Hypotheses

Derived by the Research Lead from MQL5 Code Base's EMA-Crossover_Signal: test its EMA3/EMA7 directional crossover, minimum three-point separation, avoidance of narrow ranges, and opposite-signal exit. The page credits Jason Robinson (jnrtrading). This is a completed signal specification, not a reproduction of an inspected program: the downloadable MQ4 source could not be retrieved. Research choices are completed 4-hour bars on LINKUSD.HYPERLIQUID, a scale-normalized point equal to 0.10% of the preceding close, a seven-bar range filter, scheduled FOMC/CPI entry blackouts, capital-relative sizing, and an intrabar protective stop. Thus three normalized points mean 0.30% EMA separation and ten mean a 1.00% range; these are explicit crypto adaptations of the source's unspecified original point scale. Entries execute as market orders at the signal bar's close. News-event selection uses separate scalar flags, avoiding the invalid string sequence in fixed news_event_types.

Derived by the Research Lead from Kieran Wood et al., Trading with the Momentum Transformer: test whether attention to historically similar return regimes improves long/short positioning across BCH, ETC and FIL perpetuals using daily bars. This operationalization uses the normalized kernel-attention formulation in Section II rather than reproducing the trained attention-LSTM architecture. A Gaussian similarity kernel retrieves historical regimes and their subsequently realized returns; volatility scaling and a cost hurdle translate the forecast into positions. The supplied full text was read, including the methods and appendices. No neural-model performance claims are adopted. Research Lead choices are the universe, Gaussian kernel, five-day forecast horizon, historical memory, entry hurdle and protective exits. All signal choices are registered below; risk and execution settings remain fixed. Implement entirely with NumPy and permitted strategy helpers. Orders execute at completed signal-bar closes, with intrabar protective stops matched by the platform.

Derived by the Research Lead from the FMZ KLineChart Bollinger signal example summarized in the original hypothesis. Preserve BTC spot daily bars, BOLL(20,2), long-only execution and the fixed 10% protective stop. Explicitly amend the frozen allocation contract from allocation_pct=0.20 to allocation_pct=0.10; this is a conservative, unvalidated risk revision, not a demonstrated improvement in edge. Open one position when flat and the completed daily close is strictly above the upper Bollinger band; close it when the completed daily close is strictly below the lower band. Execute market orders at the signal bar's close. Allocate 10% of current mark-to-market equity at entry, with no leverage, pyramiding or subsequent rebalancing. Maintain an intrabar protective stop 10% below the actual entry fill, matched by the platform. Entry allocation is not an ongoing exposure cap: appreciation can increase the invested share of equity. Register this amendment as a separate lineage variant and evaluate chronologically without reusing previously exposed protected holdout or exam observations as unseen evidence. Source text and figures were not supplied for renewed inspection in this revision.

Faithful signal and order-target translation of PcpArbitrageStrategy by 用Python的交易员, MIT licensed, at commit fcad8e2d86825506b9d33441ab78be853c2c3bdd. Preserve entry_level=20, price_add=5 and fixed_size=1 exactly. Use one-minute last-trade bars for a same-strike, same-expiry Binance call/put pair and a Bybit linear ETH perpetual. The perpetual replaces the source's unspecified futures contract; consequently this tests synthetic-forward versus perpetual spread convergence, not a maturity-matched, risk-free parity arbitrage. All three prices are USDT per ETH. Adapt only symbol parsing, contract multipliers and equity-relative sizing. The two option instruments exist in the local catalog; Bybit bars may be collected on demand. Preserve limit orders, persistent targets and futures-position-based branching; do not add indicators, entry filters, stops or signal-reversal rules. Source performance is unverified.

