Hypotheses

A cross-venue, long/short, multi-asset daily strategy. Signals come from Binance USD-M perp and Binance spot daily bars. They are used only as signals and never traded. Execution is on OKX USDT perpetual swaps, a venue the corpus barely uses (2.5% of experiments). Universe: LINK, ADA, AVAX, DOT, ATOM, UNI. The rule picks pre-2021 large caps that have continuous 1-DAY bars on all three venues from 2020 (OKX history verified back to 2020-03..09). Signal per coin per daily close: perp_share = perp_base_volume / (perp_base_volume + spot_base_volume), both from Binance, same coin, base units. psz is the z-score of perp_share over the trailing `share_lookback` days. ret7 is the 7-day return of the OKX swap close. Thesis: an advance carried by spot buying is financed with cash. It has no liquidation overhang, so it tends to continue. This is the mechanism behind the promoted long-only spot volume-share strategy. An advance carried disproportionately by perps is financed with leverage. Crowded longs, rising funding and liquidation fuel make it fragile, so it tends to unwind. The hypothesis tests BOTH sides of that mechanism as a long/short book on a different venue. It does not re-run the long-only spot version. Execution substitution: decisions are made on the daily close and filled by a market order at that close. On a 24/7 venue that price is the next bar's open (the registry's use_equivalent for next_bar_open_fill). Protective stops are intrabar reduce-only stop-market orders matched by the platform (supported). There is no exchange-resting requirement. Sizing: each position's notional = equity * risk_per_trade / (atr_stop_mult * ATR14%), capped at 15% of equity per coin, with at most 6 concurrent positions (one net position per coin) and leverage 1.5. Gross exposure is at most ~90% of equity times 1.5. Platform defaults apply where unstated. Funding on the OKX swaps is charged through the funding ledger: longs in rallies typically pay it and shorts receive it.

Long-short, single-instrument, pure-OHLCV strategy on the OKX USDT perpetual LINKUSDT.OKX (4H bars, history from 2020-03). OKX is at 2.4% of experiments against a >=5% quota, and LINK on OKX has essentially no history in the corpus, so this spends unused data rather than re-mining BTC/ETH on Binance. Mechanism: volatility clustering. Long stretches of compressed realized range on an alt (Bollinger bandwidth in the bottom percentile of its own trailing distribution) are followed by range expansion. The first decisive 4H close outside the compressed band points to the direction of the expansion. This is different from plain Donchian or momentum trend-following because it only trades breakouts out of a measured compression state. Otherwise it stays flat; it should be in the market roughly 25-35% of the time. Sizing: 1.5% of equity risked per trade. Notional = equity * 0.015 / (atr_stop_mult * ATR14 / price), capped at 1.5x equity notional (leverage 2 configured, so the cap binds well inside the margin). Execution substitution (per the capability registry): 'enter at the next bar open' is implemented as a market order at the signal bar's close, which on a 24/7 venue is the next bar's open. The protective trailing stop is a platform-matched intrabar reduce-only stop-market (_submit_stop_market), re-placed each bar. No exchange-resting order is required. Fee arithmetic: on LINK 4H, compression breakouts historically extend 6-12% over the following 2-6 days when they follow through. About 40-45% of them follow through; the rest stop out at roughly -1.0 to -1.5 ATR(4H) ≈ -2.5 to -3.5%. Expected gross per trade ≈ 0.42*8% - 0.58*3% ≈ +1.6%. Discounting by half for slippage and optimism gives ≈ +0.8% per trade against an OKX round trip of ~0.10% (8x cost). Firing rate: LINK has a compression state crossing the percentile gate about 2-3 times a month. Requiring a band-exit close gives roughly 20-30 trades a year, which is ~150+ over 6.5 years. It fires in both bull and bear regimes, so it does not depend on crash-only triggers. Inspiration: the TTM/Bollinger squeeze literature (Carter, 'Mastering the Trade'; John Bollinger's 'Squeeze' — many variants in the FMZ corpus under state/library/ext-strategies). No code copied.

