Hypotheses

Faithful translation of ianzeng123's FMZ open-source strategy '交割合约现货套利策略测试版' (credit to the author). The source pairs a SPOT market (exchanges[0], BTC_USDT) with USDT-margined DELIVERY futures (exchanges[1], BTC_USDT.quarter and BTC_USDT.next_quarter; allowedCoins=['BTC']). For each delivery contract it keeps a 1-minute history of spreadRate = (deliveryClose - spotClose)/spotClose (max 5000 points), runs an hourly ADF stationarity test on the last 1000 points, and when the spread is 'stationary enough' and |z| >= 1.0 it opens a hedged pair: z>0 ('positive' / 正套) = BUY spot + SELL the delivery future; z<0 ('negative' / 反套) = SELL spot + BUY the delivery future. It exits when the dollar spread returns into a target band, on a 1.5% combined stop-loss, 1 day before delivery, or after 10 days if profitable. There is no take-profit. PLATFORM MAPPING. Spot leg: BTCUSDT.BINANCE_SPOT (CASH, 0.10%/side taker). Futures leg: Binance USD-M BTCUSDT dated quarterlies (BINANCE venue, MARGIN, 0.05% taker / 0.02% maker), catalog ids BTCUSDT_YYMMDD.BINANCE, 1-MINUTE bars verified present in the catalog for every quarterly from BTCUSDT_210326 to BTCUSDT_261225 (each with ~3-6 months of 1m history before its expiry; BTCUSDT_270326 has only started). Primary instrument = BTCUSDT.BINANCE_SPOT-1-MINUTE-LAST-EXTERNAL; every quarterly is an extra_instrument with its 1-MINUTE bar type. The source discovers pairs once at startup; in a multi-year backtest the port must re-run the discovery whenever a contract expires: at any bar the 'quarter' contract is the listed contract with the nearest delivery date that still has bars, 'next_quarter' the following one; each is paired with spot independently (own price history, ADF cache, fail counter, cooldowns and position record), and a contract is skipped for NEW entries when daysToDelivery < minDaysToDelivery (7). Delivery time is parsed exactly as the source does: the YYMMDD in the symbol at 16:00 UTC (parseDeliveryDateFromString). Bars of a contract after that time are ignored. Futures leg leverage = marginLevel 10 (config leverage 10.0; spot leg is CASH, unlevered). SIZING (source default, kept): orderAmount = 0.05 BTC spot per pair; futures quantity = the same 0.05 BTC (the source's calculateContractAmount floor(spotAmount/ctVal) is a coin-to-contract conversion so the futures notional matches the spot leg; Binance USD-M quarterlies are quoted in base units with 0.001 step, so the hedge is exactly 0.05 BTC). Spot buys require USDT balance >= 0.05*price + minUSDTReserve (10 USDT). With the $100k account this uses ~2-6% of equity per pair — that is the author's sizing and is kept on purpose; per-trade results are judged in % of notional. SPOT SHORTING. The account starts with USDT only, so the source's own checkSpotBalance('sell') fails for a reverse (z<0) entry ('BTC不足', canRetry=false: skip, no cooldown, keep waiting for a positive opportunity). The port keeps that check; coins already committed to the OTHER pair's open positive-arb record are treated as unavailable (otherwise the reverse arb on one contract would sell the hedge of the other — a corruption the source would also suffer, documented as the one deliberate deviation). In practice the strategy therefore runs only the positive (long spot / short quarterly) side unless spot BTC is free. Direction overall: market-neutral (long spot + short future).

