How to read an equity curve without fooling yourself

You pull up your equity curve, see a line heading up and to the right, and decide the strategy works. Or you see a six-week sag and decide it's broken.

Both readings are usually wrong. The shape of a curve is driven as much by the order your trades happened to arrive in as by whether you have an edge. Read it without knowing which is which and you will cut a working system after a bad run, or add size to a lucky one.

Here is what the line actually contains, what you can legitimately conclude from it, and where it goes quiet.

Three decisions hide inside every equity curve#

Before you interpret a curve, you need to know how it was built. Two charts of the same account can look nothing alike.

Realized or mark-to-market. A realized curve plots cumulative closed P&L and steps only when you exit. A mark-to-market curve revalues open positions, so it moves on days you did not trade. The realized version hides how much heat you sat through; the daily version shows the heat but mixes in unrealized P&L that may never be booked.

Trade axis or calendar axis. On a trade axis a quiet month takes up no space at all. On a date axis, twenty trades crammed into one week look like a cliff. Same data, different story.

Deposits in or out. A curve that includes transfers is not a performance record. If you topped up after a bad month, the chart will show a recovery you did not trade.

Pick one definition and keep it. Comparing a realized trade-axis curve from March against a mark-to-market calendar curve from June is comparing nothing.

Pull three numbers before you look at the shape#

Shapes invite stories. Numbers do not. Extract these first.

Maximum drawdown. Track the running peak — the highest equity reached up to that point — and subtract current equity from it. The largest value that difference ever reached is your maximum drawdown. Record it in both currency and percent, and state the denominator you used: the running peak, not your starting balance, or the figure will flatter you as the account grows.

Current drawdown from peak. The same subtraction, at today's equity. This is the number that tells you where you are standing right now, which the shape of the line does not.

Recovery factor. Net profit divided by maximum drawdown. It answers a question the curve cannot: how much profit did each unit of pain buy? A system that made 800 with a 400 drawdown (factor 2.0) and one that made 800 with a 1,200 drawdown (factor 0.67) are not the same system, even when they end on the same number.

That last pair is not hypothetical.

The same trades, two completely different curves#

Take 20 trades. Win rate 40%, winners at +2.5R, losers at -1R, 1% of a 10,000 starting balance risked per trade — so a winner adds 250 and a loser costs 100.

The arithmetic is fixed before we order anything:

• 8 winners x 250 = +2,000
• 12 losers x 100 = -1,200
• Net = +800, final equity 10,800
• Expectancy = (8 x 2.5R - 12 x 1R) / 20 = 8R / 20 = +0.4R per trade

Now deal those same 20 results in two different orders.

Order A alternates for the first sixteen trades, then runs four losers at the end. Equity climbs to a peak of 11,200 at trade 16, then slides to 10,800.

• Maximum drawdown: 11,200 - 10,800 = 400, or 3.6% of peak
• Recovery factor: 800 / 400 = 2.0

Order B front-loads six winners, then delivers all twelve losers back to back, then the last two winners. Equity peaks at 11,500 at trade 6, bleeds to 10,300 by trade 18, and closes at 10,800.

• Maximum drawdown: 11,500 - 10,300 = 1,200, or 10.4% of peak
• Recovery factor: 800 / 1,200 = 0.67

Identical trades. Identical edge. One curve looks like a staircase, the other like a system that broke in the middle and got lucky at the end. A trader holding Order B would very likely have stopped trading it around trade 15 — and would have been wrong.

This is the single most useful thing to understand about equity curves: shape is a sample of the ordering, and ordering is luck. The expectancy is the thing you own. The curve is the thing that happened to you.

My own last 100 trades make the point more bluntly than the example. In R they netted -4.0R with a 26.7R maximum drawdown. That is a negative recovery factor, with nothing to recover. They also span two accounts: a tiny crypto account where I sized anywhere from 1% to 10% per trade, and my main account at a flat 1%. On the main account alone it's 33 trades, +11.7R, +$759, and a $1,070 maximum drawdown that I'm sitting in today. That is a recovery factor of 0.71. Most of that drawdown is plain losing: 14 losers in 15 trades. But one BTC trade I'd accidentally oversized lifted my balance from $10,000 to about $14,400 mid-sample, so the same 1% loss grew from $100 to about $140. The winners that built the peak paid around $99 per R; the losers that gave it back cost around $116. About a fifth of the drawdown is sizing, not trading, and the dollar curve on its own would never have told me that.

Was that flat spot decay, or was it variance?#

This is the question the curve gets asked most and answers worst. The honest approach is to work out what a normal run of bad luck looks like for your win rate, then check whether what you saw exceeds it.

