles. Calculation: selected timeframe-ATR-EMA-14 / CurrentPrice * 100. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	percent
`ATR-Wilder-14` Volatility	1m, 5m, 1d	14-period ATR using Wilder smoothing on selected-timeframe candles. Calculation: Wilder ATR(14) of True Range on selected-timeframe candles. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price range
`NATR-Wilder-14` Volatility	1m, 5m, 1d	Normalized Wilder ATR(14) on selected-timeframe candles. Calculation: selected timeframe-ATR-Wilder-14 / CurrentPrice * 100. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	percent
`TradeCount` Position-state	State / composite; see definition	Number of executed trades for the symbol since plan start. Calculation: Starts at 0 and increments by 1 after each successful BUY or SELL execution. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	integer (0+)
`TimeSinceLastBuyMin` Position-state	State / composite; see definition	Minutes elapsed since the last executed BUY. Calculation: floor((now - last BUY timestamp) / 60), null if no BUY yet. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	minutes or null
`BarsSinceLastBuy-1m` Position-state	State / composite; see definition	Elapsed 1-minute bars since the last BUY. Calculation: Whole minutes since last BUY, null if no BUY yet. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	bars or null
`TimeSinceLastSellMin` Position-state	State / composite; see definition	Minutes elapsed since the last executed SELL. Calculation: floor((now - last SELL timestamp) / 60), null if no SELL yet. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	minutes or null
`BarsSinceLastSell-1m` Position-state	State / composite; see definition	Elapsed 1-minute bars since the last SELL. Calculation: Whole minutes since last SELL, null if no SELL yet. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	bars or null
`MACD` Trend	1m, 5m, 1d, 1w	1m-MACD, 5m-MACD: MACD line on selected-timeframe closes (12/26). Calculation: EMA(12) - EMA(26) over selected-timeframe close prices. 1d-MACD: MACD line on selected-timeframe candles (EMA12 - EMA26). Calculation: Difference between EMA(12) and EMA(26) on daily close prices. 1w-MACD: MACD line on selected-timeframe candles (EMA12 - EMA26). Calculation: Difference between EMA(12) and EMA(26) on weekly close prices. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	oscillator; MACD value
`MACD-signal` Trend	1m, 5m, 1d, 1w	1m-MACD-signal, 5m-MACD-signal: MACD signal line on selected-timeframe closes (9). Calculation: EMA(9) of the selected-timeframe MACD line. 1d-MACD-signal: Signal line for the daily MACD. Calculation: EMA(9) of the daily MACD line. 1w-MACD-signal: Signal line for the weekly MACD. Calculation: EMA(9) of the weekly MACD line. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	oscillator; signal value
`MACD-hist` Trend	1m, 5m, 1d, 1w	1m-MACD-hist, 5m-MACD-hist: MACD histogram on selected-timeframe closes. Calculation: MACD line minus signal line on selected-timeframe closes. 1d-MACD-hist: Histogram value for the daily MACD setup. Calculation: Daily MACD line minus daily signal line. 1w-MACD-hist: Histogram value for the weekly MACD setup. Calculation: Weekly MACD line minus weekly signal line. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	oscillator; histogram value
`BB-upper` Volatility	1m, 5m, 1d	1m-BB-upper, 5m-BB-upper: Upper Bollinger Band on selected-timeframe closes (20, 2σ). Calculation: SMA(20) + 2 * standard deviation over selected-timeframe closes. 1d-BB-upper: Upper Bollinger Band on selected-timeframe candles. Calculation: Daily SMA(20) plus 2 standard deviations. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price
`BB-lower` Volatility	1m, 5m, 1d	1m-BB-lower, 5m-BB-lower: Lower Bollinger Band on selected-timeframe closes (20, 2σ). Calculation: SMA(20) - 2 * standard deviation over selected-timeframe closes. 1d-BB-lower: Lower Bollinger Band on selected-timeframe candles. Calculation: Daily SMA(20) minus 2 standard deviations. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price
`BB-width` Volatility	1m, 5m, 1d, 1w	1m-BB-width, 5m-BB-width: Bollinger Band width on selected-timeframe closes (20, 2σ). Calculation: Upper band minus lower band on selected-timeframe closes. 1d-BB-width: Bollinger Band width on selected-timeframe candles. Calculation: (Upper band - Lower band) / Middle band * 100 on daily data. 1w-BB-width: Bollinger Band width on selected-timeframe candles. Calculation: (Upper band - Lower band) / Middle band * 100 on weekly data. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price range; percent
`TrendEfficiency-24` Trend	5m	Directionality of the last 24 five-minute closes; high values indicate a persistent trend, while low values indicate choppy sideways movement. Calculation: Absolute net price change divided by the sum of all absolute close-to-close changes over 24 five-minute periods, multiplied by 100. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	0 to 100 (higher means more directional)
`EMA-20` Trend	1m, 5m, 1d, 1w	1m-EMA-20, 5m-EMA-20, 1d-EMA-20: 20-period EMA calculated on selected-timeframe closes. Calculation: EMA(20) over selected-timeframe close prices. 1w-EMA-20: 20-period EMA calculated on selected-timeframe closes. Calculation: EMA(20) over weekly close prices using the standard exponential smoothing factor. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price
`EMA-50` Trend	1m, 5m, 1d, 1w	1m-EMA-50, 5m-EMA-50, 1d-EMA-50: 50-period EMA calculated on selected-timeframe closes. Calculation: EMA(50) over selected-timeframe close prices. 1w-EMA-50: 50-period EMA calculated on selected-timeframe closes. Calculation: EMA(50) over weekly close prices using the standard exponential smoothing factor. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price
`EMA-100` Trend	1m, 5m, 1d	100-period EMA calculated on selected-timeframe closes. Calculation: EMA(100) over selected-timeframe close prices. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price
`EMA-200` Trend	1m, 5m, 1d	200-period EMA calculated on selected-timeframe closes. Calculation: EMA(200) over selected-timeframe close prices. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price
`BB-middle` Volatility	1d	Middle Bollinger Band on 1-day candles. Calculation: Daily SMA(20) of close prices. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price
`BB-width-Percentile-70` Volatility	1d	70th percentile threshold of daily Bollinger Band width. Calculation: Rolling 70th percentile of daily Bollinger Band width values. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	percent threshold
`SMA-200` Trend	1d	200-period simple moving average on 1-day closes. Calculation: Average of the last 200 daily close prices. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	price
`MarketRegime-1d` Trend	1d	Deterministic daily market regime (bull/sideways/bear). Calculation: EMA50 vs EMA200 + MACD histogram sign. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	bull / sideways / bear
`TrendStrength-1d` Trend	1d	Daily trend-strength score from 0 to 100. Calculation: EMA structure + MACD momentum + NATR volatility scoring. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	0-100
`StopLoss` Position-state	10m, 20m, 30m	StopLoss drop magnitude over the last selected-timeframe. Calculation: Returns abs(drop) in percent points when the window is down; formula thresholds decide the trigger. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	0 or percent drop
`composit-StopLoss` Position-state	State / composite; see definition	Composite StopLoss flag based on 10m/20m/30m percent thresholds. Calculation: True when all stoploss thresholds are exceeded. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	true / false
`market-action` Market	State / composite; see definition	Market-level action signal aggregated across symbols. Calculation: BUY/SELL if at least 70% of symbols agree; otherwise NEUTRAL. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	BUY / SELL / NEUTRAL
`market-RSI-trend` Market	State / composite; see definition	Market-level RSI trend aggregated across symbols. Calculation: BUY/SELL if at least 70% of RSI trends agree; otherwise NEUTRAL. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	BUY / SELL / NEUTRAL
`market-StopLoss` Market	State / composite; see definition	Market-level StopLoss average across symbols. Calculation: Average of per-symbol StopLoss values. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	average percent drop
`market-composit-StopLoss` Market	State / composite; see definition	Market-level composite StopLoss flag. Calculation: YES when at least 40% of symbols are in composite StopLoss. Missing values: Treat missing or stale values as unavailable evidence; never coerce them to zero.	YES / NEUTRAL

Reference source: Run snapshot

Strategy 934 — dedicated CONVERT relative-momentum rotation remains the strongest measured result

Strategy 934 · version 1

Formula and indicators

The strategy is internally coherent: BUY is selective but has a first-entry path, SELL is independent of entry price to support rotation, and StopLoss is a profit-peak guard. The latest prior change slightly relaxed non-bear and first-entry BUY thresholds after active-revision inactivity, while preserving EMA, MACD, RSI trend, volatility and cooldown gates.