Translation of QuantRocket's open-source Moonshot strategy 'ValueMomentumTrendCombined' (vmot_combined.py, Apache-2.0, credit: QuantRocket LLC; modeled on Alpha Architect's VMOT ETF). LONG-ONLY, multi-instrument US equity strategy on venue USEQ, 1-DAY bars. CODE values are used where they differ from the docstring: VALUE_TOP_N_PCT=20 and MOMENTUM_TOP_N_PCT=20 (the docstring says 10). PORTABILITY, stated plainly: steps 2 (value screen: cheapest 20% by EV/EBIT, EBIT>0, Sharadar ART) and 3 (top 50% by Piotroski F-score from 9 Sharadar fundamental components) CANNOT be translated. The factory has no point-in-time fundamentals history; Yahoo supplies prices, splits and dividends only, and src/data/yahoo_quotes.py fetches current-snapshot fundamentals for display, which would be look-ahead. Step 1b (exclude financials/ADRs/REITs) is done by construction of the fixed universe below. In the port, the momentum screen therefore ranks within the whole liquidity-screened universe, where the original ranks only within quality-value stocks. That is a material change: what is left is the momentum + trend half of VMOT, and the analyst should judge it as that. Every other rule is ported exactly. UNIVERSE (fixed, all verified present in the catalog as <SYM>.USEQ-1-DAY-LAST-EXTERNAL; non-financial, non-REIT, US-listed S&P 500 large caps, no ADRs): AAPL MSFT AMZN GOOGL META NVDA AVGO ORCL CSCO ADBE CRM INTC TXN QCOM AMD IBM ACN JNJ PFE MRK ABBV LLY UNH AMGN TMO ABT MDT BMY GILD PG KO PEP WMT COST MCD NKE HD LOW SBUX DIS CMCSA VZ T XOM CVX COP CAT DE HON GE MMM UNP UPS LMT RTX BA LIN (59 stocks), plus SPY.USEQ as the trend instrument. The original TREND_SID FIBBG000BDTBL9 is SPY; SPY is a signal only and is never traded. Primary instrument = SPY.USEQ (its bar drives the daily decision clock); the 59 stocks go in extra_instruments/extra_bar_types. RULES (per daily close t, computed on synchronized closes): (1c) avg_dollar_volume = rolling 90-bar mean of close*volume; rank descending pct across names with data (pandas rank pct, average ties); liquid = rank <= 0.60. (4) momentum = (close[t-22] - close[t-252]) / close[t-252] (NaN if close[t-252]==0 or missing); rank descending pct ONLY among liquid names; have_momentum = rank <= 0.20. (5) positive_days = count of days with close.pct_change() > 0 over the last 252 bars; rank descending pct ONLY among have_momentum names; selected = rank <= 0.50. (6) Equal weight: w_i = 1/N_selected for selected names, 0 otherwise. (7) Quarterly rebalance 'Q-NOV': the stock weights are recomputed only from the signal of the last trading day of February, May, August and November, i.e. the last value inside each Q-NOV period. They are held constant until the next such date; there is no drift rebalancing in between, because Moonshot charges turnover only on weight changes. (8-9) Trend overlay on SPY: below_zero = SPY close / SPY close[t-252] - 1 < 0; below_mavg = SPY close < SPY 252-bar SMA; n = below_zero + below_mavg. This is sampled WEEKLY ('W' = weeks ending Sunday), taking the last trading day of each week (normally Friday's close), and held through the following week. Final weight = w_i if n==0, w_i/2 if n==1, 0 if n==2; the remainder stays in cash. EXECUTION LAG (target_weights_to_positions = weights.shift()): a weight change decided at close t is executed on the next session's bar t+1 with market orders. Positions are resized only when a name's FINAL target weight changes, either at a quarterly reselection or at a weekly trend-count change. Sizing: target notional_i = final_weight_i * account equity at leverage 1.0 (USEQ margin is not used); whole shares, floor; skip orders below min notional. Fees: platform USEQ (zero commission, spread+impact modeled) replaces the original $0.005/share PerShareCommission. PERFORMANCE ENGINEERING (two recent USEQ multi-name ports died on the 300 s sandbox limit): keep per-name fixed-length deques (253 closes, 90 dollar-volumes, 252 positive-day flags with a running sum, 252 SPY closes with a running sum for the SMA). Update O(1) per bar, and do the cross-sectional ranking only once per day after all 60 bars for date t have arrived (key on the bar ts, process when the date's bar set is complete or on the SPY bar), and ONLY on rebalance/trend-sample days. Never rescan history.

