Paper trading, overfitting, win rate, and interpreting results
Covering forward validation with a validation ladder, metrics with a comparison matrix, typical failure examples, and a readiness checklist for going live.
Paper trading is applying a rule forward in the live market without risking real money. Its purpose is to catch what a backtest cannot show: how orders actually fill, delays, psychological pressure, and whether the rule stays consistent under live conditions. It is one of the most practical ways to detect overfitting.
Why it matters
A backtest looks at the past; paper trading looks at the unknown future. If a rule was overfit to the past, it breaks down quickly in paper trading. And fixating on a single metric like win rate is the most common interpretation mistake; a 70% win rate can still lose money if the losses are bigger than the wins.
The validation ladder
- Idea
An untested hypothesis.
- Backtest
Did it behave sensibly in the past?
- Out-of-sample test
Is it consistent on data it has not seen?
- Paper trading
Does it hold up real-time, without risk?
- Small live
A final check with very small real risk.
Read the metrics together
| What it says | Why it misleads alone | |
|---|---|---|
| Win rate | Share of winning trades | Ignores the size of losses |
| Average R | Average return per unit of risk | Unreliable if trade count is low |
| Max drawdown | The largest peak-to-trough fall | Does not show what you can endure |
| Trade count | Sample size | If small, every metric is suspect |
Typical failure examples
A rule overfit to the past becomes inconsistent live; a classic overfitting symptom.
Frequent small wins are erased by rare large losses; win rate misleads.
90% over 10 trades is statistically meaningless; it can be luck.
Readiness checklist for going live
- Was consistency seen over a sufficient number of trades?
- Do the results hold on out-of-sample data too?
- Can you psychologically endure the maximum drawdown?
- Could you follow the rule without manual override?
- Is the starting risk kept very small?
In Tradora’s teaching, the order is fixed: first an explicit rule, then a backtest, then forward validation. That order is the most honest way to tell whether an idea is genuinely robust or merely fitted to the past.
Even paper trading does not fully reflect the psychology of real money, and no validation guarantees future profit. Past and simulated results can differ systematically from live results.
You have completed the validation cycle. To reinforce the practice, loop back by redefining a rule or deepening the backtesting step.