A long-short panel on five long-history OKX USDT perpetual swaps (LINKUSDT.OKX, ADAUSDT.OKX, DOTUSDT.OKX, ATOMUSDT.OKX, BCHUSDT.OKX; 4H bars, history from 2019-2020). OKX is 2.4% of the corpus against a 5% quota, long-short is 18% against a 45%+ target, and none of these five series has been used on OKX. The mechanism is volatility clustering plus range release, not momentum ranking. On each 4H close, for each instrument, compute the box width over the last box_bars bars, defined as (highest high - lowest low) / close. A COMPRESSION state holds when that width's percentile rank within its trailing 180 values is at or below squeeze_pct. A position opens only when the bar after a compression state CLOSES outside the box: long on a close above the box high, short on a close below the box low. Orders are market orders at the signal bar's close, which on a 24/7 venue is the next bar's open (the registry's use_equivalent for next_bar_open_fill). Each position gets an intrabar protective stop: a reduce-only stop-market at the opposite side of the box at entry, then trailing at trail_atr x ATR(14) from the best close. It triggers inside the bar in the simulated venue, which is supported natively, and nothing is hosted on the exchange. A failed-break time stop applies: exit if the position has not gained at least 1 ATR within 6 bars (24h). Sizing is risk-parity per leg: risk 1% of equity per trade at the initial stop distance, notional capped at 25% of equity per instrument, at most 3 legs open at once (about 75% gross), leverage 1-2x. The family is a TradingView-style 'squeeze' breakout without the Bollinger/Keltner dressing, here the plain box-width percentile. No supplementary feed is used (L189). Funding is paid or received through the backtest ledger.

A dollar-neutral, cross-sectional, long-short panel strategy on OKX USDT perpetual swaps. It uses daily OHLCV only. OKX is 2.4% of experiments against a target of at least 5%. The universe is fixed. All eight coins are listed on OKX, have daily history from 2020 and have stayed liquid. They were chosen by listing date, not by performance: LINKUSDT.OKX, BCHUSDT.OKX, ADAUSDT.OKX, ETCUSDT.OKX, ATOMUSDT.OKX, XLMUSDT.OKX, TRXUSDT.OKX and AVAXUSDT.OKX. The signal is the MAX effect (Bali, Cakici & Whitelaw 2011, 'Maxing out: stocks as lotteries'; also reported in crypto cross-sections). For each coin, MAX_i is the mean of its `max_days` largest daily close-to-close returns over the trailing `max_lookback` daily bars. Retail traders looking for lottery-like payoffs bid up the coins with the highest MAX, and those coins then underperform. Levered retail is the marginal perp buyer on alts. The coins with the lowest MAX get little attention and outperform. Every `rebalance_days` daily bars (fixed at 7), rank the eligible coins by MAX. The target is short the top 2 and long the bottom 2, each leg 15% of equity notional: gross 60%, net 0, leverage 1. A leg whose target has not changed is not traded, so only rank changes cause orders. This roughly halves turnover. This is not a momentum or trend bet. MAX looks at the tail of the return distribution and is only weakly correlated with trailing return. It also differs from the live XRP skewness strategy, which is a time-series bet on one coin. This one is a relative, market-neutral bet across eight coins on a different venue. Execution: decisions are made at the daily bar close, and market orders fill at that close. On a 24/7 venue that is the faithful equivalent of a next-bar-open fill; this substitution follows the capability registry. Protection: each leg carries an intrabar protective stop (a reduce-only stop-market) at `stop_atr_mult` x 20-day ATR from entry. The platform matches it inside the bar in backtest and paper, and live trading mirrors those fills. The stop matters most for the short legs, because the lottery names are the ones most prone to squeezes. A stopped coin cannot re-enter until the next rebalance. The signal needs no funding feed; the engine charges funding in PnL as usual. Platform defaults cover anything not stated here.

