Maximum Adverse Excursion: What the Drawdown in Your Winners Actually Tells You

A trader reviewing six months of EUR/USD records discovers that 68% of winning trades went into drawdown before reaching profit, some by as much as 18 pips. That number—the largest unrealized loss a position sustained before exit—is Maximum Adverse Excursion. It measures how much heat winners had to endure, and it appears nowhere on a broker statement. John Sweeney introduced MAE in the 1990s as a tool for stop-loss optimization, plotting adverse movement against final outcome to find where recoverable drawdown ends and terminal failure begins. This article covers calculating MAE from execution records, interpreting the distribution, and the boundary conditions where it misleads: sample size floors, regime shifts, and the difference between per-trade risk and portfolio exposure.

Calculating MAE from tick data and broker statements

Most broker statements show only entry, exit, and profit. They omit the one number that matters for stop placement: how far against you the position moved before it closed. That figure requires either tick data or bar data granular enough to capture intraday extremes. Daily bars will miss a 30-pip drawdown that reversed in the London session.

What the calculation requires

MAE is the largest adverse move from entry to any point during the trade’s life, measured before exit. For a long position, it is the distance from entry to the lowest price reached while the trade was open. For a short, entry to the highest price. The measurement can be in pips, ticks, or percent depending on the instrument. Forex records typically use pips; crypto often uses percent.

The data must include every price extreme touched during the position’s life. One-minute bars capture most intraday moves in liquid pairs. Five-minute bars start to miss spike lows in volatile sessions. Anything coarser than fifteen minutes is guesswork. Broker statements rarely include this. Platform exports, API feeds, or third-party logging tools are the usual sources.

A worked example with three trades

The table below shows three constructed EUR/USD trades with tick-level data. Each row includes entry, the worst price reached, exit, final profit, and MAE in pips.

Trade	Entry	Intraday Low	Exit	Profit (pips)	MAE (pips)
1	1.1000	1.0982	1.1035	+35	18
2	1.0950	1.0938	1.0970	+20	12
3	1.1050	1.1062	1.1080	+30	0

Trade 1 went 18 pips offside before recovering to close 35 pips in profit. Trade 2 shows a smaller excursion of 12 pips. Trade 3 never moved against entry; its MAE is zero. All three were winners, but the heat they endured differed sharply.

Without the intraday low, you know only that Trade 1 made 35 pips. You cannot distinguish a trade that moved cleanly to profit from one that tested a 20-pip stop twice. That distinction determines whether your stop is too tight or your entry timing needs work. Statements that omit intraday extremes make MAE analysis impossible, which is why traders exporting from MetaTrader or using broker APIs often find the effort pays back in the first month of data.

The scatter plot: separating winners from losers by how far they went wrong

Plot each trade as a single point: the horizontal axis is MAE in pips or percent, the vertical axis is final profit or loss in the same unit. A winning trade that dropped 35 pips before closing +120 appears at (35, 120). A loser that fell 80 pips and closed there sits at (80, -80). Build this for every trade in the sample.

The pattern that emerges is rarely random. Losing trades cluster at higher MAE values, often forming a dense cloud along or below the X-axis where MAE and final loss are nearly equal—the trade moved against the position and never recovered. Winning trades spread across a wider horizontal range. Some winners endured significant adverse movement; others faced almost none. The vertical scatter among winners reflects the strategy’s profit distribution, but the horizontal spread is what matters here.

Look for the MAE threshold where the two populations begin to separate. Below a certain MAE value—say, 40 pips in a constructed example—most trades that survived went on to profit. Above that line, the majority closed as losses. This boundary is not a recommendation to place stops at 40 pips; it is evidence of where, in this specific sample under these conditions, adverse movement distinguished recoverable drawdown from terminal failure.

When no clean separation appears, the plot reveals one of two problems. Either the entry has no edge—trades are as likely to reverse from any drawdown level as any other—or stop placement is arbitrary relative to the price behaviour the strategy actually encounters. A circular cloud with winners and losers mixed at every MAE level means the record contains no information about optimal stop distance. The common belief that “tighter stops are always better” breaks here: if winners routinely need 60 pips of room and your stops sit at 25, you are cutting profitable trades at random.

