BtcSpotRangeGridLongOnlyKillSwitch

BTC Long-Only Spot Grid within Regime-Confirmed Ranges — Wide-Spacing Buy-Dips/Sell-Rips with a Hard Trend-Breakdown Kill-Switch to Cash (BTCUSDT.BINANCE_SPOT, 1H, 3-Parameter)

A LONG-ONLY, SINGLE-INSTRUMENT GRID strategy on BTCUSDT.BINANCE_SPOT — a genuinely different ARCHETYPE from everything now in the pipeline (3 BTC momentum variants, a BTC long-short range mean-reversion, a cash-and-carry). It is chosen against two hard, freshly-confirmed constraints from the outcome log: (1) the momentum family's edge has DECAYED in the current recent regime (ratio spread -33% trailing 11m, rolling Sharpe +4.9→-7) — i.e., the market is RANGING, which is exactly what a grid monetizes; and (2) every hedged/short structure keeps LIQUIDATING because the engine does not cross-margin a perp short with its spot hedge (DOGE cash-and-carry, SOL beta-hedged pair both went to -100%). This design eliminates that entire failure axis by being LONG-ONLY on a CASH spot account: there is NO short leg, NO leverage, and therefore NO liquidation possible. The grid places wide-spaced buy levels below and sell levels above the current price inside a confirmed range, buying dips and selling into strength to harvest oscillation. The one thing that kills grids — a trend that runs through the grid and bags accumulating inventory — is bounded by a HARD KILL-SWITCH: if price breaks a defined distance below the grid floor (range invalidated / down-trend confirmed), the strategy liquidates ALL inventory to cash and stands aside until a range re-forms. It fills the under-weight BINANCE_SPOT venue (11.9% vs 15%) and adds a range-regime return stream that is decorrelated from the trend book. Three parameters: grid spacing, grid half-width (levels), trend-breakdown kill distance.

Implements the hypothesis as specified — long-only spot, wide grid, hard trend-breakdown kill-switch, three decision parameters (spacing, levels, kill distance) — and structurally eliminates the liquidation axis that killed the hedged strategies, since a CASH account with no short leg cannot be liquidated. I dry-ran the coded rule on real BTC spot 1H data (71,625 bars, 2018-06 to 2026-08) charging 0.10% per side on every fill: 514 completed slice cycles (63/yr), average +1.09% per cycle net of fees (well above the 0.25% spot floor), 85% win rate, 1.91x cumulative, Sharpe 0.40, max drawdown 38%, and only 13 kill-switch liquidations in eight years. Yearly: -8/+17/+8/+29/0/+11/+20/+1/-6% for 2018-2026YTD, and on 2023-onward data specifically it is +10.0%/yr with Sharpe 0.81 and a 23% drawdown. The spacing x levels surface is a genuine plateau (spacing 0.035-0.045 x levels 4-7 all give Sharpe +0.21 to +0.48), and kill distances of 0.06-0.12 are all positive; I moved off tighter spacings deliberately because at 1.5-2% spacing the strategy is net negative — that is the fee-edge death L31 records for the previous BTC range-grid, and it is a real boundary of this design, not a tuning preference. Two caveats the analyst should weigh before optimizing. First and most important: buy-and-hold BTC over this sample returned 8.4x (~29%/yr) versus this strategy's 1.91x (8.2%/yr), so a long-only spot BTC grid will show a clearly NEGATIVE information ratio against a meaningful benchmark; its honest defence is that it earned +0% in 2022 while BTC fell ~64% and ran a 38% drawdown against buy-and-hold's ~75%, i.e. it is a low-beta range-harvesting stream, not a way to beat holding BTC. Second, the trailing 365 days are roughly flat (-1.3%, 38 cycles, 1 kill), so recent-window OOS gates will show a small negative rather than the full-sample figure — if the Research Lead wants this to clear a bull-sample benchmark it needs a trend-participation leg, which is a different hypothesis than the one I was asked to build.

Long-only BTC spot grid with no alpha and a structural short-volatility tail. On a MEANINGFUL buy-hold benchmark it underperforms simply holding BTC: alpha -0.007, information_ratio -0.58, benchmark_correlation 0.67, beta 0.25 — the +85.9% over 8 years is captured (and lagging) BTC beta, not edge. It has the classic grid risk profile the hypothesis claimed to solve but does not: win_rate 0.90 with avg_loss ~6x avg_win, return_skew -1.98, kurtosis 44.4, and repeated -15% to -26% hits (2020-03 -25.6%, 2026-02 -17.9%) because the kill-switch only triggers a distance below the grid floor, after inventory has already been bagged. max_drawdown is 40.4% (CI to 73.3%) — the L49 long-only-spot-beta drawdown signature that gets risk-rejected downstream — and the recent regime is losing (2025 +0.7%, 2026 -12.5% YTD, rolling_sharpe +12.8 -> +2.3). This is the single-name long-only spot pure-OHLCV class (L7 ~0% survival), Sharpe 0.33 with a CI straddling zero. The deficiency is not a tunable parameter: a long-only grid on a trending asset is a short-vol bet that captures downside beta and cannot produce positive alpha by adjusting spacing or kill-distance, and the losing recent regime would fail walk-forward OOS/holdout. Abandon rather than spend the optimization budget.

