Built like an experiment.
A backtest is a hypothesis. Our job is to try to kill it before the market does. A strategy that fails any of the seven gates below never trades.
Generate
Ideas come from two places: classic rules published by well-known traders and researchers, and large-scale machine generation. Both go through the same funnel, with the same costs: realistic spreads, commissions, slippage and overnight financing.
Out-of-sample
The history is split. The strategy is built on one part and judged on years it has never seen. We require profit in every period we test, not just on average.
Parameter heat map
We move every setting up and down and map the result. A real edge looks like a broad plateau. A curve-fitted one looks like a single spike. The original settings must not be much better than the median of their neighbors.
Monte Carlo
Thousands of alternative histories: trades reshuffled, some skipped, extra slippage added. We size risk on the worst 5% of those runs, not on the backtest itself. The strategy must keep at least half of its quality in that worst case.
Second source
Every survivor is rebuilt on a second platform with another broker's data and compared trade by trade. If the two versions disagree, the strategy is out.
Portfolio fit
A good strategy that moves with what we already run adds risk, not return. New strategies must have low correlation with the portfolio. Good ones that fail this test wait in reserve.
Live, in public
Strategies run on independently monitored accounts. Every month the live result is compared with the range our simulations predicted. A strategy that drifts outside that range is cut in weight or removed, and we say so.
What our strategies will not do.
No martingale, no grid, no averaging down
Losing positions are never doubled or added to.Every position has a defined exit
A stop, a target or a time limit, set before the trade opens.Risk sized on the worst case
Position sizes come from the Monte Carlo worst 5%, not from the average backtest.No promises of returns
We publish method, simulations and live results. You judge them.Many small edges, little in common.
The current portfolio combines 22 strategies across US indices, gold and currency crosses: short-term mean reversion, opening-range breakouts and multi-timeframe systems.