Are Forex Signals Profitable? The Expectancy Math That Decides It

Forex signals are profitable when the signals carry positive expectancy after spread, commission and slippage, and when the person following them executes closely enough to keep that expectancy intact. Expectancy, not win rate, is the number that decides it: win rate multiplied by average win, minus loss rate multiplied by average loss. A set that wins 40 percent of the time at 1:3 risk to reward is far more profitable than one winning 80 percent at 1:0.3.

• Win rate on its own is meaningless. Without the average risk to reward ratio attached, it cannot tell you whether a system makes or loses money. An 80 percent win rate can be a losing system.
• Expectancy per trade in R is the honest metric. E = (win rate x average win in R) minus (loss rate x average loss in R). Above zero after costs is a business. Below is a hobby that bills you.
• Cost drag is measured against your stop, not your account. An all-in round-trip cost of 0.8 pips is 6.7 percent of a 12 pip stop and 1.3 percent of a 60 pip stop. Tight-stop signals lose far more edge to costs.
• The execution gap can turn a profitable signal into an unprofitable trade. Entering three pips late on a 30 pip stop and 90 pip target cuts realized risk to reward from 3.00 to 2.64 and expectancy by roughly a quarter.
• Judge on 100 trades minimum, not 10. At 20 trades the 95 percent confidence band around a measured win rate is about plus or minus 21 points, wide enough to make a great system look broken.

The short answer: it is an expectancy question, not a signal question

"Do forex signals work" is the wrong shape of question. Signals are just instructions: buy XAUUSD at 2418.40, stop 2412.90, target 2434.90. Whether following them makes money depends on one arithmetic property of the whole set, expectancy, and on how much of it survives contact with your broker and your reaction time.

That reframing makes the question answerable. You cannot evaluate "is this provider good" in the abstract, but you can check whether a documented sequence of entries, stops and exits produced positive expected value per trade over a large enough sample, and whether your own fills tracked it closely enough to keep that value. So: some signals are profitable, most published ones are not, and the ones that are can still lose money for the person following them. Three failure points, and the arithmetic for each is below.

Why win rate alone tells you nothing

Win rate leads every signal pitch because it is the only number that reads well in isolation. It also carries the least information. Account A closed 100 trades, won 40, and every winner made three times what every loser lost. Account B closed 100 trades, won 80, and every winner made 0.3 times what every loser lost. Account B has double the win rate. Account A made fifteen times as much money.

A win rate is half the picture and the payoff ratio is the other half. Together they define a breakeven line: the breakeven win rate for any risk to reward ratio R is 1 divided by (1 + R). At 1:1 you need 50 percent. At 1:2, 33.3 percent. At 1:3, 25 percent. At 1:0.3, 76.9 percent. At 1:0.25, exactly 80 percent, which is the trap: a system advertising an 80 percent win rate on quarter-R targets sits on breakeven before you pay a single spread.

Account A's 40 percent sits 15 points above its breakeven line. Account B's 80 percent sits 3.1 points above its own. Margin above breakeven is what survives bad months, and 3.1 points is inside the noise of a hundred-trade sample. Account B does not reliably have an edge. It has a rounding error.

The expectancy formula, worked through with numbers

Expectancy is the average result of one trade expressed in R, where 1R is the amount you risk on a single position. Working in R rather than dollars is the only way to compare trades taken at different position sizes without the biggest positions drowning out the rest.

E = (W x A) - (L x B)

W is win rate as a decimal, A is average win in R, L is loss rate as a decimal, B is average loss in R (usually 1.0 if stops are honored).

System A, 40 percent win rate at 1:3:
E = (0.40 x 3.0) - (0.60 x 1.0) = 1.20 - 0.60 = +0.60R per trade

System B, 80 percent win rate at 1:0.3:
E = (0.80 x 0.3) - (0.20 x 1.0) = 0.24 - 0.20 = +0.04R per trade

Now attach money. On a $10,000 account risking 1 percent per trade, 1R equals $100. Across 100 trades, ignoring compounding and costs, System A returns 60R or $6,000. System B returns 4R or $400 on the same trade count and the same risk.

The perverse part is which one feels better. System B wins four out of five times and produces streaks of six and seven green trades. System A loses six out of ten, so most days end red and most weeks contain a stretch where you are convinced the thing has stopped working. The system that pays is the one that feels like failure, and that mismatch is why people abandon positive expectancy signals halfway through a normal drawdown.

