RSI, Drawn Out
RSI might be the most widely used and most widely misread indicator there is. The one-line summary — above 70 is overbought, below 30 is oversold — is so well known that it mostly gets repeated rather than checked. The real data in this lesson shows fairly clearly why that summary is dangerous.
The charts use the same window as the Bollinger Bands and Moving Averages lessons: 200 four-hour BTC/USDT candles from Binance, 3 August to 5 September 2026, computed with the same math the Simulator runs in a backtest.
1. What RSI actually measures
RSI doesn't measure where price is. It measures the balance of movement. Over the last 14 bars it compares the average size of the up moves against the average size of the down moves: when gains dominate it climbs toward 100, when losses dominate it falls toward 0.
The way it averages matters. RSI uses Wilder's smoothing rather than a plain average, so each new bar only shifts the running figure by about 1/14. That's why, in a one-directional stretch, RSI parks near one end of the range and stays there — which is the whole point of section 4.
Over 14 bars, RSI compares the average size of up moves to the average size of down moves, and expresses the result from 0 to 100.
- Price (close)
- RSI 14
09/05: price closed down $101 — RSI 53.1
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
The lower panel is RSI. Press play to follow how each bar's gain or loss nudges it.
2. Where 70 and 30 come from
They're conventions. Wilder proposed them and they stuck as defaults — the market doesn't do anything in particular at 70. Moving the lines just changes how much of the chart you've decided to call extreme.
70 and 30 are conventions. Move them and see how much of the window each one captures.
- Price (close)
- RSI 14
Above 70: 44 bars. Below 30: 0 bars. Out of 186 bars with an RSI value.
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
Drag the lines. Changing the number doesn't change the market, it changes what you've defined as unusual.
3. Shorter period, twitchier reading
Shorten the lookback and RSI hits both extremes far more often. Lengthen it and it barely leaves the middle. More signals doesn't mean the market offered more opportunities — it means the measuring stick got more sensitive.
A shorter lookback swings to both extremes; a longer one barely leaves the middle.
- Price (close)
- RSI 14
At 14 bars, RSI in this window ranges from 33.5 to 94.3.
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
Drag from 2 to 40. The figure below is the actual low and high RSI reached in this window at that setting.
4. Overbought is not a sell signal
If you take one thing from this lesson, take this. RSI above 70 is a statement about the past — recent gains have been strong. It isn't a forecast that price is about to fall. In a strong trend, RSI sitting above 70 for a long stretch is ordinary behavior.
The longest unbroken stretch above 70 in this window, found in the data rather than picked out by hand.
- Price (close)
- RSI 14
RSI stayed above 70 for 22 consecutive bars, 08/19 to 08/23, peaking at 94.3. Over that same stretch price went from $68,554 to $76,928 — so "overbought, expect a pullback" would have been wrong every one of those bars.
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
In this window RSI stayed above 70 for 22 consecutive bars. Price kept rising throughout.
Twenty-two four-hour bars is more than three and a half days. RSI peaked at 94.3 along the way.
Reading that as "overbought, a pullback is due" would have been wrong twenty-two bars in a row. It's the same phenomenon as walking the band in the Bollinger lesson — one market behavior, flagged two different ways by two different indicators.
5. When the rule never fires at all
It's only fair to look at the other side. Take the textbook oversold entry — RSI drops below 30, then comes back up — and apply it to this window.
It never triggers. Across all 200 bars, RSI's lowest reading was 33.5. It never got below 30 at all.
A strict oversold entry needs RSI to drop below the line and come back up. Watch what that produces here.
- Price (close)
- RSI 14
- Entry this rule would take
With the line at 30: zero entries. RSI never got below 33.5 in this entire window — the rule simply never triggered, which is a real outcome, not a broken chart.
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
The dashed line is the lowest RSI actually reached here. Raise the threshold to find where entries start appearing.
That isn't a broken chart or a useless indicator. It's just what this stretch of market was. This is how a backtest ends up reporting zero trades — not because the strategy is wrong, but because the condition it waits for didn't occur in that window. Worth remembering whenever you're looking at a backtest with a low trade count.
6. Divergence
The other common use of RSI is watching for it to disagree with price. When price makes a higher high but RSI makes a lower one, the push that produced the highest price had less force behind it than the earlier push did.
A higher high in price on a lower high in RSI — the strongest push came earlier than the highest price did.
- Price (close)
- RSI 14
08/21: price $77,222, RSI 94.3. 09/03: price $81,755 — higher — but RSI only 76.7. One instance in one window, not a rule.
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
Both points are found in the data — the highest close in the window, and the earlier bar where RSI peaked highest.
Divergence often gets introduced as a reversal signal, and it deserves more caution than that. The observation that momentum weakened is fair; what it doesn't tell you is how long the weaker state persists, or when anything happens as a result. The chart above is one instance in one window.
7. Common misreadings
- Above 70 means recent gains were strong, not that price is expensive. The stronger the move, the longer it stays there.
- 30 and 70 are conventional defaults. Changing them changes how many signals you get, not how accurate they are.
- RSI measures the balance of change, not price level. The same RSI value can occur at wildly different prices.
- No signal at all is a result. The oversold entry in this window is exactly that case.
- Divergence speaks to momentum weakening, not to timing.
8. Seeing it run in the Simulator
RSI shows up in the Simulator in two distinct roles. One template uses it to decide entries and exits outright; two optional filters use it to gate someone else's entries instead — the same indicator generating signals versus suppressing them are very different jobs.
To compare period and threshold combinations rather than guess, the Sweep tool runs a real backtest across every combination in the ranges you pick and ranks the results.
- RSI Mean-Reversion — buys as RSI climbs back out of oversold territory and sells as it falls back from overbought.
- RSI filter — an optional gate on another strategy's entries: skip buys above a given RSI, or only buy above one.
- RSI rising filter — a separate gate keyed to RSI's direction (is it higher than the previous bar) rather than its level.