Does RSI below 30 work in crypto? 10 coins tested
After RSI first fell below 30, 10 Binance coins rose a day later 54.27% of the time, but the average move matched normal hours. See fees and years.
After the first 1-hour close with RSI below 30, price was higher 24 hours later in 54.27% of 4,452 cases across 10 Binance perps. For all hours it was higher 50.60% of the time. But the average move after RSI fell below 30 was +0.25%, almost the same as the +0.23% for all hours. So price rose a bit more often, and not by more. RSI above 70 was not followed by falls on average either. These are past cases, not a forecast.
What is RSI, and what do 30 and 70 mean?
RSI stands for relative strength index. It is a number from 0 to 100. It compares the size of recent up moves with recent down moves, usually over the last 14 bars. A bar is the price record for a fixed span, such as one hour.
If most recent bars fell, RSI is low. If most rose, it is high. Many traders call RSI below 30 "oversold" and above 70 "overbought". Those are only labels. We tested whether they were followed by something unusual.
What counts as an event?
An event is the first bar that closes with RSI below 30, after the bar before was at 30 or higher. RSI often stays low for many bars. If we counted each one, a single sell-off would look like dozens of cases. So we keep the first event and skip new ones until the forward window has passed. We call this de-clustered.
We ran the test on 10 Binance USDT perps (BTC, ETH, SOL, XRP, DOGE, ADA, AVAX, LINK, BNB, TRX) from September 2019 to October 19, 2026. A perp is a futures contract that never expires.
What happened after RSI fell below 30?
"All bars" is the average move after every bar of the same coins and bar size, over the same span. The last column is a long position's average trade after a 0.05% taker fee on each side (paid when an order trades right away), 0.10% in total. It counts price moves only.
| Bars and look-ahead | Events | Higher | Average move after | All-bars average | After fees |
|---|---|---|---|---|---|
| 1-hour bars, 24 hours later | 4,452 | 54.27% | +0.25% | +0.23% | +0.15% |
| 4-hour bars, 48 hours later | 1,363 | 54.73% | +0.86% | +0.45% | +0.76% |
| Daily bars, 7 days later | 205 | 57.07% | +2.48% | +1.68% | +2.38% |
Price was higher more often than the all-bars share on 1-hour bars (54.27% against 50.60%). But for 1-hour bars the average gain matched a normal hour. The daily and 4-hour averages look better. Still, the 205 daily events fall on only 106 different days, because several coins hit RSI 30 in the same sell-off. The gap between +2.48% and +1.68% is smaller than the noise we measured for it (1.46 percentage points), so we cannot tell it from luck. For why a few hundred clustered cases are not enough, see how many trades a backtest needs.
Did it work in both halves of the data?
No. We split events by date at January 1, 2023.
| Bars | Period | Events | Higher | Average move after | All-bars average | After fees |
|---|---|---|---|---|---|---|
| 1-hour, 24 hours | 2019-2022 | 1,896 | 55.49% | +0.39% | +0.34% | +0.28% |
| 1-hour, 24 hours | 2023-2026 | 2,556 | 53.37% | +0.16% | +0.14% | +0.06% |
| 4-hour, 48 hours | 2019-2022 | 610 | 55.41% | +1.48% | +0.69% | +1.38% |
| 4-hour, 48 hours | 2023-2026 | 753 | 54.19% | +0.36% | +0.28% | +0.26% |
| Daily, 7 days | 2019-2022 | 93 | 59.14% | +4.87% | +2.63% | +4.77% |
| Daily, 7 days | 2023-2026 | 112 | 55.36% | +0.49% | +0.98% | +0.39% |
The later period is weaker in all three. On daily bars, the 2023-2026 average (+0.49%) was below the average for all days (+0.98%). The headline gain after fees (+2.38% on daily bars) comes mostly from 2019-2022. We looked at 3 bar sizes, 3 look-ahead lengths for each, 2 sides and 3 periods, so some cells will look good by chance.
Does it differ by coin?
Yes, a lot. Here are daily bars, 7 days later, for each coin. The "all days" column is that coin's middle 7-day move over every day.
| Coin | Events | Higher | Middle move after | Middle move, all days |
|---|---|---|---|---|
| BTCUSDT | 24 | 58.33% | +1.69% | +0.44% |
| ETHUSDT | 20 | 50.00% | +0.07% | +0.50% |
| SOLUSDT | 18 | 61.11% | +3.67% | −0.07% |
| XRPUSDT | 19 | 52.63% | +2.51% | −0.54% |
| DOGEUSDT | 26 | 65.38% | +2.40% | −0.94% |
| ADAUSDT | 27 | 62.96% | +2.19% | −0.68% |
| AVAXUSDT | 25 | 36.00% | −3.32% | −0.50% |
| LINKUSDT | 15 | 46.67% | −3.25% | +0.21% |
| BNBUSDT | 17 | 70.59% | +2.27% | +0.46% |
| TRXUSDT | 14 | 71.43% | +4.82% | +0.80% |
Bitcoin had only 24 daily events in seven years. Each coin has between 14 and 27, so one coin's row is a short list, not a rate. AVAX and LINK went the other way. On 1-hour bars, 8 of the 10 coins were higher more often than their own all-hours share after 24 hours. BTC and TRX were not.
