Position Size Creep: Are You Trading Bigger After a Winning Streak?
Position size creep is when the risk you take per trade drifts above your plan without a decision behind it. The most common trigger is a run of wins. Three green trades in a row, and the fourth one goes on at 1.5% of the account instead of 1%. Nobody writes that rule down. It just happens.
Most sizing guides cover how to calculate a position, and our guide to position sizing does that. This post covers a different job: finding creep in your own closed trades and testing whether the pattern holds up or is noise. The worked example uses 150 hypothetical trades from a seeded simulation.
The short version
- Measure size as risk at the stop divided by account equity at entry. Lots and contract counts hide creep.
- Split your trades by what came right before them: after 2 or more wins in a row, and everything else. Compare the median risk of the two groups.
- In the example, the median risk after 2+ wins was 1.40% against 1.01% for every other trade, a ratio of 1.38. Shuffling the risk column 10,000 times never produced a gap that wide.
- Over 10,000 hypothetical journals, creep took more total risk than the 1% plan and had a deeper drawdown in 87.0% of them. A flat 1.13% on every trade, the same average risk, ended higher than creep in most runs.
- If you want more risk, raise it on purpose for every trade. A streak is a poor signal to size on.
What counts as creep and what doesn’t
With percent-risk sizing, your dollar risk rises as the account grows. Risking 1% of $10,000 is $100. Risking 1% of $12,000 is $120. That’s the plan working as designed.
Creep is a rise in risk as a percent of equity. It shows up in three common ways:
- You add size after a run of wins because you feel in sync with the market.
- You keep the same lot size but widen the stop, so each lot carries more risk.
- You round lots up “just this once” on trades that feel strong.
The second one is why lot size is the wrong measure. Here’s a hand-checked example on EUR/USD in a USD account, where one standard lot moves $10 per pip:
| Trade | Lots | Stop distance | Risk at stop |
|---|---|---|---|
| A | 0.50 | 20 pips | $100 |
| B | 0.30 | 50 pips | $150 |
Trade B uses fewer lots and risks 50% more. A journal that only shows lots would call B the smaller trade.
The mirror image, sizing up after losses to win it back, is covered in our guide on how to stop revenge trading. The check below works for both. Only the grouping changes.
What you need from your journal
For each closed trade, in the order you took them:
- Account equity at entry. If you don’t log it, rebuild it from the starting balance and each closed result.
- Risk at the stop in money: entry price to initial stop, times position size, times the value of one point.
- The result in money and in R, where 1R is the amount you risked on that trade.
- Whether the trade was a win or a loss.
Risk percent for each trade is risk at the stop divided by equity at entry. Use the initial stop. A stop moved later changes the outcome, not the size decision you made at entry.
If some trades had no stop, keep them out of this check and count how many you dropped. A size decision with no stop has no risk number to compare.
How to check your own trades
- Add a column for the win streak before each trade: 0 after a loss, 1 after one win, 2 after two wins in a row, and so on.
- Put trades with a streak of 2 or more in one group and all other trades in the second group.
- Take the median risk percent of each group. Use the median because one oversized trade can drag an average a long way.
- Divide the first median by the second. That’s your creep ratio.
- Run the shuffle check (next section) to see whether a ratio that size could come from chance.
- Compare the average R of the two groups. If your sized-up trades don’t earn more R per trade, the extra size bought nothing.
Write the pass rule down before you look: for example, a creep ratio of 1.10 or less passes. Lot rounding alone moves risk a little from trade to trade. In the example below, a journal that stuck to 1% with ±10% rounding came out at 1.00.
Worked example: 150 hypothetical trades
The setup: a $10,000 account, a plan of 1% risk per trade, a 50% chance of winning each trade, wins paying between 0.8R and 2.4R, and losses of 1R. Every trade is independent of the ones before it, so a streak tells you nothing about the next trade.
The trader follows the plan after a loss or a single win. After 2 or more wins in a row, 70% of the time they size up: 1.4% after 2 wins, 1.8% after 3, 2.2% after 4, capped at 2.5%. Every size, plan or not, also gets ±10% of lot rounding. The same 150 outcomes are then replayed at a flat 1% to show the plan.
This journal had 81 wins out of 150 (54.0%) and a net result of +56.9R. The longest winning streak was 5 trades.
| Group | Trades | Median risk | Win rate | Average R |
|---|---|---|---|---|
| After 0 or 1 wins | 115 | 1.01% | 56.5% | 0.43R |
| After 2+ wins in a row | 35 | 1.40% | 45.7% | 0.22R |
Dividing the two medians before rounding gives a creep ratio of 1.38. Of the 35 trades after a streak, 22 went on at more than 1.1% risk.
The sized-up trades didn’t win more often. They won less often in this sample, by chance, since the example was built so that streaks carry no information. On the 19 losing trades after a streak, the extra size cost $1,346 more than the plan would have. On the 16 winners, it added $1,176.
| Creep | Plan (1%) | |
|---|---|---|
| Ending equity | $17,152 | $17,412 |
| Maximum drawdown | 5.9% | 4.9% |
| Largest single loss | −$437 (at 2.5% risk) | −$175 |
The account grew either way. Creep ended $260 lower, with a deeper drawdown and a largest loss 2.5 times the plan’s. If this trader works under a daily loss limit of 5%, one stop at 2.5% risk uses half of it.
