Normalized DB

arod23 · 10-08-2005, 12:42 PM

Arod23 NormDB-v30
by Ankit and Arod23

Version 3.0
Last revision: 6/14/2005

Version Changes:
Assigned neutral ballpark to all teams
Assigned realistic fielding ratings (see below)
Fixed incorrect spellings of some names

OVERVIEW

This database includes the best players of all time. All players' statistics have been "normalized" to the same environment in order to allow for more accurate simulations using players from across different eras. In other words, deadball era batters hit homers and deadball pitchers give up homers. Modern day pitchers with ERAs inflated by playing in the higher HR era compare more favorably with their low run environment counterparts.

A more detailed explanation follows, but in short:
All player seasons have been translated to a neutral park in a standardized league, that scores 4.5 team runs per game, with a .260 batting average, .330 on base average, and .420 slugging average, making for a nice even .750 OPS.

Players have been imported based on their normalized career totals. These figures have then been recalculated based on 500 AB for hitters and 250 IP for pitchers for ease of comparison. Note, this does not impact the simulation in any way, as OOTP is only concerned about ratios not actual totals. That is, there is no difference in OOTP's Power Rating in a player with 10 HRs in 150 ABs and a player with 40 HRs in 600 ABs.

Players fielding stats are based on the positions played during the course of their career. However, range and arm ratings for their primary position have been edited based on Bill James' Win Shares Defensive Rankings.

PLAYER EXAMPLES

Addie Joss played in the deadball era, holds the 3rd lowest career ERA of all time, dominates cross-era simulations. Active star Clemens with a career ERA that doesn't rank in the top-50 all time, clear hall of famer, yet he gives up 7x more HRs in cross-era simulations. When normalized Clemens numbers look better than Joss'.

HTML Code:

Addie Joss 1902-1910
Played in the deadball era, 3rd lowest career ERA of all time
		ERA	H/9	HR/9	BB/9	SO/9	
Actual		1.89	7.3	0.1	1.4	3.6
Normalized	3.60	8.4	0.9	1.5	5.3

Roger Clemens 1984-2004
		ERA	H/9	HR/9	BB/9	SO/9	
Actual		3.18	7.7	0.7	2.9	8.6
Normalized	2.74	7.1	0.6	2.4	9.1

Ty Cobb average falls a bit, but still a very high career average, yet he benefits from getting credited with some long balls, nearly hitting 500 over his career.

HTML Code:

Ty Cobb 1905-1928
		BA	OBA	SLG	OPS	HR	AB/HR
Actual		.366	.433	.512	.945	117	97.7
Normalized	.352	.420	.584	1.004	484	25.1

Hornsby had some deadball years as well as some years when the run environment was higher. His average suffers but his power increases, although not to the same extent as a player like Cobb.

HTML Code:

Rogers Hornsby 1915-1937
		BA	OBA	SLG	OPS	HR	AB/HR
Actual		.358	.434	.577	1.011	301	27.2
Normalized	.328	.415	.606	1.021	604	14.1

As great a player as Mantle was, his actual stats sometimes fall short of other greats because he played many years in a lower run environment. His average is now over .300 and he actually gets credit for another 85 HRs

HTML Code:

Mickey Mantle 1951-1968
		BA	OBA	SLG	OPS	HR	AB/HR
Actual		.298	.421	.557	.978	536	15.1
Normalized	.314	.427	.602	1.029	621	13.1

As dominant as Bonds is today, he is even better normalized. Why? He has played in a pitcher's park and still dominants his league, a league in a relatively high run era.

HTML Code:

Barry Bonds 1986-2004
		BA	OBA	SLG	OPS	HR	AB/HR
Actual		.300	.443	.611	1.054	703	12.9
Normalized	.312	.453	.644	1.097	774	11.6

Babe Ruth dominated the game by such an extreme that all normalization procedures have trouble dealing with him. Since he hit more HRs than some teams, in other normalizations, his annual HRs can exceed 100. In these calculations, Ruth doesn't hit more than 69 homers in any normalized year, although he does hit 60 or more seven times.

