Four kinds of result, never mixed: research candidates (what the lab found), the selection process (how well the rules picked recipes on data they hadn't seen), forward demo results (actual trading on an eToro demo account) and live results (real money). Everything is after eToro's costs. Short profitable periods are not turned into annual-return claims.
Run 2026-11-02 18:52:09 UTC · recipes chosen on the older half, results below are the newer, out-of-sample half unless noted.
15-min best-fit momentum (current) (1.5σ), stop 1.5 ATR, target 1.5 ATR · closest rejected candidate (never selected, never traded)
| Metric | Value | Why it matters |
|---|---|---|
| Net return after costs | −6.95% per market / month on the newer half · +1.26% on the older half it was chosen on | what remains after trading costs |
| Maximum drawdown | 5.8% | worst fall of an equal-weight basket of these markets, before leverage |
| Benchmark (buy & hold) | −2.63% / month | holding the same markets over the same dates |
| Trades and test dates | 202 trades · 4 markets · 2026-09-28 → 2026-11-02 | split date 2026-10-16 |
| Win rate · profit factor | 38% · 0.42 (newer half) | a low win rate can still pay if winners are larger |
| Confidence after multiple testing | day-clustered t 0.65 · needed 3.43 for 168 candidates | many recipes tried; correlated markets aren't independent evidence |
| Cost stress | with 0.02% slippage per side: −8.25% · costs ×1: −6.96% · costs ×1.5: −8.34% · costs ×2: −9.72% | newer-half return per market / month if costs were higher |
| Neighbouring settings | 0% of 8 nearby settings positive on the newer half · median −4.39% | if only one exact setting works, it is probably noise |
| Exposure and leverage | 5× default leverage · risk 0.5% of equity per trade at the stop | makes risk comparable |
Monthly returns (whole test, per market, before leverage)
15-min fade stretched moves (2σ), stop 3 ATR, target 2 ATR · closest rejected candidate (never selected, never traded)
| Metric | Value | Why it matters |
|---|---|---|
| Net return after costs | −1.53% per market / month on the newer half · +0.99% on the older half it was chosen on | what remains after trading costs |
| Maximum drawdown | 1.7% | worst fall of an equal-weight basket of these markets, before leverage |
| Benchmark (buy & hold) | +0.06% / month | holding the same markets over the same dates |
| Trades and test dates | 183 trades · 7 markets · 2026-09-28 → 2026-11-02 | split date 2026-10-15 |
| Win rate · profit factor | 45% · 0.53 (newer half) | a low win rate can still pay if winners are larger |
| Confidence after multiple testing | day-clustered t 1.06 · needed 3.43 for 168 candidates | many recipes tried; correlated markets aren't independent evidence |
| Cost stress | with 0.01% slippage per side: −1.84% · costs ×1: −1.53% · costs ×1.5: −1.59% · costs ×2: −1.65% | newer-half return per market / month if costs were higher |
| Neighbouring settings | 0% of 19 nearby settings positive on the newer half · median −0.74% | if only one exact setting works, it is probably noise |
| Exposure and leverage | 5× default leverage · risk 0.5% of equity per trade at the stop | makes risk comparable |
Monthly returns (whole test, per market, before leverage)
4-hour buy dips / sell rallies in the trend (1.5σ), stop 3 ATR, target 1.5 ATR · closest rejected candidate (never selected, never traded)
| Metric | Value | Why it matters |
|---|---|---|
| Net return after costs | −2.99% per market / month on the newer half · −0.86% on the older half it was chosen on | what remains after trading costs |
| Maximum drawdown | 23.4% | worst fall of an equal-weight basket of these markets, before leverage |
| Benchmark (buy & hold) | −3.61% / month | holding the same markets over the same dates |
| Trades and test dates | 109 trades · 8 markets · 2025-09-08 → 2026-11-02 | split date 2026-04-06 |
| Win rate · profit factor | 54% · 0.39 (newer half) | a low win rate can still pay if winners are larger |
| Confidence after multiple testing | day-clustered t -0.97 · needed 3.34 for 121 candidates | many recipes tried; correlated markets aren't independent evidence |