Expected losing streaks, so the drawdown does not surprise you

The expected longest losing streak in n trades with loss probability q is roughly ln(n) divided by ln(1/q). For System A over 100 trades with q = 0.60, that is 4.605 / 0.511, about 9 consecutive losses, which at 1 percent risk with compounding is a drawdown near 8.6 percent. For System B, q = 0.20 gives about 3. Smooth, comfortable, and worth $400 per hundred trades before costs.

40 percent at 1:3 versus 80 percent at 1:0.3, side by side

Costs and slippage decide this comparison, so the table prices them in one at a time and then together. It assumes a $10,000 account, 1 percent risk per trade, a round-trip cost of 0.05R (1.5 pips against a 30 pip stop, or a 15 cent gold spread against a $3.00 stop), and three pips of adverse entry slippage.

Read the fifth row again. A cost of 0.05R per round turn, unremarkable on any retail account, shaves 8 percent off System A and pushes System B below zero. Row seven is what a live account actually experiences, because you pay both drags: System A still banks +0.41R per trade while System B bleeds 0.10R per trade and wins four out of five of them. That is why the "high win rate" angle sells so well: technically checkable, emotionally reassuring, fully compatible with a negative edge. Our breakdown of how fake forex signals are constructed covers the other statistical tricks that travel with it.

Spread, commission and swap: where expectancy leaks out

Costs are quoted in pips or dollars, which makes them look small. Convert them to R and they stop looking small.

On a raw-spread EURUSD account you might see a 0.1 pip spread plus $3.50 commission per side per standard lot. The commission is $7.00 round turn on a 1.0 lot, which is 0.7 pips, and the spread adds another 0.1, so the all-in cost of the trade is 0.8 pips. Against a 12 pip stop that is 6.7 percent of your risk gone before price moves. Against a 60 pip stop, 1.3 percent. Same broker, same size, five times the drag, purely from stop distance.

Gold bites hardest. XAUUSD typically shows a spread of $0.15 to $0.35 during the London and New York overlap, roughly 12:00 to 16:00 UTC in summer and 13:00 to 16:00 UTC in winter, the deepest liquidity window of the day. Around the daily rollover at 00:00 broker server time, commonly 21:00 or 22:00 UTC depending on the season, spreads of $0.50 to $1.00 are routine and liquidity thins. A gold scalp with a $3.00 stop taken at rollover can pay 30 percent of its risk to the spread. The same signal in the overlap pays 5 to 12 percent. Only the clock changed.

Swap is the third leak and the quietest. Holding across the 00:00 server rollover charges financing, and the Wednesday rollover is charged triple to cover the weekend value date, so an eight day hold pays financing for roughly eight days including the days the market is closed. On a tight stop that is a real slice of 1R, so price it in.

The consequence: wide-stop signals are structurally more robust to costs than tight-stop signals, even when the headline win rate is worse. Divide a provider's typical stop distance by your all-in round-trip cost on that instrument. That ratio tells you more about whether you can harvest their edge than any equity curve screenshot.

The execution gap: how a profitable signal becomes an unprofitable trade

This is the failure mode that catches disciplined people. The signal has genuinely positive expectancy, the provider is honest, and you still lose money, because the trade in your account is not the trade in the signal.

Take a signal with a 30 pip stop and a 90 pip target, a clean 1:3. You see the message, unlock your phone, open the platform, and fill three pips worse than the posted entry. Your stop is now 33 pips away and your target 87. Size on your actual stop so you still risk exactly 1 percent, and realized risk to reward is 87 divided by 33, which is 2.64, not 3.00.

Feed that back in with the same 40 percent win rate: E = (0.40 x 2.64) - (0.60 x 1.0) = +0.45R. Three pips, roughly a quarter of the edge gone. Over 100 trades on that $10,000 account at 1 percent risk, that is about $4,500 instead of about $6,000. Three pips of reaction time cost $1,500.

Scale the same arithmetic down to System B. Stop 30 pips, target 9. Three pips late and your stop is 33 while your target is 6, a realized ratio of 0.18. E = (0.80 x 0.18) - (0.20 x 1.0) = -0.06R. The system flips from marginally positive to reliably negative on three pips of delay, and no amount of discipline fixes it, because the geometry was never robust enough to absorb friction.

Where the execution gap comes from

Rarely one big thing. Usually four small ones stacked: notification delay, seconds spent reading and typing levels, the spread you cross on a market order, and the fact that a signal posted at 13:32 UTC is often posted precisely because price is moving. The moment you are slowest is the moment price is fastest.