Why seventy percent of your winners went into drawdown first

The belief that good entries don’t get tested

A discretionary trader once submitted thirty-nine trades from a three-month EUR/USD campaign, twenty-six profitable, all taken at what he described as inflection points visible on the four-hour chart. His expectation was that winning trades would move into profit within three to five bars. The records told a different story: twenty-two of the twenty-six winners—eighty-five percent—went negative at some point before closing in profit, and the average maximum adverse excursion across those winners was 14.2 pips. His average take-profit was 41 pips, so the heat wasn’t trivial.

This pattern is not unusual. In constructed examples drawn from typical retail execution logs, roughly seventy percent of winning forex trades show some adverse excursion, and in crypto the figure is similar but the magnitude larger. The common belief that a well-timed entry moves into profit immediately is contradicted by the microstructure of execution itself, and by the noise inherent in price formation at retail timeframes.

What spread and slippage contribute

The bid-ask spread alone guarantees that most market orders begin life underwater. A two-pip spread on EUR/USD means a long market order fills at the offer and immediately shows a two-pip loss against the bid, the price at which it could be closed. That loss appears in every MAE calculation until price moves far enough to overcome it. In Bitcoin spot, where spreads routinely reach ten to fifteen basis points on retail exchanges during normal hours and widen further in thin conditions, a market entry starts eight to twelve ticks adverse before any directional move occurs.

Slippage adds to this. In the constructed example below, a trader enters long EUR/USD at 1.1000 with a limit order, intending to buy at that price. The fill comes at 1.1001 due to a fast market, the trader’s software records entry at 1.1001, and price immediately ticks down to 0.9996 before rallying to close at 1.1040 for a forty-pip gain. Maximum adverse excursion: 5.0 pips, or twelve percent of the eventual profit.

Instrument	Entry Price	First Tick After Fill	Lowest Price Reached	Exit Price	MAE (pips)	Final P/L (pips)	MAE as % of P/L
EUR/USD	1.10010	1.09996	1.09960	1.10400	5.0	39.0	12.8%
BTC/USD	42,150	42,085	41,803	43,890	347	1,740	19.9%
GBP/JPY	183.42	183.38	183.11	184.88	31	146	21.2%

The figures in the table are constructed for illustration. They reflect order-of-magnitude relationships seen in retail execution logs but are not taken from any single trader’s record.

Beyond spread and slippage, intra-bar volatility and mean-reversion within the broader trend direction account for most of the rest. A trend may be intact on the hourly chart, but one-minute bars inside that hour will oscillate, and if your entry lands in the middle of one of those oscillations rather than at its trough, the trade will move against you first. In Bitcoin, where volatility averages roughly four times that of major forex pairs when measured in percentage terms, constructed samples show MAE for winners averaging 8.3 percent of entry price, compared to 0.08 to 0.15 percent—roughly twelve to twenty-five pips on a 1.3000 entry—for EUR/USD or GBP/USD.

The boundary here is execution method. The seventy-percent figure and the magnitudes in the table assume market orders or limit orders filled in moving markets. A limit order that waits in a ranging market and fills only when price reaches it, then reverses, will show lower MAE on average because the entry itself represents a local extreme. Those trades are rarer and depend on market structure: range-bound conditions and sufficient liquidity at the limit price. For most discretionary traders taking directional trades in trending or breakout conditions, the assumption that seventy percent of winners will show adverse excursion holds across enough of the sample to matter for stop placement.

Using MAE to set stops that keep more winners alive

The MAE distribution from a set of trades typically shows two clusters: winners with modest adverse excursion and losers with deep drawdowns. The optimal stop distance sits just beyond the MAE envelope of most winners but inside the region where losers concentrate. This threshold preserves the majority of profitable outcomes while cutting losses before they compound.

