Divergence in Depth: Regular, Hidden and Why Most Go Nowhere

Reading the title carefully: “Most Go Nowhere” and the frequency wording in the headings are cautions, not a verified majority statistic. This lesson provides no defined-market study, hit rate or outcome horizon. It teaches comparison labels and their limitations; no reversal or continuation probability is certified.

Before you read: RSI and Stochastic Oscillators: What Overbought Really Means

What you will learn

• What regular divergence and hidden divergence describe.
• Why the swing points you choose decide whether a divergence exists.
• Why divergences can persist for a long time, and what they cannot show.

What is divergence?

If price makes a new high but an indicator does not, what is different between the two highs?

Divergence compares specified price swing points with indicator readings at specified times. A difference between those comparisons can be labeled without establishing future direction. The oscillator formula, price source, bar interval, pivot selection and when confirmation becomes available all matter. This lesson uses strict higher/lower inequalities; some authors use looser flat-point conventions. State the convention rather than silently changing it after seeing the outcome.

Regular and hidden divergence

Which swing points are compared in each type?

All times below are fictional session bar-close times in CST. Prices are invented points and supplied RSI(14,close) readings are invented inputs on a 0–100 scale from one unchanged configuration, not values claimed to be recomputed from a full historical price series. We compare RSI at the price-pivot timestamps, not arbitrarily chosen independent indicator pivots.

Worked regular bearish comparison — Session A: 09:05 CST high 100, RSI 70; 09:25 CST high 104, RSI 60. Price change 104−100=+4 points, while RSI change 60−70=−10 oscillator points. Higher price high/lower RSI reading satisfies the stated regular-bearish comparison. It does not mean a short position is indicated.

Worked hidden bullish comparison — separate Session B: 10:05 CST low 100, RSI 40; 10:25 CST low 102, RSI 30. Price change 102−100=+2, RSI 30−40=−10: a higher price low with a lower reading. This satisfies hidden bullish, not regular bullish. For regular bullish the second price low would have to be lower, not higher.

Near counterexample: Keep Session B price lows 100→102, but supply RSI 40→40. Equality is not a lower low under our strict rule, so hidden bullish is not satisfied. With 40→41 it also fails. A suggestive drawing cannot replace both inequalities. Labels describe the chosen observations; they do not measure trader intent or validate the names’ predictive implications.

Divergence diagram

Pivot timestamps locate observations; right-hand confirmation times locate when the comparison can be known. A later higher high is retained, not erased. Readings are supplied fictional inputs, not a reconstructed RSI history.

Why does swing selection matter?

If you pick a different earlier high, does the divergence disappear?

Use a rule whose availability can be recorded. For this example a price high pivot must be strictly higher than each of the two bars immediately before and two immediately after it; low pivots reverse those comparisons. Bars are five minutes. A point cannot be confirmed by this rule until two later bars close — ten minutes later.

These are the complete neighboring high/low inputs needed for the two worked pairs:

For A’s second pivot the two left highs are 98 and 103; both below 104. Its two right highs 103 and 102 are also below 104. The comparison using that second pivot is available at 09:35 CST, not 09:25. B’s second low 102 is below its left 104/103 and right 103/104, so that comparison is available at 10:35 CST.

An indicator that draws a marker back on 09:25 after confirming it at 09:35 has placed the marker at the pivot’s historical location; it has not supplied information at 09:25. A live bar’s RSI can change before close. Bar-close settings do not eliminate every revision, higher-timeframe or backplotting issue. Different swing rules/settings can yield different labeled pairs; choosing them after the outcome is hindsight.

Why do so many divergences go nowhere?

How many lower highs on RSI can a strong trend produce?

There is no frequency estimate here for “so many” or “most.” A formula comparison alone does not state whether, how far or when price will reverse. A slower indicator reading at a higher high can coexist with further price increases.

Completed non-reversal counterexample: In Session A, after the second high is confirmed at 09:35 CST, the next bar at 09:40 CST has a high of 106. That is 2 points above 104. The described divergence existed under the specified rule, yet price made another higher high. This single invented path demonstrates logical possibility, not how frequently reversals fail. Defining a reversal horizon and outcome before testing is necessary to ask a statistical question.

What this does not tell you about divergence

• Whether any divergence will be followed by a reversal.
• Which swing points or settings are correct. That is a choice.
• How long a divergence can persist.
• Nothing here suggests buying, selling or holding when a divergence appears.

Check yourself

1. Classify lows 100→102 with RSI 40→30, and show both differences. Hint: compare lows, and which one is higher.
2. When is the Session A second pivot usable under the stated rule?
3. Would RSI 40→40 qualify in the hidden example? What does the later 106 high establish?
4. Two readers label divergence differently on the same record. Name two choices that can explain the disagreement without assuming one has found a future reversal. Hint: swing choice and settings.
5. Why can a continuing strong price trend display several labeled divergences rather than reverse after the first? Hint: what does an oscillator measure?

Last reviewed November 8, 2026. Spotted a mistake? Report an error.