Kaufman adaptive moving average

Perry Kaufman's adaptive average speeds up when price is moving in one direction and slows down when price is noisy. An efficiency ratio compares the net move over a window with the sum of the absolute bar-to-bar moves. The average then uses a faster smoothing constant in clean trends and a slower one in chop.

Not investment advice. Nothing here is a recommendation to buy or sell anything. Backtests and historical studies are not guarantees, and a rule that worked in one sample can fail in the next. Trading can lose money. Links are for education, not endorsements. Do your own research.

Rules

The usual efficiency window is 10 bars. Smoothing constants are often mapped to a fast EMA near 2 periods and a slow EMA near 30. Trade in the direction of a rising or falling KAMA, and stand aside when the efficiency ratio is low and the average has gone flat. A volatility stop still caps the loss if the regime label is wrong.

Where AI fits

AI can compute the efficiency ratio as a regime feature and let a separate entry rule fire only in the high-efficiency state. That is a better use than asking the model to invent a new smoothing map. The fast and slow constants should be chosen on a training window and left alone on the test window.

Markets

Futures and FX, where trend and noise alternate, and liquid equity indexes.

Timeframes

Hourly to daily.

Risks

The efficiency ratio itself whipsaws on a short window. Adaptive averages can look brilliant on a trend chart and do nothing useful out of sample. Implementation details, especially how the smoothing constant is scaled, change the series.

← Back to all strategies