HTML Code:

Babe Ruth
		BA	OBA	SLG	OPS	HR	AB/HR
Actual		.342	.474	.690	1.164	714	11.8
Normalized	.330	.461	.702	1.163	932	9.4

LEAGUE SETUP

The players have already been imported into a league file. The file contains 8 teams, split into 2 leagues of 4 teams. The schedule is set for 162 games. All players have been set as rookies in year 1901 and are available for drafting. Obviously, you may modify the league structure as you see fit, given the number of players that are available. The league settings will produce a close to modern day league, slightly better batting averages and less homers, but you should run test sims and adjust to your liking.

All current ratings have been set so they match talent ratings, so everyone initally performs as dictated by their stats. (For some reason, OOTP does have some disparities between actual rating and talent ratings even when each season in a career is identical). L/R splits have been included but they have been adjusted to more realistic, less extreme ratings than OOTP defaults. Contact, Eye, Avoid K's include L/R splits, but Gap and Power are not split by L/R. Lastly, pitcher L/R splits have been removed since these are accounted for in the hitter's ratings (otherwise it seems like we would be double counting).

Each player's primary fielding position has been edited to reflect Bill James' Win Shares Defensive Rankings. We stole this idea from Garlon's All Time DB. In James' Win Shares book, he grades players from A+ to F. We assigned ratings based on these grades: A+ 100, A 90, A- 85, B+ 80, B 75, B- 70, C+ 65, C 50, C- 35, D+ 30, D 25, D- 20. These were applied to both a players fielding range and his arm rating. Only a player's primary position was modified in this fashion. For other eligible positions, the range rating is randomly assigned by OOTP, with the exception of OF. For each primary position outfielder, the other OF positions were modified to be consistent with the primary rating. We made some generalizations here: CF, being the most difficult OF position, receive the same range ratings for LF and RF; LF/RF received the same rating for the opposite field, but a lower rating for CF, fielding percentages were also adjusted down in some cases to account for high fielding percentages resulting from limited chances. Pitchers' fielding ratings were not changed.

HOW WE CHOSE THE PLAYERS

We began with Bill James' New Historical Abstract. Per James' all-time lists, we included:
1) The top-100 players of all time, any position
2) The top-100 pitchers
3) At least the top-25 players at every position: C, 1b, 2b, 3b, SS, LF, CF, RF
4) Additional players:
a) when normalized would be competitive with the top players, e.g., there is no need to include the 80th best SS of all time because he won't be used as a DH or PH, and his stats don't even come close to the 300th best OF
b) who are commonly known as greats; part of the fun of a database like this is to see how your favorite players compare
c) who are active but may have ranked low or not at all on James' lists because their career hasn't been long enough yet, e.g., If Albert Pujols stopped playing today, he's not getting into the Hall of Fame, but statistically, his first four years are up there with the best ever
d) who are active or played recently, viewed as "stars" but might not be considered one of the best of all time; again part of the fun is playing with your favorite players
5) Pre and early 1900s players whose normalized stats compare favorably with the top players

Invariably, we left someone's favorite out, but we hope you will find that we have compiled a comprehensive list of players. And, consider this version 1.0, so we can always update it later.

NORMALIZATION METHOD

Although we put this database together, the underlying statistical work is not ours and should be properly credited. This normalized database was compiled using Baseball Prospectus' Davenport Translations (http://www.baseballprospectus.com/dt/) developed by Clay Davenport.

Normalizing baseball statistics across history is arguably one of the most difficult exercises one can undertake in both computational complexity and philosophy. While there are several approaches, we feel that Clay Davenport has done the most comprehensive work. Basically, he has considered everything.

For those interested, here is a summary of what is accounted for in his translations (summarized from his own explanations):

- All player seasons have been translated to a neutral park in a standardized league:
Batting: .260 batting average, .330 on base average, and .420 slugging average, making for a nice even .750 OPS
Pitching: 4.50 ERA, 9.0 H/9, 1.0 HR/9, 3.0 BB/9, 6.0 SO/9

1) Park Adjustment

An adjustment made to account for the fact that some parks are easier to hit in than average, giving an advantage (in raw statistical terms) to hitters who play for that team. Park factors are always made relative to a league average of 1.00. The park adjustments in the BP are made only on the park factor for runs, averaged over five years.