Faithful translation of hummingbot's controllers/market_making/pmm_simple.py (PMMSimpleController, commit 9af100d6822da7d2d0291a906c730ef172284ee2, author: hummingbot, Apache-2.0; credit goes to the original authors). The controller file only overrides get_executor_config(). Every trading rule comes from the defaults in MarketMakingControllerBase / MarketMakingControllerConfigBase and from the PositionExecutor triple barrier, and each one is named below so the developer can translate them line for line. The source makes no performance claims; our backtest with real fees decides. MARKET (source default): connector binance_perpetual, trading_pair WLD-USDT, mapped to WLDUSDT.BINANCE (USD-M perpetual, maker 0.02% / taker 0.05%) on 1-MINUTE bars. Binance has 1m history for WLD back to 2023-07. BINANCE is over-represented in the corpus. A faithful port still keeps it, because the source's default connector and pair are Binance perpetual WLD-USDT. WLD itself is rarely used in the corpus, and the mechanism is under-represented on several counts: it trades long and short, works on a short horizon, and makes markets by providing passive liquidity on both sides with a triple barrier on each order. LEVELS: buy_spreads [0.01, 0.02] and sell_spreads [0.01, 0.02] give four level ids: buy_0, buy_1, sell_0 and sell_1. The optimization plan accepts only scalars, so these lists are declared as the scalar parameters buy_spread_0=0.01, buy_spread_1=0.02, sell_spread_0=0.01 and sell_spread_1=0.02. buy_amounts_pct and sell_amounts_pct default to None, which means equal distribution: total_amount_quote is split 50% buy side / 50% sell side, then equally across each side's levels, so each level gets 25% (scalars level_amount_pct_buy_0 = buy_1 = sell_0 = sell_1 = 0.25). The reference price is the mid price with spread_multiplier = 1, because pmm_simple has no custom processed_data. Level price = reference_price * (1 - spread) for buys and reference_price * (1 + spread) for sells. Level amount (base) = level quote amount / level price. SIZING (platform adaptation): the source's total_amount_quote default is 100 USDT on a 20x account. On our $100k engine, total_amount_quote = total_notional_equity_fraction (1.0) x get_account_equity() at executor creation. Each level is therefore 25% of equity notional, and the maximum one-sided exposure is 50% of equity (two levels on one side). config.leverage = 20 is the source default and within the BINANCE 20x cap; it only sets margin. Round quantities with make_qty, guard price > 0, cap notional at the computed value, and skip any level whose notional is below the $5 minimum. EXECUTION SUBSTITUTIONS (from context.execution_capabilities): (1) hedge_mode is unsupported (the source's position_mode default is HEDGE), and per_executor_deadlines are unsupported. The equivalent used here: the strategy keeps an internal ledger of up to four VIRTUAL executors, one per level id. Each has its own entry price, filled quantity, creation timestamp and barriers. Every virtual entry and exit is sent as a real order on the single NETTING position. The venue's net position therefore always equals the sum of the open virtual executors, and account PnL and fees equal the sum of the executors' PnL and fees, since in hedge mode each fill would also have been a separate order paying the same fee. Only margin usage and position reporting differ from hedge mode. (2) Mid price: this design uses no quote ticks, so the reference price is the latest 1m bar close. (3) Hummingbot ticks about once per second; we evaluate on each 1m bar close. executor_refresh_time of 300 s is 5 bars, and cooldown_time of 15 s rounds up to 1 bar, so a closed level may re-quote at the next bar close. Limit entries and the limit take-profit rest in the simulated venue and fill intrabar when a bar trades through their price. The stop loss is an intrabar protective stop: a reduce-oriented stop_market order on the net position, matched inside the bar by the platform's simulated venue and mirrored live. Time-limit and trailing exits are market orders at the close of the bar that detects them.

