Commodity Channel Index: Measuring How Far Price Strays From Average

Almost every oscillator on a standard charting platform is bounded. RSI runs 0 to 100. Stochastics run 0 to 100. Williams %R runs 0 to -100. You learn to read those scales as maps with edges, where the interesting territory is near the border. The Commodity Channel Index is not built that way. It divides a distance by a dispersion, and there is no arithmetic reason the result cannot print +340 or -410. That single structural fact is why traders who arrive at CCI from RSI keep misreading it, and why the conventional ±100 lines are something quite different from the ±100 lines they look like.

Quick Answer

Everything below is the detail that paragraph compresses: the arithmetic run over six bars you can check on a phone, what the 0.015 is actually doing and why that number and not some other, the three readings the same price move can produce depending only on how noisy the window was, the exact difference from RSI, and the two signal conventions most tutorials state backwards.

What Is the Commodity Channel Index?

A single line in its own pane under price, oscillating around a zero line, with horizontal reference lines conventionally drawn at +100 and -100. Zero is not an arbitrary midpoint the way 50 is on a bounded scale. Zero means the current typical price is sitting exactly on its own moving average, which is a real event you could verify by eye on the price chart above.

Donald Lambert published it in 1980 in Commodities magazine, which is now Futures magazine. The original design target was cyclical turns in commodity markets, which is where the name comes from and where most of the confusion about the indicator starts. Nothing in the formula is commodity-specific. It consumes a high, a low and a close, and any instrument that prints those three feeds it identically, which is why it ships as a built-in on equity, index, futures and currency charts, and why the overwhelming majority of people running it today have never traded a commodity contract. The name is an accident of where it was first published. Treat it the way you would treat any other name that stuck.

The premise underneath is that price tends to return toward its own average, and that the useful information is not the raw distance from that average but the distance graded against how far this instrument normally strays. A stock that routinely swings two dollars away from its 20-day average is not doing anything unusual when it swings two dollars away from it again. A stock that has been hugging that line for a month is. CCI is a way of asking that question with one number, and the grading step is the part that makes it genuinely different from the bounded oscillators sitting next to it in the same menu.

What Is the Commodity Channel Index Formula?

Step three is the step people get wrong, and the absolute value is the reason. Deviations from a mean always sum to zero. That is what a mean is. If you skipped the absolute value and averaged the signed deviations you would get 0.00 every single time, on every chart, forever. The absolute value is what turns a set of distances that cancel into a usable measure of how spread out the window was.

It is also worth noting what mean deviation is not. It is not the standard deviation. Standard deviation squares the distances before averaging and then takes a square root, which weights outliers more heavily. Mean deviation just averages the raw distances. Some platforms have quietly substituted standard deviation into this slot, which produces a systematically different denominator and therefore a different line, and it is one of the reasons two charts of the same ticker can show two different CCI values. Both the StockCharts ChartSchool entry on the Commodity Channel Index and the TradingView help page for CCI specify mean deviation, and that is the version the ±100 convention was built around.

Here is the whole thing over six bars with a six-period lookback, shortened from 20 so every sum stays checkable by hand. The prices are made up, picked to make the arithmetic visible rather than to describe any real security.

Mean deviation = 2.80 / 6 = 0.4667. The current typical price is 25.40 and the SMA is 24.60, so the numerator is 0.80. CCI = 0.80 / ( 0.015 × 0.4667 ) = 0.80 / 0.0070 = 114.29. Just over the +100 line, which under the conventional reading is the moment the chart becomes interesting rather than the moment it becomes a trade.

Note the signed column. It sums to exactly 0.00, as it must, which is the sanity check that tells you your SMA is right before you go any further.

What the 0.015 is doing there

The constant is the least explained part of the indicator and the easiest to reason about. It is a pure scaling factor with no market meaning, chosen so that the numbers coming out land in a range people find readable. Reference sources state the target directly: roughly 70 to 80 percent of CCI values fall between -100 and +100.

Work backwards and you can see where it comes from. Setting CCI to exactly 100 means the numerator equals 100 × 0.015 × Mean Deviation, which is 1.5 × Mean Deviation. So the ±100 lines are not thresholds in price terms at all. They are the points where typical price has strayed exactly one and a half mean deviations from its own average. If you were to assume price deviations were roughly normally distributed, mean absolute deviation runs to about 0.80 of a standard deviation, so 1.5 mean deviations is about 1.2 standard deviations, and a symmetric 1.2-sigma band captures a bit under 77 percent of a normal distribution. That is the 70 to 80 percent figure, and it is not a coincidence.

Two honest caveats on that. Price deviations are not normally distributed, they have fatter tails than the assumption allows, so the real-world hit rate drifts around the theoretical one depending on the instrument and the regime. And the whole calibration is tied to the default 20-period window. Change N and you change the typical size of the mean deviation, which changes how often the line reaches ±100, which means the inherited threshold is no longer describing the same rarity. That is the part worth carrying forward when you go fiddling with settings, and it is the same trap that catches people tuning any default-period indicator down for intraday charts.