Finding the threshold

Calculate the efficiency ratio at each potential stop level: divide the count of winners preserved by the count of losers stopped. The following constructed example uses a sample of 150 trades with MAE measured in pips:

At a 30-pip stop, 112 of 120 winners survived (93.3%) and 22 of 30 losers were stopped (73.3%). Efficiency: 112/22 = 5.09. At 40 pips, 118 winners survived (98.3%) but only 18 losers stopped (60.0%). Efficiency: 118/18 = 6.56. At 50 pips, all 120 winners survived (100%) but only 12 losers stopped (40.0%). Efficiency: 120/12 = 10.0.

The 40-pip threshold offers the best trade-off in this constructed data: it preserves nearly all winners while still eliminating 60% of losers before they reach full size. Moving to 50 pips keeps every winner but allows twice as many losers to run, increasing aggregate loss. Tightening to 30 pips cuts seven winners that would have recovered, a cost of roughly 5.8% of total winning trades.

The tradeoff between risk and premature exits

Tighter stops reduce capital at risk per trade but raise the premature-exit rate. A stop placed at the 85th percentile of winner MAE will cut approximately 15% of profitable trades. If those trades represent above-average winners—common in momentum strategies—the aggregate profit impact exceeds the percentage count. The efficiency metric isolates stop performance but ignores the size distribution of what you preserve versus what you sacrifice. Review both the count efficiency and the dollar-weighted outcome. When winners cut by a tighter stop contributed disproportionately to total profit, the optimal threshold moves outward even if the count-based efficiency peaks closer in.

MAE and position sizing: the capital actually at risk

A trader sets a 100-pip stop on EUR/USD and sizes the position so that stop represents 2% of capital. The trade wins, but the execution record shows it went 147 pips against him before reversing. His position sizing assumed 100 pips of risk; the capital actually exposed was nearly 3%. That gap is what MAE quantifies, and it matters most when the gap appears consistently.

The common belief is that stop distance defines risk. It defines the loss you accept if the stop is hit. It does not define the largest unrealized loss you will carry while the position is open. MAE measures the latter: the worst drawdown from entry to any point before exit, whether the trade ultimately wins or loses. For winners, MAE reveals how much heat you absorbed. For losers, it shows whether the stop was the binding constraint or whether the trade was already deeper underwater when you cut it.

Position sizing based on stop distance alone treats intra-trade excursion as irrelevant. That works if your platform closes the position the instant price touches the stop, with no slippage and no decision lag. It breaks in two common situations: when you use mental stops or discretionary exits, and when volatility or spread widening pushes price well past your intended level before the order fills. Crypto markets, particularly outside major pairs during thin hours, regularly produce fills 1% to 3% beyond the stop price on leveraged positions. If you sized for a 2% stop and the fill comes at 3.5%, the excess comes from somewhere—usually the next trade’s risk budget or your willingness to hold through larger drawdown than planned.

The table below shows constructed figures for a set of 50 winning trades, each entered with a planned 80-pip stop. All positions were sized to risk 1.5% of a $10,000 account at the stop level.

Percentile	MAE (pips)	Actual capital at risk	Multiple of planned risk
50th	68	1.28%	0.85×
75th	102	1.91%	1.28×
90th	134	2.51%	1.68×
95th	159	2.98%	1.99×

Half the winners stayed within the planned risk. A quarter exceeded it by 28%. The top 5% needed nearly double the capital buffer. If you run five positions concurrently, each sized to 1.5% at the stop, and two of them hit 95th-percentile MAE simultaneously, you are carrying 6% realized drawdown when your model assumed 3%. That difference shows up as margin calls in leveraged accounts or as forced exits when equity drops below maintenance thresholds.

Sizing to MAE percentiles instead of stop distance means choosing a percentile—typically 90th or 95th—from your own trade history, then treating that excursion as the risk figure in your position size formula. If your 90th-percentile MAE is 1.7 times your average stop distance, you reduce position size by that factor. The trade-off is straightforward: smaller positions, less frequent margin stress, lower peak equity when everything works. Whether that trade-off improves risk-adjusted returns depends on how often your actual drawdowns exceeded your planned risk in the past, which the MAE distribution answers directly.

