Home › Learn › Lesson 22

Momentum inertia — one school’s reading of RSI extremes as fuel, not exhaustion

By TradingPrimer · Published 1 Sep 2026

Every beginner is taught that RSI above 70 means overbought and RSI below 30 means oversold. There is an entire school of chart reading, taught in our own slide course, that throws both words away and reads a high RSI as the opposite of a warning — as evidence of fuel. This lesson lays that reading out honestly, gives it the strongest version of its own case, and then does something nobody who teaches it does: prices it. Along the way, one small piece of algebra turns RSI from an abstract oscillator into a headcount you can do on your fingers, and one modelled correction shows that the school’s own invalidation rule does not fire until the entire impulse has already been given back.

What does “momentum inertia” actually claim?

That a market which has just proved one side is overwhelming the other does not reverse that proof in a single move — so a high reading is a statement about fuel, not about exhaustion.

The claim has a specific trigger. In an up-move, RSI must close a candle at 80 or above; in a down-move, at 20 or below. When that happens, this school treats the wave as having inertia: after the correction that follows, it expects another wave back to at least the old high (or, in a downtrend, back to at least the old low). The practical payoff is that you get to draw the next wave before the correction has finished, instead of waiting for the correction to prove itself.

You have already met the ingredient this rests on. In lesson 15 we established that RSI is a ratio — average gain over average loss, squashed onto a 0–100 scale — and that RSI 80 corresponds to a 4:1 ratio. What inertia adds is a direction of reading. The classical school looks at 4:1 and says stretched. This school looks at 4:1 and says one-sided, and then asks how a market that one-sided is supposed to flip in a candle or two.

Both readings are interpretations. Neither is established by the number. What follows gives the inertia reading the strongest version of its own case, and then measures what obeying it costs — because the cost is the part nobody teaching it puts in writing.

What does an RSI close above 80 actually require of the candles?

Under one simplifying assumption, something surprisingly plain: eleven of the last fourteen closes have to be up.

Here is the piece of algebra that makes RSI legible. Suppose, just for translation, that every candle in the lookback window has the same size m. If u of the last 14 closes were up and the rest were down, then the average gain is u m / 14 and the average loss is (14−u) m / 14. The ratio is u / (14−u), and when you push that through the RSI formula the candle size cancels completely:

RSI = 100 × u ÷ 14

The oscillator becomes a headcount. Not a metaphor for one — literally the percentage of the last fourteen closes that were up, expressed as a number between 0 and 100. So:

Two honest caveats before anyone takes this to the chart. Real candles are not the same size, and the standard RSI uses Wilder’s exponential smoothing rather than a flat 14-candle average, so this is a translation device, not a measurement. It tells you what kind of stretch a reading corresponds to. It does not replace the indicator.

With that said, look at what the table does to the vocabulary. “Eighty percent of the market is buying” is a thing people say about RSI 80, and it is not true — RSI knows nothing about how many participants there are. But 80% of the last fourteen closes were up is true, exactly, under the equal-candle assumption. The folklore was reaching for the right shape and grabbing the wrong noun.

How wide is the band you are told to stand aside in?

Two point eight candles. That is the whole of it.

This school treats 40–60 as a zone with no information in it: buying pressure and selling pressure are close enough to even that the frame has no opinion, and the correct action is to look at a different frame or do nothing. Run that band through the headcount and it spans from 5.6 up closes to 8.4 up closes out of fourteen — a width of 2.8 closes.

That number is worth sitting with, because it explains something and warns about something.

It explains why so much of chart time is spent inside the band. Three ordinary candles going the same way can carry RSI across the entire no-opinion zone. Sideways price does not need to be dull for RSI to be dull; it just needs the up and down closes to be roughly balanced, which is the definition of the range state you learned to diagnose in lesson 16.

It warns that the boundaries are not cliffs. The difference between an RSI that closes at 80 and one that closes at 76 — between “inertia confirmed” and “nothing happened” under this school’s rules — is 11.2 closes against 10.64. That is 0.56 of a single candle. This is exactly the pattern lesson 18 found in the “wick at least twice the body” rule: a threshold stated to two decimal places, sitting inside a band of noise several times wider than the argument about it. Anyone insisting on 80.0 rather than 79.4 is insisting on half a candle.

What are the four stages, and which one are you actually trading?

Four, and the trade lives in the third one — which is the stage that looks least like an opportunity.

The version of this taught in our slide course puts a fast and a slow moving average on the RSI line itself rather than on price, and reads the cycle from how those two behave. The frame is worth learning even if you never trade it, because it is one of the few chart methods that names its stages explicitly instead of leaving them to be inferred:

Two details in that table do real work.