Why Does the Scale Have No Ceiling?

Because nothing in the formula caps it. A bounded oscillator gets its ceiling from its construction. RSI feeds a ratio through 100 minus 100 over one plus that ratio, and that transformation cannot produce a number outside 0 to 100 no matter what you put into it. Williams %R divides a distance by a range that contains it, so the answer is always between 0 and -100. CCI divides a distance by an average distance, and there is no rule saying the current distance has to be smaller than the average one. It usually is not, in fact, whenever anything interesting is happening.

The practical consequence is that the ±100 lines are a statistical convention rather than a boundary. A CCI of +340 is unusual, and your platform will plot it without complaint. An RSI of 140 is not unusual, it is impossible, and if you see one your data is broken. Those are different kinds of statement and reading the first as though it were the second is the core misunderstanding this indicator produces.

The more useful consequence is that CCI grades every reading against local conditions. Take the exact same 0.80 gap between typical price and its average from the worked example above, and drop it into three different windows.

Same move. Same distance from the average. Three verdicts spanning from "nothing to see" to "rarer than the harder band." That is the indicator working exactly as designed, and it is also the answer to the question people ask when they see a big number: what does +250 actually mean. It means price is 3.75 mean deviations from its own average, because +100 is 1.5 mean deviations and the scale runs linear from there. It does not mean the move was large in dollars, and it does not mean the same thing on the next chart you open.

That normalization is the clean structural split from every bounded oscillator already on this site. A reading of 82 on the volume-weighted 0 to 100 scale of the Money Flow Index means the same arithmetic thing on a $4 biotech and a $900 index tracker, because the scale is fixed by construction. A CCI of 82 does not. The same goes for the 0 to -100 close-position-in-range scale and for the 0 to 1 second derivation that runs a stochastic over the RSI line. Neither approach is better. Fixed scales are comparable across instruments and blind to local conditions. Normalized scales are the reverse. Knowing which one you are holding is the part that matters.

Commodity Channel Index vs RSI

Both get filed under momentum oscillators, both live in a pane under price, both get read against zones at the top and bottom of their range. That is roughly where the similarity ends, and treating them as interchangeable readings of the same underlying quantity is the most common way people end up running two indicators that they believe confirm each other.

RSI measures the ratio of average gain to average loss over its lookback. It is a question about the balance of the moves: have the up-bars been bigger than the down-bars, and by how much. Volume is absent, the intrabar range is absent, and the only price that enters is the close. CCI measures the distance between one typical price and an average of typical prices, divided by the average distance in that window. It is a question about position and dispersion: where is price standing relative to its own centre, and is that unusual for this tape. Those are different inputs producing answers to different questions, and neither is a noisier version of the other.

The cleanest way to see the split is a chart that separates them. A slow, relentless one-way grind, the kind where every bar closes a little higher than the last, gives RSI a huge average gain against a tiny average loss and pushes it deep into overbought. CCI stays moderate through the same move, because the 20-period average is rising underneath price at nearly the same rate, so the gap between the two never widens much. Now invert it: a tight three-week range that breaks with one violent bar. RSI barely notices, because one bar is one period out of fourteen. CCI spikes hard, because the numerator jumped while the denominator is still averaging three weeks of quiet. Same two indicators, opposite reactions, both correct about what they measure.

The row worth sitting with is the second to last. Because CCI normalizes against a local denominator and RSI does not, a CCI value is only comparable to other CCI values on the same instrument and the same timeframe, while an RSI of 72 means the same arithmetic thing wherever you find it. That makes RSI the easier indicator to build a mechanical rule around and CCI the more informative one about whether this particular move is out of character for this particular tape. If you run both, run them for the disagreement, which is the same argument as counting how many genuinely independent signals you have before committing rather than counting how many lines are on the screen.

How Do You Trade a CCI Signal?

Three named conventions, and the first one is stated backwards in most tutorials.

The trigger is the cross back out, not the tag

The conventional rule is not "buy when CCI goes below -100." It is: wait for CCI to drop below -100, then buy when it crosses back above -100. The sell side mirrors it, so you wait for a push above +100 and sell when the line crosses back below +100. The zone entry is the setup. The zone exit is the trigger.

This distinction is doing more work than it looks like. Because the scale is unbounded, entering a zone tells you nothing at all about how much further the line can go. A tag of -100 can be followed by -180, then -260, then -340, and each of those is an ordinary valid reading rather than a stretched one. Requiring the cross back out is what stops you from repeatedly buying into a move that has not finished, which is exactly what a bounded-oscillator instinct pushes you toward.

The ±200 band exists for the same reason. On more volatile instruments the line reaches ±100 often enough that the level stops being rare, and ±200 is the harder-to-reach substitute. There is no rule for when to switch. Look at your own chart, count how often the line visits each level over a few months, and pick the one that is actually uncommon on the instrument you trade.

The zero line gets used as a separate, weaker signal: a cross above zero says typical price has moved back above its own average, which some traders use as a trend-continuation filter rather than an entry. It is the mildest thing CCI says and it is worth treating that way.

