Hypotheses

This is a faithful translation of the MIT-licensed pinkfish example 'double-7s-ave-portfolio' by Farrell Aultman (https://github.com/fja05680/pinkfish/blob/18a0d557ead25e0e82de3ece86bcab6c325a59ed/examples/portfolios/double-7s-ave-portfolio/strategy.py). The credit belongs to the author. The rules come from Larry Connors and Cesar Alvarez, 'Short Term Strategies that Work'. It is a tradeable strategy, not a utility. With the default options (periods=[7], sma=200, use_regime_filter=True, margin=1.0, use_adj=False), the 'portfolio' has one synthetic symbol, SPY_7. It reduces to the classic single-ETF Double 7s: a daily mean-reversion dip-buy. The source passes the symbol in at runtime. Its examples use SPY, so we trade SPY.USEQ on 1-DAY bars. These bars have decades of history, which gives the factory's most robust backtest venue for this mechanism. USEQ is commission-free, with about 0.02-0.05% round-trip spread and impact. Dividends are not in PnL. The prices are split-adjusted, and SPY has never split, so use_adj=False is matched. The strategy is long-only with no shorts and no leverage. The source's margin=1.0 means the position is sized at 100% of equity × weight. Do NOT add any 'leverage' key to the optimization plan. The engine default (1.0) already equals the source. Any performance claimed by the source is unverified; our backtest with real costs decides. Portfolio-diversity note: this is a long-only single-symbol hypothesis. It is admitted because it is a user-requested faithful port, it is on the under-used USEQ venue (6.6%) rather than BINANCE futures, and it is not crypto/BTC.

Test whether large LTC/DOGE relative-price deviations converge over several days when their preceding relationship is stable. Trade synchronized 4-HOUR bars: LTCUSD.HYPERLIQUID is primary; DOGEUSDT.BYBIT is the equal-dollar opposite leg. Litecoin and Dogecoin share merged-mining infrastructure, providing a possible common supply channel. This does not establish a stable price ratio; convergence is the falsifiable research premise. Economic background: [Litecoin merged-mining explanation](https://litecoin.com/news/how-litecoin-and-dogecoin-created-one-of-the-most-robust-pow-networks). Library inspiration: [Pair-trading history-length discussion](https://quant.stackexchange.com/questions/80948/pair-trading-history-length-for-spread-calculation), especially its warning against selecting windows merely because they produce favorable cointegration tests. No code is copied. Use only the exact instruments' overlapping history; local availability remains unverified.

Test whether an upside breakout driven disproportionately by Hyperliquid perpetual turnover, without corresponding Binance spot participation, predicts a multi-day unwind after the breakout fails. Trade synchronized 4-HOUR bars on three fixed assets; Binance spot instruments are signal sources only. This is a research hypothesis, not a measured edge. Use executable market orders after completed-bar decisions, with actual fees, impact and funding included. No next-open execution contract, dynamic universe, option rolling or supplementary signal feed is required.

Test whether the largest positive daily return in the preceding month predicts subsequent relative performance across a fixed eight-asset Hyperliquid basket. Long the two highest-MAX assets and short the two lowest, holding for one week. Build UTC daily closes from synchronized 4-HOUR bars. This is a prospective adaptation, not a faithful reproduction, of [Ozdamar, Akdeniz and Sensoy's MAX study](https://link.springer.com/article/10.1186/s40854-021-00291-9), which reports positive weekly high-minus-low MAX returns. Evidence is horizon-dependent: [Yadav's intraday study](https://pure.jgu.edu.in/id/eprint/9270/) reports the opposite relationship at an hourly horizon. Neither establishes profitability for this fixed, liquid perpetual universe.

