If a formula tells you to risk 20% of your account on the next trade, the formula is not wrong so much as optimal for a different animal: a machine that can tolerate 50% drawdowns without changing behaviour. Full Kelly is that formula’s classic output. Half Kelly is what most discretionary traders should discuss instead.
This is a practical read of half Kelly: what it is, why full Kelly burns humans, and how to use it without turning your journal into a casino spreadsheet.
Kelly in one minute
The Kelly criterion estimates the fraction of capital to risk to maximize long-run geometric growth when you know (or estimate) win probability and payoff. In a simplified trading form, people often use something like:
f* ≈ W − (1 − W) / R
Where W is win rate and R is the ratio of average win to average loss (when losses are normalized). Exact variants differ; the point is the same: higher edge and better payoff → higher suggested fraction; thin edge → tiny fraction or zero.
TheSpeculatorsJournal computes a Kelly percentage from your win rate and average win/loss in analytics, with in-app guidance that most traders use half Kelly or less. That guidance exists for a reason.
Full Kelly optimizes the equity curve you would have if you never flinched. Half Kelly optimizes for the equity curve you can still trade.
What half Kelly is
Half Kelly means risking about half the full Kelly fraction. If full Kelly says 10% of capital per bet, half Kelly says ~5%. Quarter Kelly is even more common among cautious practitioners.
Why cut it:
- Estimation error. Your win rate and R from 80 trades are noisy. Full Kelly assumes you know the game. You do not.
- Path dependence. Full Kelly paths include deep drawdowns. Humans change systems at the bottom; the model assumes they do not.
- Costs and regime change. Fees, slippage, and dead edges make true f* smaller than the historical sample suggests.
Half Kelly is a humility discount on your own statistics.
Full Kelly vs half Kelly vs fixed risk %
- Full Kelly: theoretically aggressive growth; practically a behaviour hazard for discretionary traders.
- Half / quarter Kelly: uses your edge estimate but leaves room for error.
- Fixed risk % (e.g. 0.5–1% per trade): still the best day-to-day policy for most people. Use the free position size calculator so every stop maps to that policy.
Kelly is a ceiling research number from your journal sample. Fixed fractional risk is the operating system. If Kelly says 8% and your plan says 1%, you follow the plan. If Kelly says 0.4%, you should not be risking 1% on that edge without a hard look at expectancy and win rate quality.
How to use Kelly output in a journal review
- Pull a stable sample of one playbook (not mixed strategies).
- Read win rate, average win, average loss, and the Kelly % the analytics show.
- Halve it (or more) as a discussion band, not an order size.
- Compare that band to your actual average risk per trade. If you are sizing above half Kelly on a thin sample, you are leveraged to your own optimism.
- Revisit after regime changes. A Kelly number from last quarter’s trend market is not a permit for this quarter’s range.
Pair with risk of ruin thinking: high fraction + long losing streak = career risk, even with a positive edge on paper.
Common misuses
- Recalculating Kelly every morning on 15 trades. Noise becomes a throttle for overtrading.
- Using all-account win rate across unrelated setups. Segment or do not bother.
- Treating Kelly as a target to “get up to.” It is a cap, not a dare.
- Ignoring that prop floors redefine ruin. Evaluation accounts often need lower fractions than unrestricted personal accounts.
A monthly Kelly review ritual
Once a month, freeze a playbook sample (for example the last 100 tagged trades). Record win rate, average win, average loss, and the Kelly % from analytics. Write half Kelly and quarter Kelly beside your actual average risk per trade. If actual risk sits above half Kelly on a thin or decaying edge, you are not aggressive — you are inconsistent with your own data.
Then decide one policy for the next month: fixed 0.5%, fixed 1%, or “never above quarter Kelly of the trailing sample.” Put it in the plan. Kelly without a written policy becomes a post-hoc justification for whatever size you already wanted.
When estimates are garbage
If you changed strategies mid-month, mixed tickers with different cost structures, or have fewer than a few dozen trades, treat Kelly as undefined. Use a conservative fixed fraction and keep collecting clean tags. Forcing Kelly on garbage inputs is how people discover risk of ruin experimentally.
FAQ
Is half Kelly always best?
No. It is a widely used compromise. Some traders use quarter Kelly or stick only to fixed 0.5% risk. Consistency of policy beats fetishizing a fraction.
Does TheSpeculatorsJournal force half Kelly size?
No. It computes Kelly % from your stats and reminds that most traders use half Kelly or less. You still choose size; the calculator and plan enforce the habit.
What if Kelly is negative?
Then the sample does not support positive growth betting. Fix edge or stop trading that playbook — do not “half” a negative number into a positive size.
Should prop traders use Kelly at all?
As a sanity check only. Prop ruin is the firm floor. Size so a normal streak cannot hit it; see the prop challenge checklist.
How does this relate to R-multiples?
Kelly needs honest average win/loss. Journaling in R keeps those averages comparable; see R-multiples.
Conclusion
Full Kelly is a theoretical growth maximizer with a human-intolerable path. Half Kelly is the practical conversation: use your journal’s edge estimate, then cut it because your sample is imperfect and your behaviour is not a formula. Day to day, size from a fixed risk % and a stop. Use Kelly as a mirror in review, not as a gas pedal in the open.
This article is for educational purposes only and is not financial advice. Trading involves risk, including loss of capital, and past performance does not guarantee future results.