Port of QuantRocket's open-source Moonshot strategy 'TrendDayStrategy' (code 'trend-day', https://github.com/quantrocket-codeload/trend-day, Apache-2.0, author QuantRocket LLC, who gets the credit). ORIGINAL RULES, exactly as written: universe 'leveraged-etf' on US-stock 1-minute data; DB_TIMES 14:00 and 15:59; MIN_PCT_CHANGE = 0.06. Each day: return = (OPEN of the 14:00 bar - CLOSE of the prior session's 15:59 bar) / prior session close. Signal +1 (long) if return > +0.06, -1 (short) if return < -0.06, otherwise 0. Weights: allocate_fixed_weights_capped(signals, 0.20, cap=1.0), so each active position gets 20% of capital, and all weights are scaled down proportionally if gross exposure would exceed 100%. Positions are entered the same day as the signal with a MARKET order at 14:01 (the close of the 14:00 bar is the entry price) and exited with a market-on-close (MOC) order at the 16:00 session close. There are no stops, no take-profit and no overnight holding. The original's costs were $0.005/share commission and 3 bps slippage; ours use the USEQ model, which charges zero commission and models spread and impact. PLATFORM ADAPTATION, stated honestly: (1) The universe is the 3 leveraged/volatility ETFs that exist in our USEQ universe and catalog: TQQQ.USEQ, SQQQ.USEQ and UVXY.USEQ. Other leveraged ETFs such as SOXL, SPXL and TNA are not collected. (2) Timeframe is 1-HOUR, not 1-minute. USEQ 1-minute bars start 2026-08-06, about 38 sessions, which would give fewer than 10 signals at a 6% threshold. USEQ 1-HOUR bars cover 2024-08-14 to now, about 530 sessions. Yahoo hourly bars are close-stamped at 10:30, 11:30, 12:30, 13:30, 14:30, 15:30 and 16:00 ET, so no bar closes at 14:00. The checkpoint is therefore the CLOSE of the bar stamped 14:30 ET, 30 minutes later than the original's 14:00, and the hold is 14:30 to 16:00, not 14:01 to 16:00. On half-day sessions (13:00 close) there is no 14:30 bar, so no trade, which matches the original because its 14:00 bar is missing on half-days. (3) The prior session close is the close of the previous session's final bar (stamped 16:00 ET, or 13:00 on half-days), equivalent to the original's 15:59 bar close. Sizing: 20% of current equity notional per active instrument, as in the original. With 3 instruments, gross exposure is at most 60%, so the 1.0 cap never binds but is still implemented. Long and short are allowed; USEQ is a Reg-T margin account, run at leverage 1.0. Note that TQQQ and SQQQ are mirror images, so a strong Nasdaq day usually signals both, long one and short the other. That is effectively the same bet on two instruments, net about 40% of equity on the Nasdaq direction. The original does this too and it is kept faithfully. Pre-study on our 1h catalog (gross, close-to-close, no costs), 6% threshold, return from 14:30 to 16:00 in the signal direction: TQQQ n=45, mean +0.31%, hit rate 64%; SQQQ n=44, mean +0.41%, hit rate 66%; UVXY n=94, mean +0.98%, hit rate 70%. That totals about 183 position-days in about 2 years. Treat these numbers as unverified until our backtest runs.

Translate QuantRocket's executable GDX_GLD_Pair subclass, preserving its daily rolling Johansen weights, one-standard-deviation bands, independent persistent long/short signal states, daily target-weight recalculation and next-session-open execution. Use GDX.USEQ as primary and GLD.USEQ as the secondary trading leg, both with 1-DAY-LAST-EXTERNAL bars. Original defaults are LOOKBACK_WINDOW=20, BBAND_STD=1, DB_FIELDS=['Close','Open']; the subclass selects GDX and GLD from usstock-1d. Original USStockCommission charges $0.005/share; replace that commission with platform USEQ zero commissions and modeled spread/impact, explicitly reporting the difference. Do not claim source profitability or add filters, stop-losses, cointegration significance tests or parameter optimization.

Test the submitted moving-average ordering mechanism independently on LTC and LINK perpetuals using 4-HOUR bars. Preserve the five SMA periods, five-bar confirmation, rising ADX filter and ordering-break exit. Replace the source's fixed forex lot size with capital-relative volatility sizing. This is a source-derived replication hypothesis, not an established crypto edge. Maintain incremental indicators and bounded buffers rather than rescanning historical data.

Translate ianzeng123's executable multi-asset allocation strategy, preserving the supplied code rather than correcting it to match the prose. Trade BTC, gold, SPY and crude-oil USDT perpetuals on Bybit. The code uses PERIOD_H1, not the description's 15-minute covariance bars. Preserve signed risk-parity weights, volatility-dependent exposure, initial/timed/weight-change/direction-change rebalancing and emergency partial reductions. Source defaults are TRADE_MODE=paper, INIT_CAPITAL=10000 USDT, LOOKBACK=60, EWMA_LAMBDA=0.94, TARGET_VOL=0.15, MAX_LEVERAGE=3, REBALANCE_HOURS=24, DRIFT_THRESHOLD=0.15, POSITION_RATIO=0.95 and MIN_REBAL_USDT=5. The platform adaptation uses its actual $100000 account equity and engine accounting, with real commissions and execution costs; do not reproduce the source's fee-free simulated PnL as backtest results. Preserve the source paper branch's allocation decisions separately from engine accounting. No parameter optimization is authorized.