Test whether a failed excursion outside an established trading range predicts movement toward its opposite boundary over subsequent days. Trade HBAR, ICP and XLM perpetuals independently using 4-HOUR bars, with at most two concurrent positions. This is an original, simplified experiment inspired by the spring/upthrust concepts described in https://www.prorealcode.com/prorealtime-indicators/wyckoff-event-detection/; it does not reproduce that indicator or claim that OHLCV proves institutional accumulation. Signals use completed bars, and market orders execute after the decision bar becomes available under the factory's supported execution model. Exact 4-HOUR catalog directories were verified for all three instruments; historical depth and completeness must be checked before evaluation.

Event-driven, dollar-neutral, cross-sectional long/short on 13 Hyperliquid perpetuals using 1-DAY bars. BTCUSD.HYPERLIQUID is used only as the signal (it is never traded). The trigger is a large 3-day BTC move. At that point each alt's 3-day log return is split into an expected part (rolling 90-day beta to BTC x BTC's 3-day return) and a residual. The residual is signed by BTC's direction. A positive signed residual means the alt over-responded in BTC's direction: it rose more than its beta implies in a BTC rally, or fell less in a BTC selloff. The hypothesis is that these residuals persist rather than mean-revert. Alts that diverge from their beta in a market-wide shock are showing real information: who is being bought as the rotation target, and who is being dumped as the funding source. That information diffuses slowly over the next week or two. This is the opposite of the 'laggards catch up' intuition, and my pre-study rejected catch-up outright (see expected_edge). It is not a plain momentum rotation: there is no position unless BTC has shocked, ranking uses the beta-residual rather than raw return, and it is flat roughly 50% of the time. Venue rationale: HYPERLIQUID holds 9.1% of the corpus against a >=20% target, and long_short holds 18% against a long-only share of 82%. The design is pure OHLCV (L189) with no extra feed. Universe, chosen for multi-year HL daily history and to avoid the names in recent HL failures (SUI/INJ/SEI/TIA, AVAX/NEAR/DOT-ATOM, LINK): AAVE, BCH, LTC, UNI, CRV, FIL, ETC, XRP, ADA, BNB, DOGE, DOT, ETH. Risk: each leg is 1/6 of 75% gross, so 12.5% of equity per leg; no leverage is needed, and gross stays <=0.75x. Stop: there is no per-leg price stop, because pair P&L is what matters. A basket-level stop closes all six legs if the combined open P&L falls to -6% of equity-at-entry x gross. Max exposure is 12.5% per leg and 75% gross.

A long-only, unleveraged rotation across 10 alts that trades the coins' BINANCE_SPOT markets and uses their Binance USD-M perpetual daily bars only as a signal. Universe: SOL, ADA, AVAX, LINK, DOGE, XRP, DOT, LTC, ATOM and NEAR. BTC and ETH are left out on purpose, to reduce the factory's BTC/ETH concentration. Each day, basis = perp_close / spot_close - 1; both daily bars close at 00:00 UTC. Take the average of that basis over basis_lookback_days. Every 7 days, rank the coins from most negative average basis to least. Hold the top_n coins whose perp sits at the deepest persistent discount, equal-weight, but only if the coin's spot close is above its 50-day SMA. Mechanism: a perp that stays below spot means leveraged traders are net short while cash buyers are absorbing supply in spot. Those crowded shorts tend to unwind over days to weeks. This is the same 'cash-financed versus leverage-financed demand' idea as the surviving volume-share strategy, but it reads the imbalance from price (the basis) instead of from volume. It is pure OHLCV with no extra data feed (L189), trades on spot (L192) and targets multi-percent moves. It is not the abandoned 4H 'Perp-Led Selloff Absorption Panel', which was a sigma-triggered entry after a crash with a 12h exit and traded perps. Risk: 1x leverage, 1/top_n of equity per leg, at most 100% gross exposure, and a -15% catastrophic stop on the daily close. Fee arithmetic: these alts move about 9% in absolute terms over 7 days. With a relative edge of 1–1.5% per week, and about half of the legs held into the next week (so no fee on those), the expected net return is about 1.5–2.5% per closed leg. The spot round trip costs 0.20%, so that is more than 5x the cost. History runs from about 2020-09, roughly 260 rebalances.

