A high win rate does not by itself establish a safe position size. Kelly sizing depends on both the probability of a win and the size of an average win relative to an average loss, and its output can change sharply when those estimates move. Use the calculator as a model of a repeated, simplified bet; a positive Kelly fraction is not a guarantee that a live crypto strategy has a reliable edge or that its drawdowns will fit your account.
What the Kelly calculator uses
The tool takes win rate W as a percentage, average win A, average loss L, and optional account equity. It converts the win rate to a probability and calculates the payout ratio:
W = win rate percent ÷ 100
R = average win ÷ average loss
Kelly fraction = W − (1 − W) ÷ R
The returned Kelly result is the fraction expressed as a percentage. Half Kelly and quarter Kelly are one-half and one-quarter of that same fraction. If equity is entered, the tool's additional risk amount is equity × max(0, Kelly fraction ÷ 2): it reports half-Kelly exposure when Kelly is positive and zero when the calculated edge is zero or negative. This amount is an output of the formula, not a recommendation to risk that share on every trade.
These inputs describe an idealized sequence in which the estimated win probability and payoff ratio are meaningful for repeated outcomes. The calculator does not model correlated trades, changing market conditions, fees, funding, slippage, gaps, or a strategy's uncertainty. The Stanford paper on risk-constrained Kelly gambling studies drawdown risk as an additional constraint, which illustrates that the classical growth fraction alone does not set a trader's acceptable drawdown limit.
Worked example: 70% winners, losses twice as large
Suppose a strategy's historical win rate is 70%, its average winning trade is $100, and its average losing trade is $200. Enter the values as 70, 100, and 200. The amounts can be any consistent unit because the formula uses their ratio.
First calculate the reward-to-loss ratio:
R = $100 ÷ $200 = 0.5
Then calculate the Kelly fraction:
f = 0.70 − (1 − 0.70) ÷ 0.5
f = 0.70 − 0.60 = 0.10, or 10%
The corresponding half-Kelly result is 5%, and quarter Kelly is 2.5%. If account equity is entered as $10,000, the calculator's half-Kelly risk amount is $10,000 × 0.10 ÷ 2 = $500. That is the model output given these assumptions; it does not mean the next loss will be limited to $500 or that the assumptions are dependable.
Now keep the average win and loss amounts unchanged but reduce the estimated win rate from 70% to 65%:
f = 0.65 − (1 − 0.65) ÷ 0.5
f = 0.65 − 0.70 = −0.05, or −5%
A five percentage-point change turns the calculated edge from positive to negative. The tool still displays the negative Kelly value, while its optional risk amount becomes zero because it does not suggest a positive half-Kelly amount for negative Kelly. This sensitivity is why a strong-looking win rate cannot be interpreted without the average payoff and confidence in the estimate.
Use the calculator to compare the full, half, and quarter fractions for your assumptions. Treat the estimates as scenarios. In particular, do not input a headline win rate without pairing it with the average outcome sizes from the same strategy rules and the same measurement period.
Why the estimate can mislead
The payoff ratio may be unfavorable. A strategy can win frequently but lose much more on each losing trade. The formula subtracts expected loss exposure through (1 − W) ÷ R; when average wins are small relative to losses, a high win rate can still produce a very small or negative Kelly fraction.
A backtest win rate is an estimate. A small number of trades may not represent the rate the strategy will achieve in future conditions. Changing the sample, fees, execution assumptions, or signal rules can change both win rate and average win/loss. As the example shows, small input changes can have a large sizing effect near the break-even boundary.
Losses may cluster. The basic formula summarizes average outcomes and assumes repeated bets with stable probabilities. If losing trades are correlated, several may occur together. A sequence of losses can cause a substantial drawdown even if the long-run win rate eventually matches the input. The simple calculator does not estimate drawdown probability or the length of a losing streak.
Trade costs change the outcomes. If the historical win rate or average win and loss are recorded before commissions, funding, and slippage, the values do not represent net outcomes. Costs can turn small gross wins into net losses and change the average payoff ratio. Use realized results after costs when possible, and be consistent about the sample and trade definition.
Risk amount is not the same as leverage or margin. The optional dollar amount is half the calculated Kelly fraction times equity. It does not determine contract quantity from a stop distance, account for required margin, or guarantee a maximum loss. A stop can execute beyond its trigger, while liquidation and contract rules vary across venues.
A practical way to use the estimate
Start with a defined strategy and calculate its win rate and average win/loss from the same set of completed trades. Include the same fee and execution assumptions for each outcome. Then check a range of plausible inputs rather than relying on a single point estimate: lower the win rate, reduce average win, or raise average loss to see whether the edge persists.
Compare the full Kelly result with the half and quarter fractions, but do not assume a fraction is automatically suitable. Smaller fractions reduce the exposure implied by the model, while they do not remove the possibility that its inputs are wrong. Decide on an acceptable drawdown independently and test whether the position size can fit that constraint under losing sequences.
If translating a risk budget into a quantity, use the position size calculator with an explicit entry, stop, equity, and risk allowance. If deciding whether a strategy's average winner is large enough relative to its stop, use the risk-reward calculator. These tools answer different questions: Kelly uses estimated outcome probabilities and payoff sizes, while stop-based sizing estimates loss at a specified price under its own assumptions.
Common mistakes
- Using win rate alone. Include average win and average loss; the ratio of those amounts affects whether the formula returns a positive fraction.
- Mixing gross and net results. Include fees and other costs consistently when measuring the wins and losses used as inputs.
- Treating a small sample as a known probability. Historical win rate is an estimate and may change out of sample.
- Assuming quarter Kelly makes bad inputs safe. Fractional sizing scales the formula output but cannot correct a mismeasured edge or correlated losses.
- Reading a negative Kelly result as a short signal. It means the provided repeated-bet assumptions do not imply a positive fraction under this model; it is not a direction forecast.
- Confusing half-Kelly exposure with a stop-loss limit. It is not a guaranteed cap on dollars lost in one trade.
- Entering inconsistent outcome units. Average win and average loss must use the same basis, such as both dollars or both percentages.
- Using leverage to justify a larger fraction. Leverage does not improve the estimated edge and can increase liquidation and execution risk.
FAQ
Can a strategy with a 70% win rate have negative Kelly?
Yes. If average losses are large enough relative to average wins, the payout ratio can make the result negative despite a high win rate.
Why does the calculator show a positive Kelly but a zero risk amount?
It reports the Kelly percentage separately and calculates the optional risk amount as half Kelly times equity, floored at zero. Check that equity is positive and that the displayed output corresponds to the inputs.
Is half Kelly always safer than full Kelly?
It produces a smaller fraction under the formula, but it does not ensure an acceptable drawdown or protect against inaccurate assumptions. Risk tolerance and model reliability still matter.
Should average win and loss include trading fees?
Prefer net outcomes that include relevant fees, funding, and execution costs. Mixing net and gross results can distort both average payoff and the implied edge.
Does this calculator tell me how many contracts to trade?
No. It estimates a theoretical fraction and optional dollar amount from probability and payoff inputs. Contract quantity also depends on entry price, stop, contract specifications, and other risk constraints.
This guide is for educational purposes only and is not financial advice.