RSI Divergence — Spotting Reversals (Class A/B/C + Hidden Divergence)

Every trend reversal starts the same way: before price turns, the force pushing it weakens. The problem is you can't see that with the naked eye — candles at the top of a bull run look just as confident as the ones halfway through the move. RSI divergence is an attempt to measure that fading strength: price is still making a new extreme, but the momentum indicator is already refusing to cooperate.

It's one of the most valuable signals in technical analysis — and, at the same time, one of the most abused. In this article we cover the full toolkit: bullish and bearish divergence, classes A/B/C, hidden divergence — and something most courses stay silent on: why this signal fails repeatedly in a strong trend, and what to do about it.

What Divergence Is — Anatomy of a Mismatch

RSI (Relative Strength Index) measures the speed of price changes on a 0-100 scale; the standard is 14 periods. Normally the indicator tracks price: new price highs mean new RSI highs. Divergence is the moment that agreement breaks.

Bullish divergence (regular): in a downtrend, price prints a lower low, but RSI at the same time makes a higher low. Interpretation: the decline continues, but its pace is fading — sellers are losing steam. This signal is looked for near the end of downtrends, ideally at strong support.

Bearish divergence (regular): the mirror image — price makes a higher high, RSI makes a lower high. The advance continues on momentum's residual strength, but that momentum is running out. Looked for near the end of uptrends, at resistance.

A workshop detail that decides signal quality: you draw divergence using RSI extremes that correspond to price extremes — you compare low to low and high to high, at the same points along the time axis. Connecting random points on the indicator is the easiest way to "find" divergence everywhere.

Classes A/B/C — Not Every Divergence Is Equal

A classic taxonomy splits divergences into three strength classes. Using the bullish variant as the example (the bearish variant mirrors it exactly):

Class A — strongest. Price makes a clearly lower low, RSI a clearly higher low. A full-blooded mismatch: the market made a new low on unambiguously weaker momentum. This is the only class worth treating as the basis of a setup.

Class B — moderate. Price forms a double bottom (the second low roughly equal to the first), while RSI is already rising. Momentum is improving, but the market didn't make a new low — a weaker signal, because it isn't clear supply was actually fully tested.

Class C — weakest. Price makes a lower low, RSI is flat (the lows sit at the same level). The mismatch is barely noticeable — noise more often than signal. Note Class C as context; don't trade it on its own.

This split is a quality filter worth a single glance: before analyzing a setup, check the class. Most "divergences" that tempt on low timeframes are class B and C — and they're what's behind the opinion that "divergences don't work."

Hidden Divergence — Continuation, Not Reversal

Alongside regular divergences there are hidden ones — and confusing them with the regular kind flips a trade's direction by 180 degrees.

Hidden bullish: in an uptrend, price makes a higher low (a healthy pullback), but RSI prints a lower low. The indicator "overreacted" to the pullback more than price did — the uptrend has more fuel left. A continuation signal: look for a long in the direction of the trend.

Hidden bearish: in a downtrend, price makes a lower high, RSI a higher high. The pullback looks stronger on the indicator than it does on price — the decline is likely to resume.

A memory rule: regular divergence sits on extremes that agree with the trend and signals its end; hidden divergence sits on pullbacks and signals the trend's return. Regular divergences are played counter-trend (which is why they demand the most confirmation); hidden ones are played with the trend, which makes them a statistically friendlier setup for less experienced traders.

How to Play Divergence Step by Step

Divergence is a filter, not a trigger. Here's the sequence that turns a warning into a setup:

1. Context: trend and level. Look for regular divergence at the end of a clear move, where price reaches a significant spot — a support/resistance zone, a supply/demand zone, a previous high. A divergence "in the middle of nowhere" isn't a setup.
2. Identify it and grade the class. Compare the price and RSI(14) extremes. Class A — keep analyzing; B/C — watch, don't trade.
3. Confirmation from price. This is the most important step. Wait for a rejection candle (an engulfing candle, a pin bar) at the level, or a break of local structure on a lower timeframe in the reversal direction. Divergence alone, with no reaction from price, is still just a hypothesis.
4. Enter after the confirming candle closes. Not before — entries taken "because there's already a divergence" are knife-catching with a scientific-sounding justification.
5. Stop loss beyond the extreme that created the divergence (below the low for a bullish setup, above the high for a bearish one), with a buffer for wicks.
6. Target: the nearest opposing level. Be realistic — divergence signals a correction or a reversal, it doesn't guarantee a new trend. The first support/resistance level is a fair first target; require a minimum of 2:1 against the stop.