Test whether agreement between relative price strength and sustained willingness to pay funding predicts multi-day continuation across LTC, BCH, ETC, UNI, AAVE and FIL Hyperliquid perpetuals. Use synchronized 4-HOUR bars and instrument-specific settled funding. Local catalog directories were confirmed for all six instruments. This is an original adaptation inspired by FMZ's funding-plus-momentum framework, not a source reproduction: https://www.fmz.com/strategy/548617. Use fixed factor weights, a fixed universe and seven-day holding periods; exclude its rolling regression and dynamic weight fitting. Estimate performance only over the intersection of actual bar and funding coverage.

Translate the signal-generating strategy by je-suis-tm, supplied under Apache-2.0, without correcting its unusual conditions. The source uses GBPUSD bid/ask-average observations from an unspecified-frequency CSV; it defines neither executable orders nor position sizing. This explicitly registered platform adaptation uses DOGEUSDT.BINANCE_SPOT 1-HOUR close prices, long-only cash execution, and 10% of current equity per entry. Preserve all source signal rules and defaults: 20-observation Bollinger bands, sample standard deviation (ddof=1), multiplier 2, period=75, alpha=0.0001 and beta=0.0001. Alpha and beta remain absolute quote-price amounts, never percentages. This tests a crypto adaptation, not a reproduction of the original GBPUSD dataset or any claimed performance.

QuantRocket LLC's Apache-2.0 QuantitativeMomentum is a tradeable, long-only equity strategy modeled on Alpha Architect's QMOM ETF. Preserve its complete NYSE stock universe, financial/ADR/REIT exclusions, liquidity screen, skipped-month momentum screen, positive-day smoothness screen, equal weights and fiscal-quarter rebalancing. A faithful reproduction is not currently established as feasible: the platform's documented USEQ collection covers an S&P 500 plus ETF universe, rather than the source universe and its exclusion lists. Do not substitute present-day large caps, ETFs or crypto instruments. Establish historical universe semantics, membership, exclusions, delisted securities and comparable Close/Volume inputs before development.

Test whether unusually heavy trading without proportionate price progress identifies exhausted speculative positioning. Use synchronized 4-HOUR bars for SUI, INJ, SEI and TIA Hyperliquid perpetuals. Among assets advancing at similar rates, short the asset requiring abnormally high turnover and buy the asset advancing on ordinary turnover. Hold the matched pair for five days, targeting relative repricing rather than a small spread between exchanges. This is an original hypothesis, not a reproduction of the supplied library strategies. Trade using the supported bar-close decision path; no next-open execution contract is required.

Test whether unusually low trading activity distinguishes temporary inventory-driven relative price shocks from informative repricing. Trade a fixed six-perpetual basket using synchronized 4-HOUR bars, buying relative losers and shorting relative winners only when their preceding 24-hour dollar volume is below its historical baseline. Hold paired positions for three days. Inspired by Bianchi, Babiak and Dickerson's cryptocurrency liquidity-provision research: [Trading volume and liquidity provision in cryptocurrency markets](https://eprints.lancs.ac.uk/172093/1/Babiak_Trading_Volume.pdf). This is a new Hyperliquid experiment, not a reproduction of the paper's aggregate-exchange daily portfolios. Exact catalog coverage must be verified before coding; missing instruments must not be substituted.

Test whether deteriorating funding after sustained one-sided crowding predicts a multi-day positioning unwind in LTC. Trade LTC perpetuals on synchronized 4-HOUR closes, using Binance LTC spot prices only to confirm that the move extends beyond a local perpetual premium. This is an original directional hypothesis, not a funding-harvest or inter-exchange arbitrage strategy. Funding mechanics source: [Hyperliquid funding documentation](https://hyperliquid.gitbook.io/hyperliquid-docs/trading/funding). Fee source: [Hyperliquid fee schedule](https://hyperliquid.gitbook.io/hyperliquid-docs/trading/fees). Evaluate only the intersection of available bar and funding history; do not assume the approximately 2.3-year candle capacity establishes actual coverage.