BUY formula and interpretation

BUY
((MarketRegime-1d != bear AND 5m-TrendEfficiency-24 >= 36 AND TrendStrength-1d >= 48 AND CurrentPrice > 1d-EMA-50 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-EMA-50 > 5m-EMA-100 AND 5m-RSI >= 51 AND 5m-RSI <= 70 AND 1m-2 > 0.06 AND 1m-10 > 0.26 AND 1m-20 > 0.48) OR (MarketRegime-1d = bear AND 5m-TrendEfficiency-24 >= 42 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-RSI >= 48 AND 5m-RSI <= 62 AND 1m-2 > 0.12 AND 1m-10 > 0.38 AND 1m-20 > 0.65) OR (TradeCount = 0 AND 5m-TrendEfficiency-24 >= 28 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-RSI >= 48 AND 5m-RSI <= 68 AND 1m-2 > 0.06 AND 1m-10 > 0.20 AND 1m-20 > 0.34)) AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI-trend = buy AND 5m-NATR-Wilder-14 < 5m-NATR-EMA-14 * 1.55 AND (TimeSinceLastSellMin = null OR TimeSinceLastSellMin > 45)

The BUY formula is explicitly built as a multi-branch momentum admission gate for CONVERT rotation. The first branch applies when “MarketRegime-1d != bear” and requires 5m-TrendEfficiency-24 at least 36, daily TrendStrength at least 48, current price above 1d-EMA-50, current price above 5m-EMA-20, and a stacked 5m EMA structure where EMA-20 exceeds EMA-50 and EMA-50 exceeds EMA-100. It then requires 5m-RSI between 51 and 70, positive 1m momentum over two, ten and twenty minutes, positive 5m and 1m MACD histograms, 5m-RSI-trend equal to buy, a volatility cap where 5m-NATR-Wilder-14 is less than 1.55 times 5m-NATR-EMA-14, and a 45-minute post-sell cooldown unless no sell exists. The second branch allows bear-regime entries, but only with stricter intraday trend efficiency and momentum thresholds. The third branch is a first-trade path using TradeCount = 0 and slightly broader thresholds. This means a falling asset cannot enter simply because it is cheap; it must show EMA alignment, MACD confirmation and short-term momentum.

SELL formula and interpretation

SELL
((5m-TrendEfficiency-24 >= 34 AND CurrentPrice < 5m-EMA-20 AND 5m-EMA-20 < 5m-EMA-50 AND 5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 1m-2 < -0.10 AND 1m-10 < -0.32 AND 1m-20 < -0.55) OR (CurrentPrice < 5m-EMA-50 AND 5m-EMA-50 < 5m-EMA-100 AND 5m-RSI-trend = sell AND 5m-RSI < 45 AND 5m-MACD-hist < 0 AND 1m-10 < -0.25) OR (MarketRegime-1d = bear AND CurrentPrice < 1d-EMA-20 AND 1d-MACD-hist < 0 AND CurrentPrice < 5m-EMA-50 AND 5m-MACD-hist < 0 AND 1m-MACD-hist < 0) OR (5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 2.20 AND CurrentPrice < 5m-BB-lower AND 1m-2 < -0.35 AND 1m-10 < -0.70 AND 5m-MACD-hist < 0))

The SELL formula deliberately avoids a ProfitSinceBuyPct gate because this is a rotation strategy, not a cash-exit profit taker. It sells when deterioration is sufficiently confirmed. The first branch requires directional 5m trend efficiency, current price below 5m-EMA-20, EMA-20 below EMA-50, both 5m and 1m MACD histograms negative, and negative 1m momentum across two, ten and twenty minutes. The second branch sells when current price is below 5m-EMA-50, EMA-50 is below EMA-100, 5m-RSI-trend is sell, 5m-RSI is below 45, 5m-MACD-hist is negative and 1m-10 is below -0.25. The third branch addresses daily bear breakdown: MarketRegime-1d equals bear, price below 1d-EMA-20, 1d-MACD-hist below zero, price below 5m-EMA-50 and both intraday MACD histograms negative. The final branch captures volatility shock below the lower Bollinger band. These are deterioration filters rather than realized-profit filters.

STOPLOSS formula and interpretation

STOPLOSS
ProfitSinceBuyPct > 1.00 AND ((5m-NATR-Wilder-14 <= 5m-NATR-EMA-14 * 1.25 AND DrawdownFromHighPct > 0.95) OR (5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 1.25 AND 5m-NATR-Wilder-14 <= 5m-NATR-EMA-14 * 1.75 AND DrawdownFromHighPct > 1.35) OR (5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 1.75 AND DrawdownFromHighPct > 1.85)) AND 5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND (5m-RSI-trend = sell OR 1m-10 < -0.45)

The StopLoss formula is not a conventional stop at an entry loss. It starts with “ProfitSinceBuyPct > 1.00,” so it only protects a position that has already moved at least one percent into profit. It then uses volatility-tiered drawdown-from-high thresholds: a tighter 0.95% peak drawdown when 5m-NATR-Wilder-14 is no more than 1.25 times the EMA version, 1.35% when volatility is moderately elevated, and 1.85% when volatility is high. The exit also requires both 5m and 1m MACD histograms below zero and either 5m-RSI-trend equal to sell or 1m-10 below -0.45. This construction tries to avoid dumping a profitable holding on a normal pullback in volatile conditions, while still allowing a temporary cash move if a profitable peak breaks down. Most plans have StopLoss disabled, so evidence remains limited.

Measured performance

This is the strongest measured strategy in the supplied packet. Over 7 days it had 7 trades, 66.6667% win rate, 6.5858 profit factor, 2.857% convert increase and 7.7755% portfolio profit across five active plans. Over 30 days it had 52 trades, 72.7273% win rate, 4.9308 profit factor, 6.1074% convert increase and 8.4985% portfolio profit across six active plans. Over 90 days it had 59 trades, 74.0741% win rate, 5.0601 profit factor and 6.0862% convert increase. The recent execution diagnostics are smaller: one confirmed switch in plan 71 produced 1.271247% convert increase, and the target outperformed the source at 15, 60 and 180 minutes. The sample is too small to generalize all switch behavior, but it supports the current rotation thesis.

Window	Trades	Profit	Profit factor	Win rate	CONVERT increase	Turnover
7d	7	156.89 $	6.59×	66.67%	2.86%	4,455.18 $
30d	52	207.54 $	4.93×	72.73%	6.11%	20,413.48 $
90d	59	200.42 $	5.06×	74.07%	6.09%	21,725.96 $

Market alignment

The market was well suited to a relative-momentum rotation system: majors were positive over 90 days, but the broader symbol universe was mostly sideways. That environment rewards staying in relatively stronger assets without requiring every symbol to be in a clean daily bull state. Strategy 934’s formulas explicitly match that setting by allowing bear-market rotation only under stricter intraday confirmation and by avoiding a normal fixed-profit sell gate.

Problems found

Active switch diagnostics are small, so target-versus-source outperformance is promising but not yet robust.
Some plans have long periods without trades, creating a risk that the formula becomes too selective in choppy markets.
StopLoss evidence is sparse because only one current plan has StopLoss enabled.
The strategy lacks supplied best-alternative counterfactual data, so lock-in risk cannot be fully measured.

Adaptive change and expected behavior

The prior optimization status was insufficient evidence in the latest top-level packet, but the active formula already reflects an earlier conservative BUY relaxation. That change was justified by zero formula events over roughly 125 plan-hours under the previous active revision while the prior baseline averaged switches roughly every 54 hours. The adjustment lowered selected non-bear and first-entry trend-efficiency and momentum thresholds while retaining core quality filters. SELL and StopLoss were unchanged because there was no normal SELL evidence and prior StopLoss evidence warned about early exit risk.