A numerical example (illustrative): BTC in an uptrend on the H4 reaches a resistance zone at $104,000-$105,000. Price makes a higher high at $104,800 (the previous one was $103,900), RSI prints 64 against 71 on the previous high — class A. Two candles later, the H4 closes with a bearish engulfing candle, and the M15 breaks local structure to the downside. Short entry at $103,600, stop at $105,300 (above the high with a buffer, risking $1,700), target at $99,800 near the previous support ($3,800 of profit, RR ~2.2:1).

A note from our own work: in a pattern study on our own crypto data (a pattern ablation study across a large trade sample), bearish RSI divergence used as a short filter turned out to be the only statistically significant winner (+0.121R per trade, p=0.003). A single result on one dataset isn't universal proof — but it's consistent with the classic thesis that divergence works best as a filter for another signal, not as a standalone system.

An Honest Look: Why Divergence Fails in a Strong Trend

This is the section most divergence material skips — and it's the one that decides your bottom line.

Divergence measures fading momentum, not its reversal. That distinction sounds academic until you see it on your P&L: in a strong trend, momentum can fade, recover and rebuild — repeatedly. A BTC parabola in euphoria can print three or four bearish divergences in a row, with price going higher after every single one. Each of those divergences was "real" — momentum genuinely was fading — and each one, played as a short, was a loss.

Practical conclusions:

• Divergence against a strong trend is the most expensive version of this signal. The stronger the trend (rising volumes, no deep pullbacks, high ADX), the more confirmation you should require — or simply pass and wait for a structure break instead.
• Divergence doesn't expire on its own — price invalidates it. If, after a divergence, the market makes another extreme with RSI higher than before, the old signal is canceled. Don't "average into" a position because "the divergence is still there."
• On low timeframes, noise mass-produces divergences. M5/M15 in consolidation will generate a mismatch every dozen or so candles. The best signal-to-noise ratio comes from H4 and daily.
• In consolidation, divergence is worthless. Every bounce in a range creates a technical mismatch — with no trend, there's nothing to reverse.

Common Pitfalls

• Trading the divergence alone, without a level and confirmation. That's not an entry signal — it's a raised eyebrow. Price gives you the entry.
• Confusing regular with hidden. The mistake flips the trade's direction. Regular = trend extremes, reversal; hidden = a pullback, continuation.
• Force-fitting divergence lines. Connecting extremes that don't actually correspond, stretching a line through the middle of the indicator — after the fact, everything fits. Compare high to high, low to low.
• Changing the RSI period "for better signals." Shortening the period multiplies false mismatches; lengthening it loses real ones. RSI(14) is a sensible standard; if you want to change it, test both versions on data first.
• A stop tight against the extreme with no buffer. Divergence extremes are natural stop-hunt targets — a wick that pushes the low a fraction of a percent deeper right before the reversal is a classic.
• Ignoring the regime context. The same divergence in a consolidation, a weak trend, and a parabolic bull run are three different signals with three different success rates. Before you play it, name the regime you're in.

RSI divergence illustrates a general truth about indicators well: a signal's value doesn't come from its mere existence, but from the context and rules around it. Class A at a strong level, with confirmation from price and a stop beyond the extreme — that's craft. Shorting every mismatch in a bull run because "the indicator showed it" — that's a subscription fee paid to more patient traders. The difference isn't in RSI; it's in you.

FAQ

Rafał — Krypto Bez Ściemy

Trader and founder of Strefa Tradingu. He’s been breaking crypto down on YouTube for years — no hype, no signals, with a focus on market structure and risk management.

kryptobezsciemy.plInstagramYouTubeKlub Zero