This is a tradeable long-only equity strategy by QuantConnect, licensed Apache-2.0, rather than merely an indicator utility. Its trading universe is dynamically selected from fundamental-enabled US equities. A faithful platform translation requires historical coarse-universe membership, has_fundamental_data flags, adjusted_price observations and selection timestamps. The inspected UsEquitiesCollector supplies a configured stock/ETF universe and split-adjusted bars, which do not establish equivalence to those inputs. Substituting a fixed basket would change the strategy. Original backtest settings are 2018-01-01 through 2019-01-01, initial cash 1000000, daily universe resolution and security leverage 2. Platform starting cash would be 100000, an explicitly recorded platform adaptation.

Translate the supplied executable strategy faithfully, preserving all defaults and Boolean conditions. Trade XAGUSDT.BINANCE on XAGUSDT.BINANCE-1-HOUR-LAST-EXTERNAL, matching the author's silver-perpetual market and hourly timeframe. Use the platform's $100,000 starting equity, MARGIN account, 2.0 leverage, capital-relative sizing and actual execution costs. This is a fixed-parameter replication experiment; advertised performance is unverified. Preserve Pine indicator initialization, historical indexing, missing-value behavior, crossover semantics and order timing. With the supplied strategy declaration's default execution model, evaluate signals at completed bars and execute market orders on the next available tick, normally the next bar open; do not fill using the already-observed signal close.

Test the article's asymmetric pyramiding mechanism: enter with one of ten equal capital allocations, add only after favorable price advances, and move a shared protective stop after every completed grid step. Trade LTCUSD.HYPERLIQUID in both directions using 4-HOUR bars. Adapt the illustrative 10% stop to a baseline 4% stop, confirm entries and grid advances at completed bar closes, and impose explicit exposure, risk and duration limits. This is a trend grid that adds to winners, with no averaging into losses.

Test the source's concrete mechanism: construct five evenly spaced levels from the preceding 3–5 days' trading range, enter against excursions, take profits after half a grid interval, and liquidate on a range break. Adapt it to independent LTC and BCH perpetual grids using 4-HOUR bars. Replace exponential martingale sizing with equal, capital-relative tranches and a fixed loss budget. Each instrument may hold only one inventory direction at a time, consistent with Hyperliquid's netting account. This is an experiment in compensated inventory provision, not an assumption that grids inherently generate profits.

Test one mechanism from 发明者量化-小小梦: resting purchases below an anchor, a separate profit-taking sell for each filled node, and replacement of that node only after its inventory is sold. Adapt to LTC Binance Spot with percentage-scaled prices, equal capital allocations, wide profit targets and mandatory inventory liquidation controls. This is a research hypothesis, not a demonstrated profitable strategy. Virtual-order activation and grid relocation are execution features of the same mechanism, not additional hypotheses.

Test the source's combined breakout and reversal mechanism on LTCUSDT.BYBIT using 1-MINUTE bars, with positions held for hours rather than minute scalps. Freeze six reference levels from the preceding complete UTC day's high H, low L and final close C. Use the prose's previous-day interpretation rather than the sample's rolling two-record extrema and minimum close. Maintain daily extrema and observation flags incrementally. Implement as a FactoryStrategy subclass with bounded state and constant-time per-bar work.

Distil the source's single mechanism: maintain prefunded coin and cash inventories on two spot exchanges, buy LTC on the cheaper exchange and sell matching existing LTC on the dearer exchange, then repair actual fill imbalances. Preserve aggregate LTC inventory without borrowing, leverage or routine inter-exchange transfers. Attribute incremental trading profit against an identical passive-inventory benchmark; the prefunded LTC retains directional market risk. This is a conditional research proposal, not an established profitable strategy.