Expected behavior is moderately more rotation opportunity without allowing unconfirmed falling assets. The formula should still wait for aligned EMAs, positive MACD and RSI trend confirmation. Future review should test hypothesis H-2026-W40-04: whether modest BUY relaxation restores switch frequency while preserving 60-minute and 180-minute target-minus-source performance.

Strategy 935 — high activity, mixed BUY/SELL economics and a conservative SELL adjustment

Strategy 935 · version 1

Formula and indicators

This strategy is a BUY/SELL recovery-profit-guard system with one small CONVERT cohort. BUY is strict and cooldown-based; SELL protects profit but has now been widened; StopLoss is a separate deep-loss emergency layer. The formula is more suitable for trusted assets than for broad high-volatility rotation.

BUY formula and interpretation

BUY
(TimeSinceLastSellMin = null OR TimeSinceLastSellMin > 180) AND (TimeSinceLastBuyMin = null OR TimeSinceLastBuyMin > 180) AND MarketRegime-1d != bear AND CurrentPrice > 1d-EMA-50 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-EMA-50 > 5m-EMA-100 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI-trend = buy AND 5m-NATR-Wilder-14 < 5m-NATR-EMA-14 * 1.30 AND (((LastSellPrice != null AND CurrentPrice <= LastSellPrice * 99.5%) AND 5m-TrendEfficiency-24 >= 40 AND TrendStrength-1d >= 50 AND 5m-RSI >= 50 AND 5m-RSI <= 63 AND 1m-2 > 0.10 AND 1m-10 > 0.34 AND 1m-20 > 0.62) OR ((LastSellPrice = null OR CurrentPrice > LastSellPrice * 99.5%) AND 5m-TrendEfficiency-24 >= 44 AND TrendStrength-1d >= 54 AND 5m-RSI >= 54 AND 5m-RSI <= 67 AND 1m-2 > 0.14 AND 1m-10 > 0.42 AND 1m-20 > 0.78))

The BUY formula starts with two anti-churn gates: TimeSinceLastSellMin must be null or above 180 minutes, and TimeSinceLastBuyMin must be null or above 180 minutes. It then excludes daily bear regimes, requires current price above 1d-EMA-50, current price above 5m-EMA-20, 5m-EMA-20 above 5m-EMA-50, and 5m-EMA-50 above 5m-EMA-100. It also requires both 5m and 1m MACD histograms above zero, 5m-RSI-trend equal to buy, and 5m-NATR-Wilder-14 less than 1.30 times 5m-NATR-EMA-14. The recovery branch applies when LastSellPrice exists and price is no higher than 99.5% of it; it then demands 5m-TrendEfficiency-24 at least 40, TrendStrength-1d at least 50, 5m-RSI between 50 and 63 and positive 1m-2, 1m-10 and 1m-20 thresholds. The normal branch is stricter: trend efficiency 44, trend strength 54, RSI 54 to 67 and stronger short-term momentum. This is a selective non-bear trend-recovery formula.

SELL formula and interpretation

SELL
((ProfitSinceBuyPct >= 0.95 AND ((5m-NATR-Wilder-14 <= 5m-NATR-EMA-14 * 1.25 AND DrawdownFromHighPct > 0.80) OR (5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 1.25 AND 5m-NATR-Wilder-14 <= 5m-NATR-EMA-14 * 1.75 AND DrawdownFromHighPct > 1.15) OR (5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 1.75 AND DrawdownFromHighPct > 1.60)) AND 5m-MACD-hist < 0 AND 1m-MACD-hist < 0) OR (5m-TrendEfficiency-24 >= 42 AND CurrentPrice < 5m-EMA-50 AND 5m-EMA-20 < 5m-EMA-50 AND 5m-EMA-50 < 5m-EMA-100 AND 5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 5m-RSI-trend = sell AND 1m-2 < -0.20 AND 1m-10 < -0.55 AND 1m-20 < -0.90) OR (MarketRegime-1d = bear AND CurrentPrice < 1d-BB-lower AND 5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 1.80 AND 5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 1m-10 < -0.65))

The applied SELL formula was changed after the reporting cutoff to relax only the normal profit-guard branch. It now requires ProfitSinceBuyPct at least 0.95 before the profit-guard exits can trigger. The drawdown-from-high thresholds are tiered by volatility: 0.80 when 5m NATR is near or below its EMA reference, 1.15 when moderately elevated, and 1.60 when high. That branch also requires both 5m and 1m MACD histograms below zero. The second SELL branch remains a confirmed intraday breakdown: trend efficiency at least 42, price below 5m-EMA-50, EMA-20 below EMA-50, EMA-50 below EMA-100, both MACD histograms negative, RSI trend sell and strongly negative one-, ten- and twenty-minute momentum. The third branch is a bear-market extreme condition requiring price below 1d-BB-lower, high volatility and negative MACD. The change therefore delays profit protection but keeps danger exits intact.

STOPLOSS formula and interpretation

STOPLOSS
ProfitSinceBuyPct <= -3.0 AND (CurrentPrice <= LastBuyPrice * 95.0% OR (ProfitSinceBuyPct <= -3.25 AND CurrentPrice < 5m-BB-lower AND CurrentPrice < 5m-EMA-100 AND 5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 5m-RSI-trend = sell AND 1m-2 < -0.50 AND 1m-10 < -1.00 AND 1m-20 < -1.50) OR (5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 3.00 AND CurrentPrice < 5m-BB-lower AND 5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 1m-2 < -0.70 AND 1m-10 < -1.30 AND 1m-20 < -1.85))

The StopLoss formula is intentionally loss-tolerant compared with tight scalping stops. It first requires ProfitSinceBuyPct less than or equal to -3.0. A hard catastrophe branch exits if current price is at or below 95% of LastBuyPrice. A confirmed breakdown branch requires a loss of at least -3.25%, price below 5m-BB-lower and 5m-EMA-100, negative 5m and 1m MACD, RSI-trend sell, and strong negative 1m momentum. A volatility shock branch requires 5m-NATR-Wilder-14 above three times 5m-NATR-EMA-14, price below 5m-BB-lower and strongly negative momentum. This design avoids repeated shallow stop-outs while retaining protection for deeper or confirmed falls. In the latest active evidence, StopLoss events were too sparse to justify further modification.

Measured performance

Aggregate results are mixed. Across all plans, the 7-day window had 138 trades, 62.3188% win rate, 0.739 profit factor and -0.7758% portfolio profit. The 30-day window had 630 trades, 68.2594% win rate, profit factor 0.9987 and -0.0647% portfolio profit. The 90-day window was near breakeven with profit factor 1.0044 but still slightly negative portfolio profit. BUY/SELL dominates the sample and underperformed: 30-day BUY/SELL profit factor was 0.993. The single CONVERT plan looked much better with 30-day profit factor 14.6769 and 28.0716% portfolio profit, but that was only six trades and should not override the larger BUY/SELL evidence. Revision KPIs show the prior active change had 59 trades and profit factor 2.7587, but the aggregate daily windows show current broader weakness.

Window	Trades	Profit	Profit factor	Win rate	CONVERT increase	Turnover
7d	138	-23.03 $	0.74×	62.32%	6.43%	8,553.55 $
30d	630	-1.91 $	1×	68.26%	6.43%	34,188.13 $
90d	697	-3.9 $	1×	69.11%	6.43%	35,726.15 $

Market alignment

The formula’s non-bear daily regime requirement aligns with a market where majors are broadly positive. However, the broader CoinRobot universe was mostly sideways, so the strategy’s high trade count can still encounter chop. The evidence suggests the entry filter is not the only problem; exit timing and loss size matter because high win rate did not translate into profit factor.

Problems found

High win rate did not prevent near-zero or negative profit factor in aggregate windows.
SELL evidence showed high early-exit and quick-recovery behavior, especially in active-revision diagnostics.
BUY/SELL and CONVERT evidence are very uneven in sample size, making shared-formula conclusions difficult.
Some plans had no recent trades or inactive status, reducing clarity in portfolio-level interpretation.

