Hypotheses

Classification: (a), a concrete mechanism on a market we cannot trade. The source measures the estimated equity allocation of Chinese public mutual funds and uses changes in that aggregate positioning to time the A-share index. It also splits time into trend and range regimes with an MA system and reports that the positioning signal only times the market in ranges. We have no fund-position data and no CNEQ estate to run it on (L103). The crypto analogue of 'aggregate positioning of an identifiable participant class' is Binance's published positioning metrics. The metrics feed (data/supplementary/metrics/<SYM>/all.parquet, BinanceVisionCollector, 5-minute rows from 2021-12-01) carries long_short_ratio (all-account global ratio) and top_trader_long_short_ratio. POINT-IN-TIME PRE-STUDY (daily closes, 10 alts, 2022-04 to 2026-09, non-overlapping 5-day events, |z|>1.5 on the 60-day z-score of log long_short_ratio): (1) FOLLOWING the crowd lost in every variant, and FADING it earned a gross +0.55% per 5-day trade over 1155 events (t ~1.9). By regime: range +0.44% (n=410, t 1.0), trend +0.62% (n=745, t 1.55). (2) The source's regime split is NOT reproduced: the signal is stronger in trends than in ranges. The hypothesis therefore uses no regime gate and records the source's range-only claim as falsified here. (3) Top-trader ratio and positioning-change (residualised on price) variants showed no edge, roughly 0 +/- 0.3%, and are excluded. (4) Fading crowd LONG extremes carries most of the edge (+1.02% on n=174 in ranges), while fading crowd SHORT extremes is about flat. (5) Per year, range-regime means are 2022 +0.45%, 2023 +0.92%, 2024 +0.33%, 2025 -0.02%, 2026 +0.62%. Seven of 10 symbols are positive. This is a modest, honestly weak edge. Implementation: one independent position per symbol, at most 10 concurrent, 10% of equity notional each, leverage fixed at 1.0. The universe is capped at 10 names per L173, and the per-bar work is O(1) rolling statistics on daily bars (precompute the daily metric series once in __init__ and use bisect for as-of lookup, per the sandbox-timeout lessons).

Classification: (a), but only partly. Zero's FMZ post (2015) describes how to EXECUTE a buy, not WHEN to buy. Instead of lifting the ask, you post a bid deep inside the spread and raise it step by step ($1.05 -> $1.06 -> ... -> the ask) until an impatient counterparty hits it. On its own that gives no edge: it saves part of the spread and the maker/taker fee gap, about 0.03-0.05% per fill on an OKX perp. That is far too small to back a strategy, and lesson L167 rules out porting a sub-second tick loop as it stands. So we keep the mechanism and restate it at bar resolution, attached to a moment when the post's premise is strongest: impatient sellers pay for immediacy. On OKX USDT swaps the trigger is a 1-HOUR bar whose log return is below -k x the trailing 30-day hourly sigma, on volume at least 3x the trailing 7-day median hourly volume. That is a forced/impatient selling hour (stop-outs, liquidations). After the trigger we act as the bargain-poker. Over the next 3 bars we post a LIMIT (maker) buy that starts deep and steps up toward the market: bar 1 bid = close x exp(-1.0 x sigma x depth_scale), bar 2 = close x exp(-0.5 x sigma x depth_scale), bar 3 = the trigger close. Each unfilled bid is cancelled at the bar's close and replaced with the next, higher one (the post's cancel-and-raise loop). If nothing fills by bar 3 we skip the event and never chase with a market order, so a fill only happens when sellers come down to us. Exit by market after hold_bars (12h), or earlier on a 3-sigma-equivalent stop. Instruments: ADAUSDT.OKX (primary), DOGEUSDT.OKX, AVAXUSDT.OKX, DOTUSDT.OKX. Each one triggers independently, one position per instrument, a 24h cooldown per instrument, and equal risk per event. Universe capped at 4 names (lesson L173). Source credit: Zero, 'High-frequency trading strategy: Poke for bargain', https://www.fmz.com/bbs-topic/41. source_ref: the paragraph that starts with the $1.05 -> $1.06 -> $1.07 stepped-bid description.

Classification: (b)/(a) borderline. The source (Guotai Junan quant report #60, 2015-06-04, via BigQuant) is an abstract of a style-neutral multi-factor A-share selection strategy hedged with CSI500 index futures. Its one concrete, transferable claim is about the HEDGE: alpha is only clean when the hedge benchmark matches the universe's industry and size distribution, so hedging a mid-cap selection with a mid-cap index ('the touchstone') removes style beta that a mismatched index (CSI300) leaves in. The abstract names no specific factors, and A-share venues (CNEQ/HKEQ) have no working data estate here (lesson L103), so we restate it on USEQ. Universe: 10 long-history S&P 500 utilities (NEE, DUK, SO, D, AEP, EXC, SRE, XEL, ED, PEG). Hedge: XLU, the benchmark built from the same industry and size bucket, so industry, rate-sensitivity and size style net out by construction. This is the paper's 'touchstone' idea. Alpha factors (OHLCV only, because no fundamentals are in the catalog): the two price/volume factors that dominate A-share multi-factor models and have US evidence too. (1) Low residual volatility: std over 60 sessions of the daily residual vs XLU (the idiosyncratic-volatility anomaly, Ang et al. 2006). (2) Low abnormal turnover: 20-session mean volume divided by 250-session mean volume (high attention or turnover predicts underperformance). Composite = average cross-sectional rank. Every 21 sessions, hold the 3 best-ranked names equal-weight long and short XLU sized to the basket's 120-session beta, so the book is dollar- and beta-neutral within the sector. Cost: zero commission, ~0.02-0.05% spread+impact per leg on mega-liquid utilities. The universe is capped at 10 names (lesson L173) so per-rebalance ranking is O(10) on precomputed rolling stats. Credit: Guotai Junan Securities quant team, 'Quantitative Topic #60: CSI500 as the Alpha Touchstone', 2015-06-04, republished by BigQuant (宽邦).

Turtle Soup restated from first principles on US ETFs, the asset class it was written for, with honest gap fills. The pre-study is in the file. On OKX crypto perps the setup scores at or below drift. On the ETFs, with a realistic fill, per-trade means are about 0 (t of 0.6 or less), and the edge decays to about 0 after 2016. The full spec is written to workspace/outbox/research_lead/hypothesis_0bc51c25a4.json.

