Beyond Fixed Position Sizing
Most risk management advice recommends a fixed percentage risk per trade — say 2% of your bankroll. This is sound as a baseline, but it treats every prediction equally. In reality, some setups are significantly stronger than others.
Variable position sizing means allocating more capital to your highest-conviction setups and less to marginal ones. A high-confluence setup at a major support level with volume confirmation might warrant 2-3% risk, while a lower-quality opportunistic trade might only get 0.5-1%.
The key constraint is that your maximum risk per trade should still be capped. Variable sizing operates within the range of your minimum to maximum, not beyond your maximum.
Creating a Conviction Rating System
Develop a simple rating system — for example, A, B, and C grades — and assign pre-defined position sizes to each grade. A-grade setups (highest confluence, all criteria met) get full size. B-grade setups (most criteria met) get two-thirds. C-grade setups (minimum criteria met) get one-third.
The criteria for each grade should be specific and pre-defined. An A-grade might require: alignment with higher-timeframe trend, candlestick signal at a structural level, volume confirmation, and favorable regime.
Write down the grading criteria so that the assessment is as objective as possible. Subjectivity in grading leads to emotional sizing, which defeats the purpose.
The Mathematics of Variable Sizing
Variable sizing improves overall returns because it concentrates capital on predictions with the highest expected value. If your A-grade setups have a 65% hit rate and your C-grade setups have a 52% hit rate, allocating more to A-grade predictions is mathematically optimal.
The improvement can be substantial. A trader who allocates twice as much to A-grade predictions as C-grade predictions, assuming the hit rate differential described above, can improve portfolio returns by 20-30% compared to uniform sizing.
The risk is that your conviction rating must actually correlate with outcomes. If your A-grade predictions do not outperform your C-grade predictions, variable sizing adds complexity without benefit.
Avoiding Common Scaling Pitfalls
The most dangerous pitfall is allowing emotions to influence your conviction rating. After a winning streak, every setup looks like an A-grade. After losses, nothing seems worthy of more than a C-grade.
Guard against this by using an objective checklist. The conviction grade should be determined by the checklist results, not by your emotional state.
Another pitfall is over-sizing "revenge" trades after losses. The impulse to go big on the next prediction to recover is the opposite of disciplined scaling.
Implementing Variable Sizing Gradually
Start by tracking what your conviction level would have been for each prediction, without actually varying your size. After 50-100 tracked predictions, analyze whether your conviction ratings correlate with outcomes.
If the correlation is strong (A-grade significantly outperforms C-grade), begin implementing variable sizing with conservative differences — perhaps 1.5x for A-grade versus 1x for C-grade.
Expand the range gradually as you gain confidence in your rating system. Eventually, a 2:1 or 3:1 ratio between top-grade and bottom-grade sizing is a reasonable target.