Short answer: on gold, entering at a random moment with a 1 ATR stop and a 2:1 target won 32.7% of the time on the hourly chart. The textbook number for a market with no pattern is 33.3%. On every chart from 15-minute upward, random entries landed within a point or two of the theoretical figure, and entries at support and resistance did not beat them.
Your win rate is set by where you put your stop and your target, plus what you pay to trade, plus which way the market happened to be drifting. The entry moves it far less than almost anyone expects.
The rule, in one line
For a market with no pattern in it, the chance of reaching your target before your stop is
risk ÷ (risk + reward)
which is 50% at 1:1, 33.3% at 1:2, 25% at 1:3, 16.7% at 1:5. This is a property of the geometry, not of your analysis. The expected result is zero at every one of them, before costs. After costs it is negative, at every one of them.
Everything below is a measurement of how closely gold obeys that, and of what moves it.
How we measured it
- Gold only (XAUUSD), Deriv's feed, UTC candles, January 2017 to September 2026.
- One random entry per trading day per chart, chosen with a fixed seed so the test is reproducible, long and short.
- Stop: 1 × ATR20 (the average candle range of the previous 20 candles) from the entry. Targets: 1, 2, 3 and 5 times the risk. Maximum 200 candles, in practice fewer than 1% of trades ever reach that limit, so the result is a clean race between stop and target.
- Costs: the real spread, at least 15 points, charged once per round trip; a candle that touches both stop and target counts as the stop, because we cannot see the order inside a candle.
- Same-day support and resistance entries got the identical plan, for comparison.
Random entries, measured against the law
| Chart | Direction | 1:1 (law 50%) | 1:2 (33.3%) | 1:3 (25%) | 1:5 (16.7%) |
|---|---|---|---|---|---|
| M1 | long | 28.8% | 19.2% | 14.5% | 10.1% |
| M1 | short | 30.7% | 21.3% | 16.0% | 10.7% |
| M5 | long | 40.4% | 28.6% | 21.7% | 14.9% |
| M5 | short | 41.1% | 27.2% | 21.1% | 13.9% |
| M15 | long | 45.1% | 31.1% | 24.0% | 16.3% |
| M15 | short | 44.8% | 30.4% | 23.0% | 15.4% |
| M30 | long | 47.3% | 32.7% | 25.7% | 17.6% |
| M30 | short | 45.5% | 30.4% | 22.1% | 15.2% |
| H1 | long | 47.8% | 32.7% | 25.9% | 18.9% |
| H1 | short | 46.0% | 30.9% | 23.8% | 15.0% |
| H4 | long | 48.2% | 33.8% | 27.6% | 19.5% |
| H4 | short | 47.3% | 32.2% | 22.6% | 14.1% |
| D1 | long | 52.7% | 39.4% | 32.1% | 24.0% |
| D1 | short | 45.1% | 27.4% | 19.0% | 11.8% |
Each cell is 1,000 to 3,000 trades. Three things are visible, and all three are useful.
1. From the 15-minute chart up, the law is almost exactly right
Hourly, 1:2, long: 32.7% against a theoretical 33.3%. Hourly, 1:3: 25.9% against 25%. Nobody analysed anything. Nobody read structure. The win rate is the geometry.
If you are being sold a "90% win rate" system, you now know precisely what it is: a very close target and a very wide stop. That is not an edge, it is a ratio, and the arithmetic of the losses is waiting on the other side of it.
2. On fast charts you fall below the law, and the gap is the cost
On the 1-minute chart a 1:1 trade won 28.8% instead of 50%. Nothing is broken. A 1 ATR stop on 1-minute gold is a very small distance, and the spread, about 0.89 bp on average across this sample, eats a meaningful share of it before the race even starts. The faster the chart, the larger your costs are relative to your stop, and the further below the law you sit.
This is the single most useful number on the page for anyone scalping. Your target does not have to be wrong for you to lose; it only has to be close enough that the spread is a real fraction of the distance.
3. Drift bends the law, and it bends it both ways
On the daily chart, random longs won 39.4% at 1:2 against a theoretical 33.3%, and random shorts won 27.4%. Same instrument, same days, same geometry, opposite directions. The difference is gold's decade-long rise.
Read this carefully, because it is where people fool themselves: a random long on gold made +0.18 R per trade at 1:2 and +0.44 R at 1:5 on the daily chart. That is not a strategy and it is not skill. It is a rising market rewarding anyone who was long, and it reverses the moment the market does. Any backtest of a long-biased rule on gold over 2017 to 2026 inherits that same number without earning it, which is exactly why every test we publish subtracts it before calling anything an edge.
Did entering at a level help?
Same days, same stop, same targets, entries at the previous session's high and low, against entries at a random moment:
| Chart | At a level | 1:2 win rate | Random entry | 1:2 win rate |
|---|---|---|---|---|
| M15 | buy support | 30.6% | random long | 31.1% |
| M15 | sell resistance | 28.7% | random short | 30.4% |
| H1 | buy support | 30.7% | random long | 32.7% |
| H1 | sell resistance | 29.8% | random short | 30.9% |
| H4 | buy support | 32.3% | random long | 33.8% |
| H4 | sell resistance | 30.6% | random short | 32.2% |
Six for six to the random entry, by one to two points. We are not claiming random entries are better, the gaps are small. We are saying the measured answer to "does entering at the level improve my odds" was no.
What this means in practice
- Choosing a bigger reward-to-risk ratio does not make you money. It moves your win rate down by the same amount it moves your average win up. At 1:5 you should expect to be right about one time in six, and the arithmetic is neutral before costs, negative after.
- A win rate on its own is uninformative. Always ask what the geometry was. 70% at 1:0.4 is worse than 30% at 1:3.
- The cost of trading is not a rounding error on fast charts. It is the difference between 50% and 29%.
- Any long-only result on gold from 2017 to 2026 must have the market's own rise subtracted before it means anything. Random entries earned +0.44 R at 1:5 on the daily chart. If a strategy cannot beat that, it has not shown you anything.
- The entry is the part of trading that gets all the attention and moves this number least. Stop placement, target placement, costs and market direction did all the work in every measurement above.
What this page does not say
- It does not say entries never matter. It says these entries, on this instrument, over these years, did not beat a random moment on the same day.
- It does not test trailing stops, partial exits, breakeven moves or scaling, all of which change the geometry and therefore the win rate, in ways that need their own measurements.
- It does not test other markets. Drift bends the law, and every market drifts differently.
How we tested
- Stop and target distances are set at entry and never moved. One position at a time.
- Random entries use a fixed seed (31), one per trading day, so anyone can reproduce the same set.
- Fewer than 1% of trades hit the 200-candle limit, so the measured win rate is a genuine stop-versus-target race and not an artefact of the time limit.
- A candle touching both stop and target counts as the stop. On the plans in this study that affects 0.1 to 0.9% of resolved trades from the 15-minute chart up; splitting those 50/50 instead would move the win rates by under half a point.
Reproduce it: every rule is written out in full above, so anyone with gold price data can rebuild this test and check our numbers. The candles are the broker's and are not ours to redistribute.