RSI Indicator Explained: How to Use It for Crypto Trading
RSI is a momentum oscillator scaled 0–100 that compares average gains to average losses over a lookback period, usually 14. Readings above 70 are conventionally overbought and below 30 oversold — but in crypto those thresholds misfire often enough that many traders shift them to 80/20 and never trade RSI without a trend filter.
RSI is the most widely used and most widely misused indicator in crypto. The misuse is nearly always the same mistake: treating an overbought reading as a sell signal. This article covers what the calculation actually measures, why crypto's volatility breaks the textbook thresholds, and the four ways RSI is used by people who do not lose money with it.
- RSI measures the ratio of average gains to average losses — not price, and not value.
- In a strong trend RSI can hold above 70 for weeks. Shorting that is the most common way to lose money with RSI.
- Period length is a noise/lag trade-off: 7–9 for scalping, 14 as default, 21–25 for position trading.
- Divergence is the highest-quality RSI signal and the one most damaged by evaluating unclosed candles.
- RSI is a filter and a confirmation, not a standalone entry trigger.
What does the RSI indicator actually measure?
RSI — the Relative Strength Index, published by J. Welles Wilder in 1978 — compares the size of recent gains to the size of recent losses and expresses the result on a 0–100 scale.
RS = average gain over N periods / average loss over N periods RSI = 100 − (100 / (1 + RS))
Two properties follow directly from that formula and explain most RSI misuse:
- It is bounded. It cannot exceed 100 or fall below 0, so it compresses at extremes. A market getting twice as strong does not double the reading.
- It is relative to its own recent history, not to any notion of value. RSI 75 does not mean "expensive". It means recent up-moves have outweighed down-moves in this lookback window. That is all it means.
The name causes confusion worth clearing up: this measures an asset's strength relative to its own recent behaviour, not relative to other assets. Comparing coins requires a different tool.
Why do 70/30 levels misfire in crypto?
The 70/30 convention was calibrated on 1970s equity and commodity markets. Crypto is more volatile, trends more persistently, and runs 24/7 with no session structure to mean-revert around.
The consequence: RSI can sit above 70 for weeks during a real trend, and every reading is a "sell signal" that would have been wrong. Shorting overbought in an uptrend is the single most reliable way to lose money with this indicator, and it is what most introductory material implicitly teaches.
Two adjustments help:
Shift the thresholds. Many crypto traders use 80/20 instead of 70/30, which cuts false extremes materially at the cost of fewer signals.
Shift them by regime. More rigorous: in an established uptrend use 80/40 — overbought at 80, and 40 as the level from which pullbacks tend to resume rather than 30. In a downtrend, 60/20. RSI behaves asymmetrically depending on trend direction, and a symmetric threshold ignores that.
Which RSI period should you use?
Period length is a straight trade-off between responsiveness and noise:
| Period | Behaviour | Suits |
|---|---|---|
| 7–9 | Fast, frequent extremes, noisy | Scalping on 1m–15m |
| 14 | Wilder's default; balanced | Swing trading on 1h–4h |
| 21–25 | Smooth, lagging, fewer signals | Position trading on daily and above |
A shorter period does not give you an earlier signal in any useful sense — it gives you more signals, of which more are noise. Changing the period changes the signal-to-noise mix, not your foresight.
Multi-timeframe RSI: the same coin is overbought and oversold
A single RSI reading is close to meaningless without context. The same asset can read 78 on the 15-minute chart and 42 on the daily. Both are correct; they describe different questions.
The productive arrangement is hierarchical: a higher timeframe sets the bias, a lower timeframe sets the timing. "The 4-hour trend is up and 4-hour RSI is above 50, so I only take longs; the 15-minute RSI dropping below 35 is my entry trigger" is a rule. "RSI is oversold" is not.
This is also why single-timeframe RSI alerts generate so much noise: they fire on a condition that the timeframe above may flatly contradict.
Four RSI strategies that survive contact with the market
1. Trend-filtered mean reversion
Take oversold readings only in the direction of the higher-timeframe trend. In an uptrend, buy RSI dips below 35–40; ignore overbought entirely. This single filter removes the majority of RSI's losing signals, because it removes the trade of shorting strength.
