forex-scalping.com

Why Conviction-Based Sizing Is Dangerous

It seems sensible to risk more on trades you feel strongly about, and less on weaker ideas. The practice, called conviction sizing, is common, and it is a hazard.

The problem is that confidence is a poor predictor of outcomes. Studies of judgment show that people are often overconfident, and the correlation between feeling certain and being right is weak. In trading, the trades that feel most certain often involve crowded consensus, recent success, or emotional attachment, all of which reduce edge.

Conviction sizing also amplifies losses. A trade that felt like a sure thing, sized at three times normal, loses three times as much when it fails. Because failures happen at a rate that does not depend on how you feel, the largest losses come from your most confident moments.

Further, it undermines your statistics. If risk varies by feeling, your results mix setups with sizing decisions, and you cannot evaluate either.

A more robust method is fixed risk per trade for a given setup type, based on tested expectancy. If you want to vary size, base it on objective criteria defined in advance, such as the quality score from a checklist, with tight limits, or on the measured volatility.

Test the idea on your journal: compare results of trades you rated high conviction with others. Many find no advantage.

Practical step: add a conviction rating to your journal and check whether higher ratings correlate with better R.

The biggest losses coming from the most confident moments is so true. The trade I'm sure about is exactly when I'm least careful about everything else. I skip checking the calendar, I place the stop a bit tighter so the bigger size fits, and I'm more likely to hold through news. So the loss isn't only bigger because of size, it's often a worse trade overall. A simple rule helps: the size for each setup type is fixed in the plan, and the only allowed change is reducing it, for example around events or after a drawdown. Increasing size happens monthly based on results, never on a single trade because it feels good.

Adding a conviction rating to the journal without changing size is a good experiment. Most people find the high conviction trades perform about the same as the rest. Rate each trade 1 to 3 before entering, keep risk identical, and after 60 or so trades compare average R by rating. If the 3s really outperform by a clear margin, you've found a quality filter, but then turn it into written criteria, like trend alignment plus a tested level, and size on those. Sometimes the 3s even do worse, often because high conviction comes after a winning streak or from a strong opinion about the news, not from the chart.