Strategies

This hypothesis set out to translate Geraked's NWERSIASF v1.4 expert advisor onto a 3-symbol OKX USDT-swap basket on 2H bars, keeping every default input. That included the no-stop 1.1x martingale grid of up to 20 levels. The developer produced partial strategy code with incremental NWE, RSI and ATR-SL-Finder indicators. After 2 iterations, the coding stage hit three timeouts and the run was abandoned as 'coding_stuck_3_timeouts'. No backtest, optimization or review stage was reached, so the strategy's edge was never measured.

KeltnerKellyStopReverse was a line-by-line port of a public TradingView Keltner breakout strategy with stop-and-reverse and Kelly sizing, assigned to AVAXUSDT.OKX 4H to add long-short OKX coverage to the corpus. The developer produced a partial FactoryStrategy implementation, but the coding stage timed out three times over 2 iterations and the run was abandoned ('coding_stuck_3_timeouts'). No backtest, optimization or analyst review ran, so the Keltner breakout family still has no measured verdict.

SolTwoMomEmaSwingBracketLS was a user-requested faithful port of the 2MOMEMA expert advisor, moved onto the Bybit SOL perpetual on 1-hour bars so it could take both long and short trades. The hypothesis itself noted that faithful ports of public scripts have repeatedly failed in this pipeline and that only a backtest with real fees would decide. That test never happened: after 2 iterations the run was abandoned at coding with 'coding_stuck_3_timeouts', leaving no backtest, optimization or analyst verdict. Whether the strategy has an edge remains untested.

LtcBchCopulaMispricingPairLS tried to trade short-lived mispricings between the two Bitcoin-fork perps LTC and BCH. It used a copula fitted once a month to their trailing 365 daily returns, with beta-weighted legs sized to 50% gross equity. The strategy code was drafted, but the run went through 2 iterations and then hit three coding-stage timeouts. It was abandoned as 'coding_stuck_3_timeouts' before any backtest, optimization or analyst review, so there is no performance record.

CountryEtfLoserReversalCohorts adapted a 19-country, 36-month long/short reversal rule to the 8 single-country ETFs in USEQ, dropping the short leg and using 12 overlapping monthly cohorts that each buy the two worst negative 36-month losers. The code implemented a custom cross-leg alignment barrier and O(1) deque-based formation returns, following the hypothesis's warning about sandbox timeouts. Even so, the run was abandoned after 2 iterations with reason 'coding_stuck_3_timeouts', so it never produced backtest metrics and never reached review, optimization or analysis.

WtiBrentSpreadReversionDeribit15m adapted QuantConnect/Quantpedia's WTI-Brent spread strategy to Deribit's two USDC oil perps. It replaced the source's always-in SMA flip with a thresholded z-score entry on a rolling regression residual, with an explicit flat state. The code was written in one iteration, but the pipeline could not load any 15m bars for either leg after 5 collection retries, so no backtest was produced. The run was abandoned as data_unavailable before any evaluation stage. That fits the hypothesis's own caveat that the Deribit oil perps, whose history only started on 2026-09-17, did not yet have enough history to test.

HighVolQuintileMomentumJTCohortsLSDaily adapted QuantConnect Strategy Library #155 to crypto perps. It bet that momentum is concentrated in the highest-volatility names, trading them long-short in overlapping 6-month monthly cohorts across a point-in-time top-60 Binance USD-M universe. Across 3 iterations the code never cleared Layer 3 verification: every sandbox smoke test ran past the 300s wall-clock limit on the ~20,000-bar daily window. The run was abandoned in a verification loop with no backtest, optimization or analyst verdict, so the hypothesis itself was never tested.

SpotLiquidUniverseMonthlyPriceChangeTopQuintile was a monthly, long-only BINANCE_SPOT strategy. It ranked a point-in-time liquid universe of 30 USDT pairs by 1-month price change and held the top 6, with a per-name positive-momentum rule that moved failing slots to cash. The code used a 12-hour BTCUSDT bar as its clock and subscribed to hundreds of daily extra-instrument legs from the catalog. It timed out three times in the verification stage and was abandoned after one iteration as 'verifying_stuck_3_timeouts'. No backtest, optimization, analyst or risk review ever ran, so the idea's edge remains untested.