Classification: concrete trading strategy. Faithful translation of VolatilitySystem by freqtrade, GPL-3.0, at commit f3340ce11f5bdf62f598522e64d1f5638eaa13f5. Source: https://github.com/freqtrade/freqtrade-strategies/blob/f3340ce11f5bdf62f598522e64d1f5638eaa13f5/user_data/strategies/futures/VolatilitySystem.py. Use DOTUSDT.OKX linear perpetual, 1-HOUR bars and fixed 2.0 leverage. The supplied source specifies no instrument or exchange, and the prior contract is absent from this brief; this venue assignment is therefore explicit rather than a claimed recovery of its previous assignment. Preserve every active source rule, including its single position addition. Correct the validation error by storing minimal_roi as the finite scalar 100.0, representing the source schedule {"0": 100}; this means a 10000% return threshold, not 100%. Execute market orders at the signal bar close, the registered equivalent of next-bar-open execution on a continuous market. Performance remains unverified.

Faithful signal-rule translation of Kevin Ossenbrück's SwingHighToSky, version 1.0, INTERFACE_VERSION 3. Credits: Bloom Trading and Mohsen Hassan. Preserve 15-minute bars, long-only entries, the original static stop and elapsed-minute ROI schedule. Use the embedded buy_params and sell_params as the executable baseline; distinguish these from the IntParameter constructor defaults. No performance claim is verified. The platform supplies LINKUSDT.BINANCE_SPOT, equity-relative sizing and real fees because the source specifies neither a pair nor stake size. Signal orders execute at the completed signal bar's close, the registry's faithful equivalent of next-bar-open execution on a continuous market. Intrabar protection is platform-matched. Source: [pinned SwingHighToSky](https://github.com/freqtrade/freqtrade-strategies/blob/f3340ce11f5bdf62f598522e64d1f5638eaa13f5/user_data/strategies/SwingHighToSky.py).

Translate BbandRsi by berlinguyinca, credited in the code to Gert Wohlgemuth and converted from Mynt, under GPL-3.0. Preserve the source's 1h timeframe, RSI(14), typical-price Bollinger bands(20, 2), long-entry and exit conditions, 10% minimal ROI from minute zero, and -25% stoploss. Source: [BbandRsi pinned source](https://github.com/freqtrade/freqtrade-strategies/blob/f3340ce11f5bdf62f598522e64d1f5638eaa13f5/user_data/strategies/berlinguyinca/BbandRsi.py). The source specifies no pair or stake size; this platform baseline uses LINKUSDT.BINANCE_SPOT, long-only CASH at 1x. Signal orders execute at the completed signal bar's close, the registry's equivalent for next-bar-open execution on a continuous market. Protective stops trigger intrabar through platform matching. No performance claim is treated as verified.

Faithful translation of the supplied executable grid strategy, including its implementation quirks. Source defaults remain grid_step=2.0, profit_step=2.0, trading_size=0.5, max_pos=7.0, profit_orders_counts=3, trailing_stop_multiplier=2.0 and stop_minutes=360.0. Price steps are absolute quote-currency amounts, not percentages. Trade LTCUSDT.OKX using quotes, fill callbacks and one-second simulated-clock timer events; bars must not replace quotes or timer events. The platform adaptation is a linear OKX margin instrument with one net position, actual limit-order matching, maker fee 0.0200% and taker fee 0.0500% per fill from src/backtesting/venue_configs.py. Opposing buy/sell orders adjust one net inventory rather than independently hedged positions. Preserve the source's bid/ask-priced LIMIT liquidation orders; they are not stop-market orders or guaranteed immediate exits. The author's claims about ranging-market performance remain unverified. Retrieve GridPositionCalculator and the source timer cadence from the same pinned commit before implementation; their behavior must not be guessed.

