Hypotheses

Classification: (a). The source gives a concrete mechanism (an Engle-Granger cointegration pairs trade with a 2-sd z-score entry) but names no instrument. The author's question (re-estimate beta every bar, or freeze the in-sample beta) becomes a pre-registered categorical parameter that the backtest answers. Crypto venues are rejected for this mechanism: z-score spread fades on crypto perps keep dying on fees (mean_reversion 0/203, multi_instrument 0/148, captures below the cost of 4 legs). US equities suit it: zero commission (about 0.02-0.05% round trip for spread and impact per leg), shorting on Reg-T margin, and decades of daily split-adjusted bars, so the trade count comes from long history rather than fast trading. The universe is capped at 4 hand-picked same-industry duopoly pairs (8 names, under the ~10-name cap of lesson L173). Each pair has an economic reason to be cointegrated: KO/PEP (beverages), XOM/CVX (integrated oil, same crude and refining margins), HD/LOW (home improvement, same housing cycle), V/MA (card networks, same payments volume). Dividends are not in PnL, but yields are similar within each pair, so the omission roughly cancels across the dollar-neutral legs. The mechanism splits time into formation and trading periods (Gatev/Vidyamurthy style). Every rebalance_days sessions each pair is re-formed on the trailing formation_days of log prices: an OLS of log(y) on log(x) gives beta and the residual mean and sd. If an ADF test on the residuals gives a t-stat below adf_crit, the pair is TRADEABLE for the next trading window; otherwise it is not traded until the next formation, and any open position in it is closed. The author's question is the categorical parameter beta_mode. 'frozen' keeps the formation-window beta, mu and sigma for the whole trading window. This is the textbook out-of-sample answer: the cointegrating vector is treated as a fixed population parameter, and re-fitting it every bar lets the hedge ratio chase the dislocation and shrink the z-score exactly when it should trade. 'rolling' re-estimates beta, mu and sigma every bar on the trailing formation_days. Both run through the same code path. To avoid the Layer-3 sandbox timeout that killed the PEAD basket, keep rolling sums per pair (sum x, sum y, sum x^2, sum xy, sum y^2) so each bar costs O(1), and run the ADF only at formation dates, never per bar. Risk: each leg is capped at 25% of equity per pair, 4 pairs max, so at most 100% long plus 100% short (within 2x Reg-T). Legs are sized dollar-neutral by beta (y notional = N, x notional = beta*N). Stops: spread z beyond stop_z, or a time stop at max_hold_days.

Source classification: (a) a concrete trading idea. Zero's post describes a 10-slice fixed-percentage inventory grid: put half the capital in, buy one slice every 5% down and sell one every 5% up, and treat the floating loss as the cost of 'building an ATM'. The author's yield claims are unverified. As written this is the dead inventory/grid family: grids, ladders and constant-mix designs died 5 of 5 in the newest 50 (lesson 171), and faithful ports of FMZ scripts are dead too (lesson 167). Two parts are worth keeping. First, the WIDE step: the author insists a round trip should capture at least 5%, which is ~25x the 0.20% spot round trip, unlike the tight-rung grids that died on fees. Second, the premise that the instrument 'will not collapse 100->10'. Here that premise becomes a market observation: price above its 200-day SMA, instead of the author's assumption. The restatement keeps ONE rung. There is no inventory stack, no averaging down and no rebalance band. A single position per coin is opened when price has fallen `step` below its 10-day closing high while the long trend is up. It takes profit at entry x (1+step), the grid's 'sell one slice per step up', and is cut at entry x (1-stop_mult*step), where the original grid would have bought more. Instruments: ETHUSDT.BINANCE_SPOT and ADAUSDT.BINANCE_SPOT, traded independently on 4H bars, each with its own position and no ranking. Spot fits a long-only, no-leverage design. BINANCE_SPOT is 11% of the corpus, below its 15% target. HONEST PRE-STUDY (point-in-time, OKX perp 4H bars as a proxy for the same coins, 2020-2026, 0.12% cost per trade): UNGATED fades in both directions LOSE on every coin and step: ETH -0.65%/trade, ADA -0.39%, SOL -1.00% at step 8%. The trend gate is what is doing the work, and only on the long side. Trend-gated long trades: ETH +0.65%/trade (n=140, step 8%, stop 1.5x) and +0.89% (n=86, step 10%); ADA +0.08% (n=197) and +0.17% (n=130); SOL -0.20% / +0.09%, so SOL was dropped. Mirror-image shorts lost on ETH, so this design is long-only. Yearly sums alternate in sign (ETH 2021/2022/2025 negative). That makes this a modest, fragile edge, and the predictions below are set to match.

