CCI Indicator: Formula, ±100 Signals, and What Tests Reveal

The Commodity Channel Index (CCI) measures how far a price sits from its recent average, relative to the spread of the same prices. Traders often watch +100 and −100 for strength or extreme conditions.

Our 144,936 configuration evaluations explore lookbacks, thresholds, and three entry rules. The calculation explains an important boundary: with only two periods, CCI cannot reach +100 at all. At twenty periods, the same threshold has a very different meaning.

What does the CCI indicator measure?

A positive CCI puts the latest source price above its rolling average. A negative CCI puts it below. Larger magnitudes mean a greater distance relative to the window’s mean absolute deviation.

Despite its name, CCI can be calculated for markets beyond commodities. A twenty-period example often uses typical price, the average of a candle’s high, low, and close. Platforms can offer other sources.

TradingView’s CCI guide explains the conventional calculation and the use of extreme readings for either trend strength or potential reversal. Those interpretations need separate trading rules.

How to calculate CCI

For a window of N source prices:

Typical price = (high + low + close) / 3
Mean          = average of the N typical prices
Mean deviation = average absolute distance from that mean

CCI = (latest typical price − mean)
      ÷ (0.015 × mean deviation)

The deviation uses each price’s distance from the same current window mean. It is not the average of historical one-bar changes or a standard deviation.

For two typical prices of 100 and 110, the mean is 105 and the mean absolute deviation is 5. CCI is 5 ÷ (0.015 × 5), or about 66.67.

Use 100 and 100.10 instead: the mean and deviation both shrink, leaving CCI at the same 66.67. The oscillator normalizes the move; 66.67 does not mean a 66.67% price increase.

Why two-period CCI cannot reach +100

With two unequal source prices, the latest value is always one mean-deviation unit above or below the mean. Standard CCI is therefore approximately +66.67 or −66.67. If the prices are equal, mean deviation is zero and the ratio needs a defined missing-value convention.

There is a more general limit when the latest observation belongs to the same calculation window:

Maximum absolute CCI = N / 0.03

The mean-centered deviations sum to zero. Their positive and negative totals must balance, so the total absolute deviation is at least twice the absolute newest deviation. Dividing by N and applying the 0.015 scale gives that bound.

Every point above uses older prices of 100 and a newest price of 110, changing only the window length. The construction reaches the bound. At two observations it is 66.67; at three it is exactly 100. A rule requiring CCI strictly greater than 100 cannot qualify with three or fewer observations.

CCI is commonly called an unbounded oscillator because it has no fixed 0–100 range like RSI. That description does not remove the lookback-dependent limit of this particular same-window formula.

How the window changes a historical reading

The boundary is useful beyond a constructed example. A shorter CCI is not simply a quicker version with an otherwise identical scale.

The two-period curve stays near ±66.67 when its source values differ, while the twenty-period calculation can move beyond ±100. These are locally calculated indicator values, not backtest execution markers.

A high reading can persist as price keeps trending. A falling CCI can also result from the window average and deviation changing. It does not automatically mean the price itself fell.

Does the 0.015 constant keep 70–80% of values inside ±100?

That is a familiar description of the scaling convention, including in Fidelity’s CCI guide. It is not a distribution-free guarantee.

At a twenty-period lookback, about 60–62% of the checked historical values fell between −100 and +100 across our samples. Those observations include trending and ranging periods. The measured share belongs to those data and that formula.

Changing the 0.015 constant just rescales CCI. If a constant is multiplied by two, the same values are halved; halving the numerical threshold preserves the original inequality. Treating those paired changes as separate evidence would count the same decision twice.

Strength, fade, and recovery are different rules

The tests use a shared zero-line exit and three entries:

• Strength: buy while CCI is strictly above the positive threshold; exit below zero.
• Fade: buy while CCI is strictly below the negative threshold; exit at or above zero.
• Recovery: first require a strict below-threshold observation, then buy on recovery above it while still below zero; exit at or above zero.

In the displayed comparison, the strength rule recorded 2,388 executions at four periods, 674 at twenty, and 152 at one hundred. Neighboring lengths also produced occasional small increases in activity. Length changes the mean and mean deviation together, so it is more than a speed adjustment.

The threshold stays fixed in these additional tests using small positions. The three rules respond to different events, so their execution counts can separate even on the same indicator values. A recovery that jumps directly above zero is rejected because its exit already has priority.

The script also disables entries at thresholds at or beyond the exact bound. Those settings are valid cash outcomes by construction. Their lack of trades is not fresh empirical proof of the mathematics.

The full evidence preserves return and drawdown summaries as well as activity. The chart teaches what each rule selects; it does not identify a universally profitable CCI setting.

CCI versus RSI and useful settings

RSI summarizes smoothed upward and downward price changes on a fixed 0–100 scale. CCI compares a price level with a rolling mean and mean absolute deviation. A reading of 70 on one is not equivalent to 70 on the other.

Choose lookback and threshold together. Shortening the window can alter the available scale, the rate of extreme readings, and how often a recovery can happen. A conventional twenty-period setting is a starting point for a defined comparison, not a reason to assume ±100 has the same behavior everywhere.

Reproduce the comparison

The script defaults to twenty-period typical-price CCI and a +100 strength entry. Switch the rule to fade or strict recovery while keeping the same indicator and threshold.

//@version=6
strategy("CCI Lab: thresholds, continuation and recovery", overlay=false, initial_capital=10000, default_qty_type=strategy.percent_of_equity, default_qty_value=99, process_orders_on_close=false)

cci_length = input.int(20, "CCI lookback", minval=2, maxval=100)
threshold = input.int(100, "Absolute entry threshold", minval=25, maxval=325)
rule_mode = input.int(0, "Rule: 0 strength, 1 fade, 2 recovery", minval=0, maxval=2)

cci_value = ta.cci(hlc3, cci_length)
recovery = ta.crossover(cci_value, -threshold) and cci_value[1] < -threshold
// The newest observation belongs to the same mean-deviation window.
// A strict excursion cannot exceed the bound N / 0.03.
feasible = threshold < cci_length / 0.03 - 0.000001
ready = bar_index >= 500 and not na(cci_value) and feasible
enter_long = ready and (rule_mode == 0 ? cci_value > threshold : rule_mode == 1 ? cci_value < -threshold : recovery)
exit_long = ready and (rule_mode == 0 ? cci_value < 0 : cci_value >= 0)

plot(cci_value, "CCI", color=color.blue)
plot(threshold, "Upper threshold", color=color.orange)
plot(-threshold, "Lower threshold", color=color.green)
plot(0, "Mean", color=color.gray)
if exit_long and strategy.position_size > 0
    strategy.close("Long")
if enter_long and not exit_long and strategy.position_size == 0
    strategy.entry("Long", strategy.long)

Compare the fixed scale of RSI or the standard-deviation normalization of Bollinger Bands.