Faithful translation of DoubleMaStrategy by 用Python的交易者, MIT licensed, from the supplied howtrader source at commit fcad8e2d86825506b9d33441ab78be853c2c3bdd. Preserve fast_window=10, slow_window=20, strict crossing comparisons, close-priced limit orders, and opposite-cross reversal without additional filters, stops, targets or time exits. Use DOTUSDT.OKX linear perpetuals and one-minute bars for the source's unaggregated BarGenerator convention. Replace the source's Decimal(1) trading lot with an equity-relative research lot. Preserve pending orders: the supplied strategy never calls cancel_all. Order and stop-order callbacks perform no trading action; trade callbacks only publish state. Initialization requests 10 days of historical bars, updates the rolling array and waits for ArrayManager initialization; historical initialization must not submit executable orders. The supplied file omits the implementations of ArrayManager, BarGenerator and CtaTemplate. Attempts to retrieve those pinned helper files failed, so the conventional 100-bar array initialization, one-minute generator interval and default limit-order semantics remain dependency assumptions that the developer must verify before claiming a line-for-line reproduction. No performance study has been performed.

Faithful signal-rule port of London Breakout by je-suis-tm, Apache-2.0, at commit 611b73f2c3f577ac5b28aaa19ac8c43d3236c7a5. Adapt GBP/USD to XRPUSDT.BINANCE using completed 1-MINUTE close prices instead of sampled FX midpoint prices. Preserve risky_stop=0.01 and open_minutes=30 exactly: the executable thresholds are absolute quote-price differences, not percentages or basis points. Preserve the source's fixed UTC−05:00 clock throughout the year, without London or New York daylight-saving adjustments. Entries and exits execute as market orders at the observed bar close. Preserve branch order, including the absence of protective exits during the reference hour and opening entry window. The linked Heikin-Ashi statistics file was unavailable and contributes no rules. This tests the source's executable breakout rules on a different market; it does not reproduce GBP/USD economics or establish the author's claimed FX mechanism.

Faithful port of AtrRsiStrategy by 用Python的交易员, MIT licensed, at howtrader commit fcad8e2d86825506b9d33441ab78be853c2c3bdd. Source: https://github.com/51bitquant/howtrader/blob/fcad8e2d86825506b9d33441ab78be853c2c3bdd/examples/strategies/atr_rsi_strategy.py. Trade LINKUSDT.BYBIT long and short on completed 1-minute bars, preserving the minute-bar behavior of the source's tick-driven BarGenerator. Freeze atr_length=22, atr_ma_length=10, rsi_length=5, rsi_entry=16, trailing_percent=0.8 and source fixed_size=1. Adapt only instrument, venue, capital-relative sizing and fees. Entry orders remain marketable limit orders, and exits remain platform-matched intrabar stop-market orders. The source does not install protection in its entry branch; preserve its first protective-stop submission on the next completed bar after a position exists. No performance claim is verified. The supplied strategy text was read; retrieval of its utility dependencies failed, so the developer must verify ArrayManager initialization and indicator semantics against the pinned dependency before claiming exact numerical parity.

Derived by the Research Lead from MQL5 Code Base's Exp_AbsolutelyNoLagLwma_Range_Channel_Tm_Plus by GODZILLA: test completed-bar channel breakouts with the published TimeTrade=true and nTime=480-minute holding limit. The source describes a high/low moving-average channel; its underlying indicator page describes double LWMA smoothing. Downloadable MQL5 code could not be retrieved, so this is a disclosed reconstruction of the stated mechanism, not an exact indicator or EA reproduction. Research choices are DOTUSD.HYPERLIQUID, 4-HOUR signals, a causal twice-smoothed high/low channel, equity-relative sizing and percentage-based protective brackets. All four attached figures were inspected: the two identical trade charts show EURJPY H4, blue upper/red lower channel boundaries, colored candles and marked long/short trades; the two identical reports show 2017 balance/equity curves and deposit load. These figures supply qualitative context, not verified performance evidence. Market orders execute at the signal bar close. Protective stops trigger intrabar through platform matching; exchange-hosted protection is not required. The eight-hour deadline is retained, with execution at the first subsequent 4-HOUR bar close strictly exceeding it, explicitly approximating the source's immediate time exit on bar data.