Test whether temporary deviations of the LTC/BCH log-price ratio from a latent mean-reverting state predict economically meaningful convergence over 1–5 days. Use synchronized 4-HOUR bars and equal-dollar opposing perpetual positions. Adapt the source's state-space mechanism while explicitly defining spread polarity: a high LTC-minus-BCH spread means short LTC and long BCH. A prediction error alone is insufficient; entry also requires forecast convergence large enough to pay execution costs. No source performance claim is accepted as evidence.

Test whether sector winners over a multi-month formation period outperform sector losers over the next 21 trading sessions. Trade eight liquid US sector ETFs using synchronized daily bars, buying the two highest-ranked sectors and shorting the two lowest-ranked sectors. Exclude the most recent 21 sessions from formation returns to separate medium-term persistence from immediate reversal. This is an adaptation, not a replication of the original stock-decile study. Evaluate subsequent executable returns, never the formation-period averages shown in the source.

Test whether a breakout in the LTC/BCH log-price ratio after compressed relative volatility predicts continued relative repricing over three days. Buy the relative winner and short the relative loser with equal dollar notionals. HyperliquidCollector supplies both perpetuals' 4-HOUR OHLCV. Catalog inspection verified 5,883 synchronized observations from 2024-01-25 16:00 UTC through 2026-10-02 00:00 UTC, without missing four-hour intervals. A baseline trigger-count check found 134 compression-breakout events before economic eligibility filters; forward returns were not evaluated. Implement as a FactoryStrategy with timestamp-synchronized bounded buffers and constant-size calculations.

Cross-venue, market-neutral carry on one coin, DOGE, across two perpetual venues: DOGEUSD.HYPERLIQUID (HyperliquidCollector OHLCV plus hourly funding_history, data/supplementary/hyperliquid/funding_history/DOGE, from 2023-05-12) and DOGEUSDT.BINANCE (USD-M perp, Binance Vision OHLCV plus 8h fundingRate, data/supplementary/fundingRate/DOGEUSDT). The catalog holds 1-DAY bars for both legs (DOGEUSD.HYPERLIQUID-1-DAY-LAST-EXTERNAL and DOGEUSDT.BINANCE-1-DAY-LAST-EXTERNAL), giving about 1,200 overlapping days. The two venues serve different populations. Hyperliquid is on-chain, retail and leverage-hungry, with hourly funding. Binance is deeper and has more arbitrage capital. Funding on the two venues therefore diverges, and the divergence is persistent, not noise. Pre-study on local data (daily summed funding; spread = HL minus Binance): - When the trailing 3-day mean spread exceeded +0.05%/day, the next 10 days' summed spread averaged +1.20% (n=223 signal-days). - When it was below -0.05%/day, the next 10 days averaged -0.83% (n=92). - At a 0.08%/day threshold the figures were +1.41% and -1.04%. - The same persistence holds on SOL, LINK, AVAX, SUI and WIF (+0.6% to +1.5% over 10 days), so this is not a DOGE artefact. DOGE is chosen because it is strongest in both directions. - HL DOGE funding has 0 conflicting duplicate timestamps; the AAVE pair died on exactly that audit. Only two instruments on daily bars (~1,200 bars), so no risk of a sandbox timeout. Precompute both funding series as sorted numpy arrays once in __init__ and read them with supp_as_of / bisect only. Venue choice: a cross-venue, long-short, Hyperliquid-plus-Binance design fills the most under-represented buckets (cross-venue 7.3% vs a 15% target; long-short 16.8%; HL 7.6% vs a 20% target). It avoids BTC entirely. Not the dead 'Binance-vs-OKX funding differential' (that died no_trades with a 0.35% 3-day threshold): HL-vs-Binance gaps are 3-10x larger and the trigger here fires on ~25% of days. Not L111: no directional price bet is taken. The return is a contractual cash flow that the engine accrues on both legs.