The source EA trades the Percentage_Crossover_Channel. The midline ratchets: mid_t = close*(1-p) if close*(1-p) > mid_{t-1}; mid_t = close*(1+p) if close*(1+p) < mid_{t-1}; otherwise mid_t = mid_{t-1}. Upper = mid*(1+p), lower = mid*(1-p). DEFAULT mode is contrarian: buy when the low touches the lower border, sell when the high touches the upper border, always in the market, tested on EURUSD H1 for one year (unverified). The EA's own claims are ignored. I ran a point-in-time pandas pre-study on OKX USDT-swap 4H bars (ADA, DOT, LINK and AVAX, 2020-2026, 0.10% round trip). (1) The DEFAULT contrarian always-in mode loses money at every fixed width p in {2,4,6,10,15}% on all 4 coins, and at every vol-scaled width. Fade per-trade net ranges from -0.07% to -0.53%, annualised trade Sharpe from -0.26 to -1.17. That framing is falsified and is NOT proposed. (2) REVERSE mode turns the ratchet channel into a percent-reversal breakout: go long when close lifts the midline (close > upper border of the previous bar), go short when close pushes it down. In the pre-study, the only positive-expectancy region is a WIDE channel scaled to volatility, p = k * stdev(4h log returns, 180 bars) with k ≈ 6, i.e. a band of ~6-9%. Adding the source's midline-cross as an EXIT (flat when close crosses back through the midline against the position) is also positive there. Net results: ADA +1.23%/trade (133 trades, Sharpe 0.27), DOT +1.15% (123, 0.26), LINK +0.13% (132, 0.04), AVAX +11.5% (112, 0.61); mean hold ~9 days. Fixed-percent width (the source's literal 'Percent' input) is replaced by the vol-scaled width because one fixed % cannot fit coins whose vol differs 2x across regimes. Trading the 4 coins as independent sleeves (each 25% of equity, 1x) gives ~500 lifetime trades, which is enough sample, and stays under lesson L173's ~10-name cap. Venue: OKX (1.6% of corpus, under the >=5% quota). Direction is long-short (corpus is 84% long-only). None of these 4 instruments is BTC. CAVEAT stated up front: this is a trend/breakout mechanism, a crowded family, and the pre-study edge is thin and concentrated in AVAX. Priority is set low accordingly.

Classification: (a), a concrete idea. The source sums dictionary-scored headline sentiment for each drug manufacturer from the previous close until 30 minutes after the open. It then trades intraday in the direction of that sign and exits at the close, on Wednesdays only (Berument & Kiymaz). The source reports Sharpe 0.116 against SPY's 0.971, and only became profitable once the Wednesday filter was added. We treat both numbers as unverified. Why the idea cannot run as written: (1) Our scored news indicators barely cover the drug-manufacturer universe (LLY about 15 files and MRK about 11; PFE, ABBV, JNJ, BMY, AMGN and GILD none), so it cannot be backtested on USEQ. (2) The Wednesday filter is a date-keyed window, which the factory treats as a dead family (L179), and the source admits it is the only thing that made the strategy profitable, which is an in-sample selection. It is dropped. What we keep is the core mechanism: news sentiment accumulated over a window, traded in its sign direction for a fixed horizon of about one session. We move it to the asset with the densest per-asset LLM-scored news in our estate after BTC: ETH (about 3,000 indicator files since 2026-08-13; BTC is avoided for corpus concentration). Instrument: ETHUSDT.BINANCE USD-M perp, 1-HOUR bars, long-short, single leg. Signal: S_t = sum of (sent_avg x n_articles) over the ETH news_sentiment buckets ingested in the trailing `accum_hours` (source analogue: close to open+30m, about 17.5h). Missing buckets count as zero news. All lookups are point-in-time via supp_window/supp_as_of on ingestion time. The MACRO series merged into the ETH feed is acceptable. HONEST LIMITATION: the scored history starts 2026-08-13, about 7 weeks, so this is a short-sample exploratory test and the analyst must judge the covered span. If the result is inconclusive it should wait for the history to accrue rather than be re-tuned.

Classification: (a), a concrete trading idea. The source (QuantConnect Strategy Library, 'Dual Thrust Trading Algorithm', after Michael Chalek and Gang Wei 2012) defines one mechanism. Range = max(HH - LC, HC - LL) over the last N completed days. Buy when price >= today's open + K1 x Range, sell short when price <= open - K2 x Range, and reverse on the opposite trigger. The QC SPY port reports Sharpe -0.37 and a 65.7% drawdown. We treat that as unverified, and it is not an endorsement. The QC code also departs from the original: it has a single trigger and is always in the market, which turns it into a stop-and-reverse system that never goes flat. Chalek/Wei designed Dual Thrust as an INTRADAY system that is flat at the session close, and that is the version restated here from first principles. It is not a faithful port. Crypto has no session, so the UTC day (00:00-24:00) stands in for the session, and the day's open is the first 1h bar's open. The venue is OKX USDT swaps, which hold 1.6% of the corpus against a >=5% quota. Instrument: ETHUSDT.OKX 1-HOUR, local catalog 2019-12-25 to 2026-10-01, 59,339 bars with 0 gaps (checked 2026-10-02). It is long-short, single instrument and price-only, so it needs no supplementary feed. A point-in-time pre-study ran on local OKX 1h bars with N=4, K=0.5, a 1h-close trigger and day-end exit, and 0.10% round trip deducted. Results: ETH 720 trades (106/yr), gross +0.447% per trade, net +0.347%, per-trade-annualised Sharpe ~1.06. With K=0.7: 386 trades, net +0.544%, Sharpe ~1.28. Every calendar year 2020-2026 was net positive at K=0.7. The edge decays: 2023 ~+0.07%, 2024 ~+0.29%, 2025 ~+0.29%, 2026 YTD ~+0.15% per trade. Longs carry most of it (+0.84% net vs shorts +0.21%). The same pre-study FAILED on SOL (net +0.04% / -0.02%) and LINK (+0.10-0.15%), and DOGE was only marginal (Sharpe ~0.5). ETH was therefore chosen with that cross-symbol fragility disclosed, and it is a selection the analyst should deflate for. Source_ref: https://github.com/QuantConnect/Tutorials/tree/4a341890296f7e79e095508f06170c72ccaa629c/04%20Strategy%20Library/05%20Dual%20Thrust%20Trading%20Algorithm (sections 'Step 2: Implementation of algorithm' for the range formula and 'Step 3: Trading Implementation' for the cap/floor and reversal rules); credit QuantConnect, Michael Chalek and Gang Wei.

Restates QuantConnect's 'Option Expiration Week Effect' (Strategy Library #102, from Quantpedia and Stivers & Sun, 'Returns and option activity over the option-expiration week for S&P 100 stocks', JBF 2013). The source buys OEF (S&P 100 ETF) at the start of the week that ends on the monthly equity-option expiration (third Friday, or the Thursday before it when that Friday is an exchange holiday) and holds cash otherwise. OEF is not in our USEQ universe, so this hypothesis uses SPY.USEQ (S&P 500 ETF, 1-DAY bars, catalog span 1993-01-29 to 2026-09-30). SPY is the most heavily optioned underlying in the world, so the mechanism, if real, should be at least as strong there. Stivers & Sun also report the opposite drift in the week AFTER expiration, so the second leg (short SPY through the post-OpEx week, which Reg-T margin allows) is part of the same documented cycle. It makes the strategy long-short and doubles the sample. HONEST CAVEAT: lesson L179 lists option/VIX-expiry windows among date-keyed calendar effects that have died in the factory: SVXY pre-VIX-settlement (overfit, Sharpe 0.45) and ETH Deribit post-expiry fade (negative expectancy, 385 trades). The two differences here are the sample and the tuning. 33 years of SPY gives ~396 OpEx weeks plus ~396 post-OpEx weeks, ~790 trades, against the ~100-385 of the dead variants. And there is ZERO tuning: the plan has no tunable parameters, so there is nothing to search and the deflation step sees N=1. This is a one-shot falsification test of a published effect, not a search. If the fixed rule is not clearly positive, the family is closed for US equities as well.

