3195 - Replay A Single Historical Year

[dhill]

I have been thinking about this for awhile, how to replay a MLB year using the stats from that year, like the old table top baseball board games. The table top games gave you cards that had the stats for all of the players for a specific year. I tried this in OOTP2006 using both the use career stats and remaining stats options, but the players’ stats did not match their stats for the real year. So, I decided to play with the Lahman database to tweak it to meet my needs. I used the 1980 season.

For all players that played in 1980, I replaced their historical stats with their 1980 stats. This will keep the number of years that they played, but the overall averages will equal the 1980 stats. If a player had 30 HRs in 1980, then he had 30 HRs for every year of his career until 1980. It was a simple project of copying out the 1980 stats and using vlookup in Excel to replace all of the other stats (including fielding). I do not know if it was necessary to keep the historical stats for the 1980 players, but it seams that OOTP uses the player history to determine age factors. I also deleted all the players that did not play in 1980 and all stats after 1980. The results were very similar to the real results, with team finishing within a few (1 to 8) wins of the real season. I left injuries on to add some variability.

[rasnell]

I may not completely understand, but you realize that the Lahman database should be the real stats of every player for every season, right?

Also, if you look at the individual player in the game, there is a new tab that show's real life stats from whatever database you've used (Lahman).

Also realize that other databases like Gambo, Arod/Garlon might have changed the real stats to smooth out certain anomalies.

[pstrickert]

Originally Posted by dhill

I have been thinking about this for awhile, how to replay a MLB year using the stats from that year, like the old table top baseball board games. The table top games gave you cards that had the stats for all of the players for a specific year. I tried this in OOTP2006 using both the use career stats and remaining stats options, but the players’ stats did not match their stats for the real year. So, I decided to play with the Lahman database to tweak it to meet my needs. I used the 1980 season.

For all players that played in 1980, I replaced their historical stats with their 1980 stats. This will keep the number of years that they played, but the overall averages will equal the 1980 stats. If a player had 30 HRs in 1980, then he had 30 HRs for every year of his career until 1980. It was a simple project of copying out the 1980 stats and using vlookup in Excel to replace all of the other stats (including fielding). I do not know if it was necessary to keep the historical stats for the 1980 players, but it seams that OOTP uses the player history to determine age factors. I also deleted all the players that did not play in 1980 and all stats after 1980. The results were very similar to the real results, with team finishing within a few (1 to 8) wins of the real season. I left injuries on to add some variability.

I'm very interested in this project, dhill. Can you explain it in more detail, step by step, so I can try it? I think I understand what you've done, but I'm not proficient with Excel.

I've tried to "replay" the 1974 season after importing it from the Lahman DB. I don't sim up to the '74 season, nor do I sim after it. I simply want to see if I can do a single historical season replay. I've noticed a number of odd things:

1. The "weaken hitters" and "weaken pitchers" options do not seem to work. For example, Larry Anderson, who was a rookie in 1974 for Milwaukee, had 2-1/3 IP in real life in 1974. He didn't pitch all that much the next couple of seasons either, before fading into oblivion. Nevertheless, he imports into OOTP as a four-star pitcher. Same thing happens with Bob Sheldon. He had 17 ABs in real life in 1974, batting .118. Nevertheless, he imports into OOTP as a four-star player. The AI assigns him as the starter, rather than Pedro Garcia, the Brewers regular starter at 2B in real life.

2. The AI often ignores a player's primary position. As a result, in my 1974 import, George Scott is assigned to play 3B, even though he led the league in 1974 IRL in games played at 1B. Even if I edit his rating at 3B down to a "1", he is still assigned the starting 3B spot by the AI. (Bob Hansen, who played 3 games at 1B in 1974 IRL, is assigned the starting 1B job by the AI, instead.) IRL, Don Money was the starting third baseman for Milwaukee that year. In the 1974 import, the AI starts Money at SS. (IRL, Money never played a single game at SS in 1974, IIRC.) That causes Robin Yount, the Brewers starting SS in 1974 IRL, not to get enough ABs in the replay.

3. Sometimes, the "stars" assigned to players defies logic. Darrell Porter receives five stars, though he had only average or above average (at best) stats in 1974. Bill Castro, a rookie in 1974 IRL, sometimes receives 2-1/2 stars, 3 stars, and (once) even 5 stars when I imported the 1974 season. He's always given the closer role, too, even though IRL he had 18 IP in 1974 with no saves. I don't think he ever had the closer role during his career IRL.

