Tradiens
EssaysSign in

Tradiens Research · Data essay

What a Win Rate Leaves Out

Most of the euro–dollar trades won, and the average one still lost. The records of a currency broker's clients and the collapse of an options fund show why a win rate, on its own, says little about whether trading comes out ahead.

Version 2.0 · first published Sep 24, 2026 · rewritten Oct 4, 2026

In 2009 and 2010, 59% of the euro–dollar trades closed by clients of the currency broker FXCM made a profit.

In December 2011, DailyFX, the research unit of the online broker FXCM, published a lesson for traders that was built on its own clients' records.[1] The authors, David Rodriguez and Timothy Shea, had gone through more than 12 million trades that FXCM's clients closed on live accounts in 2009 and 2010, in the 15 currency pairs they traded most. A currency pair is the price of one currency in terms of another, so a trade in the euro against the US dollar, written EUR/USD, is a bet on which way that exchange rate will move. For each pair the article gave the win rate, which is the share of trades that closed with a gain. It also gave the average size of a winning trade and of a losing trade, here called the average win and the average loss.

The euro–dollar figures looked like good trading, because 59% of those trades won and 41% lost. By the article's own numbers, however, the average euro–dollar trade lost about 14 pips. A pip is the standard unit for a currency move: a change of 0.0001 in the exchange rate for most pairs, and of 0.01 for pairs quoted in yen. What a pip is worth in money depends on the pair and on the size of the trade, so a result in pips is not a result in dollars.

The gap between those two facts is the question this essay follows: if most of the trades won, where did the pips they gained go? They went into the losing trades, and the number that shows it is one the win rate leaves out: the size of the average loss compared with the average win.

To make that comparison, each pair's win rate has to be set next to the size of its winning and its losing trades, which is what the first figure does.

In 13 of 15 currency pairs, too few trades won to break even in pips.

  • share of trades that won
  • share that had to win to break even in pips
  • the gap between them

×1.95 = the average losing trade was 1.95 times the size of the average winning trade, in pips. Pairs are sorted by that multiple: the larger it is, the more often trades had to win.

The scale runs from 45% to 75%, not from 0%, so a dot's distance from the left edge does not show the size of the win rate. What is drawn to scale is the grey bar: its length is the gap between dot and mark, in percentage points.

In 13 of 15 currency pairs, too few trades won to break even in pips.
pairloss ÷ windot (won)mark (needed)% points vs break-even
AUD/JPY×1.0549%51.2%2.2 below
USD/CAD×1.3061%56.5%4.5 above
AUD/NZD×1.4871%59.6%11.4 above
EUR/GBP×1.7058%63.0%5.0 below
EUR/AUD×1.8561%65.0%4.0 below
USD/JPY×1.8757%65.1%8.1 below
GBP/USD×1.9464%66.0%2.0 below
NZD/USD×1.9554%66.1%12.1 below
EUR/USD×1.9559%66.1%7.1 below
EUR/CHF×2.0062%66.7%4.7 below
EUR/JPY×2.0460%67.1%7.1 below
AUD/USD×2.0459%67.1%8.1 below
USD/CHF×2.0560%67.2%7.2 below
GBP/CHF×2.0864%67.5%3.5 below
GBP/JPY×2.3566%70.1%4.1 below
Separate sample, Mar 2014 – Mar 2015, over 43 million trades (squares)
EUR/USD×1.7361%63.4%2.4 below
GBP/USD×1.9359%65.9%6.9 below

In pips, not dollars: trade size is left out; the source does not say whether trading costs were deducted. FXCM clients' closed trades, 15 most-traded pairs, 2009–10. Sources: FXCM's DailyFX, 2011, 2015; break-even shares by Tradiens.

Each row is one currency pair. A dot to the left of its mark means too few trades won to break even in pips; the right-hand column gives the gap in percentage points.

Step 1 of 3

Start with the row in bold, EUR/USD, the euro against the US dollar. The blue dot on that row sits at 59%, which is the share of euro–dollar trades that closed with a gain, so most of them won. Then look at the multiple beside the pair's name, ×1.95. It is the reason that 59% was not enough, because it says that the average losing trade was almost twice as large, in pips, as the average winning one.