This approach stops being useful below roughly 100 trades in similar conditions. Percentile estimates move wildly in small samples, and a single outlier—perhaps from a news event or a gap—can dominate the 95th percentile. It also assumes future volatility resembles the sample period. A trader who builds MAE distributions from 2022 crypto data and then trades 2023 without updating those distributions is using stale estimates of intra-trade heat. The technique does not eliminate estimation error; it shifts the error from guessing what might happen to measuring what already did, which is only useful if the regime holds.

When MAE analysis misleads: sample size, regime change, and correlated exits

A trader with forty-three closed trades and an MAE scatter plot will find patterns in the data. Most of those patterns are noise.

Sample size floors

Below roughly one hundred trades, MAE distributions are too unstable to extract reliable stop thresholds. The constructed example below demonstrates why. Suppose a strategy has been run for seventy trades with the following MAE profile for winners:

Trade count	MAE (pips)	Final P&L (pips)
10	5–15	+20 to +50
8	16–25	+30 to +60
6	26–40	+15 to +45
4	41–60	+10 to +30

The sample suggests a stop at 25 pips would have preserved eighteen winners and cut four marginal ones. Run another seventy trades and the MAE for the next batch clusters differently: eight winners now show MAE between 30 and 45 pips. The threshold that looked optimal in the first sample would have killed twelve percent of the winners in the second. This is not regime change; it is sampling variation. MAE stabilizes somewhere between one hundred fifty and two hundred fifty trades, depending on strategy hold time and how tightly the system clusters around a single setup. Discretionary traders rarely accumulate that count in a single instrument and timeframe within a year.

Regime shifts and non-stationarity

MAE optimized during one volatility regime breaks when the regime shifts. A stop threshold derived from Bitcoin trades in Q2 2023—when daily ATR averaged 3.1 percent—will cut winners prematurely when ATR doubles in a sell-off or gap event. The 2022 crypto deleveraging produced MAE distributions unrelated to those of 2023, not because the strategy changed but because the instrument did. MAE analysis assumes the distribution of adverse movement remains stationary. When it does not, the optimized stop becomes either too tight or irrelevant. Practitioners who re-run MAE quarterly and compare distributions across periods can detect this; those who optimize once and deploy indefinitely cannot.

When all trades go adverse at once

MAE assumes independent trade risk: each position experiences its own adverse excursion unrelated to others. Flash crashes, liquidity gaps, and central-bank surprises hit all open positions simultaneously. A portfolio of five EUR crosses will show correlated drawdowns when the European Central Bank surprises; MAE derived from single-trade analysis will understate portfolio heat. The technique measures per-trade behavior, not portfolio or systemic exposure. It also says nothing about whether the strategy itself has edge. A system with beautifully tight MAE and well-separated winner and loser clusters can still have a negative expectancy if execution costs or position sizing are mismanaged.

What you would do differently

Export your last one hundred fifty to two hundred trades with intraday price extremes. Calculate MAE for each. Plot MAE against final profit or loss. If winners and losers separate cleanly at some horizontal threshold, test stop distances around that boundary using the efficiency ratio: winners preserved divided by losers stopped. Compare the count-based result to the dollar-weighted outcome to ensure you are not cutting disproportionately large winners.

If your current stop sits tighter than the 75th percentile of winner MAE, you are likely cutting profitable trades. If it sits wider than the 90th percentile, you are carrying heat that adds no value. The optimal range falls between those two markers in most discretionary and momentum strategies. Range-bound or mean-reversion approaches may show tighter distributions; breakout strategies wider.

Re-run the analysis quarterly or after any month where volatility doubles. MAE thresholds derived in one regime mislead in another. Track the 90th-percentile MAE over time. When it moves by more than 30%, your position sizing assumptions are stale. Size to the updated percentile or accept that your risk exposure no longer matches your model.

MAE tells you how much heat your winners actually needed. It does not tell you whether your entries have edge, whether your exits are well-timed, or whether the strategy will continue to work. It is a diagnostic, not a guarantee. Use it to align stop placement and position sizing with the behaviour your trades already exhibit, then continue logging and reviewing. The distribution will shift, and your stops should shift with it.