Adaptive change and expected behavior

The latest applied change left BUY and StopLoss unchanged and raised the normal profit-guard SELL requirement from 0.85 to 0.95 while widening drawdown triggers. This was justified because active BUY evidence was acceptable, StopLoss evidence was sparse, and SELL exits still showed high early-exit and recovery behavior. Confirmed breakdown and bear-market SELL branches were preserved to avoid removing risk control.

Expected behavior is fewer premature profit-guard exits, with positions allowed slightly more room before profit-taking. The risk is that delayed exits can also give back more profit in fast reversals. Hypothesis H-2026-W40-01 remains the relevant test: profit factor, bad-buy rate, trade count and average loss size must improve after enough new BUY/SELL events.

Strategy 932 — strong one-plan CONVERT evidence but validation history demands restraint

Strategy 932 · version 1

Formula and indicators

This is a flexible active-momentum formula with wide entry coverage and a later, confirmation-focused StopLoss. The earlier validation-blocked attempt is important: an AI output failed because a proposed BUY formula was always false. The current applied change is StopLoss-only and passed validation.

BUY formula and interpretation

BUY
(LastSellPrice = null OR CurrentPrice <= LastSellPrice * 1.003 OR TimeSinceLastSellMin > 360) AND (5m-TrendEfficiency-24 >= 35 AND ((action = buy AND ((TimeSinceLastSellMin = null OR TimeSinceLastSellMin > 90) OR (MarketRegime-1d = bull AND 5m-TrendEfficiency-24 >= 24 AND TrendStrength-1d >= 68 AND 1m-10 > 0.65 AND 1m-20 > 1.05 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-MACD-hist > 0)) AND (HighestSinceSell = null OR CurrentPrice < HighestSinceSell * 99.35% OR (MarketRegime-1d = bull AND 5m-TrendEfficiency-24 >= 24 AND TrendStrength-1d >= 70 AND 1m-10 > 0.75 AND 1m-20 > 1.20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-EMA-50 > 5m-EMA-100)) AND ((MarketRegime-1d = bull AND 5m-TrendEfficiency-24 >= 24 AND TrendStrength-1d >= 56 AND CurrentPrice > 1d-EMA-50 AND 1d-MACD-hist > -0.05 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 1m-2 > 0.08 AND 1m-10 > 0.24 AND 1m-20 > 0.03 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI > 50 AND 5m-RSI < 67 AND 5m-NATR-Wilder-14 < 5m-NATR-EMA-14 * 1.35) OR (MarketRegime-1d = sideways AND 5m-TrendEfficiency-24 >= 40 AND TrendStrength-1d >= 38 AND TrendStrength-1d <= 58 AND CurrentPrice > 1d-BB-middle AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 >= 5m-EMA-50 AND 1m-2 > 0.06 AND 1m-10 > 0.22 AND 1m-20 > -0.05 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI > 49 AND 5m-RSI < 63 AND CurrentPrice < 5m-BB-upper) OR (MarketRegime-1d = bear AND TrendStrength-1d <= 30 AND 1d-RSI < 34 AND CurrentPrice > 1d-BB-lower AND CurrentPrice > 5m-EMA-20 AND 1m-2 > 0.10 AND 1m-10 > 0.32 AND 1m-20 > 0.10 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI > 47 AND 5m-RSI < 57)) AND (TimeSinceLastBuyMin = null OR TimeSinceLastBuyMin > 120)) OR (MarketRegime-1d != bear AND (5m-TrendEfficiency-24 >= 30 OR 1m-20 > 1.20) AND TrendStrength-1d >= 45 AND CurrentPrice > 1d-EMA-50 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 1m-2 > 0.10 AND 1m-10 > 0.45 AND 1m-20 > 0.90 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI >= 55 AND 5m-RSI <= 74 AND 5m-NATR-Wilder-14 < 5m-NATR-EMA-14 * 1.50 AND (TimeSinceLastBuyMin = null OR TimeSinceLastBuyMin > 120))) OR (MarketRegime-1d != bear AND (5m-TrendEfficiency-24 >= 30 OR 1m-20 > 1.20) AND TrendStrength-1d >= 42 AND CurrentPrice > 1d-EMA-50 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-EMA-50 > 5m-EMA-100 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 1m-2 > 0.08 AND 1m-10 > 0.30 AND 1m-20 > 0.65 AND 5m-RSI >= 53 AND 5m-RSI <= 78 AND 5m-NATR-Wilder-14 < 5m-NATR-EMA-14 * 1.65 AND (TimeSinceLastBuyMin = null OR TimeSinceLastBuyMin > 90)))

The BUY formula is broad and multi-branch. It first avoids chasing materially higher re-entry by requiring LastSellPrice to be null, current price no more than 0.3% above LastSellPrice, or TimeSinceLastSellMin above 360. It then requires 5m-TrendEfficiency-24 at least 35. The first major branch depends on action = buy, a 90-minute post-sell wait unless a strong bull continuation exception applies, and a HighestSinceSell rule that blocks buying near a higher post-sell high unless trend strength and momentum are very strong. Inside that branch, bull entries require daily bull regime, trend efficiency, trend strength, price above 1d-EMA-50, daily MACD not meaningfully negative, intraday EMA alignment, positive 1m momentum, positive MACD, RSI between 50 and 67 and a volatility cap. Sideways entries require price above 1d-BB-middle, intraday EMA support, positive momentum, MACD and RSI in a narrower range. Bear entries are only selective rebound entries with weak daily trend strength, low daily RSI and positive intraday recovery. Two additional non-bear continuation branches allow participation in accelerating moves when momentum is strong.

SELL formula and interpretation

SELL
action = sell AND ProfitSinceBuyPct > 0.35 AND (((ProfitSinceBuyPct > 0.65 AND DrawdownFromHighPct > 0.50) OR (ProfitSinceBuyPct > 1.35 AND DrawdownFromHighPct > 0.34) OR (ProfitSinceBuyPct > 2.75 AND DrawdownFromHighPct > 0.26)) OR (5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 1m-10 < -0.18 AND DrawdownFromHighPct > 0.20) OR (CurrentPrice < 5m-EMA-50 AND 1m-10 < -0.25) OR (5m-RSI > 78 AND DrawdownFromHighPct > 0.35))

The SELL formula requires action = sell and ProfitSinceBuyPct greater than 0.35. That means normal SELL is blocked when profit is too small, even if some deterioration exists. Once the profit gate is met, the first group exits on profit drawdown tiers: more than 0.65% profit and 0.50% drawdown, more than 1.35% profit and 0.34% drawdown, or more than 2.75% profit and 0.26% drawdown. A second branch exits on negative 5m and 1m MACD, 1m-10 below -0.18 and drawdown above 0.20. A third exits if price falls below 5m-EMA-50 with 1m-10 below -0.25. A fourth exits when 5m-RSI exceeds 78 and drawdown exceeds 0.35. These rules combine profit protection, momentum deterioration, EMA loss and overbought reversal logic.

STOPLOSS formula and interpretation

STOPLOSS
(ProfitSinceBuyPct <= -2.5 AND ((CurrentPrice < LastBuyPrice * 94.8%) OR (CurrentPrice < LastBuyPrice * 95.2% AND 1m-10 < -0.45 AND 5m-MACD-hist < 0) OR (MarketRegime-1d = bear AND CurrentPrice < 5m-BB-lower AND CurrentPrice < 5m-EMA-100 AND 1m-2 < -0.55 AND 1m-10 < -1.00 AND 1m-20 < -1.50 AND 1m-MACD-hist < 0 AND 5m-MACD-hist < 0) OR (5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 2.90 AND CurrentPrice < 5m-BB-lower AND 1m-2 < -0.70 AND 1m-10 < -1.20 AND 1m-MACD-hist < 0 AND 5m-MACD-hist < 0))) OR (ProfitSinceBuyPct > 0.5 AND DrawdownFromHighPct > 1.45 AND 1m-MACD-hist < 0 AND 5m-MACD-hist < 0 AND (1m-10 < -0.20 OR CurrentPrice < 5m-EMA-50))

The applied StopLoss formula is a conservative revision. It starts with ProfitSinceBuyPct <= -2.5 and then provides several emergency branches. A hard severe-loss branch exits below 94.8% of LastBuyPrice. A less severe 95.2% branch now requires additional weakness: 1m-10 below -0.45 and 5m-MACD-hist below zero. Bear-regime and volatility-breakdown branches require price below the lower Bollinger band, price below 5m-EMA-100 in the bear branch, strongly negative 1m momentum and negative MACD. A profitable-position protection branch exits when ProfitSinceBuyPct is above 0.5, drawdown from high exceeds 1.45, both MACD histograms are negative, and either 1m-10 is below -0.20 or price is below 5m-EMA-50. This was designed to reduce wick-driven stops while keeping emergency exits.

