With regards to the 'starting out of position' issue, here are some hard numbers

uruguru · 07-13-2024, 12:39 AM

I decided to run a 1962-1968 sim and collect some actual metrics on this problem to see if it is really a big issue or just some hyperventilation about some odd cases.

Taking Garlon's advice, I am testing without minor leagues, although honestly it's possible that an mlb-only sim could be worse. After all, once a team has to start someone out of position, they have fewer options to solve the problem (i.e. no minor leaguers to call up).

A quick summary of the problem: As historical sims progress through seasons, players gain or lose the ability to play certain positions based on changes in the real world. A good example would be Craig Biggio coming up as C, moving to 2B, and playing a season in CF late in his career. But there is also the more common case of older players (3B, OF and C) moving to 1B in their later career.

The problem arises in that OOTP players "remember" their playing experience at those forgotten positions, and the AI may not only occasionally play them in those positions, but make them regular starters. This is more common with stars who the AI wants to keep in the lineup. However, these guys are atrocious fielders and are clearly playing ahistorical positions.

What constitutes 'out of position'? On a scale of 1-10, if a player's maximum rating in a position is '1', I am calling it out of position. 2 and 3 are also really bad, but I want these to be clear cases. When a player is playing out of position, he becomes a defensive liability along the lines of 'all-time bad' liability.

The sim starts in 1962, no minors, fielding is set to 3 year averages, trading volume is average, and rookies enter the league as free agents.

For each season, I tabulated how many games were started out of position (and the amt for each player), and how many total innings were played out of position (starters and emergency subs)

As you would expect, the first season was almost problem-free. The problems start when recalcs start changing the fielding ratings of players.

I am not drawing any conclusions. This is just data so that everyone can see what is happening in a routine sim. You can decide for yourself if you think this is an issue or not. I suspect it would almost disappear completely if you set fielding to 'entire career', but then you'd end up with a lot of players in ahistorical positions (e.g. Hank Aaron at 2B) for most of their careers. Not sure if that is better or worse.

From 1962-1968:

1962 - 1 GS, 12 IP
3B 1 Lou Klimchock, ML1

1963 - 34 GS, 350.1 IP
3B 32 Deron Johnson, KCA
3B 2 Bobo Osborne, DET

1964 - 162 GS, 1445.2 IP
3B 15 Manny Mota, PIT
SS 1 Tom Tresh, NYY
SS 13 Felix Mantilla, NYM
SS 6 Julio Gotay, STL
CF 123 Rick Reichardt, PHI
CF 4 Ted Savage, PHI

1965 - 203 GS, 1633.2 IP
3B 2 Dom Demeter, PHI
3B 41 Deron Johnson, KCA
SS 2 Bob Johnson, BOS
SS 2 Tony Larussa, MIL
LF 156 Charley Smith, CHW

1966 - 554 GS, 6262.2 IP
C 12 Ed Kirkpatrick, BAL
1B 36 Bill Freehan, HOU
3B 134 Harmon Killebrew, MIN
3B 125 Tommy Harper, CIN
3B 2 Jim Hickman, NYM
3B 2 Lou Klimchock, ATL
SS 3 Tony Larussa, ATL
LF 9 Harmon Killebrew, MIN
LF 135 Orlando Cepeda, SFG
RF 96 Ed Kranepool, NYM

1967 - 491 GS, 4295.1 IP
C 6 Ed Kirkpatrick, BAL
2B 35 Chuck Hinton, PHI
2B 7 Frank Kostro, DET
3B 150 Harmon Killebrew, MIN
3B 37 Deron Johnson, KCA
3B 96 Tommy Harper, CIN
3B 9 Rick Joseph, HOU
3B 6 Frank Kostro, DET
SS 1 Cap Peterson, SFG
SS 19 Joey Amalfitano, HOU
LF 5 Willie McCovey, SFG
LF 120 Orlando Cepeda, SFG

1968 - 401 GS, 3599.1 IP
1B 11 Ron Swoboda, BAL
2B 26 Frank Kostro, DET
3B 62 Harmon Killebrew, MIN
3B 5 Frank Kostro, CHW
SS 8 Ernie Banks, CHC
LF 150 Willie McCovey, SFG
LF 86 Harmon Killebrew, MIN
LF 40 Cookie Rojas, CIN
CF 11 Ron Stone, ATL
CF 2 Howie Bedell, ATL

Brad K · 07-13-2024, 10:35 AM

Takes a lot of offense to make up for a 1 rating at a position.

Thanks for posting.

uruguru · 07-13-2024, 10:49 AM

Quote:

Originally Posted by Brad K

Takes a lot of offense to make up for a 1 rating at a position.

Thanks for posting.

Yes, 3B is a position that really seems to benefit in this regard. Deron Johnson, Tommy Harper and Harmon Killebrew all have earlier experience at 3B and played a LOT of games at 3B even though they were not rated for it.

McCovey and Cepeda both played full seasons in LF despite not having any outfield ratings.

fredbeene · 07-13-2024, 05:43 PM

thank you for posting. i love this type of analysis to make game better.
It would probably help if you posted your exact settings.
Recalc, Dev on of, Talent change randomness.

That said, i have always found it frustrating to have to run so many test sims to see why the AI always does the same thing.
ie in the 70s AI want to use RICO CARTY as a starting C. OK kinda cool for a random try. However everytime the AI does the same thing.
Not really adressing your issues, but there should be some RANDOMNESS to AI and it should be different for different team, not just one overall AI (not really AI)

uruguru · 07-13-2024, 06:48 PM

ok, good news! I re-ran the 1962-68 sim with fielding ratings set to "Entire Career". As a result, only 488 games were started with players out of position, as opposed to 1846 games when using 3 year averages. That's a 74% reduction from one simple change.

(settings: fielding is career, 5-year recalc, rookies enter as FA, trading is average, no TCR)

Even better, 440 of those 488 "bad" starts came from just two players. Donn Clendenon started two