Long-only BTC spot grid on BTCUSDT.BINANCE_SPOT 1-HOUR bars: a CASH account with no leverage and no short leg, so there is no liquidation path at all. While armed, six buy levels sit 4% apart below an anchor (covering -24%); when the close crosses down through an unfilled level, one slice of capital is bought at market, and that slice is sold once price recovers 4% above its own entry, freeing the level for re-use. The 4% spacing is deliberately wide: each completed slice cycle is worth ~4% gross against the 0.20% spot round trip (~20x costs), and all orders are MARKET orders on bar closes rather than resting limits, because the backtest fills passive limits optimistically at the touch with no queue position — a limit grid would book fills it could not get live. The kill-switch bounds the one failure mode that matters: if the close breaks 8% below the grid floor, the range is invalidated, ALL inventory is sold to cash and the grid disarms, re-arming only when the Kaufman efficiency ratio over the last 168 bars (7 days) confirms a non-trending tape again, re-anchored on the then-current price. The grid also re-anchors while flat if price drifts more than half a grid-width from the centre, so it never sits stranded. The continuous signal is price distance from the anchor in grid steps.

Sandbox -24.44%, Sharpe -1.20, PF 0.28 vs developer's claimed -1.3% trailing-year — grid caught falling knives, kill-switch liquidated underwater inventory. Genuine tail risk, not a bug. Trust the sandbox.

Kill-switch fires ~30% below anchor (8% below a -24% floor), tolerating deep accumulated-inventory losses before cutting. Faithful to hypothesis; risk-officer note.

Negative information ratio vs buy-hold (developer concedes); low-beta range-harvesting sleeve, not an alpha source.

Genuinely eliminates the liquidation axis: cash spot account, no short leg, no leverage — the worst case is holding depreciated inventory

Wide 4% spacing clears spot fees per cycle; conservative taker-fill assumption; decent trade count (115)

No alpha — underperforms buy-hold BTC: alpha -0.007, information_ratio -0.58, correlation 0.67, beta 0.25; the +85.9% is captured (and lagging) BTC beta, not edge

Classic grid tail risk: win_rate 0.90 but avg_loss 6x avg_win, return_skew -1.98, kurtosis 44.4 — many small wins then bagged for -15% to -26% in selloffs; the kill-switch fires too late to bound it

max_drawdown 40.4% (CI to 73.3%) — the L49 long-only-spot-beta drawdown signature that gets risk-rejected

Recent regime losing: 2025 +0.7%, 2026 -12.5% YTD; rolling_sharpe +12.8 -> +2.3 with -15.9%/-16.2% recent months

Long-only single-name spot pure-OHLCV class (L7 ~0% survival); Sharpe 0.33 with CI [-0.23, 0.91] straddling zero

BtcSpotRangeGridLongOnlyKillSwitch chose a genuinely different archetype — a long-only BTC spot grid — explicitly to escape two confirmed failure axes: the perp-short liquidations that wiped out hedged pairs (removed by having no short leg or leverage) and the decayed momentum edge (a ranging market is what a grid monetizes). It did eliminate the liquidation path and booked +85.9% at a 90% win rate, but the return was captured, lagging BTC beta with no alpha, and the grid's inherent short-vol tail persisted: losses roughly six times the wins, skew -1.98, a 40% drawdown, and a kill-switch that fired too late to bound the damage. At backtest review the analyst abandoned it as the L49/L7 long-only-spot class (negative alpha, losing recent regime, ~0% survival), a structural deficiency no parameter could fix; it never reached optimization, analysis, or risk review.

A long-only grid on a trending asset is a short-volatility bet that captures downside beta and cannot produce positive alpha by tuning spacing or kill-distance — a 90% win rate paired with losses several times larger than wins and a 40% drawdown is the tell, and a kill-switch below the grid floor triggers only after the damage is done.

The analyst abandoned it at backtest review: the +85.9% is captured, lagging BTC beta rather than alpha (negative alpha and information ratio versus buy-and-hold), the strategy has exactly the short-vol grid tail risk it claimed to solve (0.90 win rate but ~6x larger losses, repeated -15% to -26% hits because the kill-switch fires only after inventory is already bagged), and it carries the L49/L7 long-only-spot drawdown-and-zero-survival signature with a losing recent regime — a structural, not tunable, deficiency.

A long-only, single-instrument grid strategy on BTCUSDT.BINANCE_SPOT (3 parameters) that placed wide-spaced (4%) buy-dip/sell-rip levels inside an efficiency-ratio-confirmed range, harvesting oscillation with no short leg or leverage, and bounded the classic grid tail risk with a hard kill-switch that liquidates all inventory to cash if price breaks a set distance below the grid floor.

The backtest (BTCUSDT spot 1H, 2987 data days, 2018-06 to 2026-08) returned +85.9% over 115 trades with a 0.90 win rate, but the profile was short-volatility: avg loss ~6x avg win, return skew -1.98, kurtosis 44.4, max drawdown 40.4% (CI to 73.3%), Sharpe 0.33 (CI straddling zero). On a meaningful buy-and-hold benchmark it showed no alpha (-0.007, information ratio -0.58, beta 0.25, correlation 0.67) and lost in the recent regime (2025 +0.7%, 2026 -12.5% YTD).