Source: Sixk, 'Pair trading - history length for spread calculation', quant.stackexchange.com/questions/80948, plus the answer by 'THATS MY QUANT' (quant.stackexchange.com/a/80953). The question runs a cointegration z-score pair trade on crypto perps and finds the results swing with the lookback (200 vs 500 candles). The answer gives three rules: (1) do not choose the window by backtest or p-value; (2) measure whether the divergence is large enough to pay fees and slippage; (3) ask why the pair relationship exists and when it breaks down, and stand aside then. This hypothesis puts all three into one design. PAIR AND REASON: LTCUSDT.BYBIT and BCHUSDT.BYBIT linear perps. Both are 2011-2017-vintage PoW payment forks of the Bitcoin codebase. They sit in the same exchange 'legacy PoW' bucket and have a largely overlapping holder and market-maker base, with no DeFi or L2 narrative to drive one apart from the other. The relative price is therefore driven mostly by shared old-coin beta plus temporary flow imbalances. This is not the same-asset cross-venue spread that just died (SOL Bybit/OKX, HYPE spot/perp). That spread is a market-maker speed contest, which the answer explicitly warns against. This is a cross-ASSET relative-value spread with dislocations of 6-10%, not a few bps. WINDOW ROBUSTNESS: the fair-value anchor is an EWMA of log(LTC/BCH) with a half-life tuned over a broad range (60-180 4H bars, 10-30 days). The pre-study shows the sign of the edge holds across that range. No cointegration p-value is used for selection. FEE FLOOR: entry needs an ABSOLUTE log deviation of at least 6%, not just a z-score, so every trade starts about 27x above the 0.22% two-leg round trip. BREAKDOWN VETO: if the 24h volume of one leg relative to its own 30-day median is more than 1.8x the other leg's, the divergence is treated as idiosyncratic news, which is the answer's SAP-earnings example. In that case no entry is taken. Venue Bybit was chosen because it is 2.0% of the corpus, against a quota of 5% or more. LTC and BCH are nearly unused in the corpus, and Bybit 1H bars run from Oct/Dec 2020, about 5.8 years. Bars are resampled to 4H in-strategy, or 4H bars are used directly.

