How we test.
A research process built to find the flaws in a strategy before the market does.
ForexLab is a small research operation building automated trading strategies for the MetaTrader platforms. The work is unglamorous: write a hypothesis, code it, test it honestly, and throw away most of what you build.
Each Expert Advisor is tested against several currency pairs, one calendar month at a time, on tick data at high modelling quality. Results are recorded per month rather than as a single smooth multi-year curve, because a monthly breakdown is where the weaknesses show up — the flat quarters, the one bad week that ate a year of gains, the pair that only ever worked in one regime.
Every build is versioned and its results are tied to that version. If a change to the entry logic improves the numbers, the comparison is on record. If it doesn't, that's on record too. Results are cross-checked on more than one broker so we can tell a genuine edge from a quirk of one feed.
We publish losing months. A strategy that has never had one has almost certainly been fitted to its test data, and the only honest way to show that we haven't done that is to leave the bad months in.
Testing standards
- Real tick data, high modelling quality
- Spread and commission modelled per broker
- One record per EA, pair, month and timeframe
- Results validated before publication
- Cross-checked across multiple brokers
- Version history retained in full
Important. Every figure on this site comes from historical strategy-tester backtests, not from live trading. Backtested results are hypothetical, benefit from hindsight and do not account for every real-world cost or execution condition. Past performance is not a reliable indicator of future results. Nothing here is financial advice or an offer to sell a financial product. Trading leveraged instruments carries a high risk of loss.