2) Team Hitting Adjustment

An adjustment made for hitters, to account for not having to face their own pitchers. Using pitching stats, (league R * pf - team R), divided by (league IP - team IP), divided by park-adjusted league runs per inning.

3) Team Pitching Adjustment

An adjustment made for pitchers, to account for not having to face their own team's batters. Using batting stats, (league runs * pf - team runs), divided by (league PA - team PA), divided by league runs per plate appearance * pf.

4) Adjusted for season/league offense level

Corrections have been made for team's hitting or pitching level and the player's home park. The numbers have also been win-adjusted; this is equivalent to a correction for league offensive level.

5) League Difficulty Adjustment

Each league has been given a difficulty level, based on the performance of players in that league compared to the same players' performance in other seasons. The reference difficulty level was defined by the trend line of the National League from 1947 to 2002, and extended backwards to 1871. The difficulty adjustment is the ratio between the actual difficulty level and the reference level.

6) Adjustments have been made for pre-homer times

In the deadball era, doubles and triples were power. A player who had a lot of 2b and 3b in that environment would almost certainly hit a lot of home runs in a modern environment, e.g., Sam Crawford's 312 triples turn into home runs, and he ends up with 575 for his career. But that isn't always true; thanks to his speed, Ty Cobb ends up keeping a lot more of his triples, and even though he emerges as a 20-25 HR a year guy, he keeps a high average (.350) and daring baserunning (consistent double-digit triples, 50+ stolen bases).

Note: Since we compilied this database, the Davenport Translation's have changed slightly, so the figures in this DB won't exactly match. Most are within one point of batting average or 1/100th of ERA, so the differences do not impact anything.

DB can be downloaded here:
OOTP DB Downloads

Eugene Church · 10-08-2005, 03:55 PM

Thank you guys for putting this together and sharing it.

What areas will still have to be edited?

Catchers PB average?

Pitchers batting ratings?

Pitchers wild pitches and hit batsmen?

And pitchers fielding ratings? (I think you said you did not normalize those)

arod23 · 10-08-2005, 04:30 PM

Quote:

Originally Posted by Eugene Church

What areas will still have to be edited?

Catchers PB average?

Pitchers batting ratings?

Pitchers wild pitches and hit batsmen?

And pitchers fielding ratings? (I think you said you did not normalize those)

You are correct, we did not edit these. However OOTP imports is how we left them.

