The numbers, and how
they were produced
Every figure on this page comes from a walk-forward backtest over the stated period, with fees, funding and modelled slippage applied. None of it comes from customer accounts, and none of it is a forecast. The methodology is at the bottom and is the part actually worth reading.
Hypothetical results. A backtest benefits from knowing the period it runs over and reproduces none of the experience of holding a losing position with real money in it. Past performance does not indicate future results — the risk disclosure explains what that sentence actually costs.
Equity curve
120 months · walk-forward · out-of-sample segments only
Expectancy
0.31R
per trade, net
Hit rate
44%
612 trades
Max drawdown
−8.2%
peak to trough
Cost drag
−29%
of gross return
Month by month
The losing months are rendered at the same weight as the winning ones. A performance page that shades drawdown lighter than gain is arguing with its own data.
| Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov | Dec | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2024 | -1 | +2.4 | +3.8 | -1.9 | +2.6 | -0.3 | -2.6 | -2 | +4.1 | +2.8 | +3.6 | +2.8 |
| 2025 | +2.4 | +3.1 | +3.5 | -1.4 | +1.6 | -0.1 | +0.3 | +0.1 | +0.4 | +4.2 | -2.2 | -1.5 |
Attribution
An equity curve tells you something worked. Attribution tells you what — and usually that a comfortable total is one strategy in one session carrying two that are not.
- structure-v4+58%
- breakout-v2+31%
- meanrev-v1−12%
- carry-v1+23%
Execution
Measured against the decision price on every fill, not assumed.
- Close to orderp95, all stages
- 62 ms
- Realised slippagemedian, passive fills
- 1.4 bp
- Slippagep95, aggressive fills
- 6.1 bp
- Partial fill ratereconciled, not chased
- 3.2%
- Orders rejected by venueretried or abandoned
- 0.4%
Methodology
If this section did not exist, nothing above it would mean anything.
Walk-forward, not fitted
Parameters are chosen on an in-sample window and evaluated on the window after it, which is then never reused. Only out-of-sample segments appear in the curve above.
Costs applied, and shown separately
Maker and taker fees per venue, funding on perpetuals, and slippage modelled against the book depth at the decision time. Gross and net are both available above because the difference is the honest part.
Same engine as live
Backtests run through the same decision pipeline and the same risk service as production. There is no research-only shortcut that would not survive contact with the live path.
Survivorship handled
Symbols delisted during the period remain in the universe until the date they were delisted, rather than being quietly excluded because they no longer trade.
One capital assumption
Fixed fractional risk per trade at 0.5% of equity, compounding. No leverage beyond what the position sizing implies, and no pyramiding.
What is not modelled
Exchange outages, API downtime and liquidity crises deeper than the recorded book. These make live results worse than backtested ones, and are the main reason the two differ.
Run it yourself rather than taking this page's word for it. The quickstart gets a backtest running in about ten minutes, and the harness that produces these numbers is open source.