Each sizing method answers the same question, how much to risk next, in a different way. This lesson puts them side by side so you can pick one for your own strategy and tolerance for drawdowns.
The five methods side by side
| Method | Risk after losses | Risk after wins | Complexity | Main danger |
|---|---|---|---|---|
| Fixed amount | Same amount (a larger % of equity) | Same amount (a smaller % of equity) | Very low | No compounding; the % risk rises in drawdowns |
| Percentage | Falls with equity | Rises with equity | Low | A percentage that is too high for the strategy |
| Kelly (fractional) | Falls with equity; depends on estimated W and R | Rises with equity | Medium | Estimation error; full Kelly gives huge swings |
| Martingale | Multiplied (for example 2×) | Back to base | Low | Exponential growth of risk; ruin on a long streak |
| QuantRaider (adaptive) | Reduced mildly by 10-trade momentum | Raised by momentum | Medium | More risk after lucky streaks; extra parameters to overfit |
Same trades, different results
Worked example
Loss, loss, win with a reward-to-risk of 1
- Start with €10,000. Fixed risks €100: 9,900, then 9,800, then a win of €100 gives 9,900.
- Percentage risks 1%: 9,900, then 9,801, then a win of €98.01 gives 9,899.01.
- Martingale (base €100, 2×): 9,900, then a €200 loss gives 9,700, then a €400 win gives 10,100.
Martingale ends €200 ahead of the others on this three-trade sample, yet it needed a €400 stake; one more loss would have made the gap very different.
A short sample like this says nothing about long-run behaviour. The drawdowns and the largest stake matter at least as much as the final value.
How to choose
- Start from the maximum drawdown you could tolerate, then pick the risk level at which a plausible losing streak stays within it.
- Use fixed or percentage sizing for a new or untested strategy, because they need no estimate of your edge.
- Consider a fraction of Kelly only after a large sample, and expect the estimate to move.
- Treat martingale as a hazard, not as an option for accounts you cannot afford to lose.
- Test any adaptive rule against plain percentage risk on the same trades, with the same costs.
Key takeaways
- The methods differ in how risk responds to wins and losses, in complexity and in their main danger.
- Choose by the drawdown you can tolerate and the evidence you have, not by the highest final value in a simulation.
- Fixed and percentage sizing are the simplest starting points; Kelly and adaptive rules add estimates and parameters.
- Compare any method on your own trades before relying on it.