Fixes that measurably close it: use pending orders at the posted entry instead of chasing with market orders, so you get the signal's price or no trade at all. Set stop and target in the same ticket. Keep a position size calculator open so you are not doing lot math under pressure. Then measure: log the posted entry and your fill for a month. Average slippage of 0.4 pips is a non-problem. Four pips costs more than any change of provider would gain. Our guide on how to follow forex signals without leaking your edge works through each of these.

Partial closes, break even moves and why your R log needs three buckets

Real signal management is not binary. A desk running a position often moves the stop to break even after price travels a defined distance, and closes part at a first target while letting the rest run. Good risk management, and it breaks naive win rate math.

Say a signal closes half at 1R and moves the stop on the remainder to entry. If the runner reaches 3R, the result is (0.5 x 1.0) + (0.5 x 3.0) = 2.0R, not 3.0R. If the runner comes back and stops at break even, the result is (0.5 x 1.0) + (0.5 x 0.0) = +0.5R minus costs. That second case is a winner in the win rate column and a half-R gain in money. Log only wins and losses and you overstate average win while hiding how often the runner fails.

Keep three buckets: full losses at -1R, scratch outcomes near 0R after costs, and wins at their actual blended R. Then average the R column and you have expectancy that matches your statement. This is also why channels posting live management updates are auditable and entry-only channels are not: told when a position moved to break even, when part was closed and when the rest was closed, you can reconstruct the exact R of every trade. Entries plus a later screenshot of a win reconstructs nothing. That is the substance behind what verification actually means for signal providers.

What a realistic assessment period looks like: 100 trades, not 10

Sample size is where judgment breaks. Ten trades tells you almost nothing and misleads you in whichever direction the noise fell. The standard error of a measured win rate is the square root of p(1-p)/n. For a true 40 percent win rate:

• n = 20: standard error 0.110, a 95 percent band of about plus or minus 21 points. Your measured win rate can land anywhere from 19 to 61 percent with the system unchanged.
• n = 100: standard error 0.049, band of about plus or minus 9.6 points. A measured 22 percent is now evidence of a problem, not a bad fortnight.
• n = 400: standard error 0.024, band of about plus or minus 4.8 points. Here you can separate a 0.60R system from a 0.35R one.

There is a second reason 100 is the floor. The expected longest losing streak for a 40 percent system is about nine trades per hundred. Observe only 20 and your whole window may sit inside a perfectly normal losing cluster. You conclude the system is broken, you are wrong, and you used correct-looking arithmetic on a sample that could not support it.

Translate that into calendar time. A desk taking two to four positions a week reaches 100 closed trades in six to twelve months. That is the real assessment period, and the same skepticism applies to your own conclusions: three good weeks proves as little as three bad ones.

A checklist before you follow anyone

Each of these is answerable from public information, and each failure should stop you.

1. Is every trade published before the outcome is known? Entry, stop and target posted at the moment of entry, timestamped, no deletions. Gaps mean the sample you compute on is not the sample that happened, and selective deletion inflates a win rate better than any other trick.

2. Is average R published alongside the win rate? If only the win rate is quoted, compute the breakeven rate for their typical stop and target distances and see how much margin is really there.

3. Is there third-party verification of the underlying strategy? Self-reported results are an assertion, a read-only account on an independent platform is evidence. The strategy behind the JPTC calls runs on a public MyFxBook account, the kind of external record to demand from anyone, including us.

4. Do the stop distances survive your broker's costs? Divide typical stop distance by your all-in round-trip cost. Under 20 to 1 and you hand over a large share of the edge before you start.

5. Can you be at the screen when signals fire? If a provider posts mostly around the London open, 07:00 to 09:00 UTC, and you are asleep, your results will not resemble theirs however good the calls are.

6. Who pays the provider, and do you keep control? A subscription pays them whether you profit or not. A broker rebate pays them when you trade. Neither is disqualifying, but know which incentive you are inside. JPTC's model is the rebate one, which is why the JPTC signals channel carries no monthly fee, and you place every trade yourself at your own size: JPTC never touches your money and you can skip any call. Our piece on whether forex signals are worth it compares the two models.

Trading involves a significant risk of loss and past performance is not indicative of future results. The arithmetic here is generic and educational: it shows how to evaluate any signal set, including this one, using numbers you verify yourself.

Frequently asked questions