Expectancy vs Win Rate: Why 95% Wins Can Still Bankrupt You

A bot that won 95% of its trades wiped my account.

That's not a typo. Ninety-five percent. Green trade after green trade after green trade, for months, until one Thursday afternoon gold moved the wrong way, the thing martingaled into a single monster position, and a year of "wins" vanished in about an hour. I sat there staring at a number near zero, holding a strategy that, on paper, was right almost every single time.

That afternoon is the entire reason I distrust win rate as a headline number. Win rate tells you how often you're right. It tells you nothing — nothing — about whether you make money. The number that actually decides it is expectancy, measured cleanly in R-multiples. So let me give you expectancy vs win rate the way I wish someone had given it to me before I clicked buy on that bot: the real formula, both ways, with the math worked out.

This is the metric I judge every system by. Including my own.

The 95%-win-rate bot that wiped me: a beautiful number hiding a fatal edge

Here's what made it so seductive. The equity curve only went one direction. Up, up, up, a clean staircase. The win rate display read 94-something, then 95. Every screenshot looked like proof.

What the win rate hid was the shape of the rare loss. The thing won small and lost catastrophic. It would grind out twenty wins of a few dollars each, then on the twenty-first trade it would refuse to take a small loss, double down, double down again, and hold a position the size of the whole account into a move that didn't come back. The 5% it got wrong wasn't 5% of the damage. It was 100% of it.

That's the trap in one sentence: a high win rate measures frequency, not magnitude. You can be right almost always and still go broke, because going broke isn't about how often you lose. It's about how much you lose when you do, relative to how much you make when you win.

I didn't lose because the bot had a bad strategy by its own metric. I lost because I was judging it by the wrong metric entirely. If you only remember one thing here, make it that.

Risk note: no win rate, mine or anyone's, predicts a future result. Past performance is not a promise.

Win rate vs expectancy: what each number actually tells you (and what it hides)

Let me separate these cleanly, because people blur them constantly.

Win rate is the percentage of your trades that close in profit. Win 60 out of 100, that's a 60% win rate. It's a single number, it's easy to brag about, and it answers exactly one question: how often am I right? That's it. It says nothing about the size of the rights or the wrongs.

Expectancy is the average amount you can expect to make (or lose) per trade, once you account for both how often you win and how big those wins and losses are. It's the number that survives contact with reality, because it folds magnitude back in.

Here's the gap that matters. Two traders both have a 50% win rate. Trader A makes $100 on winners and loses $100 on losers — breakeven before costs, a loser after spread. Trader B makes $300 on winners and loses $100 on losers — a money machine. Identical win rate. Completely different outcome. The win rate couldn't tell them apart. Expectancy tells them apart instantly.

That's why I call win rate is a vanity metric — it flatters you while telling you almost nothing you can bank. It's the number a marketer leads with precisely because it sounds like safety. The number that decides whether you eat is expectancy. The same brutal honesty runs through everything on the results page: I lead with the worst day, not the best win rate.

R-multiples in plain English: +2.3R, -1R, and why normalizing beats raw dollars

To talk about expectancy without drowning in account-size noise, you measure everything in R.

R is simply the amount you risk on a trade. One unit. If your stop means you'll lose $100 when you're wrong, then 1R = $100. From there, every outcome gets expressed as a multiple of that risk:

• Lose the trade and hit your stop? That's -1R.
• Risk $100, make $230? That's +2.3R.
• Risk $100, make $50? That's +0.5R.

The reason this beats counting raw dollars is that it normalizes everything to a single yardstick. A $100 risk and a $1,000 risk are different dollar amounts but the same -1R when they lose. Now you can compare a trade you took on a small account in January to one on a bigger account in June, on equal footing. R strips out account size, position size, and ego, and leaves you with the only thing that matters: how big was the result relative to what you risked to get it.

And it forces the discipline that saves you. To know your R, you must define your risk before you enter — which means a hard stop, every time, set at the moment you open the trade, not bargained with later. No stop, no R, no honest accounting. That's not a coincidence. The framework and the risk control are the same habit. It's the same logic behind why every Axiom position carries a hard stop — you can see the structure on how it works.

The expectancy formula, worked both ways: in dollars and in R

Now the actual math. Two forms, same engine.

The full dollar version:

Expectancy = (Win% × Average Win) − (Loss% × Average Loss)

Take a system that wins 40% of the time. When it wins, it makes $300. When it loses, it loses $100.

