Strategy Lab
Test it the way it would really trade.
Write strategies in Python, run them against history that does not flatter them, and find out what your odds actually are.
Build
A builder that checks before it runs.
Strategies are Python on the engine, with every parameter typed and bounded. Saving checks the source and running needs a clean check, so a strategy that cannot be evaluated honestly does not get evaluated at all.
- Typed parameters with their ranges, checked on save
- No lookahead: the checks refuse a rule that reads ahead
- The same source runs in backtest and forward test
Backtest
Every trade, and where it came from.
A run gives you the curve and the trades that made it. Each one records the regime it was taken in, so you can ask whether an edge is real or whether it was one market condition that happened to repeat.
- Fills at bar closes, never inside a bar
- Every trade tagged with its regime and reason
- Parameter sweeps, with the whole surface kept
Prop firm sim
Your odds of passing, not your best run.
Walk a run through an evaluation the way a firm would score it, then resample it many times. One pass tells you very little. The distribution behind it tells you what to expect, including how deep the drawdown gets on the way.
- Evaluation rules as the firms write them
- Monte Carlo resampling around the empirical result
- Drawdown and time to pass, as distributions
Intelligence
Compare runs without fooling yourself.
Runs are ranked on a composite and on its parts, and every figure carries the percentile it sits at within the comparison. The statistics are recomputed in house from the trades rather than read off whatever produced them.
- A composite you can see the weights of
- Every metric beside its percentile in the set
- Recomputed from trades, not taken on trust
What a backtest is worth.
Nothing fills inside a bar
Stops fill at the close that crosses them, plus a tick. The same rule a live account would feel.
Labels arrive when they would have
A regime label reaches your strategy at the moment it would have been known, delay included.
Every run is reconciled
Equity is rebuilt, fills are matched to orders, and anything that does not add up is named rather than smoothed.
Free during the alpha.
We invite traders in small groups. People on the waitlist get in first.
Request access