The average losing trade was almost twice the size of the average winning trade, so breaking even in pips required 66% of trades to win.

The article supplied those two sizes itself. In the authors' words, "trader losses on EUR/USD were an average of 127 pips while profits were only an average of 65 pips." The average losing trade was therefore 1.95 times the size of the average winning trade. A trader in that position can win more often than lose and still fall behind, because losses do not need to be frequent to outweigh gains; they only need to be large enough.

The result of the average trade has two parts: how often trades won multiplied by how much they won, minus how often they lost multiplied by how much they lost. Out of 100 euro–dollar trades, 59 won an average of 65 pips, which makes 59 × 65 = 3,835 pips gained, and 41 lost an average of 127 pips, which makes 41 × 127 = 5,207 pips lost. The hundred trades therefore ended 1,372 pips down, which is 13.7 pips per trade, the loss of about 14 pips given at the start of this essay. If 66.1% of the trades had won, the average trade would have broken even: at that share the pips gained by the winners would exactly have covered the pips lost by the losers. That share is the break-even win rate, and it follows from the two average sizes alone, because it is the average loss divided by the sum of the average win and the average loss: for the euro–dollar, 127 ÷ (65 + 127) = 127 ÷ 192.

Step 2 of 3

Now look at the upright mark on the same row, at 66.1%. It shows the share of trades that would have had to win for the pips gained and the pips lost to cancel out, given how large the average win and the average loss were. The grey bar between the dot and the mark is the shortfall, 7.1 percentage points, and that shortfall is the whole reason the average euro–dollar trade lost pips even though most of the trades won.

One currency pair could be a coincidence, so the real test lies in the other fourteen, for which the same article printed the same three numbers.

In 13 of the 15 most-traded currency pairs, clients' trades won less often than the size of their losses required.

Every pair can be put through the same calculation, and each comes out with its own break-even line, because each has its own average win and average loss. On almost every pair the win rate looked healthy: in 14 of the 15, most trades won, and the one exception, the Australian dollar against the yen, came in at 49%. In all 15 pairs, however, the average losing trade was larger than the average winning one, by a factor of between 1.05 and 2.35. Because the line is the average loss's share of the average win and the average loss added together, a larger loss put every pair's line above 50%: even the smallest factor, 1.05 for the Australian dollar against the yen, gives 1.05 ÷ 2.05, or 51.2%, so more than half of a pair's trades had to win just to break even. A line above 50% did not by itself mean a loss in pips, because a pair whose win rate cleared the line still came out ahead, but in 13 pairs the win rate fell short of the pair's own line. The shortfall ranged from 2.0 percentage points for the British pound against the dollar to 12.1 for the New Zealand dollar against the dollar.

Step 3 of 3

Now read down the whole list. In 13 of the 15 pairs the dot sits to the left of its mark, which means that too few trades won to break even in pips. The two rows USD/CAD and AUD/NZD are the exceptions: their losing trades were not much larger than their winning ones, so their marks sit lower, and enough of their trades won to clear them. The squares at the bottom come from a separate sample, March 2014 to March 2015, and point the same way.

The two pairs above their lines matter as much as the thirteen below them. On trades in the Australian dollar against the New Zealand dollar (AUD/NZD), 71% won, and the average loss was only 1.48 times the average win; on the US dollar against the Canadian dollar (USD/CAD), 61% won, with losses 1.30 times the wins. Because losses there were not much larger than gains, the break-even lines sat lower, at 59.6% and 56.5%, and the win rates cleared them. The source does not say why these two pairs behaved differently, and this essay does not guess. What they do show is that a high win rate is not a bad sign in itself: where losses stayed close to gains in size, a high win rate was exactly what made the trades come out ahead in pips.

A later measurement, over a different period and a different set of trades, found the same direction. In June 2015 Rodriguez published figures from more than 43 million trades closed by FXCM clients in 2014–15, this time giving numbers for only two pairs.[2] In that sample, 61% of euro–dollar trades won, while breaking even in pips would have required 63%. The euro–dollar gap was smaller than in 2009–10 and the pound–dollar gap, the other pair reported, was larger, but both pointed the same way, so the pattern was not peculiar to one pair or to one period. Both measurements are the broker's own, unaudited teaching material, and they count trades, not clients, so they cannot say how many clients made or lost money; the Methodology section at the end explains these limits.