Measured performance

The headline thirty-day CONVERT performance is impressive but concentrated. The aggregate thirty-day window shows 18 trades, 85.7143% win rate, 2.3102 profit factor, 87.3735% convert increase and 16.9747% portfolio profit, entirely from one active plan. The 7-day window had no trades but still had positive portfolio movement. The 90-day aggregate is mixed: 368 trades, negative portfolio profit of -11.3438%, and weak or incomplete coverage for win-rate and profit-factor fields. Execution diagnostics show 23 CONVERT events, one plan, one incomplete switch pair, eight stop-loss events and no paired target-versus-source return evidence. Therefore, the strong convert increase is real in the supplied data, but attribution to switch quality is incomplete.

Window	Trades	Profit	Profit factor	Win rate	CONVERT increase	Turnover
7d	0	8.17 $	—	—	0%	0 $
30d	18	50.34 $	2.31×	85.71%	87.37%	5,654.6 $
90d	368	-283.47 $	0.88×	50%	20.88%	30,268.9 $

Market alignment

The broad BUY formula can exploit a market where majors are rising and some symbols accelerate, but the sideways-heavy broader universe makes broad entry coverage risky. The selective bear-rebound branch and anti-chase rules help, yet the strategy remains more opportunistic than Strategy 933. That is useful in fast momentum but more exposed to false starts.

Problems found

Thirty-day strength is concentrated in one active CONVERT plan, so generalization is limited.
Ninety-day aggregate performance remains negative at portfolio level despite the recent positive CONVERT signal.
Switch diagnostics lack complete paired execution evidence for target-versus-source assessment.
Prior validation failure shows that AI-generated formula changes can be invalid and must remain blocked when scenario tests fail.

Adaptive change and expected behavior

The applied change tightened StopLoss confirmation after evidence of bounce risk: 8 StopLoss exits, 75% early-exit rate, 50% near the next 60-minute low and 75% recovery within five minutes in prior evidence. BUY and SELL were not changed because formula-true entries were not clearly bad and normal SELL exits were absent in the relevant evidence.

Expected behavior is fewer wick-driven StopLoss exits, especially around intermediate loss thresholds, while hard emergency exits remain available. Future evidence should track whether recovery rates after StopLoss fall without increasing maximum adverse excursion. Because one prior optimization was validation-blocked, all future AI changes should continue to be scenario-tested before activation.

Strategy 933 — selective BUY/SELL remains profitable over 30 days but recent StopLoss evidence is problematic

Strategy 933 · version 1

Formula and indicators

Strategy 933 is a selective non-bear BUY/SELL strategy. It trades frequently but with cleaner entry requirements than Strategy 932. Its latest active revision applies a StopLoss-only change, leaving BUY and normal SELL intact to avoid stacking edits.

BUY formula and interpretation

BUY
(LastSellPrice = null OR CurrentPrice <= LastSellPrice * 1.003 OR TimeSinceLastSellMin > 360) AND MarketRegime-1d != bear AND 5m-TrendEfficiency-24 >= 36 AND TrendStrength-1d >= 48 AND CurrentPrice > 1d-EMA-50 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-NATR-Wilder-14 < 5m-NATR-EMA-14 * 1.35 AND 5m-RSI >= 53 AND 5m-RSI <= 68 AND (TimeSinceLastBuyMin = null OR TimeSinceLastBuyMin > 60) AND ((action = buy AND (TimeSinceLastSellMin = null OR TimeSinceLastSellMin > 60) AND 1m-2 > 0.08 AND 1m-10 > 0.28 AND 1m-20 > 0.20 AND (HighestSinceSell = null OR CurrentPrice < HighestSinceSell * 99.55% OR (TrendStrength-1d >= 68 AND 1m-20 > 0.95))) OR (TrendStrength-1d >= 56 AND 5m-EMA-50 > 5m-EMA-100 AND 1m-2 > 0.10 AND 1m-10 > 0.42 AND 1m-20 > 0.82 AND 5m-RSI <= 66))

The BUY formula is a cleaner, stricter cousin of Strategy 932. It begins with the same anti-chase idea: LastSellPrice must be null, current price must be no more than 0.3% above LastSellPrice, or TimeSinceLastSellMin must exceed 360. It then excludes bear regimes outright with MarketRegime-1d != bear. It requires 5m-TrendEfficiency-24 at least 36, TrendStrength-1d at least 48, current price above 1d-EMA-50, price above 5m-EMA-20, 5m-EMA-20 above 5m-EMA-50, both 5m and 1m MACD histograms positive, 5m NATR below 1.35 times its EMA reference, and 5m-RSI between 53 and 68. It also requires TimeSinceLastBuyMin null or above 60. The first entry branch needs action = buy, a 60-minute post-sell wait, positive 1m-2, 1m-10 and 1m-20, plus HighestSinceSell discipline. The continuation branch requires stronger trend strength, EMA-50 above EMA-100 and stronger momentum.

SELL formula and interpretation

SELL
action = sell AND ProfitSinceBuyPct >= 0.65 AND (((ProfitSinceBuyPct >= 0.65 AND DrawdownFromHighPct > 0.50) OR (ProfitSinceBuyPct >= 1.20 AND DrawdownFromHighPct > 0.38) OR (ProfitSinceBuyPct >= 2.20 AND DrawdownFromHighPct > 0.30)) OR (5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 1m-10 < -0.12) OR (CurrentPrice < 5m-EMA-50 AND 1m-10 < -0.18) OR (5m-RSI > 74 AND DrawdownFromHighPct > 0.28))

The SELL formula requires action = sell and ProfitSinceBuyPct at least 0.65, so it does not normally exit before a minimum profit cushion. The first group uses profit drawdown tiers: at least 0.65% profit with drawdown above 0.50, at least 1.20% profit with drawdown above 0.38, or at least 2.20% profit with drawdown above 0.30. These thresholds are already wider than the original baseline. Additional branches exit when both 5m and 1m MACD histograms are negative with 1m-10 below -0.12, when price is below 5m-EMA-50 with 1m-10 below -0.18, or when 5m-RSI is above 74 and drawdown exceeds 0.28. Normal sells were left unchanged in the latest change because StopLoss evidence should not be used to tune ordinary SELL logic and normal negative-sell evidence was limited.

STOPLOSS formula and interpretation

STOPLOSS
(ProfitSinceBuyPct <= -2.5 AND (((CurrentPrice < LastBuyPrice * 95.2%) AND ((CurrentPrice < LastBuyPrice * 94.6%) OR (1m-2 < -0.45 AND 1m-10 < -0.80 AND 1m-20 < -1.10 AND 1m-MACD-hist < 0 AND 5m-MACD-hist < 0))) OR (MarketRegime-1d = bear AND CurrentPrice < 5m-BB-lower AND CurrentPrice < 5m-EMA-100 AND 1m-2 < -0.50 AND 1m-10 < -0.90 AND 1m-20 < -1.40 AND 1m-MACD-hist < 0 AND 5m-MACD-hist < 0) OR (5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 2.80 AND CurrentPrice < 5m-BB-lower AND 1m-2 < -0.65 AND 1m-10 < -1.10 AND 1m-MACD-hist < 0 AND 5m-MACD-hist < 0))) OR (ProfitSinceBuyPct > 0.5 AND DrawdownFromHighPct > 1.25 AND 1m-MACD-hist < 0 AND 5m-MACD-hist < 0)

The latest StopLoss formula tightens the intermediate loss branch. It requires ProfitSinceBuyPct <= -2.5 and then checks whether price is below 95.2% of LastBuyPrice. If so, it exits only if price is below 94.6% or if short-term weakness is confirmed by 1m-2 below -0.45, 1m-10 below -0.80, 1m-20 below -1.10 and both MACD histograms negative. Bear-regime and volatility-breakdown emergency branches are preserved: they require price below 5m-BB-lower, price below 5m-EMA-100 in the bear branch, negative 1m momentum and negative MACD. A profitable drawdown guard remains when ProfitSinceBuyPct is above 0.5, drawdown exceeds 1.25 and both MACD histograms are negative. This attempts to reduce immediate recovery stops without removing severe exits.