Stage 3 requires the averages to re-open, not merely to cross. A cross that immediately un-crosses is noise; the cycle only counts when the two lines separate and stay separated for a few closes. This is the same distinction lesson 14 drew about crossovers on price: the cross is an event, the separation is the evidence.

Stage 4 is a checkpoint, not a finish line. The interesting question at the completion is not “did it get back to the high” but “did it print a new stage 2 on the way”. If the return to the old high fails to close RSI above 80 again, this school reads the trend on that frame as having spent itself — and treats it as a reason to take profit rather than to look for another entry.

How long does the invalidation rule take to fire?

Long enough that it is a post-mortem rather than a stop. This is the finding that changes how you should use the whole method.

The scenario has to be retirable, and this school gives a rule for retiring it. Two conditions, and it insists on both:

• RSI closes clearly below 40 during the correction; and
• price closes below the 0.5 retracement of the impulse.

The reasoning for the “and” is sound: either signal alone fires on noise, and demanding two independent kinds of evidence filters that out. But nobody who teaches the rule has measured how unequal the two conditions are. Let us do it.

Set up a model impulse. Price runs from $60,000 to $70,000 on the four-hour chart, and at the high, RSI closes at exactly 80. That last fact is not decoration — it pins down the candle size. From the headcount identity, RSI 80 means 11.2 up closes and 2.8 down closes in the window, so the net travel is 8.4 candle-lengths. Ten thousand points over 8.4 candles gives an average candle of $1,190. Everything below is derived from that, not assumed.

Now the correction. The 0.5 retracement sits at $65,000, which is $5,000 below the high — so at $1,190 a candle, price closes below it on the fifth down close. Getting RSI from 80 to below 40 is a slower business, because Wilder’s smoothing gives each new candle only one fourteenth of the weight. Simulating it with down candles of that same $1,190 size takes ten closes. And ten closes at $1,190 is $11,900, which puts price at $58,100 — below where the impulse began.

That is not a quirk of one correction speed. Vary how violent the correction is and the picture holds:

Read the last column first. Across every speed from a slow grind to a violent flush, the RSI half of the rule does not confirm until price has given back somewhere between 89% and 143% of the impulse. In four of the six rows it does not confirm until price is below where the whole move started.

Three things follow, and they are the practical core of this lesson.

The “and” is almost never binding. The price condition trips at a 50% retracement by construction. The RSI condition needs 89% or worse. So in practice the rule is “RSI below 40” with a retracement pre-filter that has already been satisfied by the time anyone checks it. The second condition is not doing the filtering work it appears to do.

This is not a stop-loss and must never be used as one. A stop placed where this rule confirms is a stop 89% to 143% of the impulse away — on our numbers, between $8,900 and $14,300 of risk on a $10,000 move. Any entry priced against that is a fraction of an R. Use the rule for what it is: a way to decide that a scenario is finished, so you stop looking for entries in that direction. Your actual stop goes somewhere structural, and much closer.

The slowness is deliberate, and you can now price it. A rule that fired at the 0.5 retracement would kill the scenario on every ordinary pullback — which is precisely what the method exists to sit through. The lag is the feature. What the table adds is the invoice: between 3 and 6 extra closes of holding an idea that the price condition already called dead, depending on how fast the correction runs. Whether that is worth paying depends on how often those corrections resolve upward, which is a question your own journal can answer and this page cannot.

If the scenario is right, what does the trade actually pay?

Against the old high, no more than 1.00R — and the ceiling is set by the school’s own validity rule, not by anything the market does.

Put the pieces together. The scenario stays alive only while the correction holds above the 0.5 retracement. So the deepest entry you are permitted to take is at f = 0.5. And from the identity established in lesson 21 — entry at retracement level f, stop at the impulse origin, target at the prior high — reward-to-risk is f ÷ (1 − f), and break-even win rate is exactly 1 − f. Which gives:

This is the collision worth understanding. The rule that keeps the scenario alive is the same rule that caps its reward. The deeper entries that pay well — the 0.618 and 0.786 that made lesson 21’s table look attractive — are exactly the ones this method says invalidate the wave. You cannot have the good price and the live scenario at the same time; the validity condition removes the bottom half of the grid.

So what are you actually betting on? Not the return to the old high — that is only a coin flip at best. The trade only becomes an ordinary 2R proposition if the completion prints a genuinely new high, half an impulse above the old one: $75,000 on these numbers. And notice that this is precisely the claim the inertia reading makes. The method is internally consistent. It is asking you to bet on the strong version of its own thesis, and the arithmetic makes that explicit instead of leaving it in the fine print.