A sector-neutral, dollar-neutral, multi-instrument LONG-SHORT short-term reversal strategy on US equities (venue USEQ, 1-DAY bars, UsEquitiesCollector Yahoo split-adjusted daily history; every leg verified in the local catalog back to at least 1999-05). Academic basis: Da, Liu & Schaumburg (2014, Management Science, 'A Closer Look at the Short-Term Return Reversal') and Hameed & Mian (2015) show that the classic 1-week reversal is weak in raw returns because industry momentum offsets it. Once the industry/sector component is removed, the remaining idiosyncratic residual reverses strongly. The reversal comes from liquidity provision to non-informational order flow. Universe: 4 sectors × 6 names. Technology: AAPL, MSFT, NVDA, ORCL, CSCO, ADBE (reference XLK). Financials: JPM, BAC, WFC, C, GS, MS (reference XLF). Energy: XOM, CVX, COP, EOG, SLB, OXY (reference XLE). Health care: JNJ, PFE, MRK, ABT, UNH, LLY (reference XLV). Sector ETFs are SIGNAL-ONLY and never traded. Each weekly rebalance, compute each stock's residual return over the last `lookback_days` sessions: resid_i = r_i − beta_i × r_sector. beta_i is a rolling 120-session OLS beta to its sector ETF, clipped to [0.3, 2.0]. Scale it by the stock's own 60-session daily residual vol to get z_i. Within EACH sector, go long the `n_per_side` most-negative z and short the `n_per_side` most-positive z, equal dollar. That gives sector-neutral and dollar-neutral construction, so market and sector beta are largely hedged out. The trigger is a market observation (cross-sectional residual rank), not a calendar date. The weekly cadence is only the evaluation clock. This is a different family from the dead crypto cross-sectional reversal: it runs on a zero-commission venue and uses residuals hedged against sector ETFs. It is also not dual-class pairs (that died at QA on a 0.04% convergence threshold) and not PEAD (that died on a sandbox timeout). IMPLEMENTATION NOTE (sandbox timeouts killed several multi-instrument daily runs): keep fixed-length deques per instrument and update rolling sums incrementally in on_bar/on_extra_bar. Do the O(N) ranking ONLY on the rebalance bar, once every instrument has printed the same session date. Never rescan history. Venue rationale: USEQ is 5.6% of the corpus, and long-short is under-represented (16.8%). The BINANCE concentration (61.9%) and BTC are avoided entirely.

An event-driven, long-short, BTC-hedged strategy on BYBIT linear USDT perps (Bybit collector, catalog ids <BASE>USDT.BYBIT, 1-DAY-LAST-EXTERNAL bars). Mechanism: tokens that launch on a perp with heavy day-one turnover (airdrops, TGEs, memecoin launches) are priced at peak attention while their float is small. Over the following weeks airdrop recipients, market makers paid in tokens, and early holders hedging locked allocations sell into the float, and attention fades. Hong, Lim and Stein-style attention decay plus IPO long-run underperformance (Ritter 1991) are the TradFi analogues. The 'listing curse' on CEX launches is widely reported (Kaiko and others, 2024-2025). LISTING DETECTION IS POINT-IN-TIME: an alt's launch day is its first daily bar in the data stream, and the strategy sees nothing before that. The universe rule is ex ante, known at entry: average daily turnover (close*volume) over the first 3 daily bars >= $100M. Every instrument below satisfies that rule and was checked on disk (first bar 2023-01 to 2026-01, >=120 daily bars, still printing at 2026-09-30). PRE-STUDY on these 76 events, gross of fees and funding: short at the close of day 3, hold 21 days, BTC-hedged 1:1 notional, close-based +50% adverse stop. Mean +13.4% per event, median +25%, win rate 75%, worst -74% (stop gapped). Mean per year was positive in each of 2023, 2024, 2025 and 2026 for the 30-day variant; the 2023 sample is small. Without the hedge the mean is about 0, so the edge is RELATIVE underperformance versus BTC, not a bear-market bet, and the hedge is mandatory. Survivorship caveat: the sample contains only listings still trading today. Delisted launches mostly collapsed, which biases the estimate AGAINST the short, but a squeezed-then-delisted name would be missing from the sample too. This differs from the abandoned Binance 'post-listing lottery-premium decay' run, which died on data_unavailable (one leg had no bars) rather than on its premise. This one uses Bybit, whose launch coverage is broader and every leg verified locally, plus a turnover-qualified ex-ante universe, a shorter 21-day clock, a hard intrabar stop, and a funding-squeeze exit. Fills under-represented buckets: BYBIT venue (2.3% vs >=5% target), long-short (16.7% vs long-only 83%), multi-instrument, event-driven non-calendar trigger. PER-TRADE ECONOMICS: gross capture about 13% of short-leg notional, against a Bybit round trip of 0.11% on the short leg and the hedge. Funding paid by shorts on crowded launches averaged a few % per 30 days, with outliers near -20% (KAITO, MYX), which the funding exit is designed to cut. A net expectation of 3-6% per short is about 30-50x the round-trip cost. PERFORMANCE: per-bar work must be O(1): keep per-instrument counters/small fixed-length arrays, never rescan history; funding via supp_as_of/supp_window only.

