What is risk of ruin — and why does a profitable system still blow up?
Most beginners evaluate a trading system by asking whether it wins. That is the wrong first question, because a system can win and still take you out. The right first question is whether you will still be solvent by the time the winning shows up — and that is a separate calculation with a separate answer. This page explains what risk of ruin is, shows what it actually measures at each risk setting, demonstrates why counting losing streaks gives an answer that is wrong by roughly five orders of magnitude, and then argues that every number on this page is still too optimistic.
KEY TAKEAWAYS
- Risk of ruin is driven by position size far more than by edge quality. Moving one break-even system from 1% to 3% risk per trade took its chance of a 50% drawdown from 0.008% to 40.4%.
- Counting losing streaks gives the wrong answer. At 2% risk you need 35 consecutive losses to lose half — about a 1 in 868,000 event over 200 trades. The real chance of losing half was 9.9%, some 86,000 times higher.
- A positive expectancy does not protect you. A system with a genuine edge (50% wins at 1.2:1) still had a 39.8% chance of a 50% drawdown at 5% risk per trade.
- The risk setting that maximised the median outcome was 8% — at which the chance of ruin was already 87.3%. Optimising for growth and optimising for survival are not the same problem.
- Every figure here assumes results are independent. Make them cluster, holding the win rate at exactly 45%, and ruin at 3% risk goes from 0.74% to 60% — 81 times worse, from sequencing alone.
What is risk of ruin, in one paragraph?
Risk of ruin is the probability that your account falls to a level it cannot practically come back from, at some point before your edge has had time to express itself. Three things decide it: your win rate, your reward-to-risk ratio (how big a win is relative to a loss), and the fraction of the account you put at risk on each trade. The first two describe the quality of your trading. The third describes how long you get to keep doing it. They are separate questions, and traders overwhelmingly spend their attention on the first two.
“Ruin” needs a definition before the number means anything, because zero is the wrong threshold. If you risk a percentage of what you currently have rather than a fixed dollar amount, the account mathematically never reaches zero — it just gets small enough to be irrelevant. Throughout this page, and in our risk of ruin simulator, ruin means a 50% drawdown from the account’s highest point. That threshold is chosen for an arithmetic reason: from minus 50% you need a 100% gain simply to get level, which is why the drawdown recovery maths turns hostile there. Most people who reach it do not trade their way back. They stop.
Every number below comes from simulating each combination 500,000 times over a 200-trade run, compounding: after a win the account is multiplied by 1 + (risk × reward ratio), after a loss by 1 − risk. Three systems are used throughout — a strong edge (45% wins at 2:1), a break-even system (40% at 1.5:1, expectancy exactly zero), and a thin but real edge (50% at 1.2:1). You can reproduce any of it in the simulator.
Why isn’t “how many losses in a row?” the same question?
This is the intuition almost everyone starts with, and it is comprehensively wrong. The reasoning goes: at 2% risk per trade I could lose 35 times in a row before I was down half, that will never happen, therefore I am safe. The first half of that sentence is correct. The conclusion does not follow.
Here is the streak calculation done properly, for the break-even system over 200 trades, next to the chance of actually losing half:
| Risk per trade | Straight losses needed to lose 50% | Chance of that streak in 200 trades | Actual chance of a 50% drawdown | Understated by |
|---|---|---|---|---|
| 1% | 69 | 1 in 38,000,000,000,000 | 0.008% | 3,200,000,000× |
| 2% | 35 | 1 in 868,000 | 9.9% | 86,000× |
| 3% | 23 | 1 in 1,764 | 40.4% | 712× |
| 5% | 14 | 1 in 17 | 84.9% | 15× |
At 2% risk the streak test says one in eight hundred and sixty-eight thousand. Reality says one in ten. The gap is not a rounding error or a modelling quibble; it is four to nine orders of magnitude, and it is in the direction that gets accounts killed.
The reason is simple once seen. Ruin is not built out of a streak. It is built out of scattered losses with small wins in between, each of which feels survivable at the time. The chart at the top of this page is one such run, pulled straight from the simulation: a break-even system at 2% risk that delivered exactly its 40% win rate — 80 wins, 120 losses, no bad luck in the count at all — rose to 1.26× its starting capital, and then bled down to 0.55×. Its longest losing streak anywhere in 200 trades was nine. Not 35. Nine. And it still lost 56.5%.
That is what a drawdown feels like from inside. There is no single catastrophic week to point at, no moment that announces itself. There is a slightly-below-average month, then another, then a recovery that stalls, and by the time it is obvious the arithmetic has already moved somewhere unpleasant.
What does it actually look like at each risk setting?