Recently, I simmed the 1974 season after importing it. The stats (compared to RL) left much to be desired. I'm wondering if your project will make it possible to have a more satisfying experience replaying a single historical season. I sure hope so!

[pstrickert]

I wonder if it's possible to set up a league of the "greatest teams" of baseball history without having to worry about the limitations of the DB import system. Each of those teams would be a single season import, right?

[swampdragon]

An interesting issue. Who did the AI play at third for Milwaukee? And what happens if you use the strategy slider to favor prospects? I'm not shocked that Yount didn't get too much playing time, because he really wasn't a great player as an 18 year old, but in view of the alternatives, I'd think a strategy of committing to prospects would see him get his 300 ABs. Garcia's performance also left a lot to be desired. The AI doesn't care who actually played (maybe that should be a factor or an option) and is remorseless about weeding out the truly incompetent starters like Garcia who got a bunch of ABs in real life. According to Baseball Encyclopedia, Garcia's AOPS was 66. The mystery may not be why the AI didn't play him, but why Del Crandall did.

It's going to be much easier to go through and delete the anomalies than to completely rebuild the database. If you're playing a single season, you can just delete Sheldon and Anderson and not worry about it.

[pstrickert]

Originally Posted by swampdragon

An interesting issue. Who did the AI play at third for Milwaukee? And what happens if you use the strategy slider to favor prospects? I'm not shocked that Yount didn't get too much playing time, because he really wasn't a great player as an 18 year old, but in view of the alternatives, I'd think a strategy of committing to prospects would see him get his 300 ABs. Garcia's performance also left a lot to be desired. The AI doesn't care who actually played (maybe that should be a factor or an option) and is remorseless about weeding out the truly incompetent starters like Garcia who got a bunch of ABs in real life. According to Baseball Encyclopedia, Garcia's AOPS was 66. The mystery may not be why the AI didn't play him, but why Del Crandall did.

It's going to be much easier to go through and delete the anomalies than to completely rebuild the database. If you're playing a single season, you can just delete Sheldon and Anderson and not worry about it.

The AI assigned G. Scott to 3B, even when I made him a "1" at 3B. He had a "15" at 1B, much higher than Bob Hansen's "4" at 1B. Don Money's highest rating was at 3B, his lowest at SS. Still, he was assigned to play SS. Sometimes (since I tried different options to see how it would affect the starting lineups), Johnny Briggs would be platooned with Bob Hansen at 1B. IRL, G. Scott won a Gold Glove award that year at 1B and (as I said), he led the league in games played at 1B.

I still don't think the "weaken batters" and "weaken pitchers" options work right. Otherwise, Bob Sheldon, who batted .118 in 17 ABs in 1974, should never have received "four stars" and the starting job at 2B. (If he had hit .300 IRL in 17 ABs, he probably would've been made a Hall of Famer in OOTP!) Even though Pedro Garcia hit only .199 in 1974, that's better than Sheldon's .118, isn't it? Garcia BTW only receives "one star" in OOTP.

[rasnell]

Originally Posted by pstrickert

I wonder if it's possible to set up a league of the "greatest teams" of baseball history without having to worry about the limitations of the DB import system. Each of those teams would be a single season import, right?

You're correct that each of those teams is a single season import. If you can confirm that the Lahman database shows the correct position and the correct debut year, but OOTP is not getting this right during the inaugural season of simming, we need to TT this for a future patch.

I'm not aware of so many problems as noted here so keep the details coming so we can figure this out.

[pstrickert]

Originally Posted by swampdragon

It's going to be much easier to go through and delete the anomalies than to completely rebuild the database. If you're playing a single season, you can just delete Sheldon and Anderson and not worry about it.

I wish it were that easy. What about the other teams? How would I know the AI has put their best players on the field? Milwaukee should've had G. Scott at 1B, Pedro Garcia at 2B, Robin Yount at SS, and D. Money at 3B. Instead, they had B. Hansen at 1B, B. Sheldon at 2B, Don Money at SS, and G. Scott at 3B. That certainly would affect their infield defense. It also would affect their offense. Pedro Garcia and Robin Yount -- assuming they were rated properly on import -- would provide better production than B. Hansen and B. Sheldon.