What happened after RSI went above 70?
Many people expect a fall. We did not find one. Here is the mirror test, with the average trade for a short position (a bet on a fall) after the same fees.
| Bars and look-ahead | Events | Higher | Average move after | All-bars average | Short, after fees |
|---|---|---|---|---|---|
| 1-hour bars, 24 hours later | 4,704 | 48.21% | +0.64% | +0.23% | −0.74% |
| 4-hour bars, 48 hours later | 1,629 | 52.18% | +1.79% | +0.45% | −1.89% |
| Daily bars, 7 days later | 354 | 55.65% | +7.52% | +1.68% | −7.62% |
After an overbought reading, the average move was above normal at every bar size. The daily average is pulled up by a few huge runs: the middle result was +1.23%, not +7.52%. On 1-hour bars, fewer than half were higher (48.21%), yet the average was above normal. A few large rises outweighed many small falls.
What does this data not show?
- It does not show that RSI causes anything. It is a summary of past prices. For a different way to measure a sell-off, see Bitcoin after 5 red days in a row, and for another classic indicator, the golden cross and death cross.
- It does not cover other RSI lengths or cutoffs.
- The coins are ones still trading today. Coins that faded away are missing.
- Fee examples leave out funding (a small regular payment between perp traders), spread and slippage.
- Events from different coins cluster in the same sell-offs, so they are not independent tests.
How we measured this
We used Binance hourly, 4-hour and daily bars for 10 perps, from September 8, 2019 to October 19, 2026. RSI is the standard 14-bar version on closing prices. Forward moves start at the close of the event bar, and an event counts only if every bar in its window exists. We compared each result with every eligible bar of the same coin and bar size. Costs are 0.05% per side. Splits use the event date.
Try this in Comparand
Connect Comparand to your AI app and ask:
- "Get completed 4-hour bars for Binance ETHUSDT from 2020 to today. Work out the 14-bar RSI, list each first close below 30, and show the move 12 bars later." The venue_bars tool returns the open, high, low and close of each completed bar at 1-hour, 4-hour or daily size, and it is free. Your agent works out the RSI. The tool does not.
- "Find the 10 closest past charts to Binance ETHUSDT's last 48 hourly bars." The shape_analogs tool compares the whole shape of a drop with past cases of the same coin on the same exchange, instead of one indicator. Each past case shows its own path and its own move 12, 24 and 48 bars later, in bp (100 bp is 1%). It costs $0.01 per search and gives no average or hit rate.
You can try 3 coins free, read the docs, or see pricing. To judge any test like this one, read about overfitting with 1,000 random rules.
The short version
- After the first 1-hour RSI close below 30, price was higher a day later 54.27% of the time (4,452 cases) against 50.60% for all hours.
- The average move matched normal hours: +0.25% against +0.23%.
- Daily results were strong in 2019-2022 (+4.87%) and weak in 2023-2026 (+0.49%).
- After RSI above 70, the average move was above normal, not below.
- Past cases, not a forecast.
Questions people ask
Does RSI below 30 work in crypto?
Only weakly in our test. After the first 1-hour close below 30 on 10 Binance perps, price was higher 24 hours later in 54.27% of 4,452 cases, against 50.60% for all hours. But the average move was +0.25%, close to the +0.23% for all hours. These are past cases, not a forecast.
What happened after RSI went above 70?
Price did not fall back on average. After the first daily close above 70, the average 7-day move was +7.52% across 354 cases, against +1.68% for all days. The middle result was only +1.23%, so a few very large rises pulled the average up.
Did the oversold bounce change over the years?
Yes. On daily bars, the average 7-day move after RSI first fell below 30 was +4.87% in 2019-2022 (93 cases) and +0.49% in 2023-2026 (112 cases). In 2023-2026 that was below the +0.98% average for all days.
Did a rule that went long when RSI fell below 30 make money after fees?
It made small averages before funding and spread. After 0.10% in round-trip fees, the average trade was +0.15% on 1-hour bars (24 hours), +0.76% on 4-hour bars (48 hours) and +2.38% on daily bars (7 days). Most of the daily gain came from 2019-2022.
What RSI settings did you use?
A 14-bar RSI on closing prices, the common setting. We did not test other lengths or other cutoffs than 30 and 70.