Polarized Fractal Efficiency (PFE): Formula, Reading as a Trend Filter, and an S&P 500 Example
Hans Hannula’s PFE compares the straight-line distance price traveled over ten bars with the path it actually took, then signs and smooths the ratio; here is the formula, why its scale depends on price units, and what it showed on a year of an S&P 500 ETF.

On this page
Polarized Fractal Efficiency (PFE) measures how efficiently price has moved. Over the last ten bars, it compares the straight-line distance from the old close to the new one with the length of the path drawn by the closes in between. A market that moved in a straight line scores near 100%; one that zigzagged scores near zero. The result takes the sign of the move, up or down, which is the polarized part, and is smoothed with a 5-bar exponential average. Hans Hannula introduced it in Technical Analysis of Stocks and Commodities in January 1994.
Polarized Fractal Efficiency on a daily S&P 500 ETF

- Price, daily candles (hollow up, filled down)
- PFE(10), EMA(5) smoothed (solid line)
How PFE is calculated
Infront’s documentation gives Hannula’s geometric approach. The length of a move is not the price difference but the length of the line joining two points of the chart, with the time axis counted in bars. Over n bars, the direct length is the square root of the squared price change plus n squared. The path is the sum, over the n bars, of the square root of each one-bar price change squared plus one. The fractal efficiency is 100 times the direct length divided by the path. It takes a plus sign if the close is above the close n bars ago and a minus sign otherwise, and the signed value is smoothed by an exponential moving average of z bars. Infront’s defaults are n = 10 and z = 5.
What Polarized Fractal Efficiency compares

The fractal in the name refers to this comparison of a coarse line with the finer path inside it, in the spirit of fractal geometry. It has nothing to do with the five-bar Williams fractals, a separate tool that marks swing highs and lows.
The time term counts one unit per bar in price units, so PFE depends on the price level. On Bitcoin, where a day’s move is hundreds of dollars, the time term is negligible and PFE is close to the net move divided by the sum of the daily moves. On the ETF of our chart, whose median daily change over the window was 3.48 dollars, the time term still weighs. LuxAlgo points out this sensitivity to scale, and Infront’s version has a price scale setting for it. One consequence: with raw prices, a market that barely moves, in steps small against one price unit, reads close to ±100, because the time term dominates; a flat close gives the minus sign.
To see the size of the effect, we ran PFE(10, 5) on the same year of the ETF with every price multiplied or divided by 10 or 100, which changes nothing about the market itself:
Days with PFE(10, 5) beyond ±50, same ETF year, price rescaled

Quoted in units a hundred times larger, the same ETF spent 73.2% of its days beyond ±50 instead of 13.6%. Bitcoin, whose daily moves are worth hundreds of price units, had the same share in all five units: 9.6% of its last 250 days. Compare PFE readings only on one market, in one price unit, on one platform.
How to read PFE
| PFE reading | What it says |
|---|---|
| Near +100 | Price rose in an almost straight line over the last n bars |
| Above +50 | An efficient rise: LuxAlgo treats readings beyond ±50 as trending |
| Around 0 | Price went back and forth |
| Below −50 | An efficient decline |
| Near −100 | Price fell in an almost straight line |
Used as a trend filter, PFE separates efficient moves from choppy ones: a trend rule would only take signals while PFE is beyond a threshold in the same direction. LuxAlgo describes this use as a regime filter; we know of no published test of it. Infront documents a different, contrarian basic trading system with lines at +40 and −40: a buy when PFE crosses −40 upward, closed when it crosses back below −40, and a sell when PFE crosses +40 downward, closed when it crosses back above.
PFE on a real chart
The full chart below the introduction shows PFE(10, 5) on a dividend-adjusted S&P 500 ETF from October 30, 2025 to October 28, 2026. It was above +50 on 34 of the 250 days (13.6%), and never below −50: its low was −47.29 on March 30, 2026. The high was 87.95 on April 20, 2026. The other 216 days, 86.4%, were between −50 and +50, and PFE changed sign 25 times. The last 250 days of Bitcoin, from February 21 to October 28, 2026, spread differently:
Share of days in each PFE(10, 5) zone, last 250 daily bars

- S&P 500 ETF
- Bitcoin
By the ±50 reading, the ETF year had a few efficient rises and no efficient decline, although the ETF lost 8.6% from its close of 687.78 on February 25 to 628.78 on March 30, 2026: the fall came with rebounds, so its path was not efficient enough to pass −50.
Does it work?
We have no long-run measurement of PFE: our bench tests chart patterns detected by rule against random entries, not indicators (see our methodology). A year of two markets is an illustration, not a test.
For other measures of trend quality, see the Trend Intensity Index, the Choppiness Index, the Directional Movement Index and the Rainbow Oscillator, or the full list of trend indicators.
Frequently asked questions
What is Polarized Fractal Efficiency?
An indicator by Hans Hannula that measures how straight the recent price path has been, from −100 to +100, with the sign of the move. High absolute values mean an efficient trend; values near zero mean choppy prices.
What is the PFE formula?
PFE = EMA(z) of 100 × √((C − Cn)² + n²) / Σ √(ΔC² + 1), signed by the direction of the n-bar move. The usual settings are n = 10 and z = 5.
What PFE level means a trend?
LuxAlgo treats readings above +50 or below −50 as trending. That is a convention, not a tested threshold. Infront’s basic trading system uses lines at +40 and −40.
Why does PFE differ between platforms?
Because the time term is counted in price units and some platforms rescale prices first, and because the smoothing may differ. The same chart can show different readings on two platforms.
