Dark Matter

Pulsar

ML-driven statistical arbitrage · Crypto perpetuals

BacktestWalk-forward out-of-sample · 181 days31 Mar 2026 – 28 Sep 2026

Equity · cumulative % from the start of the record

0%+10%+20%+30%Apr ’26MayJunJulAugSep

Statistics

Total return
+30.4%
Sharpe
5.41annualised
Sortino
9.83annualised
Max drawdown
−3.76%peak to trough, % of peak equity
Worst day
−2.34%16 Apr 2026
Worst month
−1.42%Sep 2026
Days
182

Total return, Sharpe, max drawdown and days are the record’s own figures. Sortino, worst day, worst month are computed on this page from its daily series, annualised at the series’ own 365 observations a year. CAGR is omitted: the record spans less than a year.

Drawdown · peak to trough, % of peak equity

−2%0%Apr ’26MayJunJulAugSep

Rolling Sharpe · 21-day window, annualised

010May ’26JunJulAugSep

Monthly returns

YearJanFebMarAprMayJunJulAugSepOctNovDec
2026no datano data0.0%+6.4%+2.9%+10.1%+7.8%+1.7%−1.4%no datano datano data

Outlined months are partial: the record starts or ends inside them.

How this was tested

  • This is a backtest: the strategy simulated over historical market data. No money was traded.
  • It covers 31 Mar 2026 to 28 Sep 2026, 182 days in the record.
  • Returns are measured as % of starting capital.
  • Costs: Maker fees charged on every fill. Not charged: slippage or funding. $10,000 starting capital, no compounding.
  • Out-of-sample: no day in this window was used to build or tune the strategy.

Record: Walk-forward out-of-sample · 181 days. Exported 30 Sep 2026 from platform commit 44852bc086 (ML-Stat-Arb-Signal-Service 44852bc08666 ops/oos_walkforward.py --recipe prod --fixed-keep 0.10 --size-by cal_pass --train-pools (pooled), flip exits, one position per pair; model pulsar_v3 (the live model, threshold 0.60)).