Here is the full grid. Each cell is 500,000 simulated 200-trade runs; the figure is the percentage of them that hit a 50% drawdown at any point.
| Risk per trade | Strong edge 45% wins, 2:1 | Break-even 40% wins, 1.5:1 | Thin edge 50% wins, 1.2:1 |
|---|---|---|---|
| 0.5% | 0.000% | 0.000% | 0.000% |
| 1% | 0.000% | 0.008% | 0.000% |
| 2% | 0.013% | 9.93% | 0.30% |
| 3% | 0.74% | 40.4% | 5.43% |
| 5% | 18.2% | 84.9% | 39.8% |
| 7.5% | 68.0% | 98.8% | 82.5% |
| 10% | 94.1% | 99.96% | 97.5% |
Three readings fall out of it. The first is that the response to risk per trade is violently non-linear. For the break-even system, tripling the risk from 1% to 3% multiplies the chance of ruin by roughly 5,000. Nobody who moves from “a percent” to “three percent” in their head thinks they are tripling anything important. They are not tripling it; they are detonating it.
The second is that the whole table goes flat at the top. Below 1% risk, all three systems — including the one with no edge at all — register zero ruins in half a million runs. This is the honest reason the 1–2% rule keeps being repeated by people who cannot explain it. It is not folklore. It is the region where the curve has not started climbing yet.
The third is that the columns converge at the bottom. At 10% risk the strong edge dies 94.1% of the time and the no-edge system dies 99.96% of the time. The difference between good trading and no trading at all has shrunk to almost nothing, because at that size the sequence decides everything before the edge can accumulate.
Can a system with a real edge still ruin you?
Yes, and refusing to believe it is how competent traders lose accounts.
Take the thin-edge system: it wins 50% of the time and makes 1.2 times what it risks. Expectancy is (0.50 × 1.2) − (0.50 × 1) = +0.10R per trade. That is a real, positive, tradeable edge — over 200 trades the expected gain is 20R. It is better than most retail traders actually have.
Risk 1% of the account on it and the chance of a 50% drawdown was zero in 500,000 runs. Risk 5% and it was 39.8%. Risk 10% and it was 97.5%. Same edge. Same trader. Same signals, in the same order. The only variable that moved was the size of the bet.
What has gone wrong is a confusion between two different statements. An edge is a statement about the average of a long sequence. Survival is a statement about the worst point along the path. Averages are indifferent to order; paths are not. Expectancy will tell you where you end up if you get to take all 200 trades. It says nothing whatsoever about whether you are still in a position to take trade 140, and the account does not carry on averaging after you have stopped.
So what is the biggest risk setting that still makes sense?
There is a genuine tension here, and it is worth showing honestly rather than just repeating “risk less.” Risking more does make more money, for a while. Below is the thin-edge system at every risk setting from 2% to 12%, with the median ending account value beside the chance of ruin:
| Risk per trade | Median ending account | Chance of a 50% drawdown |
|---|---|---|
| 2% | 1.42× | 0.30% |
| 3% | 1.63× | 5.4% |
| 4% | 1.83× | 19.9% |
| 5% | 2.01× | 39.8% |
| 6% | 2.15× | 60.0% |
| 7% | 2.25× | 76.6% |
| 8% | 2.29× ← peak | 87.3% |
| 9% | 2.28× | 94.0% |
| 10% | 2.22× | 97.5% |
| 12% | 1.95× | 99.6% |
The median outcome peaks at 8% risk per trade, and past that point more risk buys less money — by 12% the median has fallen below where it was at 4%. There is a mathematical ceiling on useful aggression, and beyond it the compounding works against you.
But look at what that peak costs. At 8% risk the chance of losing half the account first is 87.3%. The growth-optimal setting is a setting that ruins seven accounts out of eight on the way to its median. Meanwhile at 3% risk you keep 71% of that peak median (1.63× against 2.29×) while the chance of ruin falls from 87.3% to 5.4% — a sixteen-fold reduction in the chance of catastrophe for a 29% reduction in the median.
That asymmetry is the whole argument for trading small, and it is quantitative rather than temperamental. You are not giving up much upside. You are buying an enormous amount of survival.
Why is the real number worse than any calculator says?
Every figure above — and every figure any risk-of-ruin calculator on the internet will give you, ours included — rests on one assumption that markets do not honour: that each trade’s outcome is independent of the last one. Flip a coin, record it, flip again, the coin has no memory.
Trading results have memory, and the mechanism is not mysterious. A strategy built for trending conditions wins repeatedly while the trend runs and loses repeatedly when it stops. Correlated positions resolve together. And the trader has memory too: losses change position sizing, entry timing and patience in ways that make the next loss more likely.
That is testable. Take the strong-edge system at 3% risk. Hold the win rate at exactly 45% and the payoff at exactly 2:1 — change nothing about the edge — and only make wins and losses sticky, so that a win is more likely to be followed by a win. Here is what happens to ruin:
| How results are sequenced | Realised win rate | Chance of a 50% drawdown |
|---|---|---|
| Independent (what calculators assume) | 45.0% | 0.74% |
| Mildly clustered | 45.0% | 9.3% |
| Clustered | 45.0% | 28.3% |
| Strongly clustered | 45.0% | 60.0% |
Identical edge, identical win rate, identical position size, 81 times the ruin. Nothing changed except the order the results arrived in. Every number on this page is therefore a floor, not an estimate — the best case of a model that assumes away the thing that makes drawdowns deep.
The practical translation is uncomfortable but short: whatever number a calculator gives you, treat it as the optimistic end of a range, and give yourself more room than it says you need.