Measured performance

The medium-window evidence is favorable but recent evidence weakened. The 30-day aggregate shows 1521 trades, 78.4024% win rate, 1.3148 profit factor, realized result 51.0811 and 14.0474% portfolio profit across two active plans. The 90-day window remains positive with 1.1834 profit factor and 9.1368% portfolio profit. The 7-day window deteriorated: 227 trades, 78.4314% win rate but only 0.906 profit factor and -2.8411% portfolio profit. Plan-level evidence shows plan 69 performed much better than plan 68 over 30 days, while plan 68 was weak. Active execution analysis recorded 157 BUY/SELL events, showing enough activity for inference, but the latest StopLoss diagnostics remain a concern.

Window	Trades	Profit	Profit factor	Win rate	CONVERT increase	Turnover
7d	227	-13.72 $	0.91×	78.43%	—	4,884.59 $
30d	1,521	57.8 $	1.31×	78.4%	—	33,353.33 $
90d	2,156	39.29 $	1.18×	69.52%	—	46,916.08 $

Market alignment

The strategy aligns well with a rising but noisy market because it excludes daily bear regimes, requires daily trend strength, intraday EMA alignment and MACD confirmation. That selectivity should help when many symbols are sideways. However, the recent negative week suggests that even selective entries can be harmed by exit timing, stop placement or symbol-specific chop.

Problems found

Seven-day profit factor fell below one despite a high win rate, indicating adverse loss sizing or exit timing.
Enabled StopLoss cohort showed 12 StopLoss exits, all negative, with high early-exit and recovery rates.
Plan dispersion is large: the broad-symbol disabled plan was strong over 30 days while the enabled smaller plan struggled.
SELL early-exit evidence remains a watch item, though latest change targeted StopLoss only.

Adaptive change and expected behavior

The latest AI change was StopLoss-only. In the active revision, 12 StopLoss exits were all negative, 75% early, 58.333% near the next 60-minute low, and many recovered within one to five minutes. The change therefore required stronger short-term confirmation at the intermediate 95.2% loss threshold while preserving hard and emergency exits. This directly addresses H-2026-W40-02 but does not yet prove it.

Expected behavior is fewer StopLoss exits that immediately recover. The formula should still exit severe losses or confirmed breakdowns. The hypothesis remains unresolved until enough new StopLoss exits accumulate; normal SELL timing should be monitored separately because StopLoss evidence should not be overextended to ordinary exits.

Strategy 918 — safe-entry design has historical evidence only, so the StopLoss change is a hypothesis

Strategy 918 · version 1

Formula and indicators

Strategy 918 is a safe-entry BUY/SELL strategy with long cooldowns, strict daily and intraday trend gates and a one-percent normal profit exit. The latest change affects only StopLoss, not the high-confidence entry or normal SELL design.

BUY formula and interpretation

BUY
action = buy AND (TimeSinceLastSellMin = null OR TimeSinceLastSellMin > 360) AND (TimeSinceLastBuyMin = null OR TimeSinceLastBuyMin > 720) AND ((MarketRegime-1d = bull AND TrendStrength-1d >= 70 AND CurrentPrice > 1d-EMA-20 AND 1d-EMA-20 > 1d-EMA-50 AND 1d-MACD-hist > 0 AND 1d-RSI >= 52 AND 1d-RSI <= 68 AND CurrentPrice < 1d-BB-upper * 1.005 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-EMA-50 > 5m-EMA-100 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 1m-10 > 0.28 AND 5m-RSI >= 52 AND 5m-RSI <= 66 AND 5m-NATR-Wilder-14 < 5m-NATR-EMA-14 * 1.10) OR (MarketRegime-1d = sideways AND TrendStrength-1d >= 42 AND TrendStrength-1d <= 58 AND CurrentPrice > 1d-EMA-50 AND CurrentPrice > 1d-BB-middle AND CurrentPrice < 1d-BB-upper * 1.005 AND 1d-MACD-hist > 0 AND 1d-RSI >= 47 AND 1d-RSI <= 59 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-MACD-hist > 0 AND 1m-10 > 0.25 AND 5m-RSI >= 51 AND 5m-RSI <= 63))

The BUY formula is intentionally low-frequency and high-confidence. It requires action = buy, TimeSinceLastSellMin null or above 360, and TimeSinceLastBuyMin null or above 720. That creates a long cooldown after both exits and entries. The bull branch requires MarketRegime-1d = bull, TrendStrength-1d at least 70, current price above 1d-EMA-20, 1d-EMA-20 above 1d-EMA-50, positive 1d-MACD-hist, 1d-RSI between 52 and 68, and price below 1d-BB-upper times 1.005. It then requires current price above 5m-EMA-20, 5m-EMA-20 above 5m-EMA-50, 5m-EMA-50 above 5m-EMA-100, positive 5m and 1m MACD, 1m-10 above 0.28, 5m-RSI between 52 and 66 and very low volatility with NATR-Wilder below 1.10 times NATR-EMA. The sideways branch is less trend-strong but still requires daily and intraday support. Bear entries are not allowed.

SELL formula and interpretation

SELL
action = sell AND ProfitSinceBuyPct >= 1.0 AND (((ProfitSinceBuyPct > 1.0 AND DrawdownFromHighPct > 0.35) OR (ProfitSinceBuyPct > 1.6 AND DrawdownFromHighPct > 0.22)) OR 5m-RSI > 69 OR CurrentPrice > 5m-BB-upper OR (5m-MACD-hist < 0 AND 1m-MACD-hist < 0) OR CurrentPrice < 5m-EMA-50)

The SELL formula has a hard normal profit gate: action must equal sell and ProfitSinceBuyPct must be at least 1.0. Once that gate is met, it can exit when profit above 1.0 has a drawdown above 0.35, or profit above 1.6 has drawdown above 0.22. It can also exit if 5m-RSI exceeds 69, if current price is above 5m-BB-upper, if both 5m and 1m MACD histograms are negative, or if current price falls below 5m-EMA-50. This is a conservative normal-exit structure: it refuses to take ordinary sells below one percent profit, but once the profit gate is satisfied it responds to overextension, momentum weakening or EMA loss. Because there is no recent active trade evidence, the report cannot judge whether the one-percent gate is currently optimal.

STOPLOSS formula and interpretation

STOPLOSS
CurrentPrice < LastBuyPrice * 98.3% OR DrawdownFromHighPct > 3.1 OR (1m-2 < -0.60 AND 1m-10 < -1.05 AND 1m-20 < -1.45) OR (market-composit-StopLoss != NEUTRAL AND (1m-2 < -0.20 OR 1m-10 < -0.45 OR 5m-MACD-hist < 0))

The updated StopLoss formula widened the hard price threshold from 98.5% to 98.3% of LastBuyPrice, widened DrawdownFromHighPct from 2.9 to 3.1, tightened the pure momentum crash branch to require 1m-2 below -0.60, 1m-10 below -1.05 and 1m-20 below -1.45, and changed market-composit-StopLoss logic so that a non-neutral broad market stop must be accompanied by local weakness in 1m-2, 1m-10 or 5m-MACD-hist. This is a StopLoss-only change designed to avoid exiting solely because a broad market flag changes while the local symbol has not confirmed downside. It still preserves exits for deeper price loss, large drawdown and sharp local momentum decline.