The hypothesis restated a cross-sectional equity momentum/reversal screen as a long-short time-series rule on BNBUSDT 4H bars, gated by a GARR-style ratio of the 5-day return rate to the 60-day return rate. Across 4 iterations the developer capped gross exposure at 30% of equity and rewrote the indicators as O(1) incremental updates to fix Layer 3 sandbox timeouts. The smoke backtest still exceeded the 300s limit on the ~20,000-bar window, and the run was abandoned in the verification loop. It never produced a backtest or reached the analyst, so neither the screened strategy nor the ungated control was ever scored.

This hypothesis restated Xiaocao's always-on multi-coin BTC hedge as a flat-by-default strategy: trade one alt/BTC pair, beta-hedged, only when that alt's BTC-relative deviation from its slow EMA reaches 2.5 cross-sectional sigmas, with a point-in-time universe on 4H Binance USD-M bars. The developer coded it with BTC as the primary instrument and a fixed list of 10 alt perps as extra instruments, using incremental rolling statistics. Across 3 iterations the code never passed the Layer 3 sandbox backtest, which kept going over the 300s limit on a 2400-day 4H window, and the run was abandoned in a verification loop. Because no backtest ever completed, the economic claim was never tested and the later pipeline stages were not reached.

The strategy restated an FMZ fixed-ladder grid as an ETH spot grid with a point-in-time anchor, 2% geometric spacing (about 10x the 0.20% spot round trip), per-lot take-profits, and a floor stop to cap losses on a breakout. It never got past verification. Across 3 iterations the Layer-3 sandbox backtest kept going over the 300s limit on an 833-day 1-hour window. The developer's iteration-3 notes say per-bar cost was already O(n_nodes) and put same-day timeouts down to the runner's ParquetDataCatalog glob on a heavily loaded host. The run was abandoned with no backtest metrics, and no later stage ran.

This run was meant to test Connors' RSI(2) pullback rule on the four broad US index ETFs it was designed for, with post-2009 data as a pre-registered out-of-sample check. The strategy never got past Layer-3 sandbox verification: all three iterations hit the 300s smoke-test limit. In iteration 2 the developer made per-bar indicator work O(1). In iteration 3 profiling pointed to the NautilusTrader catalog file scan, which strategy code cannot shorten. The run was abandoned with no backtest, optimization or analyst verdict, so the hypothesis itself is still untested.

The hypothesis carried a FX momentum and mean-reversion regression over to a single-instrument, time-compressed, strictly point-in-time version on LTC daily perp bars. Iteration 2 rewrote the strategy to O(1) per-bar running sums and a closed-form 3x3 OLS solve after a Layer-3 sandbox timeout, but the timeout recurred. Iteration 3 diagnosed catalog instrument globbing in the backtest runner as the cause and left the logic unchanged. The run was abandoned in the verification loop after 3 iterations without ever producing a backtest, so the edge itself was never tested.

EthBybitKingKeltnerBreakout4H restated the FMZ King Keltner breakout (a slope-confirmed EMA midline with ATR bands, long and short, exiting on a midline recross) on ETHUSDT.BYBIT 4H. The 4H clock was chosen after a pre-study showed the 1H version could not cover fees. The code never passed Layer 3 sandbox verification: the smoke test repeatedly ran past the 300s limit on a ~20,000-bar window. After 3 iterations, including a rewrite of the ATR as an O(1) rolling sum, the run was abandoned before backtest review, optimization, analysis or risk review, so the edge was never measured inside the factory.

The hypothesis ported the Avellaneda-Lee PCA s-score reversion idea from US stocks to a point-in-time top-N universe of Bybit USDT perps on daily bars, trading both tails as unhedged single-name bets. Over 3 iterations the strategy kept failing Layer 3 sandbox verification by exceeding the 300s wall-clock limit. The last revision changed only performance: it cut the pool to 8 deep perps (BTC, ETH, SOL, XRP, DOGE, ADA, LINK, LTC) and left the signal unchanged, but it still did not pass. The run was abandoned before producing any backtest metrics.

BtcOkxTrendBarContinuation restated an FMZ beginner's dual-MA intraday script as a 4H BTC OKX trend-bar continuation strategy, pre-registered at PF 1.46 and +1.03% per trade. Over three iterations the developer removed the same-bar flip and then the bracket-order path to fix an incoherent MTM curve, but the paired daily spikes persisted. The final backtest returned +28.96% over 354 trades with PF 1.049, only 2 of 7 positive years, a 47.8% drawdown and a per-trade PnL-weighted edge near cost. The analyst abandoned it at backtest review, before optimization, as a falsified premise with no edge, and optimization, analyst, and risk stages were never reached.