Faithful signal-rule port of limin's Apache-2.0 TqSdk dualthrust.py at commit 78c99226f11056b2860c39369f453808938edde2. Preserve NDAY=5, K1=0.2, K2=0.2, strict last-price comparisons, persistent target positions and opposite-band reversals. Adapt the source's expired DCE.jd2011 instrument to BCHUSDT.BINANCE and its fixed three-contract target to equity-relative sizing. Daily reference candles use UTC days; decisions occur on changed trade prices, not daily closes. The source blog returned HTTP 404 and was not read; the supplied executable code defines this contract. No stop-loss, profit target, daily liquidation, momentum filter or optimization of source constants is added. Source TargetPosTask's execution algorithm is replaced by platform market orders toward the specified net target; identical execution prices are not claimed.

Faithful signal-rule port of Myquant's Apache-2.0 TurtleStrategy at commit 17595e6bf4a118e1fa87c90bfb0fd78afa69a60b. Source: https://github.com/myquant/strategy/blob/17595e6bf4a118e1fa87c90bfb0fd78afa69a60b/Turtle/python/turtle.py. Trade a fixed DOT/LINK/LTC Binance Spot basket using trade ticks and daily bars for initialization. Preserve the actual code rather than conventional Turtle rules: initialize each instrument's high and low once from preceding daily bars, never roll those thresholds, enter long above its frozen high, and liquidate all selected holdings when any selected instrument trades below its own frozen low. The Python source supplies no default for period, csv_file or buy_amount; its companion configuration could not be retrieved. Therefore period=20 and the chosen basket are explicitly platform selections, not verified original defaults. Preserve the source's hop fallback of 0.1 quote-currency units. Replace CSV cash amounts and 100-share stock lots with equity-relative budgets and exchange quantity increments. Entry remains a buy limit at last trade price plus hop, not a guaranteed fill at that price. No ATR, pyramiding, rolling Donchian channel, trailing stop, profit target or calendar rebalance is added. Historical initialization uses only completed daily bars before each evaluation window; thresholds then remain frozen throughout that window.

Derived by the Research Lead from "eur/usd trend catcher" by MQL5 Code Base (library) https://www.mql5.com/en/code/10977: Retain the source's 15-minute filtered moving-average trend entry and 50:20:350 distance ratio; choose SMA settings, percentage scaling, symmetric directions, position sizing and bounded holding time. Adapt the moving-average trend-catching mechanism described by sergeyrar1 in the MQL5 Code Base to DOTUSDT.BYBIT using 15-minute bars. The source specifies a 50-pip noise filter, 20-pip stop and 350-pip target, but the supplied description leaves the moving-average settings and filter implementation unspecified. Complete these with a 100-bar SMA and crossings of percentage bands around it. Replace EUR/USD pip distances with crypto-relative distances while preserving the filter:stop:target ratio of 50:20:350; baseline distances are 1.0%, 0.4% and 7.0%. This tests a crypto adaptation, not a reproduction of the EUR/USD results. The attached figure shows rising balance/equity and increasing trade sizes, without price candles, indicator settings or marked entries; it provides no verified performance evidence. The downloadable MQ4 source could not be retrieved, so no claim of code-level fidelity is made. Entries execute at signal-bar close. Protective stops trigger intrabar in the platform; no exchange-hosted protection or selectable same-bar ordering is required.

Derived by the Research Lead from "X Bar Clear Close Trend" by MQL5 Code Base (library) https://www.mql5.com/en/code/1622: Source supplies strict close-breakout signals, example length 10 and recommended range 3–15; Research Lead selects BCH perpetuals and 4H bars and completes execution, exits and risk management. Distilled from MQL5 Code Base's X Bar Clear Close Trend: preserve its strict close-above-prior-highs and close-below-prior-lows signals, with baseline pattern_bars=10. Trade BCHUSD.HYPERLIQUID on completed 4-HOUR bars, long and short. The source describes an indicator rather than a complete trading system; Research Lead additions are market execution at the signal close, opposite-signal exits, a fixed 6% intrabar protective stop, and equity-relative sizing. No moving average, price-distance filter, or supplementary feed determines the trend signal. Source performance is unverified.