Divergence

Bullish divergence: price makes a lower low while CCI makes a higher low. Bearish divergence: price makes a higher high while CCI makes a lower high. The read is that the second push covered less ground relative to its own average than the first one did, so whatever was driving it has less behind it.

Two caveats, and the second one is specific to CCI rather than generic divergence advice. First the generic one: a divergence is a description of something that already happened, with nothing in it to say how long the disagreement may run or which side resolves it, so it is a reason to start watching rather than a reason to act. Now the CCI-specific one. Because the denominator re-centres on every bar, a strong trend can produce divergence after divergence that never resolves: price keeps making new highs, the moving average keeps rising to meet it, so the gap between them stays roughly constant while the absolute price level climbs. Reference sources flag divergences as particularly misleading in a strong trend for exactly this reason. The reliable ones tend to show up where the trend is already losing structure, which is a judgement you make off the price chart and not off the oscillator.

The trend-line-break filter

This is the named technique that separates people who actually use CCI from people who read one article about it, and it is the direct answer to the whipsaw problem. Instead of only watching the ±100 crosses, you draw trend lines on the CCI line itself, connecting its own successive peaks or its own successive troughs inside the oscillator pane. A signal then requires two things: the zone cross, and a break of that trend line.

The logic is that a bare zone cross in a choppy market happens constantly and means almost nothing, while a break of a line drawn across the oscillator's own recent structure means the shape of the oscillator changed and not just its level. The drawing rules are the ordinary ones, so if you are comfortable with anchoring a line to swing points and requiring a clean break rather than a wick through it, you already know how to do this. It is the same skill applied one pane lower.

Worth being plain about the cost. Requiring two conditions rather than one means you enter later and you miss some moves entirely, because the line break often arrives after the best price. That is the trade you are making. Filters do not make signals better, they make them rarer, and whether rarer is better depends on how much the false ones were costing you.

Where the Commodity Channel Index Breaks

None of these are faults. They are what a normalized, unbounded, rolling-window statistic does when its inputs stop behaving the way the construction assumed. Most of them trace back to one thing: the denominator is a moving target, and every reading is a ratio of two numbers that both change every bar.

The first row is the expensive one and every authoritative source on the indicator states some version of it. Reaching an overbought or oversold threshold does not mean a reversal is coming. In a genuine trend CCI can hold beyond ±100 for days, weeks or months, and unlike a bounded oscillator there is no compression against a ceiling to slow it down. The conventional response is to flip the reading: a sustained move past +100 in a confirmed uptrend is treated as strength rather than exhaustion. Which means you need something outside CCI telling you whether you are in a trend at all, and a trend-strength reading that has no opinion on direction is the usual candidate precisely because it answers a question CCI structurally cannot.

Rows two and three are the pair that catches people who have learned to read the number rather than the construction. A large CCI out of a dead tape and a modest CCI out of a violent one are both honest reports. The oscillator is grading against recent norms, and when the norms shift the same price behaviour maps to a different number. If you are going to act on a level, glance at what the price chart was doing over the lookback before you decide what the level means.

Where CCI Fits on a Real Chart

It answers one question and answers it well: is price unusually far from its own recent centre, graded against how far it usually gets. That is genuinely not readable off the price chart at a glance, and it is a different question from the ones the rest of your pane is answering. What it cannot contribute is where the levels are, whether the structure is clean, what the higher timeframe is doing, or how much you should risk. The ordering that works is structure and regime first, then CCI as a stretch check on what those already suggested, which is the same layering the broader technical analysis overview walks through, and the same reason an intraday momentum playbook treats extension as a reason to wait rather than a reason to enter. For the historical framing and the original 1980 attribution, the Wikipedia entry on the commodity channel index is the compact version.

What a screenshot read can and cannot see here

Worth being straight about, since this site sells a tool. SnapPChart has no Commodity Channel Index field. It does not compute a typical price, it does not hold a 20-period simple moving average of one, it does not calculate a mean deviation, and it carries no CCI state at all, the way it does for market structure and trend, breaks of structure and changes of character, liquidity sweeps, fair value gaps and order blocks, the moving average stack, the VWAP relationship, MACD, volume participation, support and resistance, and candlestick patterns. There is no RSI field either, which is worth saying plainly given how much of this post compares the two. What the engine does is read a chart screenshot you upload. So if you plot CCI in a lower pane on your own platform before taking that screenshot, that line is part of what the analysis sees, as a shape drawn beneath the price chart. That is a picture of an oscillator, not a verified computed value: it cannot confirm your period is 20, it cannot tell you the denominator is being held up by one gap bar from three weeks ago, and it cannot know whether your platform used mean deviation or standard deviation to draw it. Which chart states a screenshot-based read genuinely carries and which it only infers from shape is the subject of the wider guide to how AI reads a chart, and a neutral description of what a single chart read covers sits on the AI chart analysis page. If CCI is the reason you are taking the trade, verifying CCI stays your job.