There is a refinement in the same course material that fits here neatly. If RSI corrects too deep during the impulse — below 40 in a rising wave — the completion is expected to weaken into merely a test of the old high rather than a break of it. Price that downgrade: a completion that stops at $70,000 instead of extending 20% beyond to $72,000 takes the 0.5 entry from 1.40R down to 1.00R, moving break-even from 41.7% to 50.0%. The downgrade costs 8.3 percentage points of win rate. That is what a piece of qualitative guidance looks like once it is denominated in something you can act on.

What does waiting for confirmation cost here?

Between 24.5 and 35.7 percentage points of break-even win rate — the highest bill this site has yet put on the word “confirmation”.

The temptation with a four-stage cycle is to wait for stage 4 before committing: let the RSI averages cross back up and re-open, then buy. It feels like discipline. Model it. Say the correction bottomed at the 0.5 level and it takes three average up closes for the RSI averages to flip and separate. Three closes at $1,190 puts your entry at $68,570, and your target is still $70,000.

Waiting for the cycle to confirm costs 24.5 points of break-even win rate even when you also tighten the stop, and 35.7 when you do not. No confirmation signal buys that much accuracy. Which yields the single sentence most worth taking off this page: inertia is a scenario tool, not an entry trigger. Its whole purpose is to let you commit earlier — during the correction, when the chart looks wrong — because you drew the next wave in advance. Using it as a trigger inverts the reason it exists.

This is now the third time the site has priced the same instinct, and the three numbers together are more useful than any one of them:

The pattern across all three: confirmation is cheap when it costs you a small fraction of the move (the candlestick case) and ruinous when the confirming event is most of the move (the neckline and the stage-4 cases). That is a rule you can apply to any confirmation signal anyone ever sells you — measure how much of the intended profit has already happened by the time it fires.

How does this compare with the standard 30/70 reading?

They disagree about what to do, not about what is measured — and each is strongest exactly where the other is weakest.

It would be dishonest to present the inertia reading as a correction of a mistake. The classical overbought/oversold reading is not stupid; it is a mean-reversion strategy, and mean reversion is a real property of ranging markets. Set the two side by side:

Notice that the failure modes are mirror images, which means the two are not really competitors: they are the same tool assigned to two different market states. That makes the question “which reading is right” a bad question, and “which state am I in” the question that actually decides. Which is lesson 16 again, arriving from a new direction — and it is why this school insists you diagnose the state before you look at any signal at all.

The inertia school does add one thing the classical reading does not have: an explicit rule against acting on a signal in the wrong state. “RSI sitting inside 40–60 means this frame has no opinion” is a rule that removes the range trades from a trend method. The classical reading has no equivalent guard, which is why beginners taught only 30/70 spend their first year shorting things that keep going up. That failure is documented on this site as one of the most common ways year one ends.

When is everything on this page wrong?

Four conditions, and the first is the one you will meet this week.

In a range. Everything here assumes an impulse with a start, an end and a correction. Inside a box, RSI rarely closes beyond 80 at all — and if it does, on the day the box finally gets stretched, the “inertia” you record belongs to a move that has nowhere to go. This method produces its worst signals at the moment a long range ends, which is also the moment it looks most exciting.

When the frames disagree. The version of this taught in our slide course asks, in a bull market, that every higher frame from the daily upward produce its own bullish inertia. That is an extremely demanding condition and it is worth saying plainly that it will almost never be fully satisfied. A more usable form is the one from lesson 17: an inertia reading on a frame above yours is permission, a reading on your frame is the setup, and a reading only on frames below yours is noise. Reading an inertia signal from a frame you are not trading is the same cost that lesson measured at up to 79.6% of the trade.

When you read it on an unclosed candle. The whole rule is built on closes, and a live candle showing RSI at 81 is not a signal — it is a candle that has not happened yet. This is not pedantry: a reading four hours from settlement can move several points in either direction, and the entire distance between “confirmed” and “nothing” is 0.56 of a close.

When the claim itself is not tested. The strongest version of this school’s case includes a high success rate for completed inertia cycles. We do not publish that number, here or anywhere, because we cannot reproduce it and neither can you — it depends entirely on how the cycle is defined, which frame it is measured on, and which market regime the sample covers. What this page has done instead is give you the arithmetic that surrounds the claim, so that if you decide to test it, you already know what hit rate the trade needs: 50.0% at the 0.5 entry against the old high, 61.8% at the 0.382.

