Strategies

[3Commas] Static Geometric Grid (by 3Commas, MPL-2.0) — faithful port on APEUSDT.BINANCE_SPOT 15m: 64 geometric levels 0.10591–0.15466, long-only close-cross grid, trailing-up OFF, date window 2026-01-12→2026-05-14

Outcome: Abandoned

ThreeCommasStaticGeometricGridApeSpot

Outcome Summary

The strategy attempted to port a 3Commas static geometric spot grid to APEUSDT.BINANCE_SPOT. No backtest or optimization results were produced because verification stalled. After three verification timeouts, the run was abandoned.

Strategy report

Hypothesis

Faithful line-for-line translation of the TradingView Pine v6 script '[3Commas] Static Geometric Grid' by 3Commas (MPL-2.0, https://www.tradingview.com/script/apaoI4Zr-3Commas-Static-Geometric-Grid/). Credit: 3Commas. It is a real tradeable strategy: a long-only static-range spot grid. The source backtested OKX:APEUSDT 15m with a 0.1% commission. Our OKX_SPOT APE history is only days long, so the port runs on APEUSDT.BINANCE_SPOT 15m. That market has continuous 15m history from 2022-03 and the same 0.10% taker fee as the source's commission_value = 0.1. Grid: 64 levels (default Grid Mode = Geometric), level_i = lowPx * (highPx/lowPx)^(i/63) for i = 0..63, with highPx = 0.15466 and lowPx = 0.10591. The rung spacing is (0.15466/0.10591)^(1/63) - 1 = 0.603%. The Arithmetic option (level_i = low + (high-low)*i/63) should be kept as a parameter but is OFF by default. Trailing-up is OFF (trailUp=false). Its inputs (threshold 3.0% above high, shift 50% of range) should be carried as inert parameters. If enabled, the code runs on any bar in the window with close > curHigh*(1+3/100): it shifts curHigh and curLow up by 50% of (curHigh-curLow) and recomputes the levels. Date window (useDates=true by default): trading only when bar time is in [2026-01-12 00:00 UTC, 2026-05-14 00:00 UTC]. Outside the window no grid orders are placed, and lots still open when the window ends are simply held. Sizing: the source uses 24.15 USDT per level on a 2000 USDT account, which is 1.2075% of initial capital. On our $100,000 account that is 1,207.5 USDT per level, qty = amount / level price. With pyramiding = 100 the maximum inventory is 64 lots (about 77% of capital, which fits a CASH account). The 'investment' input (1545.84) is reference only and is not used in the logic. Webhook, alert, stats card, watermark, grid-line drawing and the timeframe warning are display/utility code with no effect on trading, and are not ported. Sanity check on our data: within the window, APE 15m closes ranged 0.0817–0.2527 (11,808 bars) and there are about 1,169 close-based down-crossings of grid levels, which is consistent with the source's 1,026 closed trades. The source's claimed +18.29% / PF 1.95 / DD 26% is unverified; our fee-inclusive backtest decides. Flat state: with no lots filled (price above all filled levels, or outside the window) the strategy is flat. In-window it is expected to hold at least one lot about 85-95% of the time, and outside the window it is idle apart from any stranded lots.

Line-for-line translation of the Pine grid logic: identical level formula, same SELL-then-BUY pass order, same close vs close[1] cross conditions, per-lot filled flags and quantities (kept as scalar attributes _f{i}/_q{i} so restarts restore them). Platform adaptations: (1) orders are market orders filled at the signal bar's close; Pine fills strategy.entry/close at the next bar's open, and its 2-tick slippage is replaced by our fill/impact model and the real 0.10% BINANCE_SPOT fee, which matches the source's commission_value. (2) Pine `time` is bar-open time, so the window test uses ts_init minus the 15m interval. (3) Sizing: the fixed 24.15 USDT/level on a 2000 USDT account is 1.2075% of equity, anchored once on the first grid buy and kept fixed afterwards, like the source's fixed cash amount. At most 64 lots, about 77% of capital, so it fits the CASH account. (4) Sells are clamped to the net long so spot can never go short. (5) Venue: BINANCE_SPOT instead of OKX spot, because OKX APE history is only days long. Display, alert and webhook code is not ported. Caveat (L175): the venue nets every lot into one position, so the positions report books overlapping lots as one open-to-flat trade. total_trades and avg_trade_return_pct therefore count netting episodes, not the roughly 1,000 per-lot round trips, and need per-lot reconciliation in review. The date window also means optimization and holdout slices outside 2026-01-12 to 2026-05-14 will be idle by construction.

verifying_stuck_3_timeouts

Implementation

Faithful port of 3Commas' '[3Commas] Static Geometric Grid' (MPL-2.0) on APEUSDT.BINANCE_SPOT 15m. 64 geometric levels between 0.10591 and 0.15466 (0.603% rung spacing). On each confirmed bar inside the 2026-01-12 to 2026-05-14 UTC window: SELL pass closes lot i when the close crosses up through level i+1, then BUY pass buys lot i (1.2075% of initial equity / level price) when the close crosses down through level i. Long-only, no leverage, trailing-up and arithmetic mode carried as inert parameters. Outside the window no orders are placed and stranded lots are held.

Outcome Summary

Verification must complete before performance can be assessed or later pipeline stages can run.

It was abandoned during verification after three timeouts (verifying_stuck_3_timeouts). Later stages were not reached.

A long-only APEUSDT.BINANCE_SPOT 15-minute static geometric grid with 64 levels, close-cross entries and exits, and trailing-up disabled.

No backtest report or performance metrics were provided, so return, Sharpe, and trade count cannot be assessed. The strategy did not reach backtesting.

Analysis

[3Commas] Static Geometric Grid

Backtest and paper results are hypothetical. Trading involves risk of loss.