Who created the Correlation Trend Indicator?

John Ehlers, who published it in 2020 as a way to measure trend by correlating price against an ideal straight line.

What lookback should I use?

The default is around 20 bars. The window defines the trend scale being measured: half-period-of-interest choices are common, and readings from different lookbacks routinely disagree because they measure different swings.

What counts as a strong reading?

A common convention treats magnitudes above roughly 0.5 as a meaningful trend in the window. Sustained readings pinned near +1 or -1 mark unusually clean trends and are sometimes followed by consolidation.

Does a high CTI mean the trend will continue?

No. It certifies that the recent window was linear, nothing about the future. It is best used to select which playbook applies, not as a standalone entry signal.

Correlation Trend Indicator

The Correlation Trend Indicator (CTI) is John Ehlers' answer to a deceptively simple question: how much does recent price look like a straight line? It is a Trend concept. The Library holds 1 implementation, ready to run on Vela.

Ehlers CTI

Top Correlation Trend Indicator indicator

The top custom implementation, built on the original standard Correlation Trend Indicator formula.

1 total

Open the Correlation Trend Indicator implementation below on Vela, then have Quant, our coding agent, turn it into a strategy you can backtest.

Open in Vela

Correlation Trend Indicator

Indicator

What is the Correlation Trend Indicator?

The Correlation Trend Indicator (CTI) is John Ehlers' answer to a deceptively simple question: how much does recent price look like a straight line? Introduced in 2020, it computes the Pearson correlation between price and an ideal rising line over a lookback window. The output lives between -1 and +1: readings near +1 mean price has tracked a steadily rising line, readings near -1 mean a steadily falling one, and readings near zero mean the window contains no linear drift worth the name.

The elegance is that trend is defined by fit rather than by displacement. A market can be far above an old level yet trending badly (one gap followed by chop scores poorly), while a modest but relentless grind scores near the extremes. Because correlation is scale-free, thresholds transfer across instruments and timeframes far better than slope-based measures, whose units depend on price level and volatility.

CTI is closely related to the R-squared trend fit: squaring CTI gives the R-squared of the same linear regression, discarding the sign. Traders use CTI when they want strength and direction in one bounded number, and R-squared when direction is handled elsewhere. Like every windowed statistic, it is entirely backward-looking: a perfect +1 describes the last n bars, not the next n.

How it's calculated

CTI is the Pearson correlation between the source series and a straight rising line (the bar index) over the lookback:

Sx = SUM_(i=0..n-1) C_(t-i)

Sy = SUM_(i=0..n-1) y_i

Sxx = SUM_(i=0..n-1) C_(t-i)^2

Syy = SUM_(i=0..n-1) y_i^2

Sxy = SUM_(i=0..n-1) C_(t-i) * y_i

CTI_t = (n*Sxy - Sx*Sy) / sqrt((n*Sxx - Sx^2) * (n*Syy - Sy^2))

t: current bar index

n: lookback period in bars (default 20)

i: offset into the window, 0..n-1

C_(t-i): source value i bars ago, usually close

y_i: the ideal-line value paired with offset i, chosen so the line rises as bars get more recent (e.g. y_i = -i)

CTI_t: correlation value at bar t, bounded between -1 and +1

Orient the index so that a rising market yields positive CTI; implementations differ on whether they correlate against i or -i and flip the sign accordingly.

The result is identical to the signed square root of the regression R-squared over the same window.

How traders use it

As a regime gate: only take trend-following entries when CTI is above a positive threshold (commonly around +0.5) or below the negative mirror, and stand down or switch to mean-reversion logic near zero.
As a direction-plus-quality score in one number, replacing the two-step check of an average's slope plus a separate strength measure such as the ADX/DMI system.
Zero-line crossings are sometimes traded as trend-change signals; these lag the actual turn by a meaningful fraction of the window, so they suit confirmation rather than timing.
Its limitations are structural: a single large gap inside the window distorts the correlation, V-shaped reversals pass through zero late, and the reading says nothing about the magnitude of the drift, only its consistency.

CTI vs other trend-strength measures

R-squared trend fit: R-squared is CTI squared: same regression, direction discarded, output 0 to 1. CTI keeps the sign, so one series carries both strength and direction.

ADX/DMI system: ADX measures trend strength from smoothed directional movement and is famously slow to rise and fall. CTI is a direct windowed fit with a hard bound at 1, typically snappier but noisier at short lookbacks.

Kaufman efficiency ratio: The efficiency ratio scores net displacement against path length, so one big bar can dominate it. CTI scores linearity of the whole path, so it rewards steady grinds over jumpy ones.