MEASURED · 23 SEPTEMBER 2026
Does RSI work on gold?
Yes, in the opposite direction to the way it is taught. It is the first indicator we have tested that survives the statistics. Across twelve years of XAUUSD, higher RSI readings preceded higher returns, which is the opposite of what “overbought” is supposed to mean.
Why we did not test the 30/70 rule
Nearly every RSI study asks whether buying below 30 and selling above 70 works. That tests folklore rather than the indicator. Wilder picked those numbers as conventions. If RSI carries information, the useful question is where on the 0-100 scale it sits. The answer could be nowhere, or at 12, or only at the extremes.
So we mapped the whole curve. Every ten-point band gets the same treatment at three horizons on three timeframes, and the thresholds come out as a result instead of going in as an assumption. We tested the threshold approach directly once before, on the stochastic oscillator. This study asks whether a threshold is the right idea at all.
Every figure is measured against a random bar
Gold roughly tripled over this sample. The average forward return from any bar is positive, so “price rose after RSI fell below 30” is a statement about the bull market. Every figure below is an edge: the return in that band minus the return from a randomly chosen bar over the same horizon. Zero means the band told you nothing that being in the market did not.
Before the run, we checked our RSI against an independent implementation (identical to ten decimal places), confirmed it returns 100 on a monotonic rise and 0 on a fall, and tested for lookahead by checking that a reading does not change when later bars are added.
RSI band by band, one day ahead
One day ahead, on 15-minute bars, which is the largest sample here. Base rate 0.0585%.
| RSI BAND | OBSERVATIONS | MEAN RETURN | EDGE VS BASE | P |
|---|---|---|---|---|
| 0–10 | 34 | -0.4520% | -0.5104% | 0.016 |
| 10–20 | 1,120 | +0.0010% | -0.0574% | 0.091 |
| 20–30 | 8,377 | +0.0736% | +0.0151% | 0.231 |
| 30–40 | 31,292 | +0.0344% | -0.0241% | 0.000 |
| 40–50 | 65,808 | +0.0467% | -0.0118% | 0.009 |
| 50–60 | 67,415 | +0.0619% | +0.0034% | 0.446 |
| 60–70 | 36,786 | +0.0742% | +0.0158% | 0.007 |
| 70–80 | 10,739 | +0.0955% | +0.0370% | 0.002 |
| 80–90 | 1,589 | +0.2328% | +0.1743% | 0.000 |
| 90–100 | 31 | -0.1011% | -0.1596% | 0.428 |
Read the edge column from the middle outwards. From about RSI 30 upward it climbs steadily: 30–40 is negative, 60–70 positive, 70–80 more so, and 80–90 is the strongest band on the table. That is a gradient with no threshold in it, and it slopes the wrong way for the rule everyone is taught.
The results beat chance
Every previous study in this series ended with “the hits match what chance produces”. This one ended differently.
Significant at p<0.05
31 of 75
Expected by chance
3.8
Surviving Bonferroni
10
p < 0.000667
31 hits where 3.8 were expected is roughly eight times the chance rate, and 10 bands clear a threshold corrected for all 75 tests at once. That rules out the multiple-comparisons illusion that explained our seasonality and golden cross results.
Most surviving results point to momentum
| RSI BAND | HORIZON | N | EDGE | DIRECTION | VS COSTS |
|---|---|---|---|---|---|
| 20–30 | 1 month | 184 | +0.9938% | mean reversion | clears |
| 20–30 | 1 week | 2,629 | +0.1888% | mean reversion | clears |
| 80–90 | 1 day | 1,589 | +0.1743% | momentum | clears |
| 60–70 | 1 week | 9,831 | +0.0839% | momentum | clears |
| 70–80 | 1 day | 3,780 | +0.0609% | momentum | clears |
| 10–20 | 4 hours | 1,120 | +0.0567% | mean reversion | clears |
| 70–80 | 4 hours | 3,780 | +0.0267% | momentum | clears |
| 30–40 | 1 day | 31,292 | -0.0241% | momentum | clears |
| 30–40 | 4 hours | 31,301 | -0.0153% | momentum | clears |
| 60–70 | 1 hour | 36,796 | +0.0049% | momentum | inside spread |
7 of the 10 point to momentum: high RSI followed by strength, low RSI by weakness. Only 3 behave the way the textbook says, and all of them sit in the deep-oversold bands over horizons of a week or more. The single largest result in the study is one of those: RSI 20–30 preceded a 0.99% excess return over the following month, on 184 daily observations.