Why would winning trades be small and losing ones large across the board? A habit that would produce exactly this pattern has been measured elsewhere, among stock investors.

Selling winners quickly and holding on to losers raises, by simple arithmetic, both the win rate and the size of the average loss.

The best-known measurement of that habit is a 1998 study by the economist Terrance Odean of 10,000 randomly chosen accounts at a US discount broker, covering trades from 1987 to 1993.[3] He was testing what economists call the disposition effect, described in 1985 by Shefrin and Statman as "a general disposition to sell winners too early and hold losers too long".[4] On the days these investors sold anything, they sold 14.8% of the holdings that stood at a gain that day but only 9.8% of those at a loss, so gains were turned into cash about one and a half times as readily as losses.

The habit also changes the win rate, and that part follows by simple arithmetic: it is this essay's reasoning, not one of Odean's findings. Take the same trades, entered at the same moments and following the same price paths: a trader who sells each winner as soon as it shows a small gain and holds each loser ends up with many small wins and fewer, larger losses than a trader who closes both kinds alike. The win rate therefore rises, and so does the average loss.

Whether that costs money depends on what prices do next, and among Odean's investors it did: in his words, "For winners that are sold, the average excess return over the following year is 3.4 percent more than it is for losers that are not sold," the excess return being the return above a broad market index.

What has been measured, among ordinary American stock investors, is a strong preference for selling winners early, and that habit would leave the trace FXCM's figures show: many small wins and fewer, larger losses. Nobody measured it among FXCM's clients, and Odean counted decisions to sell shares two decades earlier, not winning currency trades, so FXCM's figures are consistent with the habit but cannot show that it produced them.

Some investment products collect small wins by design rather than by habit, and one, an options fund run by LJM Funds Management, kept what FXCM's figures lack: a record of its results over time, measured on the money invested rather than in pips, and beside it the firm's own estimate of how large a loss could be.

In January 2018, LJM reported positive performance in every quarter of its fund's last fiscal year, while its own risk report showed the estimated loss more than tripling.

The record in question belonged to the LJM Preservation and Growth Fund, which the US regulators later called the P&G Fund. On 11 January 2018 the fund filed its annual report for the year that ended on 31 October 2017, which opened with a shareholder letter from LJM's chief portfolio manager, Anish Parvataneni, whose first sentence read: "LJM is pleased to report consistent, positive performance for each quarter of fiscal 2017."[5]

The positive quarters followed from what the fund sold. According to the SEC, the US securities regulator, the fund "generated income" by selling "out-of-the-money options on S&P 500 futures contracts".[6] An option gives its buyer the right to buy or to sell at a fixed price, and the buyer pays the seller a fee for that right, called the premium. "Out-of-the-money" means that the fixed price lay beyond the market's current level, below it for an option to sell (a put) and above it for an option to buy (a call), so the option would be worth something at expiry only if the market moved past that price; the fund sold both kinds.[5] In plain terms, it sold options that pay out only if the S&P 500 makes a large move, in effect insurance against such a move, above all a crash. Selling options in this way is a legitimate and widely used strategy, and this essay does not present LJM as its typical outcome.

That is why the quarters kept coming out positive: in calm markets the premium is usually kept, because the large move the buyer pays for does not come. The comparison with insurance breaks down at one point, and that point explains how the fund later lost so much: an insurer sells cover against many risks that do not all arrive together, whereas a single fall in the index would make every put the fund had sold more costly at the same time, so the losses on those puts would arrive together rather than one by one.

A run of positive quarters measures how often the fund ended a quarter ahead: the same kind of number as a win rate, counted by the calendar instead of by the trade. That record drew money in: during fiscal 2017 the fund's net assets more than doubled, and about 90% of the increase was new money paid in by investors rather than gains.