Translation of Mahajan et al. (2019), 'Stochastic Spread Pairs Trading in the Indian Commodity Market' (arXiv:1907.08397, abstract: cointegrated commodity pairs, single-factor stochastic (OU) model on the LOG spread, trading-rule parameters fitted on training data). MCX spot commodities are not on our venues, so the closest backtestable commodity pair with long history is used: XLE.USEQ (integrated oil and gas equities) against USO.USEQ (front-month WTI crude ETF), both on 1-DAY bars since 2006 (~20 years, ~5,100 aligned sessions). Economic tie: XLE constituents' cash flows are driven by crude, so the log spread log(XLE) - beta*log(USO) is mean-reverting in the medium run. Short-run gaps open from equity-specific flows, crude futures roll and contango shocks, and these close as the equity market re-prices oil exposure. Point-in-time implementation of the paper's model: (1) hedge ratio beta = rolling 252-session OLS slope of daily log returns of XLE on USO, lagged one bar. (2) spread s_t = log(XLE_t) - beta_{t-1}*log(USO_t). (3) the OU parameters (theta = long-run mean, sigma_eq = stationary std, kappa from the AR(1) coefficient) are fitted by AR(1) regression of s on its lag over the last `z_window` sessions, using completed bars only. Closed-form AR(1) is used instead of the paper's differential-evolution fit; it is equivalent for a single-factor OU and adds no compute. (4) z = (s_t - theta)/sigma_eq. Long-short, dollar-neutral, 2 legs, flat most of the time. Crypto alternatives were checked first and rejected on data. Binance/OKX gold and silver perps (XAU/XAG/XPT/COPPER) have only ~7-10 months of 1h history. Gold-token spreads (XAU vs PAXG vs XAUT on Binance, XAG Binance vs OKX) have hourly spread std of 3-6 bps and lost money after fees in every variant (avg trade -0.01% to -0.06%). The pre-study on USEQ daily (rolling beta, 60-session z, entry |z|>1.5-2.0, exit at the mean, 20-session cap, 6 bps cost) gives XLE/USO about 100-158 trades, avg trade +0.51% to +0.58%, win rate ~0.6, Sharpe ~0.30-0.33, positive in 60-70% of years. HONEST CAVEAT: XLE/USO was the best of 21 commodity-ETF pairs scanned (GLD, SLV, GDX, XLE, XOP, USO, UNG). GLD/SLV and GLD/GDX, the 'textbook' commodity pairs, had no edge, so selection bias is material and the expected OOS Sharpe is modest. What the full paper would have to confirm: the exact trading-rule form (entry/exit bands in OU-sigma units, any stop-loss band), whether the bands were fixed or optimized per pair, and whether the reported Sharpe > 1.4 is net of costs. The abstract says the parameters were 'optimized by backtesting on the training period', so the 1.4 is likely an in-sample-tuned figure. source_ref: arXiv:1907.08397 abstract (method paragraph: Johansen cointegration selection + single-factor stochastic spread on log spread + trading rule tuned on the 80% training split).

Faithful translation of the open-source TradingView Pine v6 script 'FIL Grid Strategy [3Commas]' by 3Commas (MPL-2.0, https://www.tradingview.com/script/7cAVZopZ-FIL-Grid-Strategy-3Commas/). Credit: 3Commas. Classification: (a) a concrete, tradeable strategy. The visualization, table, watermark and webhook parts only display or send alerts, so they are NOT ported. The source is calibrated for BYBIT:FILUSDT.P on 15m. Our exact equivalent is FILUSDT.BYBIT (Bybit linear USDT perpetual) on 15-MINUTE bars; the catalog has it from 2021-06-29 to 2026-09-30 (184k bars, checked). Mechanism: a long-only price grid between FIXED bounds high_price=1.0151 and low_price=0.7099, with n_levels=36 levels and geometric spacing (the default). The level ratio is (0.7099/1.0151)^(1/35) ≈ 0.9898, a step of about 1.02%. Each level i=1..35 is an independent slot. Buy a fixed USDT amount when the close crosses DOWN through level i. Sell exactly that slot's quantity when the close crosses UP through level i-1, the level above. Level 0 (the top) is never bought. By the author's explicit design there is no trailing, no stop-loss and no time exit; the bounds and the bounded investment are the only risk caps. Sizing: per-slot USDT = investment / n_levels. The source uses investment = 10,000, which is 100% of its 10,000 initial capital (default_qty_value 278 ≈ 10000/36). On our $100,000 account the faithful equivalent is investment = 100% of starting equity: 2,777.78 USDT per slot, at most 35 slots = 97.2% of equity, leverage 1.0 (no leverage). Pine settings to reproduce: process_orders_on_close=true (orders fill at the close of the bar that gave the signal), calc_on_every_tick=false (evaluate once per closed bar), and pyramiding=200 (all slots may be held at once). The source charged 0.06% commission per side plus 3 ticks of slippage. Ours is the Bybit linear taker fee of 0.055% per side (0.11% round trip) from venue_configs. Caveats stated up front: (1) This family has died 5 of 5 here for grids and inventory designs (L171) and 5 of 5 for faithful ports of public scripts (L167). It is run because the user explicitly asked for a faithful port, and our real-fee backtest decides. (2) The fixed bounds are tied to FIL's 2026 price. FIL traded above 1.0151 from 2021 until Jan 2026, so the grid cannot fire at all before Jan 2026. Almost all of its activity (Jan–Sep 2026) falls inside the pipeline's final-20% holdout, and walk-forward windows on 2021–2025 will contain zero trades. That is a property of the faithful script, not a code defect. The performance claimed on TradingView is unverified.