10-08-2005, 12:42 PM	#1
arod23 Major Leagues Join Date: Apr 2004 Location: Philadelphia, PA Posts: 378	Normalized DB Arod23 NormDB-v30 by Ankit and Arod23 Version 3.0 Last revision: 6/14/2005 Version Changes: Assigned neutral ballpark to all teams Assigned realistic fielding ratings (see below) Fixed incorrect spellings of some names OVERVIEW This database includes the best players of all time. All players' statistics have been "normalized" to the same environment in order to allow for more accurate simulations using players from across different eras. In other words, deadball era batters hit homers and deadball pitchers give up homers. Modern day pitchers with ERAs inflated by playing in the higher HR era compare more favorably with their low run environment counterparts. A more detailed explanation follows, but in short: All player seasons have been translated to a neutral park in a standardized league, that scores 4.5 team runs per game, with a .260 batting average, .330 on base average, and .420 slugging average, making for a nice even .750 OPS. Players have been imported based on their normalized career totals. These figures have then been recalculated based on 500 AB for hitters and 250 IP for pitchers for ease of comparison. Note, this does not impact the simulation in any way, as OOTP is only concerned about ratios not actual totals. That is, there is no difference in OOTP's Power Rating in a player with 10 HRs in 150 ABs and a player with 40 HRs in 600 ABs. Players fielding stats are based on the positions played during the course of their career. However, range and arm ratings for their primary position have been edited based on Bill James' Win Shares Defensive Rankings. PLAYER EXAMPLES Addie Joss played in the deadball era, holds the 3rd lowest career ERA of all time, dominates cross-era simulations. Active star Clemens with a career ERA that doesn't rank in the top-50 all time, clear hall of famer, yet he gives up 7x more HRs in cross-era simulations. When normalized Clemens numbers look better than Joss'. HTML Code: Addie Joss 1902-1910 Played in the deadball era, 3rd lowest career ERA of all time ERA H/9 HR/9 BB/9 SO/9 Actual 1.89 7.3 0.1 1.4 3.6 Normalized 3.60 8.4 0.9 1.5 5.3 Roger Clemens 1984-2004 ERA H/9 HR/9 BB/9 SO/9 Actual 3.18 7.7 0.7 2.9 8.6 Normalized 2.74 7.1 0.6 2.4 9.1 Ty Cobb average falls a bit, but still a very high career average, yet he benefits from getting credited with some long balls, nearly hitting 500 over his career. HTML Code: Ty Cobb 1905-1928 BA OBA SLG OPS HR AB/HR Actual .366 .433 .512 .945 117 97.7 Normalized .352 .420 .584 1.004 484 25.1 Hornsby had some deadball years as well as some years when the run environment was higher. His average suffers but his power increases, although not to the same extent as a player like Cobb. HTML Code: Rogers Hornsby 1915-1937 BA OBA SLG OPS HR AB/HR Actual .358 .434 .577 1.011 301 27.2 Normalized .328 .415 .606 1.021 604 14.1 As great a player as Mantle was, his actual stats sometimes fall short of other greats because he played many years in a lower run environment. His average is now over .300 and he actually gets credit for another 85 HRs HTML Code: Mickey Mantle 1951-1968 BA OBA SLG OPS HR AB/HR Actual .298 .421 .557 .978 536 15.1 Normalized .314 .427 .602 1.029 621 13.1 As dominant as Bonds is today, he is even better normalized. Why? He has played in a pitcher's park and still dominants his league, a league in a relatively high run era. HTML Code: Barry Bonds 1986-2004 BA OBA SLG OPS HR AB/HR Actual .300 .443 .611 1.054 703 12.9 Normalized .312 .453 .644 1.097 774 11.6 Babe Ruth dominated the game by such an extreme that all normalization procedures have trouble dealing with him. Since he hit more HRs than some teams, in other normalizations, his annual HRs can exceed 100. In these calculations, Ruth doesn't hit more than 69 homers in any normalized year, although he does hit 60 or more seven times. HTML Code: Babe Ruth BA OBA SLG OPS HR AB/HR Actual .342 .474 .690 1.164 714 11.8 Normalized .330 .461 .702 1.163 932 9.4 LEAGUE SETUP The players have already been imported into a league file. The file contains 8 teams, split into 2 leagues of 4 teams. The schedule is set for 162 games. All players have been set as rookies in year 1901 and are available for drafting. Obviously, you may modify the league structure as you see fit, given the number of players that are available. The league settings will produce a close to modern day league, slightly better batting averages and less homers, but you should run test sims and adjust to your liking. All current ratings have been set so they match talent ratings, so everyone initally performs as dictated by their stats. (For some reason, OOTP does have some disparities between actual rating and talent ratings even when each season in a career is identical). L/R splits have been included but they have been adjusted to more realistic, less extreme ratings than OOTP defaults. Contact, Eye, Avoid K's include L/R splits, but Gap and Power are not split by L/R. Lastly, pitcher L/R splits have been removed since these are accounted for in the hitter's ratings (otherwise it seems like we would be double counting). Each player's primary fielding position has been edited to reflect Bill James' Win Shares Defensive Rankings. We stole this idea from Garlon's All Time DB. In James' Win Shares book, he grades players from A+ to F. We assigned ratings based on these grades: A+ 100, A 90, A- 85, B+ 80, B 75, B- 70, C+ 65, C 50, C- 35, D+ 30, D 25, D- 20. These were applied to both a players fielding range and his arm rating. Only a player's primary position was modified in this fashion. For other eligible positions, the