Test whether asymmetric price response to traded volume predicts relative performance after ordinary market weakness. Use synchronized 4-HOUR bars on a fixed six-asset Hyperliquid universe. Long the asset with the lowest downside-versus-upside price-impact proxy and short the asset with the highest. This tests persistent selling fragility rather than return momentum, spread convergence or capitulation reversal. All six exact 4-HOUR catalog directories were confirmed locally; usable common history must be established before evaluation. Feature-design inspiration: https://github.com/stefan-jansen/machine-learning-for-trading/tree/main/08_financial_features. This is an original hypothesis, not a reproduction or a measured discovery.

Test whether persistent one-sided funding followed by an opposing move confirmed in NEAR spot predicts a further multi-day unwind in the Hyperliquid perpetual. Trade only NEARUSD.HYPERLIQUID, both long and short, using synchronized 4-HOUR bars; NEARUSDT.BINANCE_SPOT is a signal-only instrument. Both exact bar directories exist locally. Funding is a positioning proxy, not proof of trader inventories. Official documentation establishes hourly settlement and the direction of payments: [Hyperliquid funding](https://hyperliquid.gitbook.io/hyperliquid-docs/trading/funding). The predictive relationship is an untested hypothesis.

Test whether persistent asymmetric selling pressure predicts subsequent relative underperformance between LINK and AVAX. Use synchronized 4-HOUR bars to buy the asset with the smaller downside-variance share and short the asset with the larger share, holding for several days. Catalog filenames indicate overlapping history beginning in February 2024; validate actual coverage before evaluation. Inspired by upside/downside semivariance measures in [Zikes and Barunik](https://arxiv.org/abs/1308.4276); their paper does not establish this crypto trading edge. This is an original, falsifiable adaptation rather than a source reproduction.

Test whether asymmetric exposure to market selloffs persists across AVAX, NEAR, DOT and ATOM perpetuals. Using synchronized 4-HOUR bars, buy the two assets with the smallest downside-versus-upside market sensitivity and short the two with the largest. Hold each basket for seven days rather than trading small spread fluctuations. This is an original conditional-risk hypothesis, not a source-strategy reproduction. Research on cryptocurrency low-volatility effects is mixed; conventional low-volatility sorting is not established evidence for this specific mechanism. Source: https://www.sciencedirect.com/science/article/pii/S154461232030667X. Required catalog coverage must be confirmed before implementation.

Test whether opposing volume-weighted closing pressure predicts multi-day relative repricing between AAVE and UNI before their returns diverge. Trade synchronized 4-HOUR perpetual bars, long the positive-pressure asset and short the negative-pressure asset with equal entry notionals. Local catalog filenames indicate overlapping history beginning January 2024; actual continuity must be verified. This is an original research hypothesis, not a source replication or an established edge.

Test whether high-volume excursions beyond established trading ranges that close back inside predict multi-day inventory unwinds on AAVE, LINK and LTC Hyperliquid perpetuals. Trade both directions independently using synchronized 4-HOUR OHLCV bars, with no funding-income assumption and no cross-exchange spread trade. This is an original hypothesis, not a source replication. Use only each instrument's actual available history; approximately 5000 four-hour candles provide up to 2.3 years before gaps and warm-up. Historical membership is a fixed named basket, with survivorship limitations disclosed. Fees follow the engine's VIP-0 configuration and the official schedule: [Hyperliquid fees](https://hyperliquid.gitbook.io/hyperliquid-docs/trading/fees).