Line-for-line translation of the FMZ/Pine Turtle Trading Decision System onto our platform. It is a tradeable strategy, not a utility. Venue/instrument: LTCUSDT.BINANCE (Binance USD-M perpetual), 1-DAY bars. The author backtested on 1d bars of Futures_Binance BTC_USDT. Why LTC rather than the author's BTC: BTC already dominates the corpus (44% lifetime, 62% recent), and an ETC Turtle is already in the pipeline. So the same rules run on a liquid, little-used USD-M perp with daily history from 2020-01-10 (~6.7 years). The Binance USD-M taker fee (0.05%/side) equals the script's own commission_value=0.05%, so the source's fee assumption is kept exactly. Default parameters exactly as written: Enter Mode='Mode 1'; Mode 1 entry_length=20, exit_length=10; Mode 2 entry_length=55, exit_length=20 (inactive by default); atr_period=14 (ta.atr = Wilder RMA of true range); risk_per_trade=0.02; initial_stop_atr_multiple=2; pyramid_atr_multiple=0.5; max_units=4 (Pine pyramiding=4). The date filter valid_date() always returns true, so the from/to inputs are unused. The highlighter input and all plot/fill code only draw on the chart and are NOT translated. HONESTY NOTE: lesson L167 records that faithful ports of FMZ scripts have died 5 of 5. This one is proposed because the user explicitly asked for a faithful statement. Unlike those ports, this script was written for a 0.05%/side commission on daily bars, the same fee level and decision interval we can reproduce exactly.

Adapt the source's correlation and cointegration screen to SOLUSDT.BYBIT and SOLUSDT.OKX on synchronized 1-hour bars. Fit a 90-day rolling residual model using completed bars and refresh it monthly. Trade only when correlation is at least 0.98 and the residual ADF statistic is below -3.34. Match SOL exposure across the long and short legs; cap combined gross exposure at 20% of equity.

Test a fixed, dollar-neutral EWT/EWY pair on synchronized daily bars. This preserves the source’s normalized-price distance mechanism while avoiding its 25-ETF pair search. Source returns are unverified. Prior two-leg USEQ runs timed out without metrics, so catalog coverage and backtest completion must be confirmed before assessing the edge.

Adapt QuantConnect's Pairs Trading with Stocks divergence-and-convergence rule to two prices for the same asset. On synchronized 4-hour bars, use HYPEUSDC.HYPERLIQUID_SPOT as a reference and trade only HYPEUSD.HYPERLIQUID perpetual. Define d = ln(perpetual close / spot close), with a trailing 180-bar mean and standard deviation calculated from observations before the decision bar. The pair is fixed before testing, with no retrospective pair selection. This single traded leg has directional risk; it is not market-neutral arbitrage. Use 1x leverage, cap notional at 25% of current equity, and risk at most 0.5% of equity on a 2% adverse price move.

