Kelly & Fractional Kelly: Sizing for Growth Without Ruin

Most sizing advice tells you to risk a flat 1% per trade. The Kelly criterion asks a sharper question: given how often you actually win and how big your wins are versus your losses, what bet size grows your account fastest over many trades? The answer is elegant — and the honest follow-up is that the full-strength version is far too aggressive for real trading.

The idea in plain language

Kelly comes from a 1956 paper by a Bell Labs researcher, John Kelly, who was working on betting with an information edge. The core insight has survived 70 years: if you keep betting a fixed fraction of your money, the way to grow capital fastest in the long run is to maximise the growth rate of your account, not the average size of any one win.

That distinction matters more than it sounds. Trying to maximise your average outcome quietly pushes you toward betting bigger and bigger — and if you bet too big, a normal losing streak wipes you out. Once you are at zero, the average no longer means anything. So the goal is not "biggest expected win"; it is "fastest sustainable compounding."

For a simple win-or-lose trade, Kelly boils down to a balance of two things:

• Your edge — how much you expect to make on average.
• Your odds — how much you win when right versus how much you lose when wrong.

A rough, usable form for traders: take your win rate, and subtract your loss rate divided by your reward-to-risk ratio. If you win 50% of the time and your winners are twice the size of your losers (a 2-to-1 ratio), Kelly suggests risking about a quarter of your account on that edge.

A quarter of your account on one idea. That number alone should tell you full Kelly is not a retail trading plan — it is a theoretical ceiling.

Why full Kelly hurts

The same researchers who popularised Kelly for markets — including Edward Thorp, who used it from blackjack tables to a hedge fund — are also the people who warned loudest against betting it in full. The reasons are practical:

• Your edge is a guess. Kelly assumes you know your true win rate and payoff exactly. You do not. You have a sample of past trades, and the real numbers drift as markets change. Feed an over-optimistic edge into the formula and it tells you to bet far too much.
• The drawdowns are brutal. Even when your edge is real, full Kelly produces gut-wrenching equity swings. A well-known result is that a full-Kelly bettor is as likely to halve their account before doubling it as not. Most people abandon a strategy long before that.
• It ignores your stomach. The math optimises growth, not sleep. A size that is mathematically optimal but emotionally unbearable is the wrong size, because you will not follow it.

Fractional Kelly: the version people actually use

The standard fix is to bet a fraction of the Kelly number — most commonly half-Kelly (50%), and often a quarter. This is not a fudge; it is well-grounded. Going to half-Kelly cuts your account's volatility by roughly half while giving up only a small slice of the long-run growth. You keep most of the upside and shed most of the pain.

Think of the Kelly number as a speed limit, not a target. If the formula says 25%, half-Kelly says risk about 12%, and a cautious trader sizing to survive estimation error might land near the familiar 1–2% per trade. The flat-percent rule you already know is, in effect, a very conservative fractional-Kelly setting — which is exactly why it is good default advice.

A few honest cautions before you reach for the formula:

• Kelly needs a real, positive edge. If you are not consistently profitable, Kelly's answer is to bet nothing, and no fraction of nothing helps.
• Estimate edge from a large, recent sample, net of fees and slippage. A handful of lucky trades is not an edge.
• When you hold several positions that can move together, treat the cluster as one bigger bet — correlated trades stack risk.

Put it to work in FSP

You do not have to compute Kelly by hand to use the thinking. FSP auto-computes your R-multiples from each trade's entry, exit, and stop, and your dashboard tracks win rate and average win versus average loss — the exact inputs Kelly cares about. Review them under Analytics, pick a conservative fraction you can actually hold through a drawdown, and write that sizing rule into your Strategy's risk plan so every future trade inherits it.