Machine learning on 24 indicators: no better than shuffled data
Failure type: overfit (no pattern beyond shuffle). Verdict: all three failed.
The idea
Let a model combine all 24 indicators. Three models, retrained each year from 2008 with purged data and a 20-session embargo: an elastic-net logistic model at 21 and 63 sessions, and extra-trees at depth 3 for 63 sessions. Each had a control trained on shuffled labels.
What happened
| Model | Sharpe | Exposure-matched SPY | Beat its shuffle control by 0.05? |
|---|---|---|---|
| Elastic-net, 21 sessions | 0.30 | 0.41 | No (worse than shuffle) |
| Elastic-net, 63 sessions | 0.32 | 0.41 | No |
| Extra-trees, 63 sessions | 0.25 | 0.41 | No |
Drawdowns were smaller than SPY's, which wasn't enough.
Why it failed
If a model can't beat itself trained on noise, it hasn't found a pattern. With 23 ETFs and daily bars there may simply not be enough independent data for these models to learn from.
Receipts: the plan was written down and committed first. 3 official trials plus shuffled controls.
Research and education only. These are paper results from historical data on a limited set of funds. Not investment advice, not a recommendation, not a prediction. See Legal.