Trade HYPEUSDC.HYPERLIQUID_SPOT long-only on completed four-hour bars. The hypothesis is that continuous fee-funded HYPE purchases help absorb selloffs that begin recovering within the same bar. The local catalog has 4,018 four-hour spot bars from November 2024 to September 2026. An exploratory check found 105 nonoverlapping signals, but it was used to select the rule and is not out-of-sample evidence. Source: https://hyperliquid.gitbook.io/hyperliquid-docs/trading/fees

Trade the HYPE perpetual when a large price and volume advance in Hyperliquid's native HYPE spot market signals demand for the token. Local 4-hour bars cover both exact instruments. An exploratory screen from May 2025 through September 2026 found 211 qualifying bars; the perp rose an average 1.33% over the following 16 hours, compared with 0.30% unconditionally. This is in-sample evidence, not a validated strategy result. Hyperliquid's spot and perp fee schedules are documented at https://hyperliquid.gitbook.io/hyperliquid-docs/trading/fees.

Source-fidelity hypothesis for Hummingbot's SuperTrend v1. Its default binance_perpetual connector and WLD-USDT pair map to WLDUSDT.BINANCE. Default candle connector and pair inherit the trading connector and pair. Defaults are 3-minute candles, SuperTrend length 20, multiplier 4.0, distance threshold 0.01, and 30 retained candles. Inherited defaults are a 100 USDT total position budget, two executors per side, 300-second cooldown, 20x leverage, HEDGE mode, 3% stop-loss, 2% take-profit, 2700-second time limit, LIMIT take-profit order, and no trailing stop. The source is https://github.com/hummingbot/hummingbot/blob/9af100d6822da7d2d0291a906c730ef172284ee2/controllers/directional_trading/supertrend_v1.py; inherited rules are in that revision's directional_trading_controller_base.py and controller_base.py. At the factory's $100,000 starting capital, the source's 100 USDT budget equals 0.1% of equity; preserve that fraction as equity changes. Leverage affects required margin, not the source's position quantity.

Buy a near-the-money Deribit ETH call when a forecast from prior ETH returns exceeds the option's ask-implied volatility by at least 12 annualized percentage points. Short the ETH perpetual to hedge the call's delta, adjusting the hedge only when delta moves beyond a fixed band. Select listed option contracts through the rolling-options path.

A single-instrument, long-only SPY.USEQ strategy on daily bars. Adapt the source's volatility-scaled put-strike formula into a cash-equity entry threshold. This buys SPY shares; it does not earn option premium. Use completed daily bars and a 20-session volatility estimate. SPY has decades of daily data, though dividends are excluded from backtest PnL.

A faithful trading hypothesis from 3Commas’ Pine strategy for BTCUSDT.BYBIT on 15-minute bars. Defaults: highPrice=83689.23, lowPrice=51629.65, nLevels=50, gridMode="Geometric", investment=10000 USDT. The source sets initial_capital=10000, default cash quantity=250, pyramiding=200, commission=0.055% per side, slippage=3 ticks, process_orders_on_close=true, and calc_on_every_tick=false. Each entry supplies an explicit quantity, so the default cash quantity does not size its grid trades. The platform starts with $100,000; preserving the source’s 10000 USDT investment makes each slot 200 USDT, about 0.2% of starting equity, with at most 9800 USDT across the 49 buyable levels. The default date filter accepts bars whose opening timestamp is from 2025-04-01 00:00 UTC through 2026-12-31 23:59 UTC. Visualization and 3Commas webhook settings do not alter trades. Source: https://www.tradingview.com/script/BR74LOKo-BTCUSDT-P-Grid-Bot-Long-Strategy/

Faithful research specification of Hummingbot's DManV3Controller by hummingbot, using its default WLD-USDT market as WLDUSDT.BINANCE. It reads 100 three-minute candles and calculates Bollinger Bands with length 100 and standard deviation 2.0. The inherited defaults are a $100 quote budget, 20x leverage, hedge position mode, at most two active executors per side, and a 300-second cooldown measured against active executors. Each new executor receives $50 of the quote budget, split equally across four maker DCA orders of $12.50 each. Dynamic order spread and dynamic target both default to None, so their scaling branches are inactive; activation bounds also default to None. This is a specification for testing, not a performance claim. Source: https://github.com/hummingbot/hummingbot/blob/9af100d6822da7d2d0291a906c730ef172284ee2/controllers/directional_trading/dman_v3.py

