My Lahman Database Project

[mberner]

still getting funky numbers for talent for pitching - all the same, and all crappy. Still not sure whether it's OOTP or the database.

[GoCanes]

Guys there are two R columns in the pitching.csv that CTorg sent out. You need to delete the one between the H and ER columns, not the one at the end. I don't know why, but this is how the Lahman DB has it. Hope this helps.

GoCanes

[Elendil]

One area of theoretical comment... Wouldn't this DB reduce the "natural" tail-ends of performance for youngsters and old guys? If you simply average their year with their career, then players who play a full season but are rookies and don't do very well will do much better than they should using this DB. Also, older guys in decline will do much better than they should. Sometimes (in fact, more often than not), the way people do in a year is exactly what they should do in a year given their career talent and the stage of their career.

Ideally, to get rid of Lahman's and Ankit's problems, what you should do is to have an age-adjusted performance curve based on career numbers. Take their career numbers and use this curve to create hypothetical numbers for any year along that curve, and use those hypothetical numbers for the season. Very time-consuming, even if you could come up with an uncontroversial formula for translating career averages into yearly performance at different ages.

Otherwise, I'd strongly recommend doing the year-career averages only for those players who had a limited number of PA's or IP. For everyone else, it will distort their development.

[ctorg]

To a degree, you are right. It does make those distortions. If OOTP imported better, we wouldn't have to worry about it. Also, if someone wanted to take the hundreds of hours needed to apply career curves to all the players in the DB, that would be the ultimate database.

This is basically an improvement, I feel, for the purposes of OOTP. Let's say you import a guy in a season where he hit 50 homers and he never hit more than 20 in any other. Using the regular Lahman DB, unless he's in decline, he will likely keep that 50-homer power for a few seasons. This tempers it a bit. It still gives him more power, but it doesn't give him as much. Fifty homers remains in the realm of possibility, but it becomes an anomaly, as it was in the career of the player.

I think of this database as a compromise between the straight Lahman DB and the career average DB. You still get some variance between years, which the career average doesn't really give, but you also get a more standardized player who will perform in his career more similarly to the actual player than he likely would with the straight Lahman DB.

[Elendil]

That's a good example. I just wish there were some way to distinguish the true "anomaly years" from the "years that deviate from career averages but are supposed to because of their location on the age-development curve."

[ctorg]

Originally Posted by Elendil

That's a good example. I just wish there were some way to distinguish the true "anomaly years" from the "years that deviate from career averages but are supposed to because of their location on the age-development curve."

I know what you mean. It really comes down to OOTP's ability to import and how it assigns ratings and talents at that time. While it's come a long way, it still seems to put too much emphasis on the year being imported as opposed to the career of the player. Stealing ratings, for instance, do not take the career into account at all. Fielding ratings are the same and have plenty of other flaws as well.

I guess that's why, despite the fact that I like playing with the database, I rarely use it. I mostly just play fictional now.