Performance

Two independent 11-month backtests of the same engine, on the same asset, run at different daily checkpoints — so the results aren't an artifact of when the engine happened to look. Every figure below traces to the committed dataset in this repo (lib/performance/data); nothing here is hand-typed.

Methodology & caveats

Asset: BTC_USDT perpetual

Venue: mexc

Engine: BTC Adaptive Legacy v1

Model: mistral-small-2603

Window: 2025-07-01 → 2026-05-31

Confidence floor: 50%

One asset (BTC perpetual), one venue (MEXC), one 11-month window.
A backtest, not a live track record. The app has no live track record yet.
Fees, slippage and funding are modeled — modeled is not the same as real fills.
Run A had 3 errored and 1 ambiguous checkpoint(s); Run B had 0 and 1.
Past performance does not predict future results.

Checkpoints: 12:00 / 15:00 / 23:00 UTC

2025-07-01 to 2026-05-31 · started at $10,000

Return

+88.9%

Win rate

66.4%

Profit factor

3.16

Max drawdown

-5.26%

Trades

119

79W · 40L

Expectancy

$74.67

per trade

Green months

10 / 11

Traps avoided

140

Equity curve

Monthly PnL

Tested checkpoints

Checkpoint	Trades	Win rate	PnL
12:00 UTC	29	65.5%	$2,346.04
15:00 UTCstrongest tested	57	70.2%	$4,695.67
23:00 UTC	33	60.6%	$1,843.59

Checkpoints: 07:00 / 15:00 / 23:00 UTC

2025-07-01 to 2026-05-31 · started at $10,000

Return

+87.2%

Win rate

75.5%

Profit factor

4.32

Max drawdown

-2.73%

Trades

71W · 23L