Stay with a 40% win rate, and assume each trade is independent of the last. The probability that any given eight consecutive trades are all losers is:

0.6^8 = 0.0168, or about 1.7%

That sounds rare enough to be alarming. But a 100-trade sample contains 93 overlapping windows of eight consecutive trades, so the expected number of all-loss eight-trade windows in that sample is:

93 x 0.0168 = 1.6

Those windows overlap, so a single ten-trade losing streak counts as three of them. Counted as distinct streaks, the chance of at least one run of eight or more losers in 100 trades at a 40% win rate is about 49%: a coin flip. A flat spot of that length is not evidence of anything. It is the cost of doing business at that win rate.

For the full distribution of outcomes, run your own win rate, reward-to-risk and trade count through the Monte Carlo simulator and look at where your actual drawdown sits inside the simulated spread. If your worst stretch lands inside the middle of that distribution, your curve is behaving. If it sits far out in the tail, you have something to investigate.

The companion question — how many trades you need before the curve means anything at all — is worth answering before you draw conclusions from a short sample: how many trades before you know a strategy works. And for the specific case of consecutive losers, how long a losing streak is normal does the same arithmetic in more depth.

My longest stretch under water ran 563 trades and 103 calendar days, from 25 February to 8 June 2025. At the bottom, on 28 May, I was 42.8R below my peak. I didn't cut size and I didn't stop. I did the opposite. I went from about 30 trades in the last days of February to 223 in May, with 16 on the worst day. I rotated through as many as nine strategies a month. Between March and May I took 95 trades at 5% to 20% risk, and they won 36% of the time, against 49% for my 1% trades. The recovery came at 1% on every trade: 82 trades in 11 days for +45R, and a single 12R gold trade did a quarter of that. This year I'm under water again, 91 trades and six months below an April peak. This time I stopped for six weeks and came back at a flat 1%. The small account had lost about half before I stopped; the main account is 7% off its peak. It hasn't recovered yet.

A smooth upward bend can be a sizing artefact#

Fixed fractional risk makes a curve bend upward even when nothing improved. Risk 1% of a 10,000 balance and each 1R loss costs 100. Grow the balance to 15,000 at the same 1% and that same 1R now costs 150, while winners pay 375 instead of 250. Identical R-multiples, steeper line.

The reverse trap matters more. A curve that flattens as the account grows often means risk per trade quietly fell — you got more careful, or you stopped scaling your position size with the balance. That is a sizing change masquerading as a performance decline, and the curve will not tell you which it was.

The defence is to plot in R rather than in currency, or at minimum to log risk percent on every trade so you can check whether the denominator moved. If you are not sure what your risk per trade has been, the position sizing calculator will tell you what a given balance, risk percent and stop distance should have produced — and the gap between that and what you actually traded is usually the more interesting number.

What the equity curve does not tell you#

The curve is one projection of your trading, and a lossy one. It is silent on:

• How close your winners came to stopping you out. Two trades that both closed at +2R are identical on the curve; one of them may have been 0.1R from your stop first.
• Open risk. A realized curve rising steadily tells you nothing about whether you were carrying three correlated positions at the time.
• Correlation. Four FX pairs long the dollar is one trade wearing four costumes. The curve shows four data points and understates the real risk taken.
• Whether the strategy stayed the same. If you changed entry rules at trade 60, the curve is two samples glued together, and the maximum drawdown across the join belongs to neither.
• Costs you did not record. Spread, commission, swap and slippage either come off each trade or they do not appear at all. A curve built on mid-prices is fiction.
• Regime. A curve made in one volatility environment carries no information about the next one.

None of this makes the curve useless. It makes it a starting point that has to be checked against the trade log underneath it, which is why the curve is worth less than the records that produce it.

Before I react to the line, I check one column: risk percent. If it's been a flat 1% across the window, the curve in R has no sizing story in it, and what's left is trade order or a lost edge. The simulator answers that question, not the line. If the risk moved, the curve is telling me about my sizing, not my strategy. My 2025 drawdown had 95 trades at 5% to 20% risk in it. The drawdown on my main account right now has none.

A ten-minute review that beats staring at the line#

1. State the chart definition out loud: realized or mark-to-market, trade axis or calendar, deposits stripped out.
2. Write down maximum drawdown in currency and in percent of running peak, plus current drawdown from peak.
3. Divide net profit by maximum drawdown. That is your recovery factor. Compare it to the same figure from your previous review, not to anyone else's.
4. Count the trades in your longest stretch below peak, then check it against what your win rate makes normal.
5. Open the five worst trades in the drawdown. Were they the strategy working as designed, or were they rule breaks? The curve cannot distinguish these; your notes can.
6. Check whether risk per trade was constant across the window. If it moved, the shape is partly a sizing story.

Do that and the curve stops being a verdict and becomes what it should be: an index into the trades worth re-reading.

The line going up does not prove you have an edge, and the line going sideways does not prove you lost one. Expectancy, sample size and a constant risk fraction tell you whether the system works. The curve only tells you what order it happened in.