from Amazon
Friday, February 05, 2010
What do you forecast?
Someone sent me an email about how to handle forecasting negative WAR since every year, we see negative WAR being actually generated. Here is my response:
There is a difference between OBSERVED and TRUE.
If Garret Anderson and Jacque Jones and a host of other players are all TRUE replacement level, and if you give them each 600 PA: guess what happens? Some will have a 2 WAR, some will have a 1 WAR, some will have a 0 WAR, some will have a -1 WAR and some will have a -2 WAR.
Overall, as a group, these players will have a 0 WAR. And that’s because we KNOW that they were TRUE replacement level. So, if you start with the idea that you know someone is a true 0 WAR, then it’s irrelevant what we will observe, any more than you know you have a true fair coin and you observe 60 heads in 100 flips or 30 heads in 100 flips.
And when you forecast players, it would be insane to give PA or IP > 0 for players who are below replacement level. Therefore, by definition, the lowest (true) WAR you can give someone is 0 WAR.
We are not trying to forecast observations. We are trying to establish the (unknowable) true rate. And your forecast must equal the true, just as you would ALWAYS forecast a coin to come up heads 50% of the time, regardless of how many observations you have seen or are about to see.