Derived by the Research Lead from "数字货币期货双均线拐点策略（教学）" by 发明者量化-小小梦 (fmz) https://www.fmz.com/digest-topic/8479: 来源提供完整信号与退出公式；研究负责人补全未显示的参数默认值、权益相对仓位和 5 分钟离散执行时序。 依据发明者量化-小小梦《数字货币期货双均线拐点策略（教学）》的明确代码规则构建。保留 ETHUSDT.BINANCE、1 小时 EMA、同步局部极值信号、双向交易、反向平仓后反手及绝对美元价差止盈。公开页面和附件未提供 ema1Period、ema2Period、profitTarget 的默认数值，因此本实验明确补全为 10、20、50 美元，不宣称这些是作者默认值。将原源码 500 毫秒轮询离散为每根已完成 5 分钟 K 线评估一次，属于执行采样适配，不能复现轮询间发生的交易。使用已完成小时线计算指标，市价订单在当前 5 分钟信号线收盘提交并成交。五张图片分别为教学封面、ETH Binance 收益概览、带两条 EMA 与 L/S/CL/CS 标记的小时图、另一张收益截图和运行状态日志；图表展示反复开平仓，日志可见 0.1 ETH 下单，但图片中的收益均未经本平台验证。采用权益相对仓位代替固定数量，不新增趋势过滤、止损或冷却条件。

忠实移植发明者量化-小小梦《区块链资产量化交易中的跨币种对冲策略》的可执行网格规则。采用 LTCUSDT.OKX 为 A、ETHUSDT.OKX 为 B，使用线性永续合约表达买入 A、卖出 B，替代原演示依赖初始现货库存的卖出操作；这改变了融资及库存环境，不能称为原现货回测复现。源码实际使用 ETH/LTC 的可成交报价比，而非文字介绍的 LTC/ETH；遵循源码。保留 100 个节点、绝对比值间距 0.08、节点金额每级递增 8%、固定初始锚点、升序扫描及每次最多执行一个节点动作。节点基准金额改为窗口启动权益的 0.1%，在 100,000 美元账户上仍为 100 美元。三张配图均已查看：第一张是主题封面；第二张是 OKEX LTC/ETH 账户和交易汇总；第三张显示约 2.2–3.4 的比值震荡及后段上升，并附交易日志，没有技术指标或独立止损标记。这些历史结果不作为盈利证据。使用报价事件保留买卖价区别，不以一分钟收盘价替代盘口，也不要求独立对冲账户或交易所托管止损。绘图库完整源码需要登录，未读取的部分不属于交易规则。

忠实移植发明者量化-小小梦的 TCI 教学策略，保留日线、N1=10、N2=21、麦语言 SMA(4,1)、严格交叉及 REF 一根完整日线延迟。将原 BTC/USD 季度合约适配为 DOTUSD.HYPERLIQUID 线性永续，并将固定一张合约改为权益相对仓位；因此不复现原合约的期限结构或绝对滑价。原代码反复提交可成交限价单的执行方式适配为平台收盘市场单，平仓成交后再开反向仓位。六张附件均已查看：分别为教程封面、日线 BTC/USD 季度合约与加权 SMA 的源码截图、普通 SMA 文档、递推公式示例、交易状态机骨架及指标计算注释；没有提供独立的行情形态证据。此次成功读取原文、修正后的前篇链接及完整教学源码，补足 manifest 中的读取失败。所有历史收益展示均视为未经验证，不运行预研究来决定是否接受该具体规则。

Derived by the Research Lead from "瀑布线交易策略在发明者量化平台的实现与应用" by 善 (fmz) https://www.fmz.com/digest-topic/4109: 保留文章明确给出的交易代码及15分钟周期；研究负责人选择LINK、Bybit、缺失的EMA周期基准值、EMA初始化及权益相对仓位规则。 依据善在《瀑布线交易策略在发明者量化平台的实现与应用》中“实现一个瀑布线策略”的 My语言代码提炼，保留15分钟周期、三组EMA平均线、排列入场、反向换仓及两类平仓规则。源码页面128420未能读取，文章和两张配图已检查：第一张是封面；第二张展示OKCoin BTC季度期货、PUBU三线及开多、平多、平空标记，未显示可核实的N1/N2/N3默认值，图中业绩不作为证据。正文的MACD叙述与代码不一致，本实验以可执行代码为准，不加入MACD、MFI或随机指标。平台选择LINKUSDT.BYBIT，补全缺失参数为N1=4、N2=9、N3=24，并明确允许优化；这些不是作者已核实的默认值。使用收盘市场订单和单一净持仓。风险配置由平台补全：每笔参考风险为权益的0.5%，单笔名义敞口最高20%，杠杆1倍；源码平仓是收盘判断，不改成盘中止损。