The record did not show how large a bad outcome could be, but LJM's own risk reports, which the SEC quoted in a complaint filed in 2021, did estimate it. They put a figure on what the P&G Fund would lose if the S&P 500 fell 5%, a moderate setback, together with a 50% rise in implied volatility, the market's expectation of how much prices will swing, read from the prices of options. According to the SEC, the reports put that loss at 18.6% of the fund's value at the beginning of October 2017 and at 61.4% by 31 January 2018, an estimate for one scenario, not a forecast of what would happen. LJM's own estimate of that loss had more than tripled in four months; over the nearly two years before, according to the complaint, it had averaged close to its October starting level.

The rise began in October 2017, the last month of the fiscal year whose quarters the manager called positive, and continued after that year had closed, into the months just before the fall. The SEC alleged that the rise was deliberate and that the change concerned the puts: the managers sold more of them and sold them closer to the market's current level, in an attempt, the complaint says, to collect more premium, which also raised the risk of loss.[6][7]

The same letter that announced the positive quarters also said the firm had been "rebalancing to retain a stable risk profile", which the SEC alleges was untrue. So on 11 January 2018, in one document, the good record and a description of steady risk stood side by side, while, according to the SEC, the managers had been adding risk since October.

LJM Preservation and Growth Fund (the P&G Fund). The estimates come from LJM's internal risk reports as quoted in the SEC's 2021 complaint, which sets out the regulator's allegations. Filing date: SEC EDGAR.

LJM's own loss estimate more than tripled between early October 2017 and 31 January 2018.

What LJM's own risk report showed

Estimated loss, as a share of the P&G Fund, if the S&P 500 fell 5% together with a 50% rise in implied volatility

18.6%, early October 2017

61.4%, by 31 Jan 2018 more than tripled

31 Oct 2017: fiscal 2017 ends

11 Jan 2018, annual report filed: "positive performance for each quarter of fiscal 2017" "a stable risk profile"

The two dots are LJM's own estimates at two dates, as quoted by the SEC. The complaint gives no number for the days in between, so no line joins them.

The figure shows what a run of positive quarters could not: the size of a possible loss, measured by the firm itself, growing in the very months when its written description of risk stayed the same. Then, in early February 2018, the market fell.

In early February 2018, over two trading days, the funds LJM managed lost about 80% of their value, more than $1 billion.

The fall that came was larger than the one in LJM's scenario: according to the SEC, the S&P 500 fell more than 6% in two trading days, Friday 2 February and Monday 5 February, and volatility jumped at the same time. Over two trading days that month, according to the regulators, the funds LJM managed, the P&G Fund and LJM's private funds together, lost about 80% of their value, more than $1 billion.[7][8] This is the actual loss of all those funds together, so it measures something different from the P&G Fund's 61.4% estimate and is no check on it.

The reason the loss was so large lies in what the fund owed and in how it held its positions: a fund that has sold options owes whatever it would cost to buy them back, and that cost is the options' market value. The options the fund had sold rose in value far faster than the index fell, because the drop and the jump in expected volatility made them more expensive at the same time. The fund had sold them on margin, which means that instead of setting aside the full amount it might have to pay, it had posted collateral with its broker to cover what it might owe, so as the cost of buying the options back rose, the broker required more. The positions were therefore closed out in the middle of the panic: according to the SEC, the broker ordered the liquidation after the market closed on Monday 5 February. On margin there was no waiting it out: the fund could not hold on until prices calmed down, because the broker, not the fund, decided when the positions were closed.

Both stories, then, rest on a record that counted how often things went well, and the answer to the opening question about FXCM's clients lies on the side that record left out: how large the losses were when they came.

The pips gained by the winning euro–dollar trades were used up by the losing ones, which were fewer but much larger.

Most euro–dollar trades won, and the average one still lost pips, because what the winning trades gained was used up by the losing ones. Among FXCM's euro–dollar trades the losses came often, in about four trades out of ten, and the average loss was almost twice the average win, which was enough to outweigh the winners' greater number. At LJM the heavy loss came once, in two trading days, and it was large enough to take most of the value of the funds the firm managed. The two losing sides had different shapes, but they lead to the same answer, because in both stories the reassuring number counted how often things went well and said nothing about how much the bad outcomes cost.