Classification: (a). The abstract gives a concrete mechanism: a cointegration spread, OU mean-reversion speed, half-life-based selection and look-back, and a spread built from many coins. It does not give entry or exit thresholds, so the z-entry of 2.0, the mean-cross exit and the 3x half-life time stop are standard choices, not the paper's. The full paper would need to confirm its z thresholds, its formation and trading window lengths, which coin class it found best, and whether its edge survives taker fees. It executed at quoted best bid/ask on BitMEX minute data, possibly as maker. Restated for our data: universe of 8 fixed Binance USD-M perps (ETH, SOL, ADA, DOT, LINK, AVAX, LTC, XRP), 4-HOUR bars from 2020-09. Every 6 bars, for each coin i, OLS of log(price_i) on a constant plus the log prices of the other 7 over a 180-bar (30-day) window. The residual must pass a Dickey-Fuller t < -3.9 (Engle-Granger, multi-regressor critical region), and the OU half-life must be 2-30 bars. The 7-coin basket is the paper's 'numerous coins' spread, capped at 8 names per lesson L173. PRE-STUDY (point-in-time, gross of fees, run 2026-10-01 on catalog bars): (1) basket variant, 4H, W=180, DF<-3.9: 3,244 trades, mean gross +0.014% per unit gross notional, win rate 50%; DF<-4.5: 2,121 trades, +0.025%; W=360: -0.101%. (2) The same basket on 1D bars, W=60/90/120: -0.045%, +0.008% and -0.208%. (3) Nine single pairs, 4H, with the same DF and half-life gate: pooled mean about -0.19% per trade. Only LINK/DOT was positive (+0.39% to +0.88%), which is selection out of 18 configurations and not a basis for a proposal. The round-trip cost on 2 to 8 legs is at least 0.10% of gross notional, so measured gross capture is below cost in every framing. RESEARCH LEAD RECOMMENDATION: do not spend a backtest slot on this. It is recorded so that the source idea has a documented falsification. The likely reason for the paper's reported profit is minute-level, quote-aware execution on BitMEX. That puts it in the sub-fee microstructure class this factory has already shown is dead (fee graveyard; lessons L113 and L111).

Source: Falck et al., 'Is Factor Momentum More than Stock Momentum?' (arXiv 2009.04824, 2020), abstract. The abstract's testable claim is that factor momentum exists both directionally (time-series) and cross-sectionally. It also claims that after controlling for stock momentum, only implementations that include the LAST MONTH of factor returns keep significant Sharpe ratios. Stock-level returns mean-revert at short lags, but factor returns continue at all lags. Our venues hold no anomaly-factor ETFs; the catalog was checked and MTUM/QUAL/USMV/VLUE/IWD/IWF/RSP are all absent. We therefore use the factor class we can build honestly from data on hand: industry factors, each defined as a sector ETF's return minus SPY's. The nine original Select Sector SPDRs (XLK, XLF, XLE, XLV, XLP, XLU, XLI, XLY, XLB) have USEQ 1-DAY bars from Dec-1998, about 27 years. XLRE and XLC are excluded because their history is too short. On every 21st session (counted in bars, not calendar dates), compute each spread's trailing 21-session return, f_i = ln(P_i,t/P_i,t-21) - ln(SPY_t/SPY_t-21). Time-series mode: hold sector i at weight sign(f_i)/9 x gross_fraction, and hold SPY at minus the sum of the sector weights. The SPY leg therefore nets the market out, and the book is a pure bet on industry-factor continuation. Hold for 21 sessions, then recompute. The abstract also tests a cross-sectional variant: long the top 3 and short the bottom 3 spreads by f_i, equal weight, with the SPY leg netting any residual. It is exposed as one categorical switch because both implementations come from the source; there is no other added degree of freedom. The universe is fixed at 10 names, within the ~10-name cap from lesson L173, so per-rebalance compute is trivial (10 log returns). How this differs from the pipeline's LargeSmallCapMonthlyMomentumSpread: that bet is a single size spread. This book is a nine-factor industry book, and the paper's specific claim is the 1-month lag, where stock-level reversal and factor-level continuation pull in opposite directions. What the full paper must confirm: (1) the 1-month formation window and monthly holding period, assumed here from 'last month of returns'; (2) that industry factors behave like their 72 anomaly factors (Moskowitz-Grinblatt 1999 and Arnott et al. 2021 report the same short-lag industry and factor continuation); (3) whether they skip any days between formation and holding (we skip none). Sizing: gross_fraction is FIXED at 0.9 of equity per side, with no leverage tuning and no buying-power parameter. A recent SPY strategy died because its Sharpe changed with position size; this design does not use the margin buffer at all. Net market beta is about 0 by construction. Stop rule: a catastrophic book-level stop flattens everything if the book is down 8% from the last rebalance's equity, and the book re-forms at the next scheduled rebalance.

Event-driven, beta-hedged, long-short relative-value trade on BYBIT linear USDT perps (catalog '<BASE>USDT.BYBIT', 1-HOUR bars, pure OHLCV, history 2020-10/2021-10 to 2026-10, gap-checked: BTC 57k bars 0 gaps, LINK 1 gap, XRP 0 gaps). Fixed universe of 10 pre-2021 alts (LINK, ADA, DOT, LTC, ATOM, XRP, BCH, ETC, TRX, XLM), under the lesson-173 cap of about 10 names, plus BTCUSDT.BYBIT as the market factor and hedge leg. Mechanism: a strong market-wide BTC move is a common shock that pushes market flow into every alt. An alt that does NOT follow its own trailing beta during that shock reveals idiosyncratic opposing pressure (a persistent seller during rallies, such as unlocks, treasury or miner supply, or a persistent buyer during selloffs). That pressure does not stop when the shock ends, so the residual keeps widening. This is residual CONTINUATION after non-participation, the opposite sign of the dead residual-reversal family (L113). The pre-study confirmed the sign: the naive 'laggard catches up' trade lost on every name. Pre-study, point-in-time, Bybit 1H, 2021-2026, 10 names, 4h windows, rolling 720h beta/sigma lagged one bar, entry at the next bar, non-overlapping per name. With BTC |4h ret| > 2.5 sigma and the alt's residual against the BTC direction < -1.5 sigma, the BTC-hedged residual over the next 24h averaged +0.447% of alt notional on 1457 events (naive t 4.2, overstated because events cluster on the same BTC shock). It was positive on 9 of 10 names (TRX flat), with years 2022/23/24/26 positive and 2021/25 negative. The same rule on Binance USD-M 1H (2020-2026) gave +0.395% on 1834 events, a cross-venue replication. The surface is monotone and smooth: gross rises with both thresholds (0.11% at k1.5/m1.0, 0.30% at k2.5/m1.0, 0.53% at k2.5/m2.0), so this is not a single-cell spike. Bybit is used to fill an under-represented venue (2.3% of the corpus) with clean multi-year history. No HL (L168), no COIN-M (L172), no calendar trigger (L179), no supplementary gate (L111).