Translate QuantRocket's Apache-2.0 strategy on SPY.USEQ using SPY.USEQ-1-MINUTE-LAST-EXTERNAL bars. Preserve its two strict sign tests, overnight-inclusive first return, same-day positioning, full-equity target weights and disabled VIX filter. The source security maps to SPY, as documented in [QuantRocket's data notebook](https://www.quantrocket.com/codeload/first-last/first_last/Part1-SPY-Data-Collection.ipynb.html). Source defaults are CODE='first-last', DB='usstock-1min', DB_TIMES=['10:00:00','15:00:00','15:30:00','15:59:00'], DB_FIELDS=['Open','Close'], MIN_VIX=None, benchmark=SPY at 15:59, broker commission=$0.005/share and slippage=0.5 bps. Encode disabled MIN_VIX as numeric min_vix=0, preserving the source's falsy branch without an invalid null optimization value. Platform costs replace source costs: zero USEQ commissions plus modeled spread and impact. Execution adaptation: the source uses the 15:30 minute's open for both signal and entry; that open is only delivered within a completed bar here. Preserve the exact signal inputs, submit at 15:31 when that minute closes, and fill at its close; never backdate a fill to 15:30. This one-minute delay must be reported as an adaptation. Exit at the close of the minute beginning 15:59, delivered at 16:00. All clock times use America/New_York with DST. No additional indicator, threshold, stop or VIX feed is introduced.

Faithful port of ReinforcedAverageStrategy by berlinguyinca, credited in the source to Gert Wohlgemuth, GPL-3.0, at commit f3340ce11f5bdf62f598522e64d1f5638eaa13f5. Preserve every source signal and default: 4h bars; TA-Lib EMA(8) and EMA(21); 50-period SMA of resampled 48-hour closes; plotting-only Bollinger bands with window 20 and standard-deviation multiplier 2; minimal_roi={"0":0.5}; stoploss=-0.2; trailing_stop=False; trailing_stop_positive=0.01; trailing_stop_positive_offset=0.02; trailing_only_offset_is_reached=False; process_only_new_candles=True; use_exit_signal=True; exit_profit_only=False; ignore_roi_if_entry_signal=False; INTERFACE_VERSION=3. The ROI threshold is 50%, not 0.5%, and the stop is 20%. Encode minimal_roi as the finite scalar 0.5 in optimization_plan.fixed, with minimal_roi_minutes=0, reconstructing the original schedule in code. Trade LINKUSDT.BINANCE_SPOT long-only at 1x with platform equity-relative sizing and actual engine fees. The supplied retry does not identify the prior instrument or venue, and no prior contract was found; this market selection therefore cannot be confirmed against that contract. Replace Freqtrade next-candle-open execution with market execution at the signal bar's close, the registry's supported equivalent for continuous markets. Protective stops trigger intrabar through platform matching. The source's performance commentary is unverified; no performance pre-study was conducted.

Translate Myquant's Apache-2.0 Bollinger_Bandit executable rules into FactoryStrategy on DOTUSDT.BINANCE_SPOT daily bars, long-only and unleveraged. Preserve the code rather than its occasionally inconsistent comments. The supplied Python file loads its parameters from a separate INI and contains no signal-parameter defaults. A repository mirror reports the defaults listed below, but the companion INI could not be retrieved at the supplied commit; these values require source verification before coding and must not be represented as verified pinned-commit defaults. Source: https://github.com/myquant/strategy/blob/17595e6bf4a118e1fa87c90bfb0fd78afa69a60b/BollingerBandit/python/bollinger_bandit.py. Platform adaptations are the instrument, cash-account execution, equity-relative sizing, instrument precision, and real fees. Crypto adjustment factors are identically one. Preserve the source's next-day sale eligibility by preventing exits of quantities acquired during the current UTC trading day. Market instructions execute at the signal bar's close. The source's retrospectively computed moving-average exit cannot receive a historical bar-open fill after observing that bar's close; its exact trigger is retained, but execution uses the decision-time market price, explicitly recording this execution adaptation. Do not replace it with a conventional moving-average crossover or a resting intrabar stop. No performance study has been run.