2. The 40–50 bounce
In an established uptrend, RSI often finds support in the 40–50 band without ever approaching 30. Waiting for a textbook oversold reading means never entering a strong trend. Treating the 40–50 zone as the pullback area gets you into trends that never look "cheap". Mirror-image in downtrends: rejection from 50–60.
3. Divergence
Price makes a lower low, RSI makes a higher low — the move is losing momentum. Bearish divergence is the inverse. This is the highest-quality signal RSI produces and the one most sensitive to implementation, for reasons covered below. Full treatment in setting up RSI divergence alerts.
4. RSI at a level
An oversold reading at a structural support level is a materially better entry than either signal alone. RSI supplies timing; structure supplies location. Neither is sufficient by itself, which is the general shape of every durable use of this indicator.
The mistakes that cost the most
Trading RSI alone. It is a momentum measure, not a system. Without a trend filter or a location, it produces a stream of signals with roughly coin-flip resolution.
Shorting overbought in an uptrend. Covered above. It deserves repeating because it is the dominant loss mode.
Using one timeframe. Guarantees you will regularly take trades the timeframe above contradicts.
Acting on an unclosed candle. The subtle one, and the most expensive. An RSI value computed on a forming candle can move substantially before the bar closes — a divergence visible at minute 4 of a 15-minute candle often does not exist at minute 15. Evaluate on closed bars only. Why divergence alerts fire at the worst possible time works through a concrete case.
Assuming defaults are optimal. 14 periods and 70/30 are conventions inherited from a different asset class in a different decade, not results.
Automating RSI without automating the mistakes
Manual RSI monitoring across dozens of pairs and several timeframes is not realistic, and the cases you miss are not random — you miss the ones that happen while you are asleep or busy.
Automating it correctly means encoding the filters above rather than just the threshold. A rule worth running looks like a conjunction: 1-hour RSI below 30 and 4-hour trend up and no high-impact news in the last few minutes — not "RSI below 30".
TradeFloor evaluates conditions like these on a 60-second cadence with up to five AND/OR conditions per rule, across all configured pairs, on closed candles only — the forming bar is stripped before any indicator is computed, with no fast-mode option to turn that off. RSI divergence is available as a condition in its own right, and every rule can be combined with structure, order-flow and news filters.
Frequently asked questions
What is a good RSI setting for crypto trading?
The 14-period default is a reasonable starting point, but the 70/30 thresholds misfire in crypto because trends persist. Many traders use 80/20 instead, and a more precise approach shifts thresholds by regime — 80/40 in an uptrend, 60/20 in a downtrend. Use 7–9 periods for scalping and 21–25 for position trading.
Does RSI above 70 mean I should sell?
No, and treating it that way is the most common way to lose money with RSI. In a strong crypto uptrend RSI can stay above 70 for weeks, and every one of those readings would have been a losing short. An overbought reading means momentum is strong, which in a trend is confirmation rather than a reversal signal.
What is the best RSI period for crypto?
It depends on your timeframe, and the choice is a noise-versus-lag trade-off rather than an accuracy one. 7–9 periods for scalping on 1m–15m charts, 14 as a balanced default for 1h–4h swing trading, and 21–25 for daily and above. A shorter period gives more signals, not earlier ones.
Can RSI be used on its own as a trading strategy?
Not reliably. RSI measures momentum relative to its own recent history and says nothing about trend direction or price location. Every durable use pairs it with something else — a higher-timeframe trend filter, a structural support level, or a divergence condition. Alone it produces roughly coin-flip signals.
What is RSI divergence and is it reliable?
Divergence is when price makes a new extreme and RSI does not — price makes a lower low while RSI makes a higher low, suggesting the move is losing momentum. It is the highest-quality signal RSI produces, but it is very sensitive to implementation: computed on an unclosed candle it can appear and disappear within a single bar.
Why does my RSI value differ between platforms?
Usually the smoothing method. Wilder's original uses a specific smoothed moving average, while some platforms use a simple or exponential average instead, producing slightly different values from identical data. Whether the current candle is included also matters — a platform computing on the forming bar shows a value that will change before close.
How to Set Up RSI Divergence Alerts for Crypto Trading
Regular vs hidden divergence, which settings to use, and how to turn an alert into an executable rule.
// questions or corrections · [email protected] · more essays · /blog