What risk of ruin is NOT
It is not a forecast. A 5% risk of ruin does not mean you have 95% of an account left at the end. It means that in 5% of possible futures for that system, you hit the threshold at some point. You get exactly one of those futures and no say in which.
It is not the chance of losing money. Losing money happens constantly and is priced into every system here. Risk of ruin measures the chance of reaching a specific, unrecoverable depth. Confusing the two makes small losses feel like emergencies and large ones feel like more of the same.
It is not fixed. It is recomputed against your current equity, not the equity you started with. A trader who is down 30% and keeps risking the same dollar amount per trade has quietly raised their percentage risk and therefore their risk of ruin — exactly when they can least afford it. This is the precise mechanism behind revenge trading: the recovery attempt increases the probability of the thing being recovered from.
It is not a substitute for a stop. The entire model assumes a loss costs exactly what you decided it would cost. That is an assumption about your discipline and your exchange, not about mathematics. Take the stop off one trade and the model is describing a system you are no longer running.
Where this reasoning breaks down
Five ways these numbers can mislead you, stated plainly.
1. The win rate and payoff are assumed constant, and yours are not. The simulation gives the system a fixed 45% or 40% forever. Real edges decay as conditions change, and a system measured over a favourable stretch will report a win rate it cannot sustain. If your true win rate is five points below the one you tested, every figure here understates your risk.
2. Losses are assumed to cost exactly 1R. They do not. Slippage, weekend gaps, thin books during the exact volatility that triggered your stop, and funding on a leveraged position all mean some losses cost 1.4R or 2R. A model where the downside is capped and reality where it is not are different models.
3. The 50% threshold is ours, and it may be far too generous for you. If you trade a funded account with a 10% maximum drawdown rule, your ruin is a 10% drawdown, and the corresponding probabilities are enormously higher than anything in these tables. If you know you would stop trading after losing 25%, that is your ruin threshold, and you should recompute against it. Use the number that ends your trading, not the one that ends your account.
4. Two hundred trades is a window, not a career. Ruin is cumulative: run the same system for 1,000 trades and the probability rises, because you keep drawing from the same distribution. A figure of “5% over 200 trades” is not an annual rate and should not be read as one.
5. Simultaneous positions are the loophole most people fall through. Everything here assumes one trade at a time. Three open longs on correlated altcoins at 2% each is not three 2% trades — on the day the market moves as one, it is a single 6% trade. Traders who obey the 2% rule per position and still blow up are almost always doing this. Add up what is genuinely at stake at the same moment, and use that as your risk input.
And one thing the model cannot price at all: it has no term for an exchange failing, a token depegging, a stop not filling because the venue went down. Those are not in any distribution you can simulate. They are the argument for keeping the modelled risk low enough that an unmodelled one does not finish you.
Where should you go from here?
The direct next step is to put your own numbers in rather than trusting ours: the risk of ruin simulator runs the identical model with your win rate, payoff and risk setting, and draws the individual equity curves so you can see how varied the paths are. Then translate the percentage into an order size with the position size calculator, and read position sizing for the four-step version of that calculation.
For the depth side of the same coin, drawdown recovery shows exactly why 50% is the threshold we chose, and avoiding the big loss covers the trade that turns a drawdown into an ending. The rule this page quantifies is written up in full at the 1–2% rule, and the discipline that makes any of it real lives in stop loss placement and the trading plan. If leverage is what pushed your effective risk per trade up without you noticing, start instead with leverage and margin.
FAQ
What is risk of ruin in simple terms?
Risk of ruin is the probability that a run of losses takes your account below the point where it can recover, before your edge has had time to pay. It has three inputs: how often you win, how much you win relative to what you lose, and how much of the account you put at risk on each trade. The third input matters far more than most people expect.
Is risk of ruin the same as the chance of losing money?
No. Almost every trader loses money at some point; that is ordinary. Risk of ruin measures the chance of reaching a specific depth from which returning is impractical. We use a 50% drawdown, because recovering from minus 50% requires a 100% gain. Losing a little is a cost of doing business. Losing half is a different category of event.
What is an acceptable risk of ruin?
There is no universal number, but the arithmetic points somewhere. Across all three systems simulated here, risking 1% or less of the account per trade produced a chance of a 50% drawdown below 0.01% over 200 trades. At 3% the break-even system reached 40%. The 1-2% convention that keeps being repeated is not superstition; it is roughly where the curve is still flat.
Can I have a positive edge and still go broke?
Yes, and this is the part people find hardest to accept. A system winning 50% of the time at 1.2 to 1 has genuinely positive expectancy. Risk 5% of the account on each trade and it still had a 39.8% chance of a 50% drawdown within 200 trades. An edge tells you where the average ends up. It says nothing about whether you are still solvent when the average arrives.
Why is counting losing streaks not enough?
Because ruin is not usually built from a streak. At 2% risk it takes 35 consecutive losses to lose half the account, and the chance of that in 200 trades is about 1 in 868,000. Yet the actual chance of a 50% drawdown for the same system was 9.9% - roughly 86,000 times higher. Drawdowns are assembled from scattered losses with small wins in between, which is why they arrive without ever feeling like a disaster.