THE EPD PROMISE

THE EPD PROMISE:
WHAT YOU MAY HAVE MISSED ABOUT EPDS

Have you ever sat through a presentation on EPDs and felt confused or overwhelmed? Maybe you’ve thought EPDs sound too good to be true and found yourself a little skeptical. Or maybe you’re just trying to use the tool and want a better understanding. Here are some pieces to the puzzle I consider vital to using EPDs that might have been glossed over or brushed aside by others. Those pieces I call The EPD Promise. To me they are crucial to using EPDs effectively.

Why EPDs Exist—and Why They Matter

Expected Progeny Differences (EPDs) have become an industry standard in the beef cattle business. First introduced in the 1970s and widely adopted by the 1980s, EPDs revolutionized cattle breeding by allowing producers to compare genetics beyond raw weights and visual appraisal. Before EPDs, breeding decisions were often educated guesses, relying on weights, ratios, and pedigree information without a clear way to separate genetic potential from environmental influences. EPDs provide ranchers with a data-driven tool that predicts how the average performance of the offspring of a bull or heifer compares to the average performance of the offspring of other animals in the breed.

Promise #1: EPDs Are Unbiased and Built on Science

So, what does someone need to know to use this tool effectively? Well, first you need to know if you can actually trust it. The first part of The EPD Promise: numbers don't lie. It's a promise to be unbiased or free from manipulation. It doesn't matter whether it is environmental influences or nefarious 'thumbs on the scales', EPDs are designed to negate those effects. By using across-the-breed pedigree records, performance data, and genetic information, EPDs ensure that the predictions are unbiased. All this data is input into a fairly sophisticated statistical model —a Best Linear Unbiased Prediction (BLUP) model—that is all but impossible to manipulate. Breed associations follow standardized, peer-reviewed evaluation processes to ensure fairness and accuracy. They work at making it fair. Multiple studies have shown that on average, differences in EPDs when comparing one individual's offspring to another accurately predict differences in average progeny performance when enough data is available. Did you take note of the emphasis on the word average? EPDs are about average offspring performance for the lifetime of that animal—every calf that cow or bull will have, averaged together. And they do not predict what that specific average will be but rather the difference between that average and some other animal's offspring average. It is a selection tool. Nothing more, nothing less. And the scientific, mathematical, statistical foundation makes it extremely trustworthy, numbers don't lie. 

Promise #2: EPDs Predict Performance—But Not Perfectly

That takes us to the second part of the EPD Promise. EPDs promise to be 'right,' to accurately predict those offspring average performance differences 68% of the time. Or to put it another way, EPDs promise to be 'wrong' 32% of the time! Almost one-third of the time, the actual average performance difference will fall outside of the predicted range. EPDs don't claim to be perfect. Perhaps those in their zeal to explain EPDs have given that impression but to use EPDs effectively, you need to know the chances they will be wrong. One thing is for sure, EPDs outperform any other means that we have to evaluate the genetic potential of cattle. That includes the very best, most experienced human eyeballs that have ever existed. But given the limitations of EPDs stated above, just how useful are they?

Understanding EPD Accuracy and Confidence Ranges

So, let’s take a closer look at this thing. If you can only count on EPDs 68% of the time, then I'm like you. I’m not sure I like those odds either. Fear not. For many traits, the chances of you being happy with the results of EPD predictions are actually 84% and depending on how picky you are, I know of a way to get to a 97.5% satisfaction guarantee, maybe even a 99.85% satisfaction rate! Give me a chance to explain. 

Let’s take weaning weight for example. EPDs are the expected difference between the average weaning weight of a lifetime calf crop of two individuals. Those two numbers we use to subtract and find that difference we call an individual's weaning weight EPD. Are you with me? But actually, there is a range associated with EPDs that is determined by a confidence indicator called accuracy. It’s expressed as a ± value with the EPD being the midpoint of the range. So, one of our, let’s say bulls, in this example has a 66 lbs WW EPD with an accuracy of 0.35. That corresponds to a possible change (or standard error) of ±10.2 pounds. (The Angus Association publishes a chart that shows the possible change at each accuracy level for each EPD category) The corresponding range for our example then is from 55.8 to 76.2. We don’t know the sire’s true progeny difference, but statistically, it will fall within this range 68% of the time. But here's the thing. You would be very happy if the true progeny difference was better than what that range predicts for weaning weight and that will happen 16% of the time (one half of the time that the EPD is wrong it will be better than predicted thus one half of 32 is 16). 68% of the time you are happy plus 16% of the time that you're very happy is 84% total happy time. 

If you could be happy with a broader range, say twice the + or - figure, or in this example 20.4, then you will be happy 95% of the time and 2.5% very happy which is 97.5% total happy time. Multiply the range figure by three and its 99.85% satisfaction guaranteed! At this accuracy level though I would not be happy. Why? Because the range is too broad. At triple the possible change the range would be 35.4 to 96.6 pounds and that makes the EPD virtually meaningless when comparing it to some other individual with a narrower range.  

When More Isn’t Always Better: Traits Like Milk EPDs

Change the example to another trait though, say Milk EPDs, and that's a different story. More is not always better for some traits. You'll be stuck with the 68% of the time with in the range unless you can accept a broader range doubling or tripling the possible change. The only time that makes sense is when accuracy increases to a level, for example in milk EPDs, where the possible change is only a pound or less. That’s at the .90 accuracy/confidence range or better. So how do you increase the accuracy you ask? Well, the simple answer is more data, which means more progeny, or adding genomic data into the equation, or increasing ancestral data, or all three. The percentage of times the EPD falls within or outside of the confidence range doesn’t change. It’s always 68-95-99.7 —but as accuracy improves, the EPD shifts less over time and gets closer to the true progeny difference.

If You Remember Nothing Else, Remember These Three Things

If all this has never been explained to you, or if you missed this in your EPDs for Dummies course, don't feel bad. Evidently there isn't much non-technical explanation of these concepts. (If I've missed some source that’s out there let me know because surely someone has better skills than I do to explain it.) In any case here are three quick take-a-ways that I would send with you: 

1. EPDs are based on sound, unbiased scientific principles. 

2. There are foundational statistical realities which define the degree to which we should rely on EPDs. 

3. It is still the best tool to separate genetic potential from environmental influences we currently have. 

You Can’t Afford to Ignore EPDs

Being equipped with this knowledge, in my opinion, is crucial to using EPDs effectively. And if you’re not using this powerful tool to select genetics for your herd, you are probably putting your cattle operation at a disadvantage in the pursuit of profit!