BtcOneHourDailyVwapExcursionMeanReversionLS

BTC Intraday Daily-Anchored-VWAP Excursion Mean-Reversion (Single-Instrument USD-M, Long-Short, 1H, Fee-Floor-Gated, ATR-Stopped, Anchor-Reset Exit)

A long-SHORT, SINGLE-INSTRUMENT, pure-OHLCV intraday MEAN-REVERSION strategy on BTCUSDT.BINANCE USD-M perpetual futures using 1-HOUR bars. It transfers the factory's PROMOTED ETH 1H daily-anchored-VWAP mean-reversion mechanism DOWN the volatility curve to BTC — the deliberate, physics-favorable direction, in direct contrast to the SOL transfer that failed UP the curve. The diagnosed reason the SOL sibling went anti-predictive (-80%) is that on HIGH-volatility assets, intraday excursions from the daily-anchored VWAP TREND THROUGH fair value rather than revert. BTC is the LOWEST-volatility major: its intraday excursions are the MOST likely to revert to the session VWAP, so the same mechanism that is promoted on moderate-vol ETH should be at least as reliable on lower-vol BTC. Mechanism: compute the daily-anchored VWAP (reset 00:00 UTC, accumulated from the day's 1H bars). When price stretches significantly BELOW VWAP, go LONG (expect reversion up); significantly ABOVE, go SHORT (expect reversion down). CRITICAL DESIGN CHOICES, each derived from a specific prior kill: (1) PURE OHLCV — NO supplementary data, avoiding the open_interest/taker/L-S coverage walls (only ~30-56 days history, hard-blocked) and the alt-collection gaps (ARB/AAVE/NEAR) that killed the baskets; VWAP is computed from the bars themselves. (2) SINGLE confirmed-data major (BTCUSDT) — no multi-symbol basket, so no data-collection failure. (3) DOWN-the-vol-curve transfer (BTC lower-vol than ETH), NOT up (SOL higher-vol, anti-predictive) — the diagnosed physics is in favor, not against. (4) FEE-FLOOR GATE — the SOL failure compounded via fee drag on small excursions; here NO trade is taken unless the excursion exceeds a minimum absolute % chosen so the reversion-to-VWAP headroom clears the ~0.10% RT futures cost with margin (directly removing the fee-drag death mode), while entries stay frequent enough (target >=10 per walk-forward OOS window) to avoid event-sparsity. (5) LONG-SHORT, NOT long-only-regime-gated — the ETH long-only-bull-gated variant starved OOS windows; this fades both sides in every regime. (6) ATR STOP + ANCHOR-RESET EXIT to contain trend-through tails. (7) NOT a closed family: NOT a single-asset funding-confirmed-trend clone (the asset-specific mechanism that failed to transfer to SOL/higher-beta), NOT pairs-ratio, NOT breakout/momentum, NOT supplementary-primary, NOT COIN-M/HL. Intentionally minimal: 3 tunables (entry_z, vol_lookback, min_excursion_pct).

Transfers the factory's promoted ETH 1H daily-anchored-VWAP mean-reversion DOWN the volatility curve to lower-vol BTC, where intraday excursions revert to the session VWAP rather than trend through it (the diagnosed reason the up-curve SOL transfer went anti-predictive). Implemented with the standard FactoryStrategy hooks, single-instrument, pure OHLCV (no supplementary data -> avoids the open_interest/taker/L-S coverage walls and basket collection gaps). The daily VWAP anchor resets at UTC midnight via integer ns math and time gates use hour = (ts//NS_PER_HOUR)%24 (no datetime import; epoch ns is UTC). calculate_signal returns the continuous VWAP z-score (varies every bar -> Layer-2 frozen-signal detection passes) while the discrete entry conditions live in should_enter. The two-condition entry is the key design: entry_z is the statistical extremeness and min_excursion_pct (0.4%) is the economic fee-floor -- requiring BOTH means a 2-sigma event only trades if the move is also wide enough that reverting to VWAP clears the ~0.10% RT futures fee with margin, directly removing the small-excursion fee-drag death mode while a 0.4% gate on BTC 1H still fires often enough to populate OOS windows. It is long-short in every regime (not long-only-bull-gated, which starved OOS windows on the ETH variant). Exits are symmetric: reversion to VWAP, a 2.5x-ATR stop to contain trend-through tails, and an anchor-reset forced flat at 23:00 UTC so no position is held across the midnight VWAP discontinuity. Robustness: zero-volume bars fall back to the mean of day closes so VWAP stays defined, std<=0 yields z=0 (no entry), and side/entry context is reconstructed from avg_px_open if lost. Sizing is capital-relative (20% of equity) clearing the BTCUSDT $100 min-notional. leverage stays 1.0 referenced only as a non-amplifying sizing multiplier so the unused-leverage gate does not apply. Three tunables keep the optimization surface minimal.