(0.40 × $300) − (0.60 × $100) = $120 − $60 = +$60 per trade

Forty percent win rate. Positive sixty dollars every time you click. Over 200 trades that's $12,000 of expectation, from a system that's wrong more often than it's right. Read that twice, because it's the whole point.

The cleaner R version:

When your loss is always your 1R risk, the formula collapses to something you can do in your head:

Expectancy in R = (Win% × Average win in R) − (Loss% × 1R)

Same system — winners are +3R ($300 on $100 risk), losers are -1R:

(0.40 × 3) − (0.60 × 1) = 1.2 − 0.6 = +0.6R per trade

That +0.6R is the system's edge, distilled. Every trade, on average, you expect to bank 0.6 times what you risked. Now flip it to my old bot: a 95% win rate, but winners of +0.3R and the rare loser of -20R.

(0.95 × 0.3) − (0.05 × 20) = 0.285 − 1.0 = −0.715R per trade

Negative. A 95% win rate with a negative expectancy. The math saw the bankruptcy coming that the win rate completely hid. Expectancy in R is the number I report and the number I trust — more on why over on tools.

Break-even win rates by reward-to-risk: 50% at 1:1, 25% at 1:3, 16.7% at 1:5

Here's the part that reframes everything. For any reward-to-risk ratio, there's a break-even win rate — the win percentage where expectancy is exactly zero. Win more than that, you make money. Less, you bleed. The formula:

Break-even win rate = 1 ÷ (1 + reward-to-risk ratio)

Run it across the common ratios:

• 1:1 (make 1R, risk 1R) → break-even at 50%. You have to be right half the time just to tread water.
• 1:2 → break-even at 33%. Right one in three keeps you flat.
• 1:3 → break-even at 25%. Win a quarter of your trades and you're even.
• 1:5 → break-even at 16.7%. Right one trade in six and you're not losing.

Sit with that 1:5 line. A system that's wrong 83% of the time breaks even — and anything above one-in-six wins is pure profit. Meanwhile, a 1:1 system that wins a "respectable" 55% is barely scraping by, and once spread and slippage take their cut, it might be underwater.

This is the death of win rate as a bragging right. A 25% win rate can mint money at 1:3. A 70% win rate can quietly bleed you at 1:0.3. The win rate alone is meaningless until you pair it with the reward-to-risk — and once you've paired them, you've just calculated expectancy anyway. So why lead with the vanity number at all? You can see the kind of honest accounting this produces, drawdown and worst-day included, on the results page.

Risk note: these are clean textbook numbers. Real fills suffer slippage, gaps, and widening spreads — especially on gold around news — so build a margin of safety above the break-even line, never sit right on it.

Why I report expectancy in R and treat win rate as a vanity number

So here's where I've landed, scars and all.

I don't care how often a system is right. I care what it makes per unit of risk, on average, over a meaningful sample. That's expectancy in R, and it's the single number that would have screamed at me to walk away from that 95%-win-rate bot before it took everything. The win rate whispered safe. The expectancy, had I bothered to compute it, said countdown.

Win rate is a vanity metric because it's optimized for how it feels, not for what it does. It's the number that sells courses and signal groups and martingale bots with gorgeous equity curves. Expectancy in R is unsexy, occasionally humbling, and completely honest — which is exactly why I trust it. A 40% win rate that nets +0.6R per trade will quietly get rich. A 90% win rate at -0.2R will quietly die. The feeling and the outcome point in opposite directions, and the feeling is the one trying to bankrupt you.

None of this is the edge itself, to be clear. Expectancy is the scoreboard, not the strategy. My actual edge is in reading price, the numbers underneath it, and time — that part stays private, the way any real fund guards its alpha. But the scoreboard is universal, and I'll teach it all day because teaching it costs me nothing and protects you from the exact pitch that wiped me.

That's the lens I built Axiom around: a hard stop on every trade so R is always defined, risk capped per trade, a max-drawdown ceiling, expectancy reported in R instead of a flattering win rate, and a 30-day profit-or-refund — not in net profit after 30 days, your $999 comes back in USDT. If that's the kind of honesty you've been looking for, the checkout page is right there. But whether you ever touch my software or not: stop asking how often a system wins. Start asking what it makes per unit of risk. That one swap changes everything.

Risk note: trading gold carries real risk of loss. A positive expectancy improves your odds over many trades; it guarantees no single trade and no individual result. Never risk money you can't afford to lose.

This is one trader’s opinion and education, not financial advice. Trading gold carries real risk of loss; any figures are illustrative and not a promise of results.