The shuffle check
A ratio of 1.38 from 35 trades could be noise. The shuffle check tests that without any formulas:
- Keep the streak column as it is.
- Shuffle the risk percent column randomly, so sizes no longer line up with streaks.
- Compute the creep ratio again.
- Repeat 10,000 times, then count how many shuffled ratios were as high as yours or higher.
If streaks had nothing to do with your size, a shuffled order would match your measured ratio often. In the example, 0 of 10,000 shuffles reached 1.38. A common bar is fewer than 5% of shuffles, set before you run it.
A spreadsheet can do this with a random sort column and a copy-paste loop, but 10,000 rounds is easier in a short script. The textbook alternative is the Mann-Whitney U test, which compares two groups without assuming a bell curve. The NIST Engineering Statistics Handbook walks through it. On the example data it gives z = 5.28, the same answer as the shuffle.
What the extra size does over many journals
One journal is one draw. The script also generated 10,000 journals of 150 trades from the same rules and played each one three ways: the 1% plan, creep, and a flat 1.13% on every trade. That 1.13% is creep’s average risk per trade across all runs, so the flat version takes the same total risk without tying it to streaks.
| Plan 1% | Creep | Flat 1.13% | |
|---|---|---|---|
| Median ending equity | $15,491 | $16,181 | $16,342 |
| 5th percentile ending equity | $11,901 | $11,902 | $12,144 |
| Median maximum drawdown | 7.8% | 8.5% | 8.8% |
| 95th percentile maximum drawdown | 13.7% | 14.5% | 15.3% |
Against the plan, creep ended higher in 77.1% of runs and had the deeper drawdown in 87.0%. That’s what more risk does to a system with a positive edge. It doesn’t show that streaks are a good time to size up.
Against the flat 1.13%, creep ended higher in only 43.4% of runs, and its median ending equity was $161 lower. Its drawdowns were a bit shallower, with the deeper one in 29.0% of runs, because creep tends to load up near equity highs. The extra risk earned less when it was tied to streaks than when it was spread evenly.
The flat version has one more advantage that the table can’t show. Its risk is known in advance, so a risk of ruin or drawdown stress test run at 1.13% describes the account you’re trading. With creep, the test runs at 1% and the account trades at 1% to 2.5%.
Fixes that hold up
- Compute size from the stop before every entry, using a calculator or a saved formula. Typing a lot size from memory is an easy way for creep to start.
- Set a hard cap per trade, such as 1.5 times your plan risk, and write it next to the plan.
- If you decide to raise risk, raise it for every trade from a set date, and recheck it with the same stress tests. Tie the change to a review of at least a month of trades, not to the last three results.
- After 2 wins in a row, check the size of the next order against your plan before sending it. In the example, all of the extra size went on those trades.
- Rerun the creep ratio once a month on the trades since the last check.
Limits of this method
- The example assumes every trade is independent. Your trades may not be. If your edge rises in some market conditions, a rule that sizes up on those conditions can make sense. A streak is a noisy stand-in for them, and the average R check in step 6 tells you whether it worked.
- 35 trades in a group is a small sample. The shuffle check guards against reading too much into it, but a ratio measured on 35 trades can still drift on the next 35. See how many trades you need before evaluating a strategy for more on sample size.
- Rebuilt equity is only as accurate as your logged results. Deposits, withdrawals and fees between trades change the denominator, so log them.
- Trades without an initial stop drop out, which can hide creep on exactly the trades where you skipped the stop.
FAQ
Is it creep if my position size grows with my account?
No. With percent-risk sizing, a bigger account means bigger dollar risk at the same percent. Creep is when the percent itself goes up. Measure risk at the stop divided by equity at entry and the two separate cleanly.
How many trades do I need for this check?
There’s no fixed number. Use the shuffle check to see whether the gap could be chance, and rerun it as trades come in. In the example, 35 trades after a streak and 115 other trades were enough for 0 of 10,000 shuffles to match the measured ratio, but a smaller gap would need more trades.
What if some of my trades had no stop loss?
Leave them out of the creep ratio and count how many you dropped. Without a stop there’s no risk number at entry. If many trades have no stop, fix that first, because it hides creep and every other risk check.
Does a winning streak make my next trade more likely to win?
Your own journal can answer it. Compare the win rate and average R of trades after 2 or more wins with your other trades, and run the shuffle check on that gap too. In the example, trades after a streak won 45.7% of the time against 56.5% for the rest, and that gap was chance by design.
Run it on your own trades
UltraTrader AI auto-syncs trades from MT4, MT5, cTrader and crypto exchanges over a read-only connection, so your journal fills in without retyping. On Premium, the AI review of your MT5 trades shows where your profit slipped away, ranked by cost, and oversized losses are one of the leaks it looks for. Try UltraTrader AI.
For MT5 auto-sync portfolios. Historical analysis, no guarantees.
The trades in this post are hypothetical. Results depend on the inputs and are not a prediction of future performance. This is educational content, not financial advice.