Prop Firm Position Sizing — What 20,000 Back-Tested Trades Taught Me About Risk Per Trade

On a live account you fund yourself, 1% is genuinely good advice. There's one way to die — run out of money — and small size buys you a long time to be wrong. Fine.

A prop firm account has three ways to die. You can break the firm's rules. You can breach the maximum drawdown. And you can breach the daily loss limit, which can end you on a single ordinary afternoon with the account still well in profit for the month.

That last one is the whole game, and generic risk advice doesn't know it exists. The rule was written for a different account with a different failure condition, and it got copied across anyway.

I believed a second version of the same mistake for a long time: take fewer trades, stop after one win, protect the day. Sensible-sounding. Also just someone's opinion. My data said the opposite, and the data doesn't have a course to sell.

Once you have a real sample, sizing stops being a philosophy and becomes arithmetic. Three inputs:

• Your win rate. It sets how long your realistic losing streaks are. A high win rate makes long streaks rare, which buys you room to size up. A modest one makes them routine, and routine streaks have to be survivable.
• Your trade frequency. Specifically per day, not per month. Five trades a day means five chances to stack losses inside one daily limit. One trade a day means the limit is almost irrelevant to you.
• The firm's rules. The daily loss limit and the maximum drawdown are hard ceilings someone else set. They don't care about your edge.

Nobody handing out sizing advice on the internet knows any of your three. That's why the advice is generic — it has to be. Yours doesn't.

Here's the arithmetic, with a worked example. Take a trader who:

• takes 5 trades a day,
• wins 60% of them,
• trades a firm with a 5% daily loss limit.

At the "correct" 1% risk, five losses in a row is exactly 5%. One bad day — not a catastrophe, not a tilt spiral, just the worst version of a completely normal day — and the account is done.

And five losses in a row at a 60% win rate isn't exotic. Run it: over 75 trades in a month, the chance of hitting a five-loss streak at some point is roughly 1 in 3. Over a quarter it's closer to 3 in 4. That's not a tail risk. That's a Tuesday you should be planning for.

Your risk per trade × your worst realistic daily streak has to fit inside the daily loss limit. Everything else is decoration.

This is the part that sounds reckless until you write it down, so let's write it down.

A prop firm account isn't your capital. The fee is. You've already paid it, it's already gone, and the only question left is what to do with the account it bought you. Bigger size needs fewer winning trades to reach the target, which means you get to a payout faster and more often — while also breaching more often.

So the real question isn't "will this size cost me accounts?" It will. The question is whether the payouts come out ahead of the fees across many attempts. If a larger size takes you from getting payouts rarely to getting payouts regularly — enough that the payouts more than cover the fees of the accounts you lost along the way — the aggressive number is the correct number, even though it looks worse trade by trade.

You cannot eyeball that. It's a probability question with a price tag on it, which is exactly what the ROI & risk-of-ruin calculator exists to answer, and what the pass-probability simulator feeds it: your edge, run thousands of times against a firm's real drawdown and daily-loss rules.

One warning, because this is where it goes wrong. This logic only holds if your edge is measured. Sizing up a strategy you haven't proven doesn't buy you a faster payout. It buys you a faster answer to a question you should have asked before paying.

Sizing isn't only about how much you put on. It's about what shape the outcome takes, and my back-test made me change my mind about that too.

I tested my exits properly. Targeting a logical pool of liquidity — the approach I'd learned and trusted — produced the highest expected value of anything I ran. It won. Clearly.

I don't use it for prop firm trading.

I take a fixed 1:1.5 instead, and its EV per trade is slightly lower. Here's why: the liquidity-target exit produced a lot of breakevens. Trades that consumed a day, consumed attention, and moved the balance nowhere.

And breakevens don't just cost you a winner — they cost you time. Every trade that closes flat is another day the account sits exactly as far from the profit target as it started. The 1:1.5 exit resolves. More trades close as real winners, each one smaller, but they stack in a straight line — so I reach the target in far less calendar time than the higher-EV exit takes to get me there.

With a clock and a drawdown running, how fast an edge resolves matters as much as how large it is. That's a trade-off you can only see if you've measured both versions. If you want to compare two exits on your own numbers, the expected-value calculator takes about a minute per version.

And one thing 20,000 scripted trades could not tell me: in a back-test, time stands still. You get to check every condition before you commit. Live, you don't. When I forward-tested, I couldn't reproduce my own back-tested win rate, because the version of the strategy I'd measured required more time than a live market gives you. So I stripped it down — simpler rules, slightly lower win rate, executable in real time. A worse strategy on paper that I can actually run beats a better one I can't. Back-testing tells you what the edge is. Forward-testing tells you which parts of it survive contact with a live chart.

Four of them. Not a system to copy — an example of what falls out when you let data decide.