For FXCM's clients that number was the win rate; for LJM it was the run of positive quarters, which counted good quarters the way a win rate counts winning trades. For an option seller the losses that matter may not have happened yet, so the question is how large the worst loss could be, and a run of positive quarters does not show it. For a record of closed trades the check is simpler, because the break-even win rate is the average loss divided by the sum of the average win and the average loss, and it can be worked out from anyone's own averages. For FXCM's clients, that arithmetic gives the answer: what mattered was how much the losing trades cost, and the win rate did not show it.

Methodology

The currency figures come from one broker's own, unaudited publication.

The main figure uses two measurements published by DailyFX, the research unit of FXCM, an online currency broker:[9] more than 12 million client trades closed in 2009 and 2010, and more than 43 million closed in a later sample, in both cases in the 15 currency pairs FXCM's clients traded most.[1][2] For each pair the source gives the win rate and the average size, in pips, of winning and losing trades; the later article does so for two pairs only, and we have not read values off its unlabeled charts for the others. The later sample is dated in its text from the second quarter of 2014 to the first quarter of 2015 and in its charts' data note from 1 March 2014 to 31 March 2015, the window the figure uses. The 2009–10 article's "Average" column is not used, because it has no matching win rate. A 2017 regulatory finding against FXCM, which concerned execution rather than these statistics, is described in the note to source 10.[10]

How the break-even line is calculated, and what pips leave out.

We calculated each pair's break-even win rate as the average loss divided by the sum of the average win and the average loss, and its average result per trade as the win rate times the average win minus the loss rate times the average loss. The source gives win rates in whole percent and average sizes in whole pips, so to test the rounding we moved each win rate up and down by half a percentage point and each average by half a pip, and no pair's dot moved to the other side of its mark. A pip is a fixed step in the quoted exchange rate, 0.0001 for most pairs and 0.01 for pairs quoted in yen, so its value in money depends on the currency pair, on the size of the trade and, for pairs not quoted in dollars, on the exchange rate at the time, and the figures do not include trade size. Pip results of different pairs therefore cannot be added or averaged. Because of that, and because the source does not say whether costs such as the spread between buying and selling prices were deducted, the break-even line holds in pips and could lie elsewhere in dollars. All the numbers describe trades, not clients, so they say nothing about how many clients made or lost money.

The habit was measured among stock investors, not currency traders.

Odean's sample was drawn at random from accounts active in 1987 at a nationwide US discount broker. On each day an investor sold something, he compared the share of available gains that were realized, that is sold, with the share of available losses; accounts closed during the period were not replaced, which could bias the sample toward survivors.[3] The step from this habit to a higher win rate and a larger average loss is our own arithmetic, and no study in our sources measured the habit among currency traders.

The LJM figures come from the fund's own filing and from the regulators' allegations.

The positive quarters, the fund's assets and its holdings come from its annual report for the fiscal year ended 31 October 2017, filed with the SEC on 11 January 2018,[5] which lists the puts and calls on S&P 500 futures that the fund had written and the options it had bought. The quarters rest on the statement of the letter's signer, LJM's chief portfolio manager, because the report gives returns only by fiscal year; the same letter names December 2016 and February 2017 as months with negative returns, which a positive quarter can contain. For Class I shares, net of fees, the fiscal-year returns were −10.20% for the first, shorter year (9 January to 31 October 2013, not annualized), then +2.00%, +12.23%, +15.88% and +8.85% for fiscal 2014 to 2017.

Everything about LJM's risk reports, its managers' decisions and the February losses comes from the regulators' complaints of May 2021 and their press releases,[6][7][8] which are allegations, not findings; the CFTC's consent order of July 2025 is a settlement.[11]

The complaint dates the first estimate in Figure 2 only as "at the beginning of October", so its short bar from 1 to 10 October 2017 is our convention. The second is the complaint's reading "by January 31st", which it does not present as the highest, and we have not read values off its daily chart. The complaint describes "+50% Vol" as "a 50% increase in implied volatility" and gives no starting level.

The SEC dates the two losing days to Friday 2 February and Monday 5 February 2018, and the CFTC to 5 and 6 February. The body therefore names the SEC's two days only together with the SEC, and otherwise writes "over two trading days" and "about" 80%, since the sources' wording differs. That loss covers all the funds LJM managed, the P&G Fund and LJM's private funds together, so it cannot be compared with the P&G Fund's own estimates or returns.