EthSpotPingPongDailyLimitPair tried to turn a single-trade public FMZ script into a repeatable daily limit-pair strategy on ETH spot, betting that its +2.56% target, about 13x the spot round-trip cost, would survive fees. The last iteration changed only order-cache access, adding a defensive _cache_orders helper. Across 3 iterations the code never passed Layer 1 static analysis; the final failure was a syntax error at line 1, and the run was abandoned as a verification loop. The strategy never reached backtesting, so the hypothesis is still untested.

This hypothesis ported an FMZ rolling linear-SVM next-bar direction classifier to the Binance COIN-M BTC perpetual at 1H, and it pre-registered a near-zero-edge prediction. After 2 iterations the backtest produced 1,408 trades averaging -0.129% per trade, with PF 0.80 and Sharpe -0.38, and the account was liquidated for a -100% return while buy-and-hold BTC rose about 3x. The analyst abandoned it at backtest review before optimization, because a zero gross edge cannot clear trading costs. The analyst also flagged implausible COIN-M daily MTM swings and conflicting liquidation dates for engine review, noting they do not change the sign of the edge.

SolSpotVolScaledStepLadder restated an FMZ price-step ladder as a vol-scaled, 25%-slice, long-only SOL spot trend follower on 4H bars. Its base backtest over about 5.7 years showed a large +2581% return, but with Sharpe 0.33, a 78.7% max drawdown, beta 0.63 and an information ratio of -0.54, meaning it underperformed simply holding SOL. After 2 iterations the analyst abandoned it at backtest review as beta without a trend edge, so optimization, analysis and risk review were never reached. The analyst also flagged a -7.43% avg_trade_return_pct that disagrees in sign with the positive PF and USD PnL, and asked for a harness check.

XrpHlReturnShockContinuation12H tested whether large 3-day XRP return shocks on Hyperliquid keep drifting for 6 days in either direction. Its single backtest looked good at first, with +125.9% and a 2.75% average trade, but the median trade was negative (-0.43%), the edge disappeared once the three biggest 2024-rally winners were removed, shorts lost money and the drawdown reached 65%. The analyst abandoned it at backtest review in iteration 1, so optimization, the post-optimization analyst and risk review never ran. The review also flagged a metrics artifact: paired cancelling prints in daily_returns inflate the volatility and VaR figures.

This strategy ported a 2014 FMZ order-book-depth market-making bot to 1-minute SOLUSDT perpetuals, quoting ±10bp around the touch with all its inventory and no stop, even though the hypothesis itself called this class a fee graveyard. Across 3 iterations it never got past verification. The sandbox backtest went to -100% with liquidation, 2,311 trades and a 0% win rate, because sizing let exposure reach about 252% of equity and the market impact per trade outweighed the 20bp spread. QA also flagged a trade ledger that doesn't reconcile and warned that the touch-fill model doesn't capture adverse selection. The run was abandoned as a code defect before the backtest review, optimization, analyst and risk stages.

This strategy ported the fee-free OKCoin 'Leeks Reaper' burst-chase and rebalance-to-neutral logic to hourly bars on the COIN-M BTC perpetual, keeping every rule and rescaling only the burst threshold by sqrt(time). After 2 iterations, the backtest showed negative expectancy from the first months (Sep to Nov 2020 were already losing). Exposure ran at 94% instead of the planned 35-45%, and the account was liquidated in February 2021 (-100%, Sharpe -3.71, PF 0.80, -0.26% per trade). The analyst abandoned it at backtest review without optimizing, and noted a possible, unverified inverse-sizing anomaly that did not change the sign of the edge.

This strategy tried to faithfully replicate a source composite-factor long/short tercile book on a point-in-time top-15 Hyperliquid universe; the hypothesis itself noted that multi-leg neutral books had a 0/150 record here. Over three iterations, the rebalance moved from daily to weekly and gross exposure was cut, but the final 809-day backtest still lost 3.06% (Sharpe -0.04, PF 0.975, -$6.54 per leg), and costs were roughly equal to the loss. The analyst abandoned it at backtest review for having no edge, noting that its positive per-trade % was a ledger artifact. The analyst also flagged an unverified weekly -2.5%/+2.5% return pattern around rebalance flips as a possible engine/ledger issue with NETTING flips.