HomePortfolioMarketsSignals
Loading…
Out of sample — the last 30%, never used to design the rules
| Strategy | Return | Per year | Worst fall | Sharpe | Trades | Costs |
|---|
What £10,000 would have become
Versus doing nothing
In sample — the first 70%, for comparison
| Strategy | Return | Per year | Worst fall | Sharpe | Trades | Costs |
|---|
If a rule looks good here and bad above, it was fitted to the past rather than to the market. That is the usual outcome, and it is the reason for the split.
How to read this
Beating buy-and-hold is the only test that counts. A rule returning 8% a year looks fine until the benchmark returned 10% while you did nothing.
Win rate is the least useful number here. A rule can win 70% of its trades and still lose money, if the losses are larger than the wins. Profit factor and worst fall matter more.
Every result already pays costs — commission, the spread crossed on each trade, and slippage. Backtests that skip those look far better than reality.
This is history, not prediction. One asset, one period. A rule that won here can lose next year.
No results yet
Waiting for the first set of results to be published.