Event-plus-clock, long-only, multi-instrument trade on 10 Binance SPOT pairs (ETH, SOL, XRP, DOGE, ADA, LINK, AVAX, LTC, DOT, BNB vs USDT; BINANCE_SPOT 4-HOUR bars, pure OHLCV, history 2020-08 onward, all 10 pairs present in the catalog). Each coin is handled independently. On every closed 4H bar: r = log return; z = sum(r over last 6 bars) / (std(r over last 180 bars) * sqrt(6)). That is the 24h move scaled to the coin's own 30-day volatility. A shock fires when z >= shock_z (+2.5 baseline). After a shock we do NOT buy. We wait entry_delay_bars = 6 bars (24h), then buy at the close of that bar. We exit on a fixed clock hold_bars = 12 bars (48h) later. Per coin, a new shock is ignored within 72h of the last one. Only upside shocks are traded; downside shocks are not traded, because the pre-study found no edge on them. Point-in-time pre-study on Binance USD-M 4H closes, 2020-06 to 2026-09, 10 coins, using this exact rule. 660 upside events. Mean log return from +24h to +72h was +1.25% per event, with t = 2.51 after clustering same-day events. At 10% notional per event, net of 0.10% cost, calendar Sharpe was about 1.07. With 0 to 24h included the window was weaker, and the first 24h after a shock is the absorption and profit-taking leg. 3-sigma shocks: +1.67% per event on 375 events, calendar Sharpe 0.98. Upside, +24h to +72h, 3-sigma by year: 2020 +1.23%, 2021 +6.06%, 2022 +0.65%, 2023 +0.39%, 2024 +1.82%, 2025 +0.01%, 2026 +0.76% (no negative year). The 2.5-sigma 2025+ subset was flat (-0.10%, about 180 events). That is the main known weakness and the reason for the modest OOS prediction. Downside shocks, the beta-hedged residual version against BTC, and spot-vs-perp volume-share and basis variants were all tested and found null (cluster t < 1), so they are excluded. Routed to BINANCE_SPOT, not USD-M: the trade is long-only and unleveraged (venue routing rule), spot is under its 15% quota, and a 1.25% gross move clears the 0.20% spot round trip about 6x. Expected lifetime trades on spot history: about 550-650.

Cross-venue (BINANCE_SPOT signal legs + BINANCE USD-M execution legs), market-neutral, event-driven. Universe is fixed in advance as the 10 largest non-BTC coins that have both a Binance USD-M perp and a Binance spot pair with 1H history from 2020: ETH, BNB, SOL, XRP, DOGE, ADA, AVAX, LINK, DOT, LTC. This is the cap from L173. On every closed 1H bar, for each coin: (1) beta = rolling 30-day (720 x 1H) OLS beta of the alt perp's hourly log return on the BTCUSDT perp's, lagged one bar. (2) The idiosyncratic 24h move is r24_res = (24h alt log return) - beta * (24h BTC log return), and z_res = r24_res / its rolling 60-day std, lagged. (3) Spot share = 24h spot quote volume / (24h spot + 24h perp quote volume), using close*volume from the *USDT.BINANCE_SPOT and *USDT.BINANCE 1H bars, and share_z = its z-score against the trailing 30 days, lagged 24h. The trigger is z_res <= -z_threshold with share_z >= +share_z_min. That is an alt-specific crash whose selling was unusually spot-routed. Economic claim: a spot-routed idiosyncratic dump is inventory being sold outright by holders, not leveraged positioning. Examples are unlock recipients, treasury or exchange-wallet disposals and spot-only market sells. Spot liquidity on these books is thinner than perp liquidity, so the temporary price impact overshoots. Over the next half-day the perp, arbitraged back by basis traders and market makers who absorbed the flow, recovers part of the move relative to BTC. Hedging with beta x BTC strips out the market factor, which is what killed earlier unhedged capitulation-buy variants. PRE-STUDY (point-in-time, run 2026-10-01 on the local catalog, 1H bars 2020-08 to 2026-09, 10 coins, one event per coin per 48h, entry at the close of the bar after the trigger, residual payoff = long alt minus beta*BTC). At z>=2.5 / share_z>=1: n=146, mean 12h residual return +0.64% gross, about +0.44% after ~0.20% two-leg round trip, win rate 54%, positive in all 7 calendar years (2020-2026), positive on 7 of 10 coins. The edge is monotone in the threshold: z>=2.0 gives n=256 and +0.44% gross; z>=3.0 gives n=97 and +0.74% gross. The same-size drops WITHOUT spot-led volume rebound less (+0.38% at 12h on z>=2.5), so the spot-share condition adds about +0.25%. HONEST CAVEATS from the same pre-study: t-stat is only ~1.7-2.0; on the 6 largest coins alone the spot-led cell gives +0.59% at 12h but n=85 and t=1.74. The 6h horizon is weaker (+0.26-0.35%) and 24h decays (+0.1%), consistent with L166 (information fades within a day), so the clock is set at the 12h peak, not chosen for time-in-market. An adjacent cell in the same pre-study looked strong (perp-led UP shocks reversing at 6h, t=-3.2) but vanished at the 1.5 and 2.5 thresholds. It was discarded as a single-cell artifact and is NOT part of this hypothesis. Library prior art: none specific. The idea is restated from first principles, not a port.