Common mistakes

Frequently asked questions

Does RSI 80 mean the market is overbought and about to fall?

It means the average up-move over the lookback is four times the average down-move, and nothing at all about what happens next. The headcount translation makes that concrete: under equal-sized candles, RSI equals 100 times the share of the last 14 closes that were up, so RSI 80 is 11.2 up closes out of 14 and RSI 40 is 5.6. That is a description of the recent past, not a forecast. Two schools read the same description in opposite directions — the classical one sees stretch and looks for a reversal, the inertia one sees one-sidedness and plans for continuation after a correction. Neither reading is proved by the number. The number is a measurement; everything else is an interpretation laid on top of it, and knowing which is which is most of what separates a trader from a person repeating a rule.

How wide is the 40 to 60 band that traders are told to stand aside in?

Two point eight closes out of fourteen. RSI 40 is 5.6 up closes and RSI 60 is 8.4, so the entire no-opinion zone spans 2.8 candles. That cuts two ways. It explains why chart time pools inside the band — three ordinary candles going the same way cross the whole thing — and it warns that the edges are not cliffs. The difference between a close at 80 and a close at 76, which under this school’s rules is the difference between a confirmed signal and nothing whatsoever, is 0.56 of one candle. Anyone drawing a hard line at 80.0 is drawing it inside noise several times wider than the distinction they are making, which is the same problem lesson 18 found in the “wick twice the body” rule.

How long does the inertia invalidation rule take to actually fire?

Long enough that it is a post-mortem rather than a stop. It requires RSI closing below 40 and price closing below the 0.5 retracement, and those two conditions are wildly unequal in speed. On a modelled impulse from $60,000 to $70,000 with RSI closing at exactly 80 — which pins the average candle at $1,190 — the 0.5 level at $65,000 breaks on the fifth down close, while RSI needs ten closes and does not confirm until price is at $58,100, which is 119% of the impulse retraced and below where the whole move began. Vary the correction from a slow grind to a violent flush and the RSI condition still waits until 89% to 143% has been given back. So: use it to decide a scenario is finished and stop hunting entries in that direction; put your real stop on structure, and put it much closer.

If the inertia scenario is right, what does the trade actually pay?

Against the old high, at most 1.00R — and that ceiling comes from the method’s own validity rule. Because the correction must hold above the 0.5 retracement, the entry can be no deeper than 0.5; and from the identity in lesson 21, an entry at level f with the stop at the impulse origin and the target at the prior high pays f ÷ (1 − f), which at 0.5 is exactly 1.00R and breaks even at 50%. The shallower entry this school calls better evidence pays less — 0.618R at the 0.382 level, needing 61.8%. The setup only becomes an ordinary 2R trade if the completion prints a new high half an impulse above the old one, $75,000 on these numbers. Which is worth knowing before you take the trade, because it names precisely which claim your money is on.

Educational content only — not financial advice, and not a trade recommendation. Every figure on this page comes from worked models built for this lesson and each can be reproduced from the numbers given: an impulse from $60,000 to $70,000 with RSI closing at exactly 80 at the high, a 0.5 retracement at $65,000, and entries at $65,000, $66,180 and $68,570 against stops at $60,000 and $64,400. Three identities do the work. First, under the assumption that every candle in the lookback has the same size, RSI = 100 × u ÷ 14 where u is the number of up closes — a translation device, not a measurement, since real candles differ in size and standard RSI uses Wilder’s exponential smoothing. Second, that same assumption fixes the average candle at $1,190, since RSI 80 implies 11.2 up and 2.8 down closes and therefore a net travel of 8.4 candle-lengths across a $10,000 impulse. Third, for a retracement entry at level f with the stop at the impulse origin and the target at the prior high, R:R = f ÷ (1 − f) and break-even win rate = 1 − f, carried over from lesson 21. The invalidation table was produced by simulating Wilder smoothing forward from an RSI of 80 with constant-size down candles at each stated multiple, and the stage-4 entry assumes three average up closes for the RSI averages to cross and separate — a stated modelling assumption, not a measurement. No success rate for completed inertia cycles is quoted anywhere on this page, because the figure taught with the method cannot be reproduced without knowing its exact test design. The rules themselves — the 80 and 20 close thresholds, the four stages, the fast and slow averages placed on RSI, the 40–60 no-opinion band, the two-part invalidation condition, and the reading of a too-deep correction as a downgrade to a test of the high — follow the classical rules taught in our own slide course, and are presented here as one school’s reading rather than as established fact. Sources: our own arithmetic, stated inline. Published 1 Sep 2026.