Some lahman bugs

[bababui]

This is OT but Markus are you related to Tommy Heinsohn of the Celtics?

[CBL-Commish]

Originally Posted by Markus Heinsohn

I use the data that's available (which excludes groundball/flyball data )... and the results are rather good, but of course not excellent because of the missing stats....

You can estimate G/F data at the team level by comparing the ratio of outfield putouts and infield assists...

[Malleus Dei]

Originally Posted by CBL-Commish

You can estimate G/F data at the team level by comparing the ratio of outfield putouts and infield assists...

Yeah, that will give you a rough figure.

[BigCity]

Originally Posted by CBL-Commish

You can estimate G/F data at the team level by comparing the ratio of outfield putouts and infield assists...

...and totally wrong if the player gets traded, or a few key players making up those figures do. I think we're losing sight of OOTP's strength, and inherent design, that once the players are imported, things start to change - in a pattern not related at all to history.

[Malleus Dei]

Originally Posted by BigCity

...and totally wrong if the player gets traded, or a few key players making up those figures do. I think we're losing sight of OOTP's strength, and inherent design, that once the players are imported, things start to change - in a pattern not related at all to history.

Those reasons, and a few others, are why it's rough.

But do you know a better way to get a rough estimate of it? I don't.

[JustPats]

Originally Posted by Markus Heinsohn

I use the data that's available (which excludes groundball/flyball data )... and the results are rather good, but of course not excellent because of the missing stats....

Actually you have pretty good GB/FB data in Lahman if you just look at the team assist and putout totals for each season. You can also do historical adjustments, park adjustments, and adjusting for lefty pitchers with same methodology. It's what they did in Baseball Mogul this year and it's pretty straightforward. I have Win Shares by Bill James and it's all spelled out step by step in Section II. It's how Bill James came up with ratings that are much better than the ones you get from Linear Weights (which are done with unadjusted fielding stats).

Jeff

[JustPats]

Originally Posted by BigCity

...and totally wrong if the player gets traded, or a few key players making up those figures do. I think we're losing sight of OOTP's strength, and inherent design, that once the players are imported, things start to change - in a pattern not related at all to history.

I don't think CBL is disagreeing with you. Calculating defensive ratings based on stats generated by the sim AFTER the import could certainly lead to some bad results especially if the player ratings or the sim itself have some holes. His point is that for historical players loaded in from Lahman, at the exact point that they are loaded in from Lahman, the defensive ratings should be as accurate as possible given the data available. I haven't double-checked but I'm pretty sure that all the stats used in Win Shares are available in Lahman. It's actually rather brilliant how James squeezes accurate ratings out of the defensive stats that happened to be recorded throughout the 20th century.

[BigCity]

(repeat post)

[BigCity]

Originally Posted by JustPats

I don't think CBL is disagreeing with you. Calculating defensive ratings based on stats generated by the sim AFTER the import could certainly lead to some bad results especially if the player ratings or the sim itself have some holes. His point is that for historical players loaded in from Lahman, at the exact point that they are loaded in from Lahman, the defensive ratings should be as accurate as possible given the data available. I haven't double-checked but I'm pretty sure that all the stats used in Win Shares are available in Lahman. It's actually rather brilliant how James squeezes accurate ratings out of the defensive stats that happened to be recorded throughout the 20th century.

But when you import a player, your importing his "career stats" or his "remaining career stats" in terms of talent. If you import him on a "career year", his following years will certainly be less impressive as his "talent ratings" which will direct his OOTP career. If your going to do this "correctly" you need to do the same for his fielding stats - not just try for 100% accuracy for the year you import him from. Again, this is a "baseball simulation", not a "replay simulation".

To better express what I mean, take Ozzie in his best year and move him to the worst team in the league - then recalculate his value using James' formulas. Besides the fact that excercise doesn't make a lot of sense since Ozzie didn't PLAY for that team, his rating will change drastically. To me, this confirms the "snapshot" in time of one season should NOT be replicated in OOTP since it will be the ONLY point in time the player's ratings will be accurate anyway - and could cause significant variation if the player gets traded.

Personally, I think Markus' approach is a decent compromise.

Having said that, I don't know why someone here got a "6" for Ozzie on import. I tried 3 different scenarios and never got anything near that low. I tried again importing through the 60s for Brooks Robinson and always got realistic results.

[Malleus Dei]

"It's deja vu all over again!"

[CBL-Commish]

Originally Posted by BigCity

But when you import a player, your importing his "career stats" or his "remaining career stats" in terms of talent. If you import him on a "career year", his following years will certainly be less impressive as hs "talent ratings" will direct his OOTP career. If your going to do this "correctly" you need to do the same for his fielding stats - not just try for 100% accuracy for the year you import him from. Again, this is a "baseball simulation", not a "replay simulation".

