MAE (Mean Absolute Error)
The average distance between your predictions and reality. Lower is better.
Mean Absolute Error answers the simplest question about predictions: on average, how far off are you? Take every prediction, subtract the actual value, ignore the negative sign (that’s the “absolute” part), and average them all. The units match whatever you’re predicting : dollars, degrees, percentages : making results immediately intuitive. Unlike RMSE, MAE treats every unit of error equally: a $5 miss counts exactly five times more than a $1 miss, with no hidden squaring.
How It Works
MAE = mean(|predicted − actual|). An MAE of $0 means perfect predictions. The scale matches your target variable, so “MAE = $0.83” means “off by 83 cents on average.”
Example
The oil model achieved a one-step MAE of $0.833 on WTI crude futures : 13% better than the prior version. With oil trading $60–85/barrel, being off by less than a dollar is strong performance. Measured in Oil v16 Sell Model.