What Is the Best RSI Oversold Level? We Tested 1 Through 50

Everyone knows the rule:

RSI below 30 means oversold.

It appears in trading books, charting platforms, courses, YouTube videos, and countless articles.

But there is an obvious question that is asked surprisingly rarely:

Why 30? Why not 29? Or 20? Or 10?

We decided to test it.

Instead of accepting the traditional RSI oversold level, we tested every threshold from 1 through 50 on SPY using more than 33 years of daily market data.

The result was surprisingly clear. RSI 30 worked. But it was nowhere near the most efficient oversold level.

And there was another surprise buried inside the results.

The highest profit factor did not necessarily identify the best RSI level. That distinction matters!

The RSI Test

We wanted to isolate one variable: the RSI oversold threshold.

The basic test used:

SPY daily data
RSI with a 2-day lookback
RSI thresholds from 1 through 50
SPY above its 200-day moving average
Entry at the next session’s open
Exit after a short-term rebound
No overlapping positions

Everything remained the same except the RSI threshold.

That means RSI below 1 was one backtest.

RSI below 2 was another.

Then 3, 4, 5, and so on until RSI below 50.

In total, we examined 50 different definitions of “oversold.”

The SPY data covers February 1993 through September 2026.

Backtest report dashboard detailing SPY RSI(2) oversold strategy performance metrics. — Strategy performance report displaying a 76.8% win rate and equity growth for SPY RSI.

At First, RSI 1 Appeared to Win

When we ranked the tests by profit factor, one result immediately jumped off the screen.

RSI below 1 produced a profit factor of 6.56.

That sounds extraordinary. But there was a problem. It generated only 13 trades during more than three decades.

That is nowhere near enough evidence to conclude that RSI 1 is some magical threshold.

If the sample is tiny, one or two trades can dramatically change the result.

So rather than declaring RSI 1 the winner, we looked for a region where three things came together:

A strong profit factor
A reasonable number of trades
Similar results at neighboring RSI levels

That led us somewhere very different.

The Practical Sweet Spot Was Much Higher Than 1, But Much Lower Than 30

The strongest practical area appeared roughly between RSI 5 and RSI 16.

Here are some representative results:

Threshold comparison table detailing RSI levels, trade counts, and win rates. — Backtest results table evaluating strategy efficiency across multiple RSI threshold parameters.

The pattern is more important than any one number.

As we loosened the definition of oversold, we got more trades.

But the quality of those trades generally deteriorated.

That is exactly what we would expect if very low RSI readings contain information that ordinary “RSI below 30” readings do not.

RSI 9 Was the Full-Sample Standout

Among thresholds with a meaningful sample size, RSI below 9 produced one of the strongest combinations.

The test generated:

233 trades
76.8% winners
0.59% average gain per trade
Profit factor of 2.59
Maximum drawdown of 10.9%
Only about 10% time in the market

Compare that with the traditional RSI 30 threshold.

RSI below 30

613 trades
70.3% winners
0.33% average gain per trade
Profit factor of 1.84
Maximum drawdown of 13.0%
27.4% time in the market

RSI 30 produced far more signals.

But those signals were weaker.

The average trade at RSI 9 was approximately 81% larger than at RSI 30.

That is a substantial difference.

Comparative backtest report panel analyzing RSI 9, RSI 15, and RSI 30 thresholds. — Strategy performance comparison chart highlighting RSI 9 as the optimal full-sample efficiency threshold.

Does That Mean RSI 9 Is the “Best” RSI Level?

Not necessarily.

This is where the results become more useful.

RSI 9 maximized one type of efficiency.

But a trader who wants more opportunities might reasonably prefer something around RSI 14 to 16.

Consider RSI 15:

383 trades
74.9% winners
0.47% average trade
Profit factor of 2.34
Maximum drawdown of only 9.4%

Compared with RSI 9, you get about 64% more trades.

The individual trade is weaker, but the overall profile remains strong.

This suggests there is no single magical oversold number.

Instead, there appears to be a historical sweet spot.

Very low RSI levels provide stronger selectivity. Slightly higher thresholds provide more trading opportunities.

Once we move toward the traditional 30 level and beyond, the extra trades increasingly appear to dilute the edge.

The Most Important Lesson: “Oversold” Is Not Binary

Trading textbooks often present RSI as though there were two states:

Above 30: normal

Below 30: oversold

Our results suggest that this loses useful information.

An RSI reading of 29 and an RSI reading of 4 may both technically be “oversold” under the traditional definition.

Historically, however, they have not behaved the same way.