So both stories are in the data, at different depths and different speeds. What is not in the data is the version printed in every textbook: that a reading above 70 means price is due to fall.
What the spread does to these numbers
Gold’s median spread is $0.20 on a median price of $1,827, which is 0.0113%. Any edge smaller than that is inside the cost of capturing it. 9 of the 10 clear that floor. The one-hour result does not, and it is the one with the most observations behind it.
Clearing the spread is only the first hurdle. None of this was tested as a strategy with entries, exits, position sizing or slippage. It measures what followed a state, which is a much weaker claim than “this makes money”. Our risk-reward study and ATR stop study both found that how you manage a trade changes the shape of results and leaves their sign alone.
The confound we cannot remove
RSI is calculated from recent price changes. A high reading is a description of recent strength, so discovering that high readings precede further gains is uncomfortably close to discovering that gold trended. It did, roughly tripling over the sample.
The base rate controls for the overall drift. It does not control for trend persistence, and no version of this test can, because the indicator and the thing being predicted are both built from the same returns. So the reading is narrower than the headline: gold’s momentum persisted over this period, and RSI is a momentum measure, so RSI appears to predict. That is a fact about gold in 2014–2026, and no hidden property of the indicator. It is the reason we have not built anything on it.
The number that would have made a better headline
RSI above 90 preceded a -0.14% one-hour move with a p-value of 0.006. “Extreme overbought crashes gold” is a far better story than anything above, and it rests on 31 observations in twelve years. The 0–10 band has 34. Both fail the corrected threshold, both are excluded from every conclusion here, and both are exactly the kind of number that gets screenshotted. The 4-hour Fibonacci level that ranked first of 131 on seven observations is the same trap.
Limits
One instrument, one broker, 2014-01-14 to 2026-09-23, in a period when gold mostly rose. A momentum result from a trending sample is the least surprising thing this could have produced, and it would be worth far more if it held in a flat or falling decade. We tested RSI(14) only, and other periods would shift the bands.
Forward windows overlap heavily, which inflates significance in a naive test. The bootstrap control draws with the same structure, and overlapping data still means fewer independent observations than the row counts suggest, particularly on the daily chart where some bands hold under 200. The full table for all three timeframes is published with our other studies, including the base rates and p-values that make each row checkable.
Common questions
Does RSI actually work on gold?
It carries real information, more than most indicators we have tested. Across 75 tests on twelve years of XAUUSD, 31 beat the base rate at p<0.05 where only 3.8 would be expected by chance, and 10 survived a strict correction for testing that many buckets. The catch is the direction: most of those results say higher RSI leads to higher returns, which is the reverse of the overbought/oversold rule.
Is RSI below 30 a buy signal on gold?
Over short horizons, no, and the data leans the other way. Over longer horizons it is the strongest single result in the study: RSI between 20 and 30 preceded a 0.99% excess return over the following month, against the return from a randomly chosen day. That rests on 184 daily observations, so treat it as suggestive.
Is RSI above 70 a sell signal on gold?
The measurement says the opposite. RSI 70-80 preceded excess returns of +0.0609% over a day and the 80-90 bucket +0.1743%, both beating the base rate and both surviving correction. On gold over this period, "overbought" was closer to a continuation signal than a reversal one.
What RSI setting is best for gold?
No single threshold stood out. This study mapped every ten-point band instead of testing 30 and 70, and no cut-off separated good outcomes from bad ones. It found a gradient: from roughly RSI 30 upward, higher readings preceded better returns, and the deep-oversold band behaved differently over multi-week horizons.
Are the RSI effects big enough to trade?
Some of them clear the spread. Gold's median spread is 0.0113% of price, so anything smaller than that is inside the cost of trading it. 9 of the 10 strongest results clear that floor, and the one-hour result does not. Clearing the spread is only the first hurdle. It says nothing about slippage, and none of this was tested as a strategy with entries, exits and position sizing.
Why does this contradict what RSI is taught as?
RSI is calculated from recent price changes, so a high reading describes recent strength. Finding that high readings precede further gains is close to finding that gold trended, and it did, roughly tripling across the sample. This measures momentum persistence in gold. It reveals no hidden property of the indicator.
Related reading
- Stochastic oscillator gold
- Blog
- Does scalping work on gold
- Gold Trading HoursWhen the gold market opens, closes and breaks — and which of those hours are actually wo…
- Gold Trading StrategyWe backtested 34 gold trading strategies on identical XAUUSD data. What survived, what f…
- How to Choose the Best Gold Trading BotThere is no objectively best gold bot, but there are five questions that separate a meas…