Test whether investors demand compensation for asymmetric exposure to crypto market losses. Use synchronized 4-HOUR bars across LINK, AVAX, XRP and DOGE perpetuals, selecting one long-short pair with similar ordinary market betas but different downside betas. Long the more downside-sensitive asset; short the less sensitive asset. This deliberately accepts downside risk rather than forecasting price reversal or ranking momentum. Source: [Is downside risk priced in cryptocurrency market?](https://www.sciencedirect.com/science/article/pii/S1057521923004635), which reports a positive downside-beta premium in historical cryptocurrency samples; transfer to these four Hyperliquid perpetuals remains unproven. HyperliquidCollector supplies exact perpetual OHLCV. Local parquet timestamps were checked: common coverage begins 2024-02-07T12:00:00Z and ends 2026-10-02T00:00:00Z, with isolated gaps in three legs. Trade only on complete synchronized observations and restart warm-up after gaps.

Test whether unusually large, high-participation altcoin advances reverse over the following three days after volume contracts and price retreats. Trade independent short positions on INJ, NEAR, OP and SEI perpetuals using 4-HOUR bars. HyperliquidCollector supplies the OHLCV data. Reading actual parquet timestamps confirmed approximately 5,881–5,901 bars per instrument, with common coverage from 2024-01-25 through 2026-10-02 and one or two gaps per series; reset indicators after gaps. A development-only trigger-frequency check on the first 80% of each series found 49 candidate events combined, approximately 23 annually, without measuring subsequent returns. Inspiration: [Karbalaii's price-and-volume anomaly detection research](https://arxiv.org/abs/2503.08692). Its Poloniex detection results motivate the trigger, but do not establish profitability on these perpetuals.

Cross-sectional long-short factor on 14 OKX USDT perpetual swaps using daily OHLCV bars. Each coin has contiguous daily history from 2020 or early 2021 through 2026-10-02 (checked in the catalog: BTC/ETH/BCH/TRX/ETC from 2020-01-02, SOL from 2021-01-24, all with a max gap of 1 day). The source is the BigQuant price-volume correlation factor (https://bigquant.com/wiki/doc/I5oStgChtF): factor = -1 x corr_20d(close/close[-1], log(volume/volume[-1] + 1)). In each coin, it measures whether up-days come with volume surges (attention- or FOMO-driven buying, high correlation) or whether volume surges come on down-days or are unrelated to direction (low or negative correlation). Every rebalance_days daily bars, rank the 14 coins by PVC_i = Pearson corr over corr_lookback days of (daily simple return, daily log volume change). Go long the n_per_side coins with the LOWEST PVC and short the n_per_side coins with the HIGHEST PVC. The legs are equal-weighted, dollar-neutral, and gross exposure equals 1.0x equity (each side 50%). The factor sign is pre-registered and fixed, not tunable, so the optimizer cannot flip it. Coins that remain in the same bucket at a rebalance are kept, not re-traded, which cuts turnover. This is a new mechanism family for the corpus: the only related hypothesis was a single-asset BTC Wyckoff divergence. It is not momentum, not calendar-keyed, not a supplementary-feed gate on one perp, and not options. It also fills under-represented buckets: OKX venue (2.0% vs >=5% target), long_short (16.7%), and multi-instrument. Execution: market orders at the daily close bar, positions held for about 1 week. OKX funding is accrued as a cost or credit by the engine; it is not a signal. Risk: each leg has a fixed 25% adverse catastrophe stop (relevant for short squeezes), and a stopped leg stays flat until the next rebalance. Max single-leg exposure is 1/(2*n_per_side) of equity (16.7% at the baseline n=3). Leverage is set via config.leverage = 1.0 and is NOT an optimization_plan parameter. Implementation note: keep a rolling deque of the last corr_lookback+1 closes and volumes per instrument, updated in on_bar/on_extra_bar, and compute correlations only on rebalance bars. This is O(14 x lookback) per rebalance; never rescan history (avoids the sandbox timeout).