Faithful translation of MultiMa V2 by @Mablue (Masoud Azizi), GPL-3.0, from the supplied Freqtrade source at commit f3340ce11f5bdf62f598522e64d1f5638eaa13f5. Use DOTUSDT.BINANCE_SPOT with 4-HOUR bars, long-only cash execution and no leverage. Effective source parameters are buy_ma_count=4, buy_ma_gap=15, sell_ma_count=12 and sell_ma_gap=68, loaded from buy_params/sell_params. The IntParameter constructor fallback defaults are separately 7, 7, 7 and 94; do not mistake these for the loaded baseline. Preserve the source's strict comparisons, asymmetric AND-entry/OR-exit aggregation, ROI schedule and -34.5% stoploss. Source performance comments are unverified. The attached figure is a portrait, with no chart, instrument, timeframe or trading annotations. Platform adaptations are instrument selection, equity-relative sizing, real fees and market execution at the signal bar's close in place of next-bar-open execution. Freqtrade parameter loading and signal-collision behavior are documented at https://docs.freqtrade.io/en/2023.12/hyperopt/ and https://www.freqtrade.io/en/latest/strategy-customization/. No performance pre-study was conducted.

Derived by the Research Lead from "SkyPark" by Myquant (myquant) https://github.com/myquant/strategy/blob/17595e6bf4a118e1fa87c90bfb0fd78afa69a60b/SkyPark/python/SkyPark.py: Preserved Myquant's literal signal and exit rules; completed unavailable INI settings and adapted instrument, daily initialization, sizing and execution. Distilled translation of SkyPark by Myquant, Apache-2.0, from the supplied commit. Preserve the executable comparisons, first-minute high/low boundaries, flat-only entry and timed liquidation. The Python file specifies OPEN_VOL=5 but reads the instrument, clock times and gap thresholds from SkyPark.ini without defaults. That companion file could not be retrieved; the values below are explicitly platform-selected completions, not claimed source defaults. Use DOTUSDT.BINANCE one-minute bars and trade ticks, UTC daily research sessions, a first-minute interval of 00:00–00:01 and an exit threshold of 23:55. Replace five source contracts with capital-relative exposure. Reinitialize once per UTC day, corresponding to restarting the source for each trading day; the original code itself has no daily reset. Decisions remain bar-triggered, using the latest observed trade price. Market orders execute at the decision bar's close. No price stop or profit target is added. The requested outbox file could not be written because this workspace is read-only.

忠实移植发明者量化-小小梦《数字货币期货多品种ATR策略（教学）》中的原始麦语言规则，采用文中默认参数：ATR周期10、典型价格EMA周期10、通道倍数1、收盘止损7%，两个品种独立交易。明确以原始麦语言公式为准：ATR使用TR的EMA，突破允许当前收盘等于通道边界；JavaScript原型使用TA.ATR且突破为严格不等式，两者不混用。源码页面回测配置中的EMA20、通道倍数2不是本实验默认值。将示例ETH/LTC迁移至支持双向净持仓的Hyperliquid，保留日线周期，并将每次固定10币改为权益相对仓位。所有交易和止损均在完整日线收盘判定并提交市价单，保留作者的收盘止损，不改成盘中保护。已查看三张附件：第一张为教学封面；第二张展示ETH_USDT、LTC_USDT日线蜡烛、EMA20中线、ATR10上下轨及开平标记；第三张为两品种固定数量10的开平仓日志，包含中线或止损退出。图片和历史表现仅解释来源，不作为盈利证据。公开源码页面登录后部分未取得，交易规则依据提交的完整文章代码。openRatio在原型中未启用，本移植也不把它当作原作者已实现的仓位规则。