Position Sizing: “Should Size Ever Change?” Is the Wrong Question

Ask ten traders whether position size should stay identical across every trade and you'll get ten different answers, most of them citing a different famous name. “Keep it flat, always” is common, disciplined, boring advice. “Add to your winners” gets Jesse Livermore's name attached to it. “Size to your conviction” sounds sophisticated and quantitative. All three get repeated as if they contradict each other. Once you actually look at what each camp means in practice, they mostly don't.

The Fixed-Size Case

The argument for identical size, trade after trade, is the one most prop-firm risk desks would recognize immediately: unpredictable position sizing is one of the clearest tells of emotional trading — sizing up on a hunch, sizing down after a scare, both driven by feeling rather than plan. Consistent size makes results legible. If size never changes, a losing streak is unambiguously information about the edge, not muddied by “was I also trading bigger that week.” It's the easiest rule in trading to state and, by most honest traders' own account, one of the hardest to actually keep.

Livermore's Pyramid

Jesse Livermore built the opposite reputation — famous for turning small stakes into fortunes by adding to winners — and it's usually flattened into “bet more when you're winning.” What he actually described was far narrower. Start with a small, exploratory position. Add only after the market has already moved in your favor and confirmed it — typically a new high following a pause, not just green on the screen. Each addition is smaller than the last, not bigger, so the position is heaviest at the entries with the most confirmation behind them and lightest at the most speculative one. Never add to a position that's losing, full stop — that's a different act entirely, and Livermore treated averaging down as one of the fastest ways to blow up an account.

The instruction isn't “size to how good this feels.” It's a fixed sequence, decided in advance, that happens to add size — triggered by price confirming the thesis, not by the trader feeling more sure of themselves as the position goes green.

Van Tharp's Percentage Risk

Van Tharp's model, laid out in his position-sizing work, is neither of the above: risk a fixed percentage of current capital on every trade — commonly cited around 1%, sometimes with a hard cap of 6–10% at risk across all open positions at once — rather than a fixed dollar amount or contract count. The dollar size still moves, but only because the account balance moved, on a schedule the trader set before any of these trades existed. Win, and the next size ticks up slightly because the base grew. Lose, and it ticks down. This is often called anti-martingale sizing precisely because it does the opposite of chasing losses with bigger bets — it shrinks exposure exactly when a losing stretch would tempt someone to do the reverse.

Kelly Criterion, and Why It's the One to Be Careful With

The mathematically purest version of “size to your edge” is the Kelly Criterion, a formula for the exact fraction of capital to risk given your true win probability and payoff ratio, built to maximize long-run growth. It's real math, not folklore — and it comes with two problems that matter more in practice than the elegance of the formula. First, it requires knowing your actual win probability with precision, which a systematic strategy with a large sample can estimate and a discretionary trader almost never can. Second, full Kelly sizing produces genuinely brutal equity swings even when the edge is real, which is exactly the kind of variance most traders can't sit through without deviating from the plan that was supposed to justify the size in the first place.

This is also where “size to conviction” quietly turns dangerous for a discretionary trader, because conviction isn't measured by anything external. It's a feeling, reported by the same mind that's currently inside the trade — and a feeling of certainty is exactly what overconfidence produces right before it's wrong. There's no instrument that tells a trader “this confidence is well-calibrated” versus “this confidence is the last good trade talking.” Both feel identical from the inside.

The Actual Dividing Line

Line these four up and the real disagreement isn't “same size or different size.” Fixed sizing, Livermore's pyramid, and Van Tharp's percentage model all vary in mechanics but agree on the same underlying rule: the size — or the rule that produces the size — was decided before the trader was inside the trade, by a condition anyone could check from the outside. Price confirmed a new high. The account balance changed. The percentage was set weeks ago. None of that requires trusting how confident you feel right now.

Confidence-weighted sizing is the one method here that asks a trader to measure something no one, including the trader, can measure reliably in real time. That's not a reason it never works — it's the reason it's the one method on this list that fails silently, in exactly the moment it's wrong.

If keeping size genuinely flat is hard to hold — and by most honest traders' account, it is — the fix isn't necessarily forcing it to stay flat forever. It's making sure whatever rule allows it to change was written on a day you weren't in a trade, not decided by how sure you feel about this one.