Research Ledger

Every Test We Ran. Published.

17 strategy families. 15 research campaigns. 7,229,294 simulated trades. The verdicts below are the raw output of our research log — including every failure.

7,229,294

Simulated trades

17

Strategy families tested

0

Survived beta-strip

Table 1

Strategy Family Verdicts

Per-trade net returns across our full coin universe, after fees (0.56% round-trip). The arbiter is excess net after beta-strip: market return is subtracted first, so a strategy only passes if it beats simply holding the market. Result:17 of 17 failed.

Strategy family verdicts after beta-strip
Strategy familyTradesWin rateRaw net / tradeExcess net / trade (beta-strip)Verdict
Momentum1,524,66548%-0.640%-0.578%FAIL (beta)
MA Cross30,17747%-0.794%-0.586%FAIL
MACD119,40450%-0.432%-0.563%FAIL
RSI Oversold166,26147%+0.012%-0.506%FAIL (beta)
Donchian Breakout139,69247%-0.635%-0.572%FAIL
Mean Reversion179,09553%-0.344%-0.565%FAIL
Bollinger Bands171,28854%-0.354%-0.560%FAIL
Keltner Channel396,77340%-0.769%-0.580%FAIL
ATR Breakout150,00448%-0.620%-0.567%FAIL
Supertrend8,45637%-2.608%-1.036%FAIL
Ichimoku Cloud976,24649%-0.612%-0.543%FAIL
Heikin Ashi1,527,42950%-0.514%-0.554%FAIL
Stochastic RSI197,06251%-0.576%-0.556%FAIL
ADX Trend1,526,34248%-0.654%-0.581%FAIL
Trendline Touch116,40039%-0.225%-0.062%FAIL
Triangle BreakoutFAIL
Multi-Signal Consensus55%-1.170%FAIL (beta)

FAIL (beta) = raw return may even be positive, but it disappears once market return is subtracted — the strategy was riding the market, not beating it. Note the trap: Mean Reversion and Bollinger Bands win 53–54% of trades andstill lose money. Win rate is not edge. Cells marked — were judged on separate grids (verdict unchanged: no alpha above cost).

Table 2

Research Campaigns

Beyond single indicators, we ran 15 structured research campaigns across five axes — directional, cross-sectional, carry, options, and non-price data. Each hypothesis got the same 9-gate test battery. Here is where each one stands.

Research campaign statuses
CampaignAxisStatusDate
Single-coin indicator timing (8 indicators)DirectionalBETA2026-06-20
Chart patterns: triangles & trendline touchesDirectionalBETA2026-06-20
Multi-signal consensus / voting ensemblesDirectionalBETA2026-06-19
Bollinger + Ichimoku combinationsDirectionalBETA2026-06-20
Indicator + filter × timeframe tuning (36 cells)DirectionalBETA2026-06-20
Cross-sectional momentum (winners long / losers short)Cross-sectionalARTIFACT2026-06-20
Cross-sectional short-term reversalCross-sectionalARTIFACT2026-06-20
Low-volume premiumCross-sectionalARTIFACT2026-06-20
Low-risk anomaly (betting against beta)Cross-sectionalBETA2026-06-20
Volatility risk premium (options)OptionsSTRONG LEAD2026-06-20
Delta-neutral funding carryCarryARMED / WAITING2026-06-14
Funding / OI / long-short-ratio extremesNon-price dataGATED2026-06-14
Mechanical accumulation (DCA)Beta accumulationLIVE2026-06-20
Signal-flip exits (stop-and-reverse holding)DirectionalMANAGED BETA2026-06-21
Entry × exit strategy cross-matrixDirectionalBETA2026-06-21
BETA
Market exposure, not alpha — excess return negative after beta-strip.
ARTIFACT
Collapsed under liquidity / in-sample isolation — statistical artifact, not edge.
STRONG LEAD
Passed the full test battery — still pre-live, no real capital deployed.
ARMED / WAITING
Conditionally armed — waiting for the opportunity threshold to trigger.
GATED
Not enough data yet — still collecting, verdict pending.
LIVE
Running live — mechanical beta accumulation, never claimed as alpha.
MANAGED BETA
Not alpha — but real risk-management value (drawdown reduction).

How verdicts are decided

The Test Battery

Arbiter: beta-strip

Every directional result has the market's own return subtracted first. Positive excess = alpha. Negative excess = you were just long the market. This single gate kills most "profitable" backtests.

Costs: 0.56% round-trip

All per-trade numbers are net of a 0.56% round-trip cost (fees + slippage). A strategy that can't survive real costs was never a strategy.

9 statistical gates

Beta-strip, in-sample/out-of-sample isolation, cost stress, liquid-universe check, walk-forward, multiplicity control, capacity check, sanity checks, and long/short leg decomposition. Pass all or no verdict.

Full simulation methodology (data, fills, fees, metrics) is documented on themethodology page. Live execution quality is published on thetrust dashboard. What survived — risk management, not prediction — is therisk-managed portfolio.

Generated 2026-07-12T00:33:54Z from strategy_verdicts.json + research_ledger.json · verdicts updated 2026-06-20 · campaigns updated 2026-06-21 ·raw JSON