Translate MACDStrategy by Gert Wohlgemuth (berlinguyinca), GPL-3.0, from the supplied source at commit f3340ce11f5bdf62f598522e64d1f5638eaa13f5. Use LINKUSDT.BINANCE_SPOT, long-only, 1x, 5-MINUTE bars. Preserve the MACD/CCI predicates, positive-volume guards, ROI schedule and -30% stoploss. TA-Lib defaults are MACD fastperiod=12, slowperiod=26, signalperiod=9 and CCI timeperiod=14; calculate macdhist even though it is unused. Parameter declarations specify buy_cci=-50 and sell_cci=100, but embedded buy_params=-48 and sell_params=687 are the effective baseline under default parameter loading. This precedence is documented in [Freqtrade Hyperopt](https://www.freqtrade.io/en/stable/hyperopt/). Preserve both facts; test the effective baseline first. Position sizing and market-order execution are explicit platform additions because the supplied class does not specify stake size or order types. Signal orders fill at the signal bar close, the registry's faithful equivalent to next-bar-open execution on a continuous market. Performance remains unverified. The validation repair replaces the nested minimal_roi optimization setting with four finite scalar settings without changing its schedule.

Faithful translation of ASDTSRockwellTrading by berlinguyinca, credited in its docstring to Gert Wohlgemuth, GPL-3.0. Preserve the 5-minute timeframe, TA-Lib MACD defaults, long entry and exit predicates, timed minimal-ROI schedule and -0.30 stoploss. The source specifies no instrument or stake size; LINKUSDT.BINANCE_SPOT and equity-relative sizing are platform additions. Execute signal orders at the completed bar's close, the registry's faithful equivalent for next-bar-open execution on a continuous market. Intrabar protection is platform-matched. Source: [pinned strategy](https://github.com/freqtrade/freqtrade-strategies/blob/f3340ce11f5bdf62f598522e64d1f5638eaa13f5/user_data/strategies/berlinguyinca/ASDTSRockwellTrading.py). No performance claims are accepted as evidence.

Classification: concrete trading strategy. Faithful translation of TDSequentialStrategy from berlinguyinca's Freqtrade collection, with the source docstring crediting @bmoulkaf; GPL-3.0. Source: https://github.com/freqtrade/freqtrade-strategies/blob/f3340ce11f5bdf62f598522e64d1f5638eaa13f5/user_data/strategies/berlinguyinca/TDSequentialStrategy.py. Use the supplied Python as the authoritative source, including its ninth-count assignment behavior rather than correcting it to match the prose. Preserve 1h bars, 30 startup candles, long-only operation, limit entries and exits, GTC entry/exit orders, stoploss=-0.05, trailing_stop=False, use_exit_signal=True, exit_profit_only=False and ignore_roi_if_entry_signal=False. The source ROI schedule {'0': 5} means a 500% profit threshold from minute zero, not 5%. Correct the validation error by storing minimal_roi as the finite scalar 5.0 and roi_start_minutes as 0; reconstruct the original schedule in implementation. The source provides no pair, stake allocation or limit-price selection; platform additions are LINKUSDT.BINANCE_SPOT, 1x cash execution, fixed equity-relative sizing and signal-close limit prices. The source's stoploss_on_exchange=False is implemented through a platform-matched intrabar protective stop at 5% below entry, with no exchange-hosted protection requirement. This explicitly substitutes the platform's intrabar stop for the source's locally managed stoploss order mapping of 'limit'. No performance claim is verified. The outbox file workspace/outbox/research_lead/hypothesis_88d0a4cd75.json could not be written because this session permits read-only filesystem access.

Test whether unusually positive asset-specific return skewness attracts speculative lottery demand that subsequently unwinds, while negatively skewed peers recover from neglected downside shocks. Use synchronized 4-HOUR bars across six Hyperliquid perpetuals. Buy the lowest residual-skewness asset and short the highest, hold for seven days, and remain flat between qualifying observations. This is a cross-sectional tail-shape hypothesis, not a price-ratio convergence or momentum rule. Market orders execute at the synchronized signal bars' closes; intrabar protective stops are matched by the platform. No exchange-hosted protection, independent opposing positions on one instrument, selectable same-bar policy, or next-session execution is required. Use only the actual common historical coverage; Hyperliquid's approximately 5000-candle limit bounds 4-hour history to roughly 2.3 years. The proposal has not been screened with a preliminary backtest.