Faithful research hypothesis from jagadeeshmanne’s [open-source TradingView strategy](https://www.tradingview.com/script/PFdze7mB-Adaptive-S-R-Reversal-RSI-Divergence-DCA-BE-Stop-5m/) for BTCUSDT perpetual futures on 5-minute bars. Trade both directions without a trend filter; the script’s 1-hour EMA is display-only. Preserve these defaults: 2x leverage cap, 6% nominal risk per cycle, two legs, 0.85% DCA spacing, 2% stop beyond the worst entry, 0.05% support/resistance zone, volume at least 1.1 times its 20-bar average, RSI(14), 5-left/5-right pivots, 20-bar divergence validity, two-day maximum support/resistance lookback when the previous day’s range is below 2%, hybrid take profit, 4% post-DCA target, 0.1% midpoint offset, breakeven activation at 1% favorable with a 0.25% buffer, 20:00 UTC flattening unless 1.5% favorable, and one cycle per UTC day. The source sets $5,000 initial capital, 0.04% commission per side and three ticks of slippage. The platform uses $100,000 initial capital and Binance USD-M fees of 0.02% maker or 0.05% taker per side. The source advances some trade-state variables when submitting limit orders, before fill confirmation; record that behavior when assessing translation fidelity.

Faithful translation of QuantConnect's DailyAlgorithm, credited to QuantConnect. The source sets a 2013-01-01 start, 2014-01-01 end, $100,000 cash, daily SPY data, and hourly IBM data with leverage 1. It calculates SPY MACD(12,26,9) using Wilder moving averages on daily closes and an IBM 90-hour EMA of Field.SevenBar, defined as (2*open + high + low + 3*close)/7. SPY supplies a signal but is never traded. The source's description mentions an SMA cross and EMA stop, but its code implements neither. Platform instruments are IBM.USEQ for trading and SPY.USEQ for the signal. USEQ charges zero commissions; spread and impact still apply. The platform's approximately two years of hourly equity history cannot reproduce the original 2013–2014 window. Source: https://github.com/QuantConnect/Lean/blob/856327ff33869cba98393781000c4ab1d2af423d/Algorithm.Python/DailyAlgorithm.py

Credit: Hummingbot, PMMister, Apache-2.0. The source is a tradeable, one-level inventory-managed market maker. Its default connector is binance and its pair is BTC-USDT, mapped here to BTCUSDT.BINANCE_SPOT. Defaults are: portfolio_allocation=0.1; target_base_pct=0.5; min_base_pct=0.3; max_base_pct=0.7; buy_spreads=[0.0005]; sell_spreads=[0.0005]; buy_amounts_pct=[1]; sell_amounts_pct=[1]; executor_refresh_time=30s; buy_cooldown_time=60s; sell_cooldown_time=60s; buy_position_effectivization_time=120s; sell_position_effectivization_time=120s; price_distance_tolerance=0.0005; refresh_tolerance=0.0005; tolerance_scaling=1.2; leverage=20; position_mode=ONEWAY; position_side=BUY; take_profit=0.0001; open_order_type=LIMIT_MAKER; take_profit_order_type=LIMIT_MAKER; max_active_executors_by_level=4; tick_mode=false; position_profit_protection=false; min_skew=1.0; global_take_profit=0.03; global_stop_loss=0.05; global_tp_enabled=false; global_sl_enabled=false; global_tp_activation_from=min_base; global_sl_activation_from=target_base; global_pnl_reference=position. At $100,000 equity, the default equal buy and sell weights make each proposed order $5,000 before quantization. The source's 20x leverage default cannot operate on this platform's Binance spot cash account; effective leverage is 1x. A 20x port would require changing the source's default venue.