A cross-venue, long-short event trade. The signal is the gap between the price of the same coin's USDT perpetual on Bybit and on Binance USD-M. We do not trade that gap (it is only a few basis points, sd 3-14 bp, far below a two-leg round trip). We read it as a crowding gauge. Bybit's flow skews toward retail and high-leverage traders. When Bybit's perp stays rich to Binance's for a full day, aggressive one-sided demand on Bybit is pushing price faster than arbitrageurs move it back. That positioning tends to unwind, and the coin underperforms over the next half-day. A persistent Bybit discount is the mirror case. The position is a single leg on the BYBIT perp of that coin; Binance USD-M bars are a signal source only. Universe, fixed in advance at 7 names, below the ~10-name cap of L173: SOL, XRP, DOGE, LINK, ADA, AVAX, LTC. BTC and ETH are excluded because the Bybit-Binance basis on the two largest books is arbitraged by large desks and is pure noise (ETH premium sd 2.6 bp). DISCLOSURE: ETH was in the pre-study and showed the opposite sign (-0.37%/trade, n=109), so this exclusion is partly ex-post. Data: 4-HOUR bars from the catalog for <COIN>USDT.BYBIT (Bybit linear perp collector, gap-free from 2020-10 to 2021-10 depending on coin, through 2026-09-30) and <COIN>USDT.BINANCE (Binance Vision klines, 0-2 single-bar gaps). Both were checked in the catalog on 2026-10-01. No supplementary feeds are needed. Point-in-time pre-study (4H closes; enter at the next bar; z = 6-bar mean of premium_bp minus its 180-bar mean, divided by the 180-bar sd; |z|>1.5; 12h clock; non-overlapping) gave these mean contrarian returns per trade: SOL +1.21% (n=77), ADA +1.92% (47), LINK +0.43% (60), LTC +0.39% (36), AVAX +0.27% (33), XRP +0.17% (43), DOGE +0.02% (68). The pooled mean is about +0.66% on 364 trades, against a 0.11% Bybit round trip. The effect is concentrated in SOL and ADA. Pooled t is about 2, which is modest, and the predictions below are set accordingly. This is a new economic claim: venue-composition crowding read off a cross-exchange price gap. It is not a funding, OI or taker gate on a trend trade (L111). It is not a convergence arb (the gap itself is not traded). It is not a symbol swap of the TSMOM survivor (L112). It also adds to the BYBIT (2.2%), cross-venue (7.2%) and long-short (15.6%) buckets, all of which are under quota.

This is a two-instrument lead-lag trade on US ETFs (venue USEQ, 1-DAY bars, history from 2006-05-22, 5121 joint sessions in the catalog: GDX.USEQ and GLD.USEQ, both checked). Gold-miner equities are a levered claim on the gold price: rolling 250-day beta of GDX daily log returns on GLD has a median of about 1.79. Equity investors who hold miners are a different clientele from bullion holders, and they absorb commodity-price news slowly. This is the producer-equity version of the documented commodity-to-equity underreaction (Driesprong, Jacobsen & Maat 2008, 'Striking Oil: Another Puzzle?', JFE). On each daily close, compute the 3-session residual gap = beta_250(t-1) * GLD 3-day log return - GDX 3-day log return. Z-score that gap against its own trailing 250-session mean and standard deviation, using only data up to t. If z > +entry_z, miners have lagged what bullion implies: go LONG GDX and SHORT beta x GLD. If z < -entry_z, miners have overshot bullion: go SHORT GDX and LONG beta x GLD. The position is beta-hedged against the gold price, so the P&L is the miner residual alone. Pre-study (point-in-time beta and z, daily closes 2006-2026, forward window starting the next session): with entry_z=1.5, the forward 3-session hedged residual averaged +0.36% after the long signal (284 signal days) and +0.60% after the short signal (273 signal days). With entry_z=2.0 it averaged +0.87% (152 days) and +1.02% (112 days). Over 5 sessions it was +0.58% / +0.74% at z 1.5. The edge decays by 10 sessions on the long side, so the clock is set at the peak (3-5 sessions), following lesson L166. This is not a calendar trigger, not a crypto mechanism, and not a cross-sectional basket: it is a single economically linked pair with a specific information-diffusion cause. It fills under-represented buckets: long-short (15.6% vs a 45%+ target), pairs, USEQ (4.9%), and a non-BTC data estate.

A directional timing rule for the ETH USDT perpetual swap on OKX. The signal comes from a different market, the Deribit ETH options market. Each day we compute the implied-minus-realized vol spread: vrp = DVOL (Deribit 30-day ETH implied vol index, supplementary key `deribit_dvol`, loaded by supp_loaders.load_deribit_dvol; hourly from 2021-03-24) minus RV, where RV is the annualized close-to-close realized vol over the last 14 daily ETH bars. We z-score vrp against its own trailing 365 daily values. When z > +entry_z, options buyers are paying an unusually large fear premium over what the market actually delivers, so we go LONG for hold_days. When z < -entry_z, options are complacent relative to realized turbulence, so we go SHORT. The economic cause is the dealer/hedger channel: a rich premium means dealers are short puts, and as the premium bleeds their hedges are bought back (vanna/charm flows), so the underlying drifts up. An IV-below-RV state marks exhausted, under-hedged moves that tend to give back. This is a new mechanism family. Every prior DVOL/options run in the corpus TRADED options (dead per L102); this one uses options only as information and trades the cheap linear swap. ETHUSDT.OKX 1-DAY bars come from OkxCollector and are complete from 2020-01-02 to 2026-10-01 with 0 missing bars. PRE-STUDY (point-in-time, DVOL lagged one full day, entry_z=0.5, 7-day non-overlapping hold, net of 0.10% RT): 164 trades, avg +1.33%/trade, win 52%. By year (count, mean net per trade): 2022 26/+2.4%, 2023 31/+0.1%, 2024 42/+0.9%, 2025 37/+1.1%, 2026 28/+2.5%. Longs +1.7%, shorts +0.9%. The RV14 neighborhood is stable (Sharpe 0.53-0.78). RV30 variants were weak. Rejected before writing this: HL-vs-Binance funding carry (decayed by 2026), new-listing short (mean ~0), and spot-perp premium carry (fires only 2020-21).

Event-driven, long-short, multi-instrument trade on 6 OKX USDT perpetual swaps: ETH, SOL, DOGE, LINK, ADA and SUI. It uses 1-HOUR bars and pure OHLCV. Each coin is traded independently with a 1/6-equity sleeve. TRIGGER (a market observation, not a calendar date): a closed 1H bar that meets three conditions. Its close breaks the prior 72-bar high (or low). Its volume is >= 3x the prior 72-bar median volume. Its 1h log return is >= 2.5 sigma (or <= -2.5 sigma) of the prior 720-bar std of 1h returns. THE NEW CLAIM IS THE TIMING, not breakout-following itself. A pre-study on our own OKX catalog (2021-01 to 2026-09, non-overlapping events) measured the forward return in the breakout direction. It is flat to negative over the first 1-4 hours (SOL -0.24% and DOGE -0.21% at h4), because liquidity providers and stop-run absorption fade the first impulse. It then rises to a peak around 18h: ETH +0.62%, DOGE +0.75%, LINK +0.59%, ADA +0.59%, SUI +0.55%, SOL +0.33%. It decays by 36-48h. So the trade waits 4 bars after the trigger and then holds 14 bars, exiting at trigger+18h. Per lesson L166, the clock is set at the measured peak, not at a time-in-market target. PRE-STUDY SETUP: rolling statistics are point-in-time. A 0.05% taker fee is charged on every position change. Each coin gets 1/6 of equity at 1x. PRE-STUDY RESULTS: - Trades: 1,781 (~310/yr). - Average gross return: 0.56% per trade, against a 0.10% OKX round trip (5.6x). Net: 0.46% per trade. - Calendar Sharpe 1.05, max drawdown -23%. - Sharpe was positive in every calendar year: 2021 1.14, 2022 1.13, 2023 0.30, 2024 1.67, 2025 1.10, 2026 0.51. ROBUSTNESS (a plateau, not one tuned point): spot checks across the grid (lookback 48/72/96, volume multiple 2/3/4, sigma 2/2.5/3) gave Sharpe 0.51-1.11, with nearly all years positive in each cell. The same rule on the same coins' Binance USD-M bars gives Sharpe 1.12. Those are the same coins traded on a different venue, so it is not an independent price series. HONEST CAVEAT: on 7 OTHER OKX coins (AVAX, DOT, ATOM, NEAR, FIL, TRX, BCH) the rule reaches only Sharpe 0.45. The edge is concentrated in the most liquid, most retail-traded alts. The 6-name universe was fixed up front: names with OKX history from 2023 or earlier and the fewest catalog gaps. BNB was dropped for having the shortest history and 19 missing bars. VENUE RATIONALE: OKX is 1.5% of the corpus against a >=5% target, while BINANCE is 63%. OKX USDT-swap 1H history is long (ETH from 2019-12, LINK and ADA from 2020-03, DOGE 2020-07, SOL 2021-01, SUI 2023-05). Lesson L168's short-history objection to non-Binance venues therefore does not apply. DATA AND COMPUTE: OkxCollector USDT-swap OHLCV in the ParquetDataCatalog (<BASE>USDT.OKX), with no supplementary feeds. The universe is capped at 6 names per lesson L173. There is no cross-sectional ranking. Each coin's rule is independent, so per-bar work is O(1) per instrument using rolling buffers.

