Interesting study on career trajectories...

[Matter2003]

Not sure if anyone has read this---found it to be pretty interesting and perhaps something of use in player modeling...

Career Trajectories in Baseball

[Vogon Poet]

Originally Posted by Matter2003

Not sure if anyone has read this---found it to be pretty interesting and perhaps something of use in player modeling...

Career Trajectories in Baseball

There's a ton of selection bias in that study. I also don't like when academics talk in terms of standard deviations rather than actual baseball stats. It has less meaning, in my opinion.

What you want to do for figuring aging patterns is find every player who played back-to-back years (e.g. when he was 24 & 25, 31&32, etc.) and compare how he did one year to the next. That ensures you're using looking at the exact same players in each pool. If you don't do that, there's a huge survivor effect and you'll see some crazy results forp layers in their 30s.

Now there's still a survivor effect, even with this method. Say you have two 35-year-old players who are both "true" .250 hitters. Because of luck though, one hit .280 and the other hit .220. What's going to happen the next year? The .280 hitter is going to come back while the .220 hitter will be released and probably retire. So in each matched pair, we're expecting the first year to be full of "lucky" players.

What happens to the lucky players' stats the next season? Well, we'd expect the .280 hitter to regress back to his "true" .250 talent level. Let's say the effect of aging is 5 points and it drops him to .245. But he hit .280 the year before! So we're seeing a 35 point drop when in reality, only 5 of those points can be attributed to getting older.

So really, you want to regress a player's stats to the mean before looking at his next year's stats, to eliminate the selection bias.

There was an article on The Hardball Times a couple years ago doing exactly that, but I can't find it at the moment. Here's a couple charts showing aging by different components (homers, walks, singles, stolen bases, etc.): http://www.tangotiger.net/agepatterns.txt

And pitching components: Forecasting Pitchers - Adjacent Seasons

This is a cool one: http://www.insidethebook.com/ee/inde...g_patterns/#27

That shows wOBA (linear weights per plate appearance, but on the same scale as OBP) for pitchers and batters by age. The cool thing is how he did it. He did the year-over-year thing as above, but didn't stop at the same hitter each year. Instead it's the same hitter v. the same pitcher in the same park. So he looked at what Derek Jeter did v. Josh Beckett in Fenway Park in both 2006 and 2007, then repeated it for every batter/pitcher/park/year combo and combined the results.

[Vogon Poet]

I found this post from two years ago about OOTP: http://www.insidethebook.com/ee/inde...seball_game/#7

Marc or Markus or whoever over there must be a reader of my site, and saw my aging charts, and implemented those. I’m not sure what else of mine they did. I know they implemented something about DIPS, though don’t know what.

So I would guess the dev team is pretty aware of Tango's newer research as well.

[RonCo]

Trying not to speak too highly of myself...but using those were my doing. V2007 had hitters pretty well modeled, pitchers not quite so much. To me v9 was a step backward. Don't know about X.