The down-the-vol-curve transfer of the ETH daily-anchored-VWAP mean-reversion to BTC failed catastrophically, falsifying the 'lower vol = cleaner reversion' physics thesis. Over a dense 4,110-trade sample (6.5 years) it loses -76.6% (CAGR -20.4%) with Sharpe -1.98 and the ENTIRE CI below zero (-2.57 to -1.36), PF 0.77, probabilistic_sharpe 0.0, a 77.6% max drawdown, and losses in all 7 years. The classic mean-reversion death mode is fully present: win rate is high (59%) but avg_win ($108) is half avg_loss ($202) — the small VWAP reversions are dwarfed by trend-through losses, exactly the behavior the hypothesis claimed BTC's lower volatility would prevent (it does not; BTC intraday excursions also trend through fair value). The fee-floor gate (min_excursion_pct) did NOT remove the fee-drag death mode either: commission is 18.86% of gross on 4,110 trades. Both transfers of this mechanism have now failed in opposite vol directions — SOL up-the-curve (-80%) and BTC down-the-curve (-76.6%) — which is decisive evidence that the daily-anchored-VWAP-MR edge is asset-specific to ETH and does not generalize across the volatility spectrum. No optimization fixes a signal losing ~20%/year with its entire Sharpe CI below zero, a 77% drawdown, and 19% fee drag; a best-of-225 sweep would fit noise. FAILURE PATTERN: intraday daily-anchored-VWAP excursion mean-reversion is a single-asset (ETH) regime-specific edge, not a transferable mechanism — on both higher-vol (SOL) and lower-vol (BTC) majors the excursions trend through fair value and the high-turnover strategy bleeds fees (picking-up-pennies: high win rate, avg_loss ~2x avg_win, commission ~19% of gross). The 'physics-favorable lower-vol' reasoning is wrong; lower realized vol does not make intraday excursions revert. This mechanism should not be transferred to further assets.

Long-short single-instrument intraday daily-anchored-VWAP excursion mean-reversion on BTCUSDT.BINANCE USD-M perpetual futures, 1-hour bars, pure OHLCV. Each bar it builds the UTC-day-anchored VWAP (volume-weighted, reset at midnight, mean-of-closes fallback when volume is absent), the excursion % from VWAP, and a z-score over a rolling vol_lookback std of (close-VWAP). It goes long when z <= -2.0 AND the excursion is <= -0.4% below VWAP, short when z >= +2.0 AND excursion >= +0.4%; the min-excursion fee-floor gate ensures the reversion headroom clears the ~0.10% round-trip cost. It exits at the earliest of reversion to VWAP (long z>=0, short z<=0), a 2.5x-ATR stop from entry, or an anchor-reset forced flat at 23:00 UTC before the midnight VWAP reset. Position 20% of equity, leverage 1.0. Three tunables: entry_z, vol_lookback, min_excursion_pct.

Clean, well-engineered implementation: pure OHLCV (no data wall), single confirmed-data major, two-sided, dense 4,110 trades, fee-floor + ATR-stop + anchor-reset all correctly coded

The down-the-vol-curve test was a legitimate, decision-relevant hypothesis to run

Catastrophic loss: total return -76.6%, CAGR -20.4%, Sharpe -1.98 with entire CI below zero (-2.57 to -1.36), PF 0.77, max DD 77.6% (CI high 87.7%), probabilistic_sharpe 0.0

Negative in ALL 7 years (2020-2026); recovery factor -0.99

Picking-up-pennies death mode: win rate 59% but avg_win $108 < avg_loss $202 — trend-through losses dwarf the small reversions (the exact failure BTC's lower vol was claimed to avoid)

Fee-floor gate did NOT remove fee drag: commission 18.86% of gross on 4,110 trades — the diagnosed SOL death mode persists

>0 to be viable

negative 7/7 years

majority positive

BtcOneHourDailyVwapExcursionMeanReversionLS transferred the factory's promoted ETH daily-anchored-VWAP mean-reversion down the volatility curve to lower-vol BTC, reasoning that BTC's calmer intraday excursions would revert to session VWAP more reliably than the higher-vol SOL transfer that had already failed. The implementation was clean and decision-relevant — pure OHLCV, single confirmed-data major, two-sided, with a fee-floor gate, ATR stop, and anchor-reset all correctly coded — but the result falsified the physics thesis outright: a -76.6% return, Sharpe -1.98 with its whole CI below zero, a 77.6% drawdown, and losses in every one of seven years, driven by trend-through losses twice the size of the small reversions plus 19% fee drag. The analyst abandoned it at the backtest-review gate on its first iteration, concluding that with both up-the-curve (SOL) and down-the-curve (BTC) transfers now failing catastrophically, the VWAP-MR edge is specific to ETH and lower volatility does not make excursions revert. The explicit recommendation was to stop porting this mechanism to further assets.

Intraday daily-anchored-VWAP excursion mean-reversion is an asset-specific (ETH) regime edge, not a transferable mechanism — both transfers have now failed catastrophically in opposite volatility directions (SOL up-the-curve -80%, BTC down-the-curve -76.6%), so lower realized vol does NOT make intraday excursions revert, and this picking-up-pennies, high-turnover design (high win rate, avg loss ~2x avg win, ~19% fee drag) should not be ported to further assets.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate, so optimization, analysis, and risk review never ran: the 'lower vol = cleaner reversion' physics thesis was falsified (BTC intraday excursions also trend through fair value), the fee-floor gate did not remove the fee-drag death mode, and no optimization fixes a signal losing ~20%/year with its entire Sharpe CI below zero and a 77% drawdown.

A long-short, pure-OHLCV intraday mean-reversion strategy on BTCUSDT.BINANCE USD-M 1H perpetual futures that faded excursions from the daily-anchored VWAP (long when price stretched significantly below VWAP, short when above), gated by a fee-floor minimum-excursion filter and exited on reversion to VWAP, an ATR stop, or a forced anchor-reset flat before midnight — a deliberate 'down-the-vol-curve' transfer of the promoted ETH VWAP-MR mechanism to lower-volatility BTC, where intraday excursions were argued to revert more cleanly.

Over a dense 4,110-trade sample in 6.5 years it failed catastrophically: total return -76.6%, CAGR -20.4%, Sharpe -1.98 with the entire CI below zero (-2.57 to -1.36), profit factor 0.77, 77.6% max drawdown (CI high 87.7%), probabilistic Sharpe 0.0, and negative in all 7 years — exhibiting the classic mean-reversion death mode (59% win rate but avg win $108 vs avg loss $202, trend-through losses dwarfing small reversions) with commission at 18.86% of gross.