Event-plus-clock liquidity-provision trade on OKX USDT perpetual swaps (OkxCollector, USDT-swap OHLCV, catalog '<BASE>USDT.OKX', history 2020-2021 onward). The universe is fixed in advance as the 10 largest non-BTC coins by market cap that have continuous OKX 1H history: ETH, SOL, DOGE, ADA, LINK, AVAX, DOT, BCH, TRX, BNB. They were picked by market cap, not by back-test results; BNB OKX history starts 2022-12. XRP and LTC were left out only because their OKX 1H catalog history begins 2026-09. There are 10 names, under the L173 cap. On each closed 1H bar, for each coin independently: rz = the bar's log return divided by the rolling 720-bar (30-day) std of hourly log returns, and vz = the z-score of log(volume) against its own rolling 720-bar mean and std. A trigger fires when rz < -shock_sigma (default 4) AND vz > volume_sigma (default 2.5). The strategy then buys that coin's swap at market on the next bar and closes it at market exactly 24 bars later. It holds one position per coin, with no re-entry on that coin while the position is open. Each position is 10% of equity, so up to 10 can be open (gross 1.0x) in a market-wide cascade. The mechanism is not a calendar window and not a supplementary-feed gate. It is a forced-seller overshoot: a single-hour 4-sigma drop on 2.5-sigma volume is dominated by liquidations and stop-outs, and price-insensitive sellers push the price below where discretionary liquidity is willing to clear. The rebound takes about a day to arrive. POINT-IN-TIME PRE-STUDY (OKX 1H; rolling stats use only past bars; entry is at the trigger bar's close, which slightly overstates fills versus the next open; 0.10% RT cost charged): (1) Pooled across 14 coins from 2021-11 onward, with non-overlapping events per coin: 1,120 events, mean 24h forward return +0.57%, win rate 58%, per-event t = 3.0. The forward-return curve by hour is +0.18% at 12h, +0.57% at 24h and +0.39% at 36h, so the 24h clock sits where the forward return peaks (L166). (2) The result is monotone in the threshold: mean 24h forward return is +0.34% at 3-sigma, +0.47% at 3.5, +0.57% at 4 and +0.93% at 5 (volume z > 2.5). (3) The same rule on BINANCE USD-M 2019-06 to 2021-11, a period outside the OKX sample, gives +3.2% mean 24h, win rate 68%, t = 4.4. (4) Calendar-time portfolio on this exact 10-coin list, 10% per position, net of 0.10% RT: Sharpe 1.11, 19.9%/yr, max DD 22% (2021-01 to 2026-09). By year: 2021 +55%, 2022 -5%, 2023 +31%, 2024 +15%, 2025 +13%, 2026 YTD +6%. From 2021-11 only, Sharpe is 0.76. CAVEATS STATED UP FRONT: most events are market-wide (BTC also shocks in the same hour), so clustered by 6h window there are only 365 independent episodes, with mean +0.51% and t = 1.66. The edge depends on letting several coins load up together in a cascade: capping gross at 30% drops Sharpe to about 0.4-0.8. A single-instrument or equal-weight-basket breadth version was weak (t < 1.2), so this is NOT a 'buy the index after a crash' trade. The clock is sensitive: Sharpe is 0.78 at 18h, 1.11 at 24h and 0.73 at 30h, so the clock is FIXED at 24h and not searched.

Cross-venue, market-neutral event trade. For each of 10 liquid alts (SOL, DOGE, XRP, LINK, AVAX, ADA, ETH, BNB, LTC, DOT), on every closed 1H bar compute two numbers. (a) Venue share = 24h base-unit volume on <COIN>USDT.BYBIT / (24h volume on <COIN>USDT.BYBIT + 24h volume on <COIN>USDT.BINANCE), z-scored against its own trailing 720-bar (30-day) mean and std. (b) Shock = 24h log return of the Bybit perp / (std of 1h log returns over 720 bars * sqrt(24)). When shock <= -shock_z AND share_z >= share_z_min, buy the Bybit alt perp at the next bar. At the same time, short BTCUSDT.BYBIT sized at beta * alt notional (beta = 720-bar hourly-return covariance / BTC variance, clipped to [0.3, 2.0]). Close both legs after hold_hours. One event per coin per hold window. Economic cause (a new claim, not a ticker swap): Bybit's flow is disproportionately retail and high-leverage. When a coin's selloff runs with an outsized Bybit share, the selling is liquidation or retail-panic flow, not informed repricing led by Binance. It overshoots the coin's own fair value and partly reverts within ~12-24h. The BTC hedge removes the market-wide crash-rebound part (on its own only Sharpe ~0.6) and keeps the coin-specific overshoot. POINT-IN-TIME PRE-STUDY (local catalog, 2021-10 to 2026-09, both venues' 1H bars, entry at next bar, 10% per alt leg, 0.22% two-leg RT cost): share_z>=1.0, shock<=-2.5, 12h clock gave 560 events, mean BTC-hedged residual +0.87% per event gross, calendar Sharpe 1.09, positive in 6/6 years. The 24h clock gave 400 events, +1.29%, Sharpe 1.02, positive in 5/6 years. CONTROL without the venue-share filter: Sharpe 0.55-0.63, so the Bybit-share condition is what lifts the edge. The unhedged long-only version only reached Sharpe 0.67-0.79 and was rejected. Fading upside shocks showed no edge and is excluded. Data: Bybit linear perp OHLCV (Bybit collector, 1H from 2021) plus Binance USD-M 1H OHLCV (Binance Vision), both confirmed present locally for all 10 coins and BTCUSDT.BYBIT. Universe capped at 10 names per L173. Per-bar work must use O(1) rolling updates.

