There’s some confusion about the term “accuracy” when it comes to EPD’s. It would make common sense that if something had a low accuracy that it would not be very reliable…correct? No, that is not how the term ” is used with EPD’s. EPD Accuracy simply describes the Trait Value prediction in greater detail. A very important concept to remember is that EPD’s are correct, or better than predicted, 85% of the time….regardless of the Accuracy. Accuracy defines a range, plus or minus the trait value that can be expected. Just like Las Vegas, EPDs don’t like to be wrong. The Accuracy Range looks more like the point spread used in sports betting. A “low accuracy” represents a wider range of acceptable outcomes. For example, according to the Table 7 of 2013 EPD Producers Report (Table 8 for suris) an accuracy of .3 (low) indicates that the result could be .82 of a micron less than or greater than the trait value predicted. In other words a relatively “Low Accuracy” prediction is just as reliable as a high accuracy prediction. The only difference is the low accuracy prediction is providing a wider range of what can be expected.
So, I think the next logical question is “How reliable are EPD predictions anyway?” As I stated above, my personal, short, unscientific, quick and simple answer is that EPD’s trait values will be as good or better than predicted at least 85% of the time. However, if you were to ask your college statistics professor he would say that EPD predictions are calculated to within 2 standard deviations (68%). The remaining 32% will result in an outcome that is better or worse than the prediction. I frequently refer to this “unknown group of 32%” as the “freaks”. There are good freaks and poor freaks. Being the optimist that I am I say that at least half of the 32% will be good freaks and feel confident with expecting 85% reliability. This is the link to the 2013 Producers Report.
