Every strategy I build gets published here — the wins, the marginals, and the failures. Full transparency. Sort, filter, and explore the data behind 433 algorithmic trading strategies.
Methodology
Every strategy in this archive is coded as a standalone Expert Advisor in MQL5 and pushed through a 10-phase validation pipeline. I use MetaTrader 5 with Model 4 (Every Tick based on real ticks) across 9 years of data (2017–2025) on 23 symbols. Phases include baseline screening, parameter sweeps, walk-forward validation on unseen data, real ECN commission testing ($7/lot RT), medium and harsh stress testing, multi-seed overfitting detection, and full statistical validation (Deflated Sharpe Ratio, Monte Carlo, FDR correction).
Gross profit divided by gross loss. PF > 1.30 required at baseline. PF > 1.0 required after commissions and stress testing. Values above 5.0 with many trades are flagged as suspicious.
Largest peak-to-trough equity decline. Must stay below 12% at baseline, 15% under harsh stress. Critical for prop firm compliance where drawdown limits are strict.
4+ overlapping anchored folds, each with a different out-of-sample window. The strategy must PASS all windows — not just one lucky split. Eliminates strategies that only worked in one specific market period.
MEDIUM: 2pip slip, 2x spread, $14 comm. HARSH: 5pip slip, 3x spread, $20 comm, 10% random trade rejection. If a strategy can't survive harsh conditions, it shouldn't trade real money.
Run each strategy with 5 different random seeds. PF variance must be <20%. If changing the seed turns a winner into a loser, the edge was noise, not signal.
100–200+ parameter combinations tested per strategy. Top performers pass 80–93% of configs. Proves the edge is robust across settings, not curve-fit to one lucky combination.
| ID ▲ | Name ▲ | Edge Type ▲ | Source ▲ | Best Symbol ▲ | PF ▼ | Trades ▲ | DD% ▲ | Status ▲ | Pipeline ▲ |
|---|