Test whether unusually large, volume-backed relative advances between BCH and LTC partially reverse after the advance stops accelerating. Trade synchronized 4-HOUR perpetual bars, buying the relative laggard and shorting the leader with equal dollar notionals. Both exact bar directories exist locally. This is a conditional sector-relative convergence experiment, not an assumption that the price ratio is permanently cointegrated. Inspiration: [QuantInsti's Principal Component Analysis in Trading](https://blog.quantinsti.com/principal-component-analysis-trading/), specifically separating shared exposure from asset-specific movement; this proposal uses a transparent fixed log ratio rather than fitted PCA. Evaluate only the instruments' overlapping history.

Test whether a large BCH-versus-LTC relative price advance reverses after the outperformer's abnormal trading volume subsides and relative price leadership breaks. Trade both Hyperliquid perpetuals using synchronized 4-HOUR bars; both exact bar directories exist locally. This is a falsifiable attention-exhaustion hypothesis, not an assumption of cointegration. Use only the overlapping history actually available, without fabricating missing bars or historical universe membership. No funding income is assumed. Hyperliquid base-tier fees are documented at [Hyperliquid Fees](https://hyperliquid.gitbook.io/hyperliquid-docs/trading/fees).

Concrete trading idea: translate Farrell Aultman's MIT-licensed Pinkfish sma-percent-band strategy, preserving the supplied rules and defaults. Source: https://github.com/fja05680/pinkfish/blob/18a0d557ead25e0e82de3ece86bcab6c325a59ed/examples/strategies/sma-percent-band/strategy.py . Trade LTCUSDT.BINANCE_SPOT using 1-DAY close-stamped bars, long-only, without leverage. Source defaults are use_adj=False, use_cache=True, sma=200 and band=0; the explanatory 5% band is an example, not the default. Fetch/cache and adjustment settings are data-preparation options, not trading filters. Calculate pf.CROSSOVER with timeperiod_fast=1, timeperiod_slow=200 and band=0; drop indicator-invalid rows before trading. Record daily marked-to-market equity. The supplied code delegates regime calculation and default purchase sizing to Pinkfish helpers; their pinned implementations could not be retrieved here and must be verified before claiming an exact translation. No performance claim is verified.

Faithful reproduction of the supplied MIT-licensed PairTradingStrategy by 用Python的交易员, commit fcad8e2d86825506b9d33441ab78be853c2c3bdd. Platform mapping: LTCUSDT.BYBIT is leg 1 and BCHUSDT.BYBIT is leg 2, both linear perpetuals in a NETTING MARGIN account with $100,000 initial equity and configured leverage 1.0. Preserve price_add=5, boll_window=20, boll_dev=2, fixed_size=1, leg1_ratio=1 and leg2_ratio=1. Crucially, fixed_size is never used: target positions are literally signed one-unit quantities, not capital-relative allocations. Ratios affect the spread only, never order quantities. This is a source-reproduction experiment, not a claim of a fee-viable edge. Do not add hedge estimation, normalized prices, filters, stops, adaptive sizing or optimization.

Concrete trading idea: faithful translation of KingKeltnerStrategy by 用Python的交易员, MIT licensed, at commit fcad8e2d86825506b9d33441ab78be853c2c3bdd. Source: https://github.com/51bitquant/howtrader/blob/fcad8e2d86825506b9d33441ab78be853c2c3bdd/examples/strategies/king_keltner_strategy.py. Preserve source defaults kk_length=11, kk_dev=1.6, trailing_percent=0.8 and fixed_size=1. Source ticks generate minute bars; minute bars aggregate into calendar-aligned five-minute bars. Initialization requests 10 days of historical bars and trading waits until the default ArrayManager is initialized. Resolve the pinned ArrayManager implementation before coding: its exact initialization length and keltner calculation are dependencies, not specified in the supplied strategy. This is a fixed-rule reproduction experiment, with no claim of verified profitability. The platform adaptation replaces one source quantity unit with one capital-relative allocation unit; every signal and exit rule remains unchanged.