Faithful signal-rule translation of DoubleMaStrategy by 用Python的交易员, MIT licensed, from the supplied vn.py source at commit 69dd4922dda8acde095b85a2f0613288e401ff7f. Trade LTCUSD.HYPERLIQUID using 4-hour bars. The source specifies no instrument or backtest interval; these are platform research choices. Preserve fast_window=10, slow_window=20, strict crossings, cancellation on every bar, and separate closing/opening limit orders on reversals. The source submits quantity 1 for every order. For capital-relative platform sizing, explicitly translate that into one fixed research lot sized at initialization to 20% of starting equity, rounded down to instrument precision; keep its quantity unchanged throughout each evaluation window. This sizing translation is a declared departure from literal one-LTC orders. Do not add signal filters, stops, targets, trailing exits or parameter optimization. Performance is unverified.

Translate Ringo's Apache-2.0 TqSdk momentum demonstration faithfully at the signal and decision level. Source defaults are SYMBOL='SHFE.au2012', N=15, daily candle duration=86400 seconds, and target positions of +100/-100 contracts. The 15-row source buffer includes the developing candle, so AR uses exactly 14 completed daily candles. Platform adaptations are explicitly limited to replacing the expired gold contract with LTCUSD.HYPERLIQUID, using UTC crypto daily boundaries, capital-relative sizing, and platform execution with real fees. This is a crypto adaptation, not a reproduction of historical gold performance. Preserve all threshold inequalities, price-change event gating, and flat-versus-positioned branching. Do not introduce indicators, filters, price stops, profit targets, trailing stops, or holding limits.

Faithful fixed-default reproduction of MAGen by the Jesse community, supplied under the MIT licence at https://github.com/jesse-ai/example-strategies/blob/7c91e0a37bf62165790120d730442e4f6eb00364/MAGen/__init__.py. Trade LINKUSD.HYPERLIQUID using LINKUSD.HYPERLIQUID-4-HOUR-LAST-EXTERNAL bars, one position at a time, both long and short. Instrument and interval are platform research choices; the supplied source specifies neither. Preserve all twelve source defaults, its balance-relative sizing, fixed entry-time ATR brackets, strict comparisons and unconditional pending-entry cancellation. This is a reproduction experiment, not a claim of demonstrated profitability. No additional trend, volume, funding, cooldown or time-exit filters are permitted.

Translate Jesse community's MIT-licensed KDJstrategy faithfully onto LTCUSDT.BINANCE_SPOT using 1-DAY bars, a CASH account and no leverage. Preserve executable Python semantics rather than correcting the comments. KDJ defaults are fastk_period=9, slowk_period=3, slowk_matype=0, slowd_period=3, slowd_matype=0 and sequential=False. Compute the nine-bar stochastic, K as its three-bar SMA, D as K's three-bar SMA, and J=3*K-2*D. Calculate previous J using candles excluding the current candle. ATR period is 14. Preserve Jesse's indicator initialization and ATR smoothing; do not substitute EMA-based KDJ. Source: [KDJstrategy, pinned commit](https://github.com/jesse-ai/example-strategies/blob/7c91e0a37bf62165790120d730442e4f6eb00364/KDJstrategy/__init__.py). Defaults verified against [Jesse KDJ implementation](https://raw.githubusercontent.com/jesse-ai/jesse/master/jesse/indicators/kdj.py); the strategy does not pin its Jesse dependency version.