| Cost stress | with 0.1% slippage per side: −3.26% · costs ×1: −2.99% · costs ×1.5: −4.44% · costs ×2: −5.89% | newer-half return per market / month if costs were higher |
| Neighbouring settings | 0% of 3 nearby settings positive on the newer half · median −4.97% | if only one exact setting works, it is probably noise |
| Exposure and leverage | 1× default leverage · risk 0.5% of equity per trade at the stop | makes risk comparable |
Monthly returns (whole test, per market, before leverage)
1-hour trend breakout (1.5σ), stop 3 ATR, target 1.5 ATR · closest rejected candidate (never selected, never traded)
| Metric | Value | Why it matters |
|---|---|---|
| Net return after costs | −2.88% per market / month on the newer half · −1.44% on the older half it was chosen on | what remains after trading costs |
| Maximum drawdown | 9.1% | worst fall of an equal-weight basket of these markets, before leverage |
| Benchmark (buy & hold) | +2.40% / month | holding the same markets over the same dates |
| Trades and test dates | 109 trades · 4 markets · 2026-07-06 → 2026-11-02 | split date 2026-09-01 |
| Win rate · profit factor | 60% · 0.28 (newer half) | a low win rate can still pay if winners are larger |
| Confidence after multiple testing | day-clustered t -1.52 · needed 3.39 for 144 candidates | many recipes tried; correlated markets aren't independent evidence |
| Cost stress | with 0.03% slippage per side: −3.34% · costs ×1: −2.88% · costs ×1.5: −4.06% · costs ×2: −5.24% | newer-half return per market / month if costs were higher |
| Neighbouring settings | 0% of 1 nearby settings positive on the newer half · median −2.29% | if only one exact setting works, it is probably noise |
| Exposure and leverage | 1× default leverage · risk 0.5% of equity per trade at the stop | makes risk comparable |
Monthly returns (whole test, per market, before leverage)
15-min fade stretched moves (2σ), stop 3 ATR, target 1 ATR · closest rejected candidate (never selected, never traded)
| Metric | Value | Why it matters |
|---|---|---|
| Net return after costs | +0.37% per market / month on the newer half · +0.13% on the older half it was chosen on | what remains after trading costs |
| Maximum drawdown | 0.2% | worst fall of an equal-weight basket of these markets, before leverage |
| Benchmark (buy & hold) | −0.80% / month | holding the same markets over the same dates |
| Trades and test dates | 347 trades · 12 markets · 2026-09-28 → 2026-11-02 | split date 2026-10-14 |
| Win rate · profit factor | 80% · 1.52 (newer half) | a low win rate can still pay if winners are larger |
| Confidence after multiple testing | day-clustered t 0.52 · needed 3.43 for 168 candidates | many recipes tried; correlated markets aren't independent evidence |
| Cost stress | with 0.005% slippage per side: +0.17% · costs ×1: +0.37% · costs ×1.5: +0.34% · costs ×2: +0.32% | newer-half return per market / month if costs were higher |
| Neighbouring settings | 100% of 15 nearby settings positive on the newer half · median +0.43% | if only one exact setting works, it is probably noise |
| Exposure and leverage | 5× default leverage · risk 0.5% of equity per trade at the stop | makes risk comparable |
Monthly returns (whole test, per market, before leverage)
| Asset class | Folds | Folds with a pick | Trades | Out-of-sample / month |
|---|---|---|---|---|
| commodities | 4 | 0 | 0 | 0.00% |
| indices | 4 | 1 | 38 | −0.78% |
| crypto | 4 | 0 | 0 | 0.00% |
| stocks | 4 | 0 | 0 | 0.00% |
| fx | 4 | 2 | 64 | +0.01% |
Each fold picks a recipe using only data before it (trades still open at the fold boundary are dropped), then trades the next block with that pick frozen. This measures the selection rules, not any single recipe. A t-statistic threshold is a screening rule, not proof of an edge: many recipes are tested, trades overlap in time and markets move together, so ordinary per-trade t-statistics overstate confidence.
Actual trades on the operator's eToro demo account (virtual money), only when the operator chooses to publish them.
The operator's demo track record isn't published yet.
No live track record yet. Real-money results will appear here separately once there are some.