WhitmanTrading

What Is a Kagi Chart?

Kagi chart draws a single line that extends while price continues in one direction and turns only when price reverses by a preset amount. Time is not on the horizontal axis at all, so quiet periods occupy no space and the reversal setting decides the entire appearance of the chart.

Most charts put time on the horizontal axis. A Kagi chart does not, and everything unusual about it follows from that one decision.

How to read it

A price series redrawn as a continuous line with sharp reversals.
A kagi chart ignores time completely. Illustrative chart - not real market data.

The line extends in the direction price is moving. As long as price keeps going the same way, the line grows. Nothing else is recorded.

A steady series where the line only turns after a threshold move.
The line only turns after a set reversal. Illustrative chart - not real market data.

It turns only when price reverses by a preset amount — a fixed percentage, a fixed price, or a multiple of average range. Below that threshold, nothing is drawn at all.

A rising series where the line thickens above a prior high.
Thick line when price takes out the prior high. Illustrative chart - not real market data.

Thickness carries the signal. The line is drawn thick when price exceeds the previous high and thin when it breaks the previous low. That switch is the traditional buy-and-sell cue.

A falling series where the line thins below a prior low.
Thin line when it breaks the prior low. Illustrative chart - not real market data.

So the chart is making a structural claim, not an indicator calculation — thick means higher highs are being taken out, thin means lower lows are.

The reversal setting is the chart

A slow series showing how the setting reshapes the drawing.
The reversal size is the only setting that matters. Illustrative chart - not real market data.

One parameter decides everything. A small reversal threshold produces a busy chart with many turns; a large one produces a few clean swings. Same price history, two completely different pictures.

A calm series where the chart barely changes at all.
Change it and you change the whole chart. Illustrative chart - not real market data.

That makes it unusually easy to fit to the past. Adjusting the threshold until the historical chart looks clean is a single-dial optimisation, and a chart optimised that way tells you about the setting rather than about the market.

The honest way to use it is to fix the threshold first, tie it to something measured like average range, and leave it alone.

A worked example

A choppy series where most movement is never drawn.
Small moves never appear at all. Illustrative chart - not real market data.

Take this site’s shared series. Direction runs average 2.01 bars with a longest of 11. Median bar range is 0.493 and the ninetieth percentile is 1.101.

Set the reversal at one median bar — 0.493. Most two-bar runs will not clear it, so the great majority of the series produces no drawing at all. The chart becomes a handful of segments covering months.

Set it at a quarter of that — about 0.12. Now ordinary single bars trigger reversals, the line turns constantly, and the thick-thin switching happens often enough to be meaningless.

Neither is wrong; they are different instruments. And the choice between them is made by a person, before any signal exists.

A falling series with a stop level marked.
A stop fills where the market is, not where you asked. Illustrative chart - not real market data.

The lag is the filter

The thickness change happens after price has already exceeded a prior extreme, which is necessarily after the move that got it there. That delay is not a defect being tolerated — it is the mechanism doing its job.

Every noise filter has the same trade. Refusing to draw small moves means refusing to signal on them, which means signalling later on the ones that matter. You cannot have the calm without the delay; they are the same property viewed from two sides.

The original data

On this site’s shared series: direction runs average 2.01 bars, longest 11. Median bar range 0.493, ninetieth percentile 1.101. A round trip costs 0.0098, about 2% of the median bar.

Those run lengths explain why the reversal setting dominates. With an average run of two bars, any threshold larger than about one bar’s range removes most of the series from the chart entirely — which is either exactly what you wanted or a catastrophic loss of information, depending on what you are trying to see.

A candlestick chart annotated with the cost of a round trip.
And a round trip costs a share of a bar. Illustrative chart - not real market data.
A price series with volume shown beneath.
Volume and price are different measurements. Illustrative chart - not real market data.

And volume is absent. Kagi charts have no volume dimension, because there is no time period to attach it to.

Where it came from and what it assumes

Kagi charts originated in Japanese rice markets, alongside candlesticks, and they encode an assumption worth stating: that price movement matters and elapsed time does not.

That assumption is defensible in some contexts and not others. A market that moves 2% over two days and one that moves 2% over two months produce the same drawing, and whether those two situations are genuinely equivalent depends entirely on what you are trading and why.

For a position held over months it is often reasonable. The path is what matters and the calendar is incidental.

For anything with a deadline it is not. An option expires on a date, financing is charged per night, and a chart with no time axis cannot tell you how many of those have passed. Combining a Kagi reading with any dated instrument means reintroducing the dimension the chart deliberately removed — which is possible, and has to be done deliberately rather than forgotten.

And it pairs badly with a stop. The chart shows no intraday range, so any level taken from it is chosen without knowing how far price actually travelled inside a segment. On this series the largest single bar covered 2.338 against a median of 0.493 — a factor the drawing never displays.

When it fails

The characteristic failure is reading the clean appearance as reliability. A Kagi chart of the past shows a small number of decisive turns with no clutter, and that legibility is easily mistaken for a method that catches turns well. It does not catch them well — it draws only the ones large enough to clear the threshold, after they have cleared it, and omits every failure that did not reach the line. The chart is clean because the misses were never drawn.

A candlestick series with a gap through a level.
A gap skips the level entirely. Illustrative chart - not real market data.

A second failure is tuning the reversal on historical data. One dial, fitted to the past, is the easiest over-fit available on any chart type.

A third is losing track of when a signal actually printed. Without a time axis it is genuinely hard to tell how many sessions passed inside one segment.

A fourth is combining it with time-based indicators, which assume evenly spaced periods the chart does not have.

A declining series cut short at a decision point.
A clean kagi signal. When did it actually print? Illustrative chart - not real market data.

And a fifth is trading it without backtesting the threshold you actually intend to use, since the setting and the strategy are inseparable here.

Market trend covers the higher-high structure the thickness is claiming. Technical analysis covers the tradition it belongs to. And volatility covers the bar sizes the reversal threshold has to be set against.

What I actually do

Kagi charts look decisive in a way ordinary charts do not, and that is the thing to be careful about. The clarity is manufactured by refusing to draw anything below the reversal threshold — so a clean-looking turn may have been printed several bars after the actual high.

— Michael Whitman

This page is educational, not financial advice. Test every idea on your own charts before risking money.