THT Forecasts update

As some of you may know, some of the Oliver projections for minor leaguers did not look too…realistic. Jesus Montero, for example, was projected to hit .314/.361/.542 this year, .351/.400/.676 by 2014. Now, Montero is absolutely one of the best prospects in baseball, but it’s probably too early to predict that he will be the second coming of Albert Pujols.

The basic problem was the way we were constructing Minor League Equivalences (MLEs) for Oliver. An MLE, for those of you who don’t know, is meant to tell you how a player’s statistics from a given level would translate to the major leagues. Obviously, if a players hits .300 in Double-A, he will not hit that well in the major leagues.

The way Oliver was constructing its MLE translations was by comparing a player’s statistics at a given minor league level to his numbers in his first few years in the major leagues. This is not the traditional way of doing it—traditionally, analysts have seen how players in Single-A do when they go to Double-A, and how players in Double-A do when they get called up to Triple-A, and finally how players in Triple-A do when they get called up to the majors, and then multiply all those numbers through—but it has significant advantages.

Just to give you an example of how the two methods differ, let’s say we take all A-ball hitters in a given year and see how they did when called up to Double-A. Maybe those hitters combined for a .300 batting average in Single-A, but hit only .270 in Double-A (note: all numbers are completely made up). So then the translation factor from Single-A to Double-A would be .270/.300 = 0.9. In other words, hitters lose 10 percent of their batting average when they go from Single-A to Double-A.

Next, we would look at all Double-A hitters who got called up to Triple-A, and see how their statistics changed. Perhaps, those players combined to hit .300 in Double-A, but only hit .280 in Triple-A. In that case, the translation factor from Double-A to Triple-A would be .280/.300 = 0.93. In other words, hitters lose 7 percent of their batting average when they go from Double-A to Triple-A (remember, all numbers are completely made up, I have no idea what the true translation factor is).

Finally, we would look at all Triple-A hitters who got called up to the major leagues. Maybe those guys hit .295 in Triple-A, but only .260 in the majors. In that case, the translation factor would be .260/.295 = 0.88. That means that hitters lose around 12 percent of their batting average when they go from Triple-A to the majors.

So let’s say we had a hitter who batted .330 in Single-A. What would we expect that to translate to in the major leagues? Based on the traditional method, his MLE would be .330*0.9*0.93*0.88 = .243. But that method is imperfect, since the players who go from A-ball to Double-A, and from Double-A to Triple-A, and from Triple-A to the MLB are not the same. In other words, we’re comparing apples and oranges and bananas and hoping it all evens out. It would be much better to track a player from A-ball and see how he ends up doing when he gets called up to the major leagues.

That’s what Oliver does, and the method has indeed been shown to be more accurate than the traditional MLE. The problem our projections had is that they were assuming this jump would occur in one year, when often it can take many. So in essence, our 2010 projections were really “when this player reaches the major leagues” projections. The projections for future years were then adding in steep aging curves for the young hitters, meaning that five or six years out, top prospects ended up looking like Babe Ruth.

Essentially, we were double counting aging. A concrete example might help here: Let’s say we look at all Single-A players who eventually reach the major leagues. And let’s say that as a group, these hitters bat .300 in A-ball, but only .270 in the major leagues. The MLE factor is then .270/.300 = 0.9, meaning that the hitters lose 10 percent of their batting average between A-ball and the major leagues. But, and this is the important part, the hitters also get older!

Maybe, on average, the hitters in that group are 20 years old when they play in A-ball, but 24 when they reach the major leagues. In that case, there isn’t actually a 10 percent difference between hitting in Single-A and hitting in the major leagues—there’s a 10 percent difference between hitting in Single-A at 20 years old and hitting in the major leagues at 24. A correct MLE would factor out that aging, and find that the translation from A-ball to the major leagues is much harsher than just 10 percent.

Until now, we weren’t doing that. And because of that, our 2010 projections were too high for top prospects and our 2014 projections were simply astronomical. We’ve since fixed the MLE process to factor out this problem, and now, the projections for top prospects look about right. Montero, for example, is now projected to hit .292/.334/.460 in 2010, and get all the way up to .323/.368/.559 in 2014. Those numbers are still really good, but they are no longer Pujols-ian, and they really speak to how great a hitter Montero looks to be.

So yes, all this is a long-winded way of telling you that the Oliver MLEs and projections for minor league hitters have been fixed, and pitchers will be fixed soon too (though they weren’t as big a problem to begin with). If you haven’t subscribed to THT Forecasts yet, now is a good time. As you can see, we are constantly working to improve our projections (which were already very accurate), and we promise to be upfront about any problems that may arise.

We also have added leader boards, for easier browsing. You can see that Ryan Howard is projected to lead the majors in home runs with 51 or that Tim Lincecum is projected to strike out the most hitters, at 241. You can search the leader boards by team, league or level, and we have separate leader boards for the depth chart adjusted projections and the default computer numbers. In other words, there are plenty of options to choose from. And soon, you will be able to automatically export any view of the leader boards to Excel. How’s that for convenience?

So once again, if you have yet to subscribe to THT Forecasts, please consider doing so now. You won’t regret it.

(Also, one final note. There’s a bug in how we are calculating our catcher fielding projections, so those have been more or less zeroed-out for now. We’re working on a fix, and should have it up very soon.)

Print Friendly
 Share on Facebook0Tweet about this on Twitter0Share on Google+0Share on Reddit0Email this to someone
« Previous: Projecting Strasburg
Next: THT Forecasts update, the short version »


  1. David Gassko said...

    Hey Josh,

    The article got posted accidentally before I was done writing it. Should be good now.

  2. Dave said...

    I hope you spend a minimum amount of time shilling your for-pay product as the season progresses. It’s not why I come to THT. I’m sure a lot of people feel similar.y

  3. Jay said...

    Really, Dave, people want to get things for free and not have to pay for them in any way? One of the main reasons people are motivated to provide value for free is the chance that it will pay off financially at some point. It just always bothers me when people want value from others without giving anything in return. This site provides a ton of value for free. Reading an occasional plug for a product is a small price to pay for that…

  4. Nick Steiner said...

    Hey Dave, you don’t have to read the occasional blog post that David or Brian puts up explaining changes in the projection system.  You can just skip them and move on to one of the other 3-4 free articles posted each day!

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>