range rating is randomly assigned by OOTP, with the exception of OF. For each primary position outfielder, the other OF positions were modified to be consistent with the primary rating. We made some generalizations here: CF, being the most difficult OF position, receive the same range ratings for LF and RF; LF/RF received the same rating for the opposite field, but a lower rating for CF, fielding percentages were also adjusted down in some cases to account for high fielding percentages resulting from limited chances. Pitchers' fielding ratings were not changed. HOW WE CHOSE THE PLAYERS We began with Bill James' New Historical Abstract. Per James' all-time lists, we included: 1) The top-100 players of all time, any position 2) The top-100 pitchers 3) At least the top-25 players at every position: C, 1b, 2b, 3b, SS, LF, CF, RF 4) Additional players: a) when normalized would be competitive with the top players, e.g., there is no need to include the 80th best SS of all time because he won't be used as a DH or PH, and his stats don't even come close to the 300th best OF b) who are commonly known as greats; part of the fun of a database like this is to see how your favorite players compare c) who are active but may have ranked low or not at all on James' lists because their career hasn't been long enough yet, e.g., If Albert Pujols stopped playing today, he's not getting into the Hall of Fame, but statistically, his first four years are up there with the best ever d) who are active or played recently, viewed as "stars" but might not be considered one of the best of all time; again part of the fun is playing with your favorite players 5) Pre and early 1900s players whose normalized stats compare favorably with the top players Invariably, we left someone's favorite out, but we hope you will find that we have compiled a comprehensive list of players. And, consider this version 1.0, so we can always update it later. NORMALIZATION METHOD Although we put this database together, the underlying statistical work is not ours and should be properly credited. This normalized database was compiled using Baseball Prospectus' Davenport Translations (http://www.baseballprospectus.com/dt/) developed by Clay Davenport. Normalizing baseball statistics across history is arguably one of the most difficult exercises one can undertake in both computational complexity and philosophy. While there are several approaches, we feel that Clay Davenport has done the most comprehensive work. Basically, he has considered everything. For those interested, here is a summary of what is accounted for in his translations (summarized from his own explanations): - All player seasons have been translated to a neutral park in a standardized league: Batting: .260 batting average, .330 on base average, and .420 slugging average, making for a nice even .750 OPS Pitching: 4.50 ERA, 9.0 H/9, 1.0 HR/9, 3.0 BB/9, 6.0 SO/9 1) Park Adjustment An adjustment made to account for the fact that some parks are easier to hit in than average, giving an advantage (in raw statistical terms) to hitters who play for that team. Park factors are always made relative to a league average of 1.00. The park adjustments in the BP are made only on the park factor for runs, averaged over five years. 2) Team Hitting Adjustment An adjustment made for hitters, to account for not having to face their own pitchers. Using pitching stats, (league R * pf - team R), divided by (league IP - team IP), divided by park-adjusted league runs per inning. 3) Team Pitching Adjustment An adjustment made for pitchers, to account for not having to face their own team's batters. Using batting stats, (league runs * pf - team runs), divided by (league PA - team PA), divided by league runs per plate appearance * pf. 4) Adjusted for season/league offense level Corrections have been made for team's hitting or pitching level and the player's home park. The numbers have also been win-adjusted; this is equivalent to a correction for league offensive level. 5) League Difficulty Adjustment Each league has been given a difficulty level, based on the performance of players in that league compared to the same players' performance in other seasons. The reference difficulty level was defined by the trend line of the National League from 1947 to 2002, and extended backwards to 1871. The difficulty adjustment is the ratio between the actual difficulty level and the reference level. 6) Adjustments have been made for pre-homer times In the deadball era, doubles and triples were power. A player who had a lot of 2b and 3b in that environment would almost certainly hit a lot of home runs in a modern environment, e.g., Sam Crawford's 312 triples turn into home runs, and he ends up with 575 for his career. But that isn't always true; thanks to his speed, Ty Cobb ends up keeping a lot more of his triples, and even though he emerges as a 20-25 HR a year guy, he keeps a high average (.350) and daring baserunning (consistent double-digit triples, 50+ stolen bases). Note: Since we compilied this database, the Davenport Translation's have changed slightly, so the figures in this DB won't exactly match. Most are within one point of batting average or 1/100th of ERA, so the differences do not impact anything. DB can be downloaded here: OOTP DB Downloads

10-08-2005, 03:55 PM	#2
Eugene Church Hall Of Famer Join Date: Aug 2002 Posts: 36,205	Thank you guys for putting this together and sharing it. What areas will still have to be edited? Catchers PB average? Pitchers batting ratings? Pitchers wild pitches and hit batsmen? And pitchers fielding ratings? (I think you said you did not normalize those) Last edited by Eugene Church; 10-08-2005 at 04:00 PM.