Monday, March 12, 2007

2007 Predictions - Introduction

Since Ross is about to start going through his 2007 predictions, I thought I'd get going on mine as well. I'm going to try something a bit different this year as a way to come up with my team win-loss predictions. I think that purely statistical projections do having something to offer - they're a good baseline to start from, since they'll adjust numbers for players switching parks (and leagues), take into account the effects of the player's age, etc. But clearly there are factors that the projections have no access to, and I'd like to think we could do better than a statistical projection alone.

My normal method would be to take all the numbers, look at them for a few minutes, and then adjust them up or down, but there's a problem: once I see the projections, my adjustments are biased based on what I've just seen. If I'm convinced ahead of time that Carlos Delgado is going to hit 40 home runs, then it doesn't matter what the projection says, because I'm just going to pencil him in for 40 home runs. But if I'm going to do that, what's the point of looking at the numbers in the first place?

So my idea for this year is to avoid looking at any numbers at first, and come up with adjustments first. Using information that projections don't have access to, I'll try to decide what players are likely to do better than their projection, and which should do worse. Then, when I'm done, I can actually look up the projections and adjust them. So what types of things are purely statistical projections likely to get wrong?

1. Injuries. This is a big one. Projections really only use playing time as a proxy for injury, and don't know anything specific about whether a player is likely to be injured this season. For example, Hideki Matsui had a freak injury last year (after never missing a game for years), but the projection can't tell the difference between that and a back problem that just surfaced.

2. Park adjustments. Statistical projections will use a park factor, but it's usually one that applies equally to all hitters. If a player switches parks, and there's a particular reason to think his old or new parks effect him differently from the average, then we'd need to adjust for that. For example, your opinion on J.D. Drew's ability to hit line drives might impact your thinking of how Fenway will affect his stats.

3. Changes in tools. Did a player start throwing harder? Develop (and effectively use) a new pitch? Bulk up or slim down?

4. Changes in environment. Did a player enter or leave a pressure situation, like New York or Boston? Did they have to switch countries, like Matsuzaka?

5. Scouting information. This isn't very important for established players, but for guys whose projections include translations of minor league stats, it is important, particularly for pitchers. Did the guy dominate in the minors because of great stuff (which should translate well to the majors), or because of deception (which won't translate as well)?

Hopefully you guys can offer suggestions about where I go wrong, so that we can fight about the adjustments before we actually look at the stastical projections themselves.


Ross said...

Well, you know I'm always more than happy to tell you where you've gone wrong...

Sully said...

I'm just amazed that you haven't looked at any projections yet.