Measured performance

Current evidence is stale. There were no active seven-day or thirty-day trades, and both plans are not executing. The aggregate ninety-day window shows 126 trades, 3016.8855 turnover and 0.1953% portfolio profit, but aggregate win-rate and profit-factor fields are unavailable. Revision KPIs provide a more useful historical view: 126 trades, 29 winning trades, 28 losing trades, 5 breakeven trades and profit factor 1.0793. Plan 23 had a modest positive ninety-day profit percentage, while plan 63 was negative. Because last trades were in late August and the analysis date is early October, the applied StopLoss change is not supported by fresh post-change activity.

Window	Trades	Profit	Profit factor	Win rate	CONVERT increase	Turnover
7d	0	0 $	—	—	—	0 $
30d	0	0 $	—	—	—	0 $
90d	126	0.58 $	—	—	—	3,016.89 $

Market alignment

The formula is selective enough for a market with strong majors but sideways breadth. It should avoid many choppy entries because it requires bull or structured sideways regimes, daily MACD support, RSI bands, EMA alignment and low volatility. The trade-off is inactivity: in fast-changing markets, long cooldowns and strict gates can miss opportunities.

Problems found

No seven-day or thirty-day active evidence is available, so current behavior cannot be measured.
Both plans are inactive, making the latest adaptive change a hypothesis rather than a live result.
Historical StopLoss evidence showed high recovery rates, but it is old relative to the analysis date.
Normal SELL effectiveness cannot be reassessed because recent trades are absent.

Adaptive change and expected behavior

Historical enabled StopLoss evidence showed 34 exits, early-exit rate around 76%, recovery above exit around 68% within one to two minutes and about 71% within five minutes. This justified widening the StopLoss layer and requiring local confirmation for broad market composite signals. BUY and SELL were left unchanged to preserve the safe-entry intent and one-percent profit gate.

If the strategy becomes active again, the revised StopLoss should reduce whipsaw exits caused by brief market-wide stress flags while still responding to deeper price losses or local momentum breakdown. The first review should require new enabled-plan StopLoss exits before judging success.

Strategy 930 — OKX fee-aware clone tightened continuation entries after weak historical entry evidence

Strategy 930 · version 1

Formula and indicators

Strategy 930 is an OKX-specific, fee-aware active-momentum clone. Its distinguishing feature is not just momentum selection but fee-aware sell thresholds. The latest change only tightened the continuation BUY branch, leaving SELL and StopLoss intact.

BUY formula and interpretation

BUY
(action = buy AND 5m-TrendEfficiency-24 >= 35 AND ((TimeSinceLastSellMin = null OR TimeSinceLastSellMin > 90) OR (MarketRegime-1d = bull AND 5m-TrendEfficiency-24 >= 24 AND TrendStrength-1d >= 68 AND 1m-10 > 0.65 AND 1m-20 > 1.05 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-MACD-hist > 0)) AND (HighestSinceSell = null OR CurrentPrice < HighestSinceSell * 99.35% OR (MarketRegime-1d = bull AND 5m-TrendEfficiency-24 >= 24 AND TrendStrength-1d >= 70 AND 1m-10 > 0.75 AND 1m-20 > 1.20 AND 5m-EMA-20 > 5m-EMA-50 AND 5m-EMA-50 > 5m-EMA-100)) AND ((MarketRegime-1d = bull AND 5m-TrendEfficiency-24 >= 24 AND TrendStrength-1d >= 56 AND CurrentPrice > 1d-EMA-50 AND 1d-MACD-hist > -0.05 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 1m-2 > 0.08 AND 1m-10 > 0.24 AND 1m-20 > 0.03 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI > 50 AND 5m-RSI < 67 AND 5m-NATR-Wilder-14 < 5m-NATR-EMA-14 * 1.35) OR (MarketRegime-1d = sideways AND 5m-TrendEfficiency-24 >= 40 AND TrendStrength-1d >= 38 AND TrendStrength-1d <= 58 AND CurrentPrice > 1d-BB-middle AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 >= 5m-EMA-50 AND 1m-2 > 0.06 AND 1m-10 > 0.22 AND 1m-20 > -0.05 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI > 49 AND 5m-RSI < 63 AND CurrentPrice < 5m-BB-upper) OR (MarketRegime-1d = bear AND TrendStrength-1d <= 30 AND 1d-RSI < 34 AND CurrentPrice > 1d-BB-lower AND CurrentPrice > 5m-EMA-20 AND 1m-2 > 0.10 AND 1m-10 > 0.32 AND 1m-20 > 0.10 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI > 47 AND 5m-RSI < 57)) AND (TimeSinceLastBuyMin = null OR TimeSinceLastBuyMin > 120)) OR (MarketRegime-1d != bear AND (5m-TrendEfficiency-24 >= 35 OR 1m-20 > 1.40) AND TrendStrength-1d >= 45 AND CurrentPrice > 1d-EMA-50 AND CurrentPrice > 5m-EMA-20 AND 5m-EMA-20 > 5m-EMA-50 AND 1m-2 > 0.10 AND 1m-10 > 0.45 AND 1m-20 > 0.90 AND 5m-MACD-hist > 0 AND 1m-MACD-hist > 0 AND 5m-RSI >= 55 AND 5m-RSI <= 72 AND 5m-NATR-Wilder-14 < 5m-NATR-EMA-14 * 1.50 AND (TimeSinceLastBuyMin = null OR TimeSinceLastBuyMin > 120))

The BUY formula has two main parts. The first part requires action = buy and 5m-TrendEfficiency-24 at least 35. It then allows entry if enough time has passed since the last sell, or if a strong bull continuation exception exists with trend efficiency, daily trend strength, strong 1m-10 and 1m-20 momentum, 5m-EMA-20 above 5m-EMA-50 and positive 5m MACD. It also blocks chasing unless HighestSinceSell is null, current price is sufficiently below the high since sell, or a very strong bull trend exception applies. The bull, sideways and bear-rebound sub-branches then mirror a trend-pullback structure using daily regime, daily trend strength, daily EMA or Bollinger context, intraday EMA support, positive short momentum, MACD confirmation, RSI ranges and volatility caps. The second part is a non-bear continuation entry that does not wait for action = buy; the latest change tightened it by requiring 5m-TrendEfficiency-24 at least 35 or 1m-20 above 1.40, and reduced the 5m-RSI ceiling to 72.

SELL formula and interpretation

SELL
action = sell AND (((ProfitSinceBuyPct > 1.10 AND DrawdownFromHighPct > 0.90) OR (ProfitSinceBuyPct > 1.80 AND DrawdownFromHighPct > 0.85) OR (ProfitSinceBuyPct > 3.00 AND DrawdownFromHighPct > 0.80)) OR (1m-2 < -0.45 AND 1m-10 < -0.85 AND 1m-20 < -1.20 AND 1m-2 < 1m-10 AND 1m-10 < 1m-20) OR (MarketRegime-1d = bull AND (((5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 1m-2 < -0.15 AND 1m-10 < -0.12) OR (CurrentPrice < 5m-EMA-50 AND 1m-10 < -0.22)) OR (DrawdownFromHighPct > 0.90 AND DrawdownFromHighPct > 5m-NATR-Wilder-14 * 1.50))) OR (MarketRegime-1d = sideways AND ((5m-RSI > 70 AND CurrentPrice > 5m-BB-upper) OR (5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 1m-2 < -0.20 AND 1m-10 < -0.15) OR (DrawdownFromHighPct > 0.90 AND DrawdownFromHighPct > 5m-NATR-Wilder-14 * 1.30))) OR (MarketRegime-1d = bear AND (((CurrentPrice < 5m-EMA-20 AND 1m-10 < -0.18) OR (5m-MACD-hist < 0 AND 1m-MACD-hist < 0 AND 1m-2 < -0.12) OR 1m-10 < -0.45) OR (DrawdownFromHighPct > 0.90 AND DrawdownFromHighPct > 5m-NATR-Wilder-14 * 1.00))))

The SELL formula is explicitly fee-aware. It requires action = sell and then allows profit-trailing exits only after enough gross profit to overcome the observed OKX fee drag: ProfitSinceBuyPct above 1.10 with drawdown above 0.90, above 1.80 with drawdown above 0.85, or above 3.00 with drawdown above 0.80. Additional exits include a steep intraday momentum cascade where 1m-2, 1m-10 and 1m-20 are all strongly negative and worsening. Regime-specific exits then apply. In bull regimes, negative MACD plus short momentum or loss of 5m-EMA-50 can trigger, as can a drawdown larger than 5m NATR times 1.50. Sideways regimes use overbought Bollinger/RSI reversal, MACD weakness or NATR-adjusted drawdown. Bear regimes use EMA loss, negative MACD or 1m-10 below -0.45, plus looser drawdown sensitivity. SELL was not changed in the latest revision.