A single-instrument volatility-timing strategy on QQQ.USEQ daily bars. Every 21 trading sessions, the strategy closes its QQQ position and reopens it at a new notional: target_vol / realized_vol_21d of equity, capped at 1.8x on the 2x Reg-T margin account. Realized vol is the annualized standard deviation of the trailing N daily close-to-close returns (sqrt(252)). The mechanism comes from Moreira & Muir (2017), 'Volatility-Managed Portfolios', Journal of Finance. Equity-index volatility clusters strongly, but expected returns do not rise in proportion to it. Cutting exposure when recent vol is high and adding it when vol is low therefore raises the return per unit of risk. The economic cause is the leverage effect plus volatility persistence. It is not price momentum: no trend, breakout or signal on price direction is used. The 21-session re-strike is a fixed bar count from the first tradable bar, NOT a calendar date (lesson L179), and it exists only so each holding period is a separately measurable trade. POINT-IN-TIME PRE-STUDY on the local catalog (QQQ.USEQ 1-DAY, 1999-2026, net of 3 bp of turnover cost): buy-and-hold had Sharpe 0.50 and peak-to-trough DD ~83% (2000-02). VM21 with a 15% target had Sharpe 0.72 full-sample (0.50 pre-2016, 1.04 post-2016) and max DD ~22%. Lookbacks of 10, 21 and 42 sessions all gave 0.68-0.72, so there is no parameter cliff. On SPY the same rule raised Sharpe from 0.55 to 0.61, which confirms direction but is weaker, so QQQ (higher vol clustering) is the primary. NEGATIVE CONTROLS from the same pre-study, which I am NOT proposing: dollar-neutral sector-ETF 3-12 month momentum (9 SPDR sectors) gave Sharpe -0.25 to +0.05 everywhere. Adding a 150/200-SMA trend gate, a TLT defensive sleeve or a short leg to VM21 did not improve it (0.52-0.67). The edge is the vol scaling alone, so only that is proposed. Venue rationale: USEQ is 4.8% of the corpus; QQQ has decades of daily history; execution is commission-free. It is long-only (an over-represented bucket) because the pre-study showed the short-side variants lose Sharpe.

Market-neutral, cross-venue convergence trade between the SAME coin's USDT perpetual on OKX and on Binance USD-M, restricted to 8 long-tail alts where the two venues' order books are thin and their participant mix differs (CVX, API3, GLM, MAGIC, NMR, AXS, GAS, ZETA). The premium is p_t = ln(close_OKX / close_BINANCE) on aligned 1H bars. A dislocation is dev_t = p_t - mean(p, 168h), with z_t = dev_t / std(p, 168h). When |z| >= z_entry AND |dev| >= min_dev (a HARD floor that the dead SolCrossVenueResidualConvergence lacked), wait ONE bar (stale-print guard). If |dev| is still >= 0.2% at that close, short the rich venue and long the cheap venue in equal base quantity. Exit when dev crosses back through zero or |dev| < 0.1%, or at a max_hold clock. POINT-IN-TIME PRE-STUDY (this session, catalog 1H bars, 2024-06 to 2026-09). A first scan of all 157 coins with >= 8000 aligned bars found 887 events at |z|>3 and |dev|>0.4%, with a pooled 24h convergence mean of 0.88%. In-sample the convergence sign is positive for nearly every name. On the 10-name panel with delayed entry and the still-dislocated filter: n=168, mean capture 0.745%, median 0.534%, 78% of trades > 0.2%, mean hold 8.9h. Yearly mean is positive in 2023 (+1.29%), 2024 (+1.10%), 2025 (+0.74%) and 2026 (+0.56%). The median hourly dollar volume on the thinner venue at event time is ~$2.9M. Selection-bias disclosure: the 8 names were picked by in-sample EVENT COUNT among those with >= ~20k aligned bars, not by in-sample PnL. Each name's convergence mean was positive, as it was across the full 157-name scan. It is a capped universe (<= 10 names, per L173) with O(1) rolling statistics per pair: no ranking, PCA or beta re-estimation, so the compute stays small. Why this differs from the dead cross-venue family: (1) I measured the same mechanism on majors (LINK, BCH, ETC, UNI, DOT, AVAX). Their 24h convergence at |z|>3 is only 0.13-0.25%, which is at or below the 0.20% two-leg cost. The edge exists only in the long tail, where gaps are 3-5x larger. (2) There is a hard minimum-gap floor and a 1-bar confirmation delay. (3) Funding-spread carry is NOT part of the thesis. I measured the deduplicated Binance-OKX 7-day funding differential at 0.1-0.25%, too small to matter. Rationale for the venue: OKX is 1.4% of the corpus and cross-venue is 7.1% (quota >= 15%). Long-short market-neutral is 15.5%. None of these 8 coins appears in the recent corpus. Data: OkxCollector USDT-swap OHLCV (<BASE>USDT.OKX) plus Binance Vision USD-M 1H klines, both already in the catalog for all 16 instruments.

A cross-venue, market-neutral carry trade on one coin at a time. Each coin is held on two venues at once: the Binance USD-M perpetual (<COIN>USDT.BINANCE) and the OKX USDT swap (<COIN>USDT.OKX). Both settle funding every 8h, but their crowds are different, so their funding rates diverge. On mid-cap alts the divergence is large and persistent. When the trailing 3-day sum of (OKX funding - Binance funding) is beyond the threshold, short the perp on the venue paying the higher funding and long the other venue's perp in equal notional. Price exposure cancels, because both legs are the same coin. What is collected is the funding differential, plus any convergence of the cross-venue price premium, which usually points the same way as the rich-funding venue. This is NOT a directional perp trade gated on funding (lesson L111). There is no price bet at all: the P&L source is the cash flow itself. It is also not the dead 12-name same-venue funding-carry basket, which earned the funding LEVEL on one venue. This trade earns the cross-venue funding SPREAD with both legs on the same coin. Point-in-time pre-study (funding_rates table, rows de-duplicated per (symbol, exchange, timestamp) because Binance rows are stored up to 5x; daily sums; 2022-02 to 2026-09). Across ~160 common USDT symbols, sign(trailing-3d gap) x next-10d gap rises monotonically with the size of the gap: 0.06% for |gap| < 0.1%, 0.19% for 0.1-0.2%, 0.29% for 0.2-0.3%, 0.43% for 0.3-0.4%, 0.49-0.61% above that. The gap therefore persists rather than mean-reverting immediately. Majors (BTC, ETH, SOL, XRP, DOGE) almost never reach a 0.25% 3-day gap, so the edge lives in mid-caps. Universe capped at 5 coins / 10 instruments (lesson L173). The 5 were chosen from coins that (a) fire most often and (b) have OKX 1H bars in the catalog from 2022 or earlier (ENJ/ONT/FIL OKX bars start 2024-2026, so they were excluded). This selection is in-sample, and the analyst should know that. Episode study on these 5 (first day |3d gap| crosses 0.35%): 84 episodes. Mean next-7-day captured differential 0.674%, median 0.234%, 52% of episodes capture more than 0.2%. By year: 2022 0.66% (n=28), 2023 0.43% (26), 2024 -0.36% (6), 2025 0.61% (8), 2026 1.52% (16). The distribution is right-skewed and 2024 was a bad year. Expect a modest Sharpe, not a spectacular one. Data: bars from the ParquetDataCatalog for the 10 instruments at 1-HOUR. Binance funding comes from BinanceFuturesDataCollector/FundingCollector, OKX funding from OkxCollector (API + daily CDN from 2022-01-17). Both are in PG funding_rates under exchange binance/okx. The strategy MUST read each leg's own series via the per-leg supp `funding_rates_by_instrument` (supp_collector loads it for every linear-perp leg), accessed point-in-time (supp_as_of / supp_window semantics: observations at or before the current bar only). Funding accrual on held positions is applied by funding_accrual for both BINANCE and OKX, so the backtest P&L includes the carry itself.