Here's something else to think about . . .

The last time I imported the 1974 season, I decided to sim the whole season to check the results. Two last place teams and one fifth-place team IRL actually won their divisions in the sim. The other division winner was a fourth-place finisher IRL. Sure, it was only one sim one time -- but how out of whack is that? John Montefusco, who pitched for SF as a rookie in 1974, had a 3-2 record with a 4.81 ERA in 39.1 IP in real life. In the sim, he was 24-6 with a 2.07 ERA in 273.1 IP. I can accept variation from real life stats, but differences like that are quite hard to accept.

[pstrickert]

This time, I decided to do a simple Lahman import of the 1975 season. I didn't spend a lot of time with it, other than to look at the Milwaukee roster again. G. Scott shows up at 3B again in 3 of the 4 lineups (he's at 1B in the other lineup). Scott's rating at 1B = 13, 3B = 3. Don Money shows up at SS again. Robin Yount's the odd-man out. How many Brewers fans would be happy about that?

I noticed, too, that Johnny Briggs is the starter in CF, even though he's rated a "3" there. IRL, he never played a game in CF in 1975. His primary position is LF (rating = 16), not CF (rating = 3). The AI assigns Briggs to bat leadoff in the Versus LHP+DH lineup. I doubt that ever happened IRL. Bill Castro (as in the 1974 import) is given the closer's role. IRL, he had 0 saves in 1974 and 1 save in 1975. Tom Murphy saved 20 games in both 1974 and 1975 IRL. He's a MR in those two imports in OOTP.

I haven't checked other teams, but I imagine I'd find things "messed up" there, too.

[rasnell]

I'm going to TT that something is wrong with the inaugural season import and the way that OOTP is reading from the Lahman database. You've found enough examples worthy of note. I'll link to this thread. Keep posting your sims and results so Markus can take a look and I'll see if we can get this entire thread moved over to the technical support forum.

PStrickert already TT'd as #3195

[pstrickert]

Imagine my surprise when I discovered -- this morning -- that Markus marked this TT as "I - Can't Reproduce". So, I uninstalled OOTP, rebooted, re-installed and applied the new beta patch. Then I did my simple Lahman import of the 1974 season, "unchecked" the scouts and coaches options, and looked at the Milwaukee roster. I get the same results I've gotten before. Larry Anderson, who had 2.1 IP in 1974 IRL, gets 4 stars. Bill Castro, who had 18 IP in 1974 IRL and 0 saves, is assigned the closer role (even though Tom Murphy saved 20 games in 1974 IRL). As for lineups, Darrell Porter bats leadoff vs. RHP -- Darrell Porter, who receives 5 stars, though his stats look "above average" at best. The IF again has Bob Hansen at 1B (though he played 3 games at 1B in 1974 IRL). Bob Sheldon -- with 2-1/2 stars -- gets the start at 2B (even though he batted .118 in 1974 IRL with only 17 ABs). Don Money starts at SS, even though he played no games at SS in 1974 IRL. George Scott starts at 3B, even though he played no games at 3B in 1974 IRL. Scott also has vastly better ratings at 1B than 3B. And IRL Scott led the league in games played at 1B in 1974.

Now let's take a quick look at the San Francisco roster? John Montefusco receives a "5-star" rating and so does Ed Halicki. IRL in 1974, Montefusco went 3-2 with a 4.81 ERA and Halicki 1-8 with a 4.24 ERA. No wonder in the last replay I simmed, San Francisco finished in first place (even though IRL in 1974 they finished in fifth place).

Yet Markus marks this TT "I - Cannot Reproduce". Somebody help me out here! What am I doing wrong?!!

[pstrickert]

I tried it again. This time, though, I "unchecked" the recalc option. I assumed this would have no affect whatsoever, since it "recalcs" after the season is completed. I only wanted to import the 1974 season, that's all.

Well . . .

