In a market with no edge, how often does a 2R target get hit before the stop?

[Fairman]

In a market with no edge, how often does a 2R target get hit before the stop?

Take a market that just wanders, with no edge at all. Each step is up or down with equal chance. You enter, put the stop 6 pips away and the target 12 pips away. What are the odds the target gets hit first? It isn't 50%. For a driftless walk it's the stop distance divided by the total distance, so 6 divided by 18, which is one in three.

The same sum for other shapes, all with a 6 pip stop:
Target 3 pips: hit first 66.7% of the time.
Target 6 pips: 50%.
Target 9 pips: 40%.
Target 18 pips: 25%.

None of these makes or loses money on average before costs. After costs they all lose the same amount per trade. With a 0.8 pip combined spread and commission, that's 0.8 pips every time, or 0.13R on a 6 pip stop, whatever the target.

That's why a backtest quoting a 66% win rate on a 3 pip target and a 6 pip stop doesn't impress me. Coin flips give you that. A high hit rate on a small target proves nothing by itself. The real test is how far your hit rate sits above the baseline for that particular shape. A 1:2 setup that wins 38% of the time against a 33% baseline has something going on. One that wins 34% is mostly noise plus costs.

Real prices aren't pure random walks, since they trend and snap back in patches, but the baseline takes ten seconds to work out and gives you a line to compare against.

[Fairman]

The gap between a 33% baseline and a 38% result is harder to detect than it looks. Over 100 trades at a true 33.3% rate, the standard deviation is about 4.7 wins, so 38 wins is only one standard deviation above chance and turns up by luck roughly one time in six. To tell a true 38% from 33% at around two standard deviations you need something like 400 trades, where the standard deviation shrinks to about 2.4 points. So 60 trades of a 1:2 setup showing 40% winners isn't proof of anything yet. Put the sample size you have next to the size of the edge you think you see, and be honest about which one is bigger.

[Fairman]

Costs move the line you actually have to clear. Using the 0.8 pip cost from the opening post, a winner nets 11.2 pips and a loser costs 6.8, so the break-even win rate on a 1:2 shape is 6.8 divided by 18, which is 37.8%. Now two numbers matter and they differ: the coin flip baseline of 33.3% and the break-even rate of 37.8%. A setup that hits 36% is better than random and still loses money. One at 39% is both. The stretch between those two lines is where a lot of reasonably real patterns die, because the edge exists but is smaller than the toll. Cutting the cost by 0.2 pip moves break-even to 36.7%.

[Fairman]

You can use the baseline to read someone else's report in about thirty seconds. Take the average win, the average loss and the win rate. Say a tester shows an average win of 9 pips, an average loss of 6 pips and 41% winners. The baseline for a 6 pip stop and 9 pip target is 6 divided by 15, which is 40%. So the strategy is winning one point above coin flips, and the expectancy is 0.41 times 9 minus 0.59 times 6, about 0.15 pip per trade before any cost. That isn't an edge, it's a rounding error with a decent equity curve if the sample is long enough. Averages hide the shape of trailing exits, so the check is rougher when exits vary, but a flat result here says look harder before going live.

[Amanhasnoname]

This baseline changed how I read my own backtests. I had a system with a 5 pip stop and 10 pip target that showed a 36% win rate over a few hundred trades, and I was ready to call it an edge. Then I worked out that the coin flip figure for that shape is 33.3%, and my break-even after a 0.7 pip round trip cost was close to 38%. I was better than random and still losing money.

What I ended up doing was keeping the entry logic and changing the shape. Widening the target to 15 pips dropped the random baseline to 25% while my win rate on the same signals fell only to about 30%. The gap above chance actually grew a little, and the cost per trade became a smaller share of the target.

The sample size reply is the important warning, though. My improved version only has around 250 trades so far, which on these numbers isn't enough to be confident. I'm trading it at a quarter of my usual size until it reaches a sample worth trusting.

[Fairman]

Partial exits change the arithmetic but not the conclusion. Try half off at 1R and half at 3R with the stop left where it was. In a driftless market, price reaches +1R before -1R half the time. From +1R the stop is 2R below and the 3R target is 2R above, so it's another coin flip. That gives: 50% of trades lose 1R. 25% reach +1R and then stop out, which is +0.5R on the first half and -0.5R on the second, zero in total. 25% reach 3R, which is 0.5R plus 1.5R, so +2R. Add it up and you get -0.5, 0 and +0.5, so zero again. Half the trades touched +1R, so the hit rate looks like 50%, which is exactly why a split exit flatters win rate statistics.

[FXS HongKong Scalper]

The stop over total distance result is useful to have in mind. With a 6 pip stop and 12 pip target, one in three is the break-even rate, so a 2R target that hits 33 percent of the time is pure noise.

One thing to add for real trading: costs turn the zero into a negative. In a driftless walk the hit rate always matches the break-even rate exactly, whatever the target, so before costs every shape has an expectancy of zero. The spread and commission are then paid on every trade, and with a 6 pip stop a 1 pip round-trip cost is about minus 0.17R per trade on average. A strategy needs to beat that baseline, not the 33 percent figure on its own.