Cross-venue signal, hedged long-short execution. For 11 liquid alts (LINK, AVAX, DOT, ATOM, NEAR, FIL, ADA, LTC, UNI, AAVE, INJ) compute the daily spot volume share = spot_volume(<ALT>USDT.BINANCE_SPOT) / (spot_volume + perp_volume(<ALT>USDT.BINANCE)), both in base units from 1-DAY bars. A 3-day mean of the share is z-scored against the prior 60 days, lagged 3 days so the event window is excluded. An event is a 3-day alt log return MINUS the BTC perp 3-day log return below -move_k * sigma_daily(30d) * sqrt(3) while share_z > share_z_entry. The relative selloff came through the spot book rather than leveraged perps. That points to holders, unlock recipients or treasuries dumping inventory, which is non-informational and pushes price below where derivatives traders value the coin. Perp-led selloffs carry information (leveraged shorts, liquidations) and do NOT rebound in the pre-study. On each event: buy the alt perp and short an equal dollar amount of BTCUSDT.BINANCE, so the trade is beta-neutral to the crypto market, and hold for hold_days. PRE-STUDY (run 2026-10-02 on local catalog daily bars, 2020-2026, 11 coins, gross and before costs): 142 events at k=1.5, z>1. Mean 3-day BTC-relative return +2.51% (overlap-adjusted t 2.08), positive in every calendar year 2021-2026 (2021 +3.7%, 2022 +0.3%, 2023 +2.1%, 2024 +4.2%, 2025 +1.6%, 2026 +1.8%). The only losing year was 2020, with a few events at -4.6%. Controls: perp-led relative selloffs (share z<-1) +0.74% (t 0.49); perp-led relative rallies faded -0.24%; the mirror 'spot-led rally continuation' worked only in 2021 and lost in 2024-2026, so it was rejected and is NOT part of this hypothesis. The edge sits where the market observation (venue composition of selling) differs, not in a calendar or a funding/OI gate. Data: BinanceVisionCollector klines for both BINANCE (USD-M) and BINANCE_SPOT. All 23 1-DAY bar series were verified present in data/catalog: perps from 2020-01 (LINK, ATOM) through 2020-09 (AVAX), spot from 2019-2020, all through 2026-10-01. Daily bars keep the run at about 55k bars total, well inside the sandbox budget. Risk: 12% of equity notional per alt leg plus an equal BTC hedge leg, at most 4 concurrent events (96% gross long + 96% gross short at 2x account leverage), one open event per coin.

A dollar-neutral, multi-instrument, cross-sectional long-short on 10 liquid cross-asset ETFs on USEQ daily bars, OHLCV only: SPY (primary), EFA, EEM, EWJ, EWZ (equity regions), TLT, IEF, LQD (rates/credit), GLD, SLV (metals). Every name has daily data in the catalog through 2026-10-01; the common overlap starts 2006-04-28 (SLV), giving about 20 years. The signal is George & Hwang (2004) 52-week-high anchoring, not trailing return: proximity_i = close_i / max(close_i over the last lookback_high sessions). Investors anchor on the 52-week high, so they under-react to news that pushes an asset toward it and over-sell assets far below it. Proximity has been shown to subsume plain momentum and does not suffer the same momentum crashes. The cross-ASSET version applies because bond, metal and regional-equity ETFs carry the same retail anchoring as stocks. Every hold_bars sessions, counted by primary-bar index rather than a calendar date (lesson L179 is respected because the trigger is a ranking, not a date window), rank the 10 by proximity. Go long the top n_legs and short the bottom n_legs. Each leg is weighted by inverse 60-session realized volatility, normalized so that the long book = gross_exposure/2 of equity and the short book = gross_exposure/2. Only legs whose membership or weight changes by more than 25% relative are traded. This differs from the in-pipeline IndustryFactorShortLagMomentumSpyHedged, which ranks 9 equity sectors on trailing return versus SPY: here the universe crosses asset classes and the signal is distance-to-high, which is weakly correlated with return momentum across asset classes. Precompute: keep a rolling deque per instrument (max 252) and update on each bar, giving O(1) per bar and avoiding the sandbox timeouts seen on prior USEQ baskets. Rows are aligned only when all 10 ETFs have a bar with the same ts_init. Risk: book-level stop at 12% equity drawdown from the last rebalance peak, which flattens all legs until the next rebalance. Sizing uses get_account_equity() with mark-to-market legs. Fixed risk budget: gross_exposure 0.9 (0.45 long / 0.45 short; within Reg-T 2x and safe at engine leverage 1). Max single-leg notional is 25% of equity. The engine's configured leverage is left at its default and is NOT pinned in the optimization plan.