This time, I saw quite a few differences. Larry Anderson, for example, received a 1-star rating and ended up on the reserve roster. Bob Sheldon also received a 1-star rating and ended up on the reserve roster. That's OK, since both barely played in 1974 IRL. I think that means the "make pitchers bad" and "make hitters bad" worked this time. Unfortunately, George Scott still was assigned to play 3B. Yount and Money platooned at SS. Money and Garcia platooned at 2B. Bob (I played only 3 games at 1B IRL) got the starting job at 1B. Darrell Porter was rated 4 stars (which makes more sense than 5 stars). Tom Murphy got the closer's role (which matches real life).

However, Jim Colborn was now listed as an MR. IRL in 1974, he started 36 times (10-13, 4.06 ERA). IRL in 1973, as a matter of fact, he won 20 games as a starter. I don't know how he got bumped from the starting rotation. I also looked at the rosters for other teams in the league. Rollie Fingers is listed on the Oakland roster as a CL. But in the pitching "depth chart", he's listed as the #5 starter. Jim Palmer (BAL) is given 1-1/2 stars, even though IRL he was 7-12 with a 3.27 ERA. Tom Seaver (NYM), by contrast, received 4-1/2 stars, though IRL he was 11-11 with a 3.20 ERA.

Some other oddities: The Milwaukee reserve roster included Gorman Thomas, Felipe Alou, and Rob Ellis. I have no problem with that. But each of them was rated a "20" (out of 20) in the Avoid K category. IRL, Thomas struck out 15 times in 46 ABs in 1974. IRL, Ellis struck out 11 tims in 48 ABs. IRL, Alou struck out 2 times in 3 ABs. So where do these ratings come from? And, aside from that, why would it make any difference if I unchecked the recalc option?

If the import DB process is working properly, then how do you explain these results? I'd appreciate it if someone would try to reproduce these results -- or tell me what I'm doing wrong, if anything. Thanks.

[rasnell]

I would suggest that you go back to your TT and select Open/Regress and note a link to this thread and that you can reproduce the problem.

[pstrickert]

Done. Thanks, R.

[swampdragon]

The AI seems to be willing to play fielders out of position in an overly cavalier fashion, even when the offensive improvement is marginal. I've seen numerous examples of this. Even in my 1876 league where the AI has very few choices, it can frequently find a way to play players out of position.

[chriskelly]

I can't find the thread reporting the issue w/ Layman importing ratings in an inaccurate way. I'm wondering if the answer is that the ratings themselves are OK, but the stars are flukey? I've noticed that when I expand from a historical league, the star ratings don't seem to accurately reflect the stats / ratings, and I'm using scouts "off".

[pstrickert]

Originally Posted by chriskelly

I can't find the thread reporting the issue w/ Layman importing ratings in an inaccurate way. I'm wondering if the answer is that the ratings themselves are OK, but the stars are flukey? I've noticed that when I expand from a historical league, the star ratings don't seem to accurately reflect the stats / ratings, and I'm using scouts "off".

Here it is: TT3195

http://www.ootpdevelopments.com/boar...d.php?t=143695

[battists]

Merging this with the related thread.

[pstrickert]

Originally Posted by chriskelly

I can't find the thread reporting the issue w/ Layman importing ratings in an inaccurate way. I'm wondering if the answer is that the ratings themselves are OK, but the stars are flukey? I've noticed that when I expand from a historical league, the star ratings don't seem to accurately reflect the stats / ratings, and I'm using scouts "off".

I'm not sure. Here's an example from an import of the 1982 MLB season. Pete Vuckovich (MILW) won the Cy Young Award IRL that year (18-6, 3.34). Moose Haas (MILW) had a mediocre season (11-8, 4.47). Well, the Lahman DB import results in the following:

Vuckovich (1 star)
stuff - 63
movement - 161
control - 86

Haas (4-1/2 stars)
stuff - 68
movement - 144
control - 162

How do you explain that?

[battists]

Originally Posted by pstrickert

I'm not sure. Here's an example from an import of the 1982 MLB season. Pete Vuckovich (MILW) won the Cy Young Award IRL that year (18-6, 3.34). Moose Haas (MILW) had a mediocre season (11-8, 4.47). Well, the Lahman DB import results in the following:

Vuckovich (1 star)
stuff - 63
movement - 161
control - 86

Haas (4-1/2 stars)
stuff - 68
movement - 144
control - 162

How do you explain that?

What are the "recalc" settings for this import? If you have it set for a 5-year recalc, that would make a big difference...