Faithful reproduction of Connor McDonald's MACD_EMA, credited as MIT-licensed in the submission. Use LTCUSDT.BINANCE_SPOT with 1-hour OHLCV bars, long-only, no leverage. Preserve defaults ema=100, fastperiod=12, slowperiod=26 and signalperiod=9. The executable rules test indicator states, despite comments describing crossings. Preserve all-balance sizing and the absence of protective stops, profit targets, trailing stops and holding deadlines. Performance is unverified. Source: [pinned MACD_EMA code](https://github.com/jesse-ai/example-strategies/blob/7c91e0a37bf62165790120d730442e4f6eb00364/MACD_EMA/__init__.py).

Test the posted Dual Thrust range-breakout mechanism independently on LINKUSD, LTCUSD and BCHUSD Hyperliquid perpetuals using 4-HOUR bars. Exact catalog directories exist for all three. Use completed-bar decisions, prior completed bars for range formation, market orders and capital-relative sizing. This is an explicit adaptation: the source polls intrabar, trades quarterly futures at 10x margin and uses unspecified external parameter defaults. This experiment instead uses perpetuals, 1x account leverage, a symmetric coefficient, bounded exposure and protective exits. It does not claim to reproduce the original execution or returns.

Adapt 善's executable grid mechanism to synchronized LTC/BCH 4-hour perpetual bars: add equal-sized hedged nodes against widening relative-price deviations, cover the newest node first on reversal, and close the oldest node when capacity is exceeded. This is an explicitly modified experiment, not a faithful reproduction of the source's 500ms bid/ask execution. No martingale sizing or funding-income assumption.

Test a cost-aware adaptation of 发明者量化-小小梦's Icebreaker mechanism on LTCUSDT.BINANCE_SPOT using 1-HOUR bars. Start long, add below average entry, and exit the entire inventory above its updated average. Preserve the source's average-price targeting formula and absence of direction reversal. Explicit adaptations are percentage price distances, capital-relative sizing, bounded additions, a campaign-wide loss limit, and decisions executed at completed-bar closes. This is not a faithful reproduction of the original one-second resting-order execution. Claimed source performance is unverified.

Faithful translation of the Apache-2.0 open-source strategy 'ADX_DMI_STOCK' by Myquant (掘金). Credit goes to the original author: https://github.com/myquant/strategy/blob/17595e6bf4a118e1fa87c90bfb0fd78afa69a60b/ADX_DMI_STOCK/python/adx_dmi_stock.py. Defaults come from adx_dmi_stock.ini and the universe from stock_pool.csv, both at the same commit. This is a tradeable strategy, not a utility: a long-only, per-stock dual-moving-average crossover that uses ADX direction and the DMI lines as a trend-strength filter. The source runs on Shanghai A-shares with daily bars (ini bar_type=86400), so the port runs on our CNEQ venue (CASH, long-only, 1x, ~0.05% per side) on 1-DAY bars. The market does not change. Universe: stock_pool.csv lists 99 SHSE large caps, and the 64 that exist in our CNEQ catalog are used. The 35 missing names are dropped, not substituted. Each stock is processed on its own daily bar, exactly like the source's per-symbol on_bar. Do not hold back one stock's processing until the other legs print. Indicators: per stock, keep a rolling 30-bar (hist_size) high/low/close buffer whose last slot is the current bar. Every bar, recompute talib-style Wilder ADX(14), PLUS_DI(14), MINUS_DI(14), SMA(close,5) and SMA(close,20) from scratch on that window; do not use a full-history running ADX. Order each bar: (1) movement stop, (2) fixed stop (disabled), (3) signal check, only if no open/close action has happened on this stock this bar. Sizing: the source buys open_vol=5000 shares on 1,000,000 initial cash, or floor(cash/close/100)*100 shares when cash is short, and skips at 0. On our 100,000 account open_vol is scaled to 500 shares, which keeps the same capital fraction per entry. The lot fallback is kept as written. Real CNEQ fees replace the source's 0.0003 commission and 0.00246 slippage. Any performance the source claims is unverified.