The deeper SPY became oversold, the stronger its short-term mean-reversion tendency generally became.

That can also be seen without using the strategy rules at all.

We simply grouped RSI readings and measured SPY’s subsequent returns while SPY was above its 200-day moving average.

Statistical analysis table comparing RSI ranges with multi-day forward returns. — Backtest results table evaluating short-term forward returns across different RSI ranges.

The strongest five-day return occurred after RSI readings between 5 and 10.

Again, this points toward the same general area.

Forward return statistics table and bar chart evaluating SPY performance across different RSI ranges. — Statistical analysis report demonstrating how deeper oversold RSI readings improve short-term mean reversion returns.

But Could This Just Be an Optimized Exit?

This is an important objection.

Perhaps RSI 9 looks unusually good because it happens to fit our particular exit rule.

So we changed the exit.

Then we changed it again.

And again.

The exact winning threshold moved around, which is what we would expect.

But the important part did not change.

Low RSI thresholds continued to perform well.

For example:

Three-day holding period

RSI 5 produced:

133 trades
Profit factor 2.50

Five-day holding period

RSI 6 produced:

164 trades
Profit factor 2.94

RSI 8 produced:

211 trades
Profit factor 2.62

RSI 9 produced:

238 trades
Profit factor 2.57

Exit when RSI(2) rises above 50

RSI 6 produced:

163 trades
Profit factor 2.77

RSI 9 produced:

241 trades
Profit factor 2.54

Exit when RSI(2) rises above 70

The results became even stronger.

RSI 8 produced:

205 trades
Profit factor 3.51

RSI 9 produced:

229 trades
Profit factor 3.48

The exact optimum changed depending on the exit.

But the broader conclusion survived.

The strongest thresholds repeatedly clustered well below RSI 30.

That is more convincing than finding one isolated optimized number.

Robustness analysis report comparing different exit types and RSI thresholds. — Strategy robustness dashboard demonstrating that low RSI threshold efficiency persists across various exit rules.

What Happens If We Remove the 200-Day Moving Average?

The 200-day moving average filter was another possible source of bias.

Maybe the RSI effect only exists because we excluded bear markets.

So we removed the filter.

RSI below 9 without the 200-day moving average produced approximately:

357 trades
Profit factor 2.15
0.58% average trade
Maximum drawdown of 14.6%

The result became weaker.

But it did not disappear.

That suggests the 200-day moving average improves the quality of the setup, but it does not create the entire RSI effect.

This makes intuitive sense.

Short-term mean reversion has historically behaved differently depending on the broader market regime.

Buying an oversold pullback during a long-term uptrend is not the same situation as buying an oversold market during a major bear trend.

Then We Found Something Even More Interesting

Up to this point, you might conclude that RSI 9 is simply the best level.

But when we split the backtest into different periods, the story changed.

1993 to 2004

The strongest practical area was around RSI 8 to 9.

RSI 9 produced:

75 trades
Profit factor 4.90
88% winners

That is an exceptionally strong historical result.

2005 to 2014

The edge became weaker.

RSI 9 produced:

65 trades
Profit factor 1.97

Still profitable, but nowhere near the first period.

2015 to 2026

Now something changed.

The strongest region shifted upward.

RSI 14 produced:

142 trades
Profit factor 2.37

RSI 16 produced:

152 trades
Profit factor 2.29

The strongest practical area in the recent period appeared closer to RSI 14 to 18.

Temporal analysis chart showing how optimal RSI oversold levels shifted across time periods. — Backtest summary report detailing the historical evolution of optimal RSI thresholds from 1993 to 2026.

The “Best” RSI Oversold Level Is Not Static

This is perhaps the most important finding in the entire test.

If we optimized the full history and stopped there, RSI 9 would look like the obvious answer.

But markets change.

The SPY market of the 1990s is not identical to the SPY market of the 2020s.

Liquidity changes.

Market structure changes.

Volatility changes.

The speed at which traders react changes.

So it should not surprise us that the exact threshold changes as well.

The robust conclusion is therefore not:

RSI 9 is the magic number.

A better conclusion is:

Historically, SPY’s strongest RSI(2) mean-reversion zone has generally appeared much deeper into oversold territory than the traditional RSI 30 level.

And within that broad zone, the best exact threshold has varied over time.

So Is RSI 30 Wrong?

No.

That would be too strong a conclusion.

RSI 30 still produced profitable historical results in this test.

It also generated far more trades.

But if the objective is to identify unusually stretched short-term conditions in SPY, RSI 30 appears to be a relatively loose definition of oversold.

The difference is substantial.