Further reading: Ben-David and Hirshleifer (2012) on why investors sell;[12] Fischbacher, Hoffmann and Schudy (2017), a laboratory experiment on automatic sale orders;[13] Wilshire Analytics (2016), commissioned by CBOE, on long-run returns of option-selling indexes.[14]

Sources

  1. David Rodriguez and Timothy Shea, "What is the Number One Mistake Forex Traders Make?", DailyFX, 8 December 2011. Internet Archive capture of 7 January 2012 (20120107175343); series introduction: Timothy Shea, "What Are the Traits of Successful Forex Traders?", DailyFX, 26 December 2011 (capture 20120108212008). We cite this 2012 copy because a version archived in August 2024, when DailyFX was under different ownership, replaces "FXCM" with "a major FX broker".
  2. David Rodríguez, "Why do Many Forex Traders Lose Money? Here is the Number 1 Mistake", DailyFX, syndicated by FXStreet, 26 June 2015. Internet Archive capture 20150716082231.
  3. Terrance Odean, "Are Investors Reluctant to Realize Their Losses?", The Journal of Finance 53(5), October 1998, 1775–1798. doi:10.1111/0022-1082.00072.
  4. Hersh Shefrin and Meir Statman, "The Disposition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence", The Journal of Finance 40(3), July 1985, 777–790.
  5. Two Roads Shared Trust, Form N-CSR, LJM Preservation and Growth Fund, annual report for the fiscal year ended 31 October 2017, filed 11 January 2018 (SEC EDGAR, accession 0001580642-18-000199).
  6. SEC v. Caine et al., N.D. Ill., No. 1:21-cv-02859, complaint filed 27 May 2021.
  7. US Securities and Exchange Commission, press release 2021-89, 27 May 2021.
  8. US Commodity Futures Trading Commission, press release 8392-21, 27 May 2021; CFTC v. LJM Funds Management et al., N.D. Ill., No. 1:21-cv-02863, complaint filed 27 May 2021.
  9. FXCM Inc., Form 10-K for 2010 (filed 31 March 2011) and for 2014 (filed 16 March 2015), SEC EDGAR.
  10. US Commodity Futures Trading Commission, press release 7528-17, 6 February 2017. In the order this release announced, issued with FXCM's consent, the Commission found that from September 2009 through at least 2014 FXCM "engaged in false and misleading solicitations" of its retail customers by concealing its relationship with its most important market maker; the finding concerned execution, not the trade statistics this essay uses, but it covers both data windows.
  11. US Commodity Futures Trading Commission, press release 9092-25, 1 July 2025.
  12. Itzhak Ben-David and David Hirshleifer, "Are Investors Really Reluctant to Realize Their Losses? Trading Responses to Past Returns and the Disposition Effect", The Review of Financial Studies 25(8), 2012, 2485–2532. doi:10.1093/rfs/hhs077.
  13. Urs Fischbacher, Gerson Hoffmann and Simeon Schudy, "The Causal Effect of Stop-Loss and Take-Gain Orders on the Disposition Effect", The Review of Financial Studies 30(6), 2017, 2110–2129. doi:10.1093/rfs/hhx016.
  14. Chicago Board Options Exchange (CBOE) and Wilshire Analytics, press release, 21 September 2016, on Wilshire Analytics Applied Research Group, "Three Decades of Options-Based Benchmark Indices with Premium Selling or Buying: A Performance Analysis", commissioned by CBOE.

How to cite this essay

Tradiens Research. (2026). What a win rate leaves out (Version 2.0). Tradiens Data Essays. https://tradiens.com/essays/win-rate-lie

@misc{tradiens2026winrate,
  author       = {{Tradiens Research}},
  title        = {What a Win Rate Leaves Out},
  year         = {2026},
  version      = {2.0},
  howpublished = {Tradiens Data Essays},
  url          = {https://tradiens.com/essays/win-rate-lie}
}
Tradiens Research · Data essay · v2.0 Trade like you know.