STOPLOSS formula and interpretation

STOPLOSS
CurrentPrice < LastBuyPrice * 98.8% OR (1m-2 < -0.55 AND 1m-10 < -1.00 AND 1m-20 < -1.45 AND 1m-2 < 1m-10 AND 1m-10 < 1m-20) OR (MarketRegime-1d = bear AND CurrentPrice < 5m-BB-lower AND 5m-MACD-hist < 0) OR 5m-NATR-Wilder-14 > 5m-NATR-EMA-14 * 2.35 OR DrawdownFromHighPct > 3.2

The StopLoss formula is unchanged. It exits if current price falls below LastBuyPrice times 98.8%, if there is a severe short-term momentum cascade across 1m-2, 1m-10 and 1m-20 with worsening order, if the daily regime is bear and price is below 5m-BB-lower with negative 5m-MACD-hist, if 5m-NATR-Wilder-14 exceeds 2.35 times 5m-NATR-EMA-14, or if DrawdownFromHighPct exceeds 3.2. This is a broad emergency layer with price, momentum, bear-regime volatility and peak-drawdown routes. The evidence was mixed: bounce rates existed, but some longer follow-up deteriorated, so the latest AI left StopLoss unchanged rather than loosening risk protection.

Measured performance

The strategy has no active seven-day or thirty-day trades, so recent live inference is unavailable. The ninety-day aggregate shows 165 trades, 6766.0622 turnover, 14.972 profit amount and 7.1224% portfolio profit across four plans. However, aggregate win-rate and profit-factor fields are unavailable. The BUY/SELL ninety-day cohort lost at the portfolio level, while the CONVERT cohort showed 11.4806% portfolio profit but limited convert-increase coverage and a -0.133% convert increase. Revision history shows 65 evaluated events, 29 buys, 15 sells, 17 stoplosses and a weak revision KPI profit factor of 0.0605 over 67 trades. The latest entry change is therefore a risk-reduction hypothesis based on historical weak-buy evidence, not a proven improvement.

Window	Trades	Profit	Profit factor	Win rate	CONVERT increase	Turnover
7d	0	0 $	—	—	—	0 $
30d	0	0 $	—	—	—	0 $
90d	165	14.97 $	—	—	-0.13%	6,766.06 $

Market alignment

The market environment favored momentum but also contained much sideways breadth. An OKX fee-aware strategy must be more selective than a no-fee model because small gross wins may be erased by costs. The formula’s profit-trailing thresholds acknowledge this, while the tightened continuation branch tries to avoid entering late or weak acceleration in a choppy universe.

Problems found

Plans are inactive, so the latest applied change lacks current post-change evidence.
Historical bad-buy rate around 31% and weak revision KPI profit factor indicate entry quality problems.
Exit evidence is noisy: SELL negative-exit rates were high, but follow-up sometimes justified caution.
CONVERT portfolio profit was positive, but convert increase was slightly negative and coverage was incomplete.

Adaptive change and expected behavior

The applied AI change tightened only the volatility-capped continuation BUY path. It asked for cleaner 5m trend efficiency or stronger 20-minute momentum before entering without the slower action signal, and slightly lowered the 5m-RSI ceiling. SELL and StopLoss were left unchanged because their evidence was mixed and because fee-aware exit logic is central to the strategy design.

Expected behavior is fewer weak continuation entries, especially in sideways conditions where short momentum is not sufficiently directional. The next review should measure bad-buy rate, average 60-minute return after formula-true buys, and whether trade count collapses when plans become active again.

Hypotheses and next review

H-2026-W40-01

For Strategy 935, tighter BUY thresholds and wider profit-guard exits will improve realized profit factor without collapsing trade count below a usable sample.

Measurement: Compare post-change BUY/SELL profit factor, bad-buy rate, trade count and average loss size with the active baseline after the latest SELL revision.
Review: After at least 100 new BUY/SELL events or two full weeks, whichever comes later.

H-2026-W40-02

For Strategy 933, the StopLoss-only confirmation change will reduce quick recovery after StopLoss exits while preserving emergency protection.

Measurement: Track enabled-plan StopLoss exit count, early-exit rate, exit-near-60m-low rate, 1m/2m/5m recovery rates and maximum adverse excursion after non-exit signals.
Review: After at least 10 new StopLoss exits or 30 calendar days.

H-2026-W40-03

For Strategy 934, modest BUY relaxation will restore rotation frequency without materially worsening target-versus-source post-switch returns.

Measurement: Track switch count, average hours between switches, convert increase and 60m/180m target-minus-source returns.
Review: After at least 10 confirmed switches.

H-2026-W40-04

For Strategy 932, the StopLoss-only adjustment will reduce wick-driven stop exits without degrading emergency protection.

Measurement: Compare enabled-plan StopLoss recovery rates, negative-exit rates and post-non-exit adverse excursion after the revised formula.
Review: After at least 10 new StopLoss exits.

H-2026-W40-05

For Strategy 930, tightening the continuation BUY branch will reduce bad-buy rate when plans resume execution.

Measurement: Measure formula-true BUY count, bad-buy rate and average 60-minute return after new continuation entries.
Review: After at least 50 new BUY events.

Conclusion

The evidence supports conservative, formula-specific iteration rather than broad claims about automation. The strongest result was Strategy 934’s CONVERT performance, which combined a high profit factor, positive convert increase and positive portfolio profit over thirty days across multiple active plans. That finding fits the market: large-cap crypto momentum was strong over the longer lookback, but the broader CoinRobot universe remained mostly sideways, making relative momentum rotation attractive. Still, the active switch sample is small, so this is not a guarantee that every future rotation will outperform.

The BUY/SELL strategies show a different lesson. Strategy 933 had good thirty-day and ninety-day aggregate results, but its seven-day window weakened and StopLoss exits showed recovery risk. Strategy 935 showed that a high win rate can coexist with near-zero or negative profit factor when average losses or exit timing offset many small wins. Strategy 918 and Strategy 930 had mostly historical or inactive evidence, so their applied changes must be treated as hypotheses. Strategy 932 demonstrated both the promise of one strong CONVERT plan and the importance of validation, because prior invalid AI output was blocked before deployment.

The educational theme is clear: customizable trading intelligence is valuable when formulas are explicit and measurable. Users and community authors can encode beliefs about trend, momentum, volatility and risk, but every threshold has consequences. AI assistance is useful only when it narrows a testable hypothesis, preserves risk controls and passes scenario validation. The next review should prioritize post-change StopLoss recovery, Strategy 934 switch quality, Strategy 935 profit-factor repair and Strategy 930 entry quality after resumed activity.

Limitations

The report uses only supplied data; crypto news and macro feeds were disabled.
Some global and exchange summary fields may reflect collection time rather than exact historical closing snapshots.
Several strategies have inactive plans, zero recent trades or historical-only evidence, limiting current inference.
CONVERT diagnostics often lack complete paired target-versus-source evidence, especially for Strategy 932.
Portfolio-profit comparisons can be distorted by different plan eligibility, start dates, coverage and execution models.
StopLoss evidence is sparse for some strategies and should not be generalized across unrelated formulas.
Event calendar entries are not causally linked to price action because trade-to-event temporal evidence was not supplied.

Reproducibility appendix

Run ID: 20261005-011001-cb0cc459. Snapshot generated at 2026-10-05T01:10:01+00:00. The machine-readable run record is retained privately by CoinRobot.AI.