A faithful port of ianzeng123's open-source FMZ strategy 'TradFi多品种网格（纯网格版）' (https://www.fmz.com/strategy/540030, Python; credit to the author). The source filters on OKX's instCategory != 1 field (tickSz/lotSz/minSz/ctVal are OKX instrument fields), so its native venue is OKX's TradFi USDT perpetual swaps. Our catalog holds these as <BASE>USDT.OKX bars from OkxCollector, starting between 2026-01-28 and 2026-03-11 (about 7 months; 15m bar counts checked: XAG 23.5k, XPT/XPD 23.1k, TSLA/MSTR/HOOD 20.9k, INTC/AMZN/COIN 20.7k, NVDA/QQQ/CL 20.1-20.2k). XAU and SPY start only on 2026-09-24, so they are excluded. The source scans every TradFi swap. Lesson L173 says rescanning a large universe times out, so the scan is replaced by a FIXED list of the 10 OKX TradFi swaps with the longest history: XAGUSDT, XPTUSDT, XPDUSDT, CLUSDT, TSLAUSDT, NVDAUSDT, MSTRUSDT, COINUSDT, HOODUSDT, QQQUSDT. NATGAS and CRCL are excluded, matching the source default EXCLUDE_SYMBOLS='NATGAS,CRCL'. CATEGORY_FILTER='ALL'. Execution uses 15-MINUTE bars, a coarser version of the 30s polling loop: the strategy fills its resting limit orders against bar highs and lows. The daily-range score is computed internally from UTC-day aggregates of the same 15m bars, so no daily-bar subscription is needed. Every rule in the source is translated, quirks included; see entry_logic and exit_logic. Default parameters exactly as in the source: TOP_N=3, REBALANCE_HOURS=48, HYSTERESIS=0.20, LEVERAGE=3, GRID_RATIO=0.015, GRID_VALUE=50 USDT, LOWER_RANGE=0.10 (it sets both sides; UPPER_RANGE is declared in the description but unused in the code), STOP_LOSS_RATIO=0.3, LOOP_INTERVAL=30s, KLINE_COUNT=20. The only adaptation is GRID_VALUE: 50 USDT is meaningless on our $100,000 account, so it is scaled to a FIXED 1,500 USDT per cell. This is a constant notional that does not compound, which keeps the source's 'fixed per-cell amount' rule. Performance claims in the source are treated as unverified. RESEARCH LEAD CAVEAT, stated honestly: lesson L171 records inventory and grid designs as 5/5 dead, and L167 records faithful FMZ ports as 5/5 dead. This hypothesis exists because the user asked for this exact port. It differs from past grid deaths in three ways. It is on a futures MARGIN account with maker-only fills (0.02% per side), not spot at a 0.20% round trip. Its rung is 1.5%, about 37x the 0.04% maker round trip. It runs on TradFi underlyings, which are untested here. The main risk is not fees but inventory stranded below the grid on trending moves. Priority is set low accordingly.

Faithful translation of the open-source TradingView Pine v6 script 'TRX Grid Long Strategy [3Commas]' by 3Commas (https://www.tradingview.com/script/tgRchEEL-TRX-Grid-Long-Strategy-3Commas/, MPL-2.0; credit to 3Commas). The author calibrated it for BYBIT:TRXUSDT.P on 15m, so it runs on our TRXUSDT.BYBIT linear perpetual, 15-MINUTE bars (catalog TRXUSDT.BYBIT-15-MINUTE-LAST-EXTERNAL covers 2021-08-31 to 2026-09-30). It is a long-only price grid between FIXED ABSOLUTE bounds high_price=0.3713 and low_price=0.2660, with n_levels=35 levels. Level i (i=0..n_levels-1) is at pct=i/(n_levels-1). Geometric mode (the default) puts it at high_price*(low_price/high_price)^pct. Arithmetic mode puts it at high_price-(high_price-low_price)*pct. Level 0 is the top (0.3713) and level 34 is the bottom (0.2660). Geometric step = (0.2660/0.3713)^(1/34) - 1, about -0.98% per level. Each level i in 1..34 is a 'slot' with an owned flag (all false at start). Every bar, for i=1..n_levels-1 in ascending order: BUY slot i if it is not owned, prev_close > level_i and close <= level_i. SELL (close slot i) if it was owned at the start of that iteration, prev_close < level_{i-1} and close >= level_{i-1}. Slot size is investment/n_levels USDT, converted to quantity at the LEVEL price (qty = usdt_per_level/level_i, not the fill price). Source default: 10,000 USDT of 10,000 capital = 286 USDT per slot. Ported: investment = 100% of starting equity ($100,000), so each slot is 100000/35 = $2,857 notional at 1x leverage. Pyramiding up to 200 entries is allowed, so many slots can be open at once. There is NO stop-loss, NO trailing and NO take-profit other than the next-level-up sell, by the author's explicit design. Orders fill at the bar close (process_orders_on_close=true, calc_on_every_tick=false). Port note on the date window: the source's useDateFilter (2025-08-13 to 2026-06-22) is a TradingView backtest-range control, not a trading rule. Keeping it would make the strategy dead outside a past window and in live trading, so it is NOT ported, which is equivalent to useDateFilter=false. All visualization, labels, tables, watermark and 3Commas webhook alert strings are display or ops code and are not ported. HONEST PRIOR: this belongs to the inventory-grid family that lesson L171 records as 5 of 5 dead. A sibling 3Commas FIL Bybit geometric grid is already in the pipeline. The fixed bounds are hindsight-calibrated to TRX's 2025-26 range. TRX traded below 0.266 from 2021 until about Dec 2024. Until price first rises above levels and then crosses back down, NO slot can be bought. The effectively active history is therefore about Dec 2024 to Sep 2026 (roughly 22 months), and the earlier years produce zero trades.