To better express what I mean, take Ozzie in his best year and move him to the worst team in the league - then recalculate his value using James' formulas. Besides the fact that excercise doesn't make a lot of sense since Ozzie didn't PLAY for that team, his rating will change drastically. To me, this confirms the "snapshot" in time of one season should NOT be replicated in OOTP since it will be the ONLY point in time the player's ratings will be accurate anyway - and could cause significant variation if the player gets traded.

Personally, I think Markus' approach is a decent compromise.

Having said that, I don't know why someone here got a "6" for Ozzie on import. I tried 3 different scenarios and never got anything near that low. I tried again importing through the 60s for Brooks Robinson and always got realistic results.

All I'm saying is that you'll get better results if you know what the G/F ratios of the teams he played for are. Something along the lines of "Aloysius Smith, a shortstop, played most of his career behind flyball staffs, so OOTP will rate him with a 80 in range instead of the 70 his raw numbers would imply." Nothing radical, just a tweak to make things a little more realistic for the historical simmers.

Maybe the effort needed to code the import algorithm is more difficult than it's worth for the benefits, but maybe not.

[Garlon]

BigCity - One of the things Arod23 and I are planning for our DB is to convert the entire fielding csv and fieldingOF csv to a career average format. The entry for every season in a player's career will have his average stats for each position he played. Ankit did the same thing with his DB. This should get reduce the effect of just happening to import a guy in a season in which his fielding stats look particularly good.

[Biggio3000]

Originally Posted by BigCity

But when you import a player, your importing his "career stats" or his "remaining career stats" in terms of talent. If you import him on a "career year", his following years will certainly be less impressive as his "talent ratings" which will direct his OOTP career. If your going to do this "correctly" you need to do the same for his fielding stats - not just try for 100% accuracy for the year you import him from. Again, this is a "baseball simulation", not a "replay simulation".

To better express what I mean, take Ozzie in his best year and move him to the worst team in the league - then recalculate his value using James' formulas. Besides the fact that excercise doesn't make a lot of sense since Ozzie didn't PLAY for that team, his rating will change drastically. To me, this confirms the "snapshot" in time of one season should NOT be replicated in OOTP since it will be the ONLY point in time the player's ratings will be accurate anyway - and could cause significant variation if the player gets traded.

But when James rates players defensively he's using career average data. Since we have all of Lahman to work with, Markus should be able to do the same thing.

[JustPats]

Originally Posted by Garlon

BigCity - One of the things Arod23 and I are planning for our DB is to convert the entire fielding csv and fieldingOF csv to a career average format. The entry for every season in a player's career will have his average stats for each position he played. Ankit did the same thing with his DB. This should get reduce the effect of just happening to import a guy in a season in which his fielding stats look particularly good.

I THINK I can save you guys a lot of trouble. Start a game in Baseball Mogul which creates essentially career average fielding data for every player with adjustments for team and era. Then export these projections (called 'predicted' stats) in the Lahman format and copy them over. There's no free demo yet but there should be next week so you could do it for free (or find someone to copy from).

[kagnew35]

andymac, you're one of the most lucid posters I've seen, but you have obviously never heard real music.

my music
1 serenade- mario lanza
2 san francisco -judy garland
3 hello dolly-louis and barbra
4 with a song in my heart- jane froman
5 five pennies saints-louis and danny kaye

[andymac]

Originally Posted by kagnew35

andymac, you're one of the most lucid posters I've seen, but you have obviously never heard real music.

my music
1 serenade- mario lanza
2 san francisco -judy garland
3 hello dolly-louis and barbra
4 with a song in my heart- jane froman
5 five pennies saints-louis and danny kaye

[bababui]

This thread leads the league in banned posters.

[bababui]

Originally Posted by kagnew35

andymac, you're one of the most lucid posters I've seen, but you have obviously never heard real music.

my music
1 serenade- mario lanza
2 san francisco -judy garland
3 hello dolly-louis and barbra
4 with a song in my heart- jane froman
5 five pennies saints-louis and danny kaye

You forgot the Liza

[kagnew35]

new york, new york- liza
cabaret- liza
once in love with amy- ray bolger
rock a bye= al jolson
deep in my heart- mario lanza

september song- jimmy durante
back home again in Indiana- kate smith

[bababui]

Originally Posted by kagnew35

new york, new york- liza
cabaret- liza
once in love with amy- ray bolger
rock a bye= al jolson
deep in my heart- mario lanza

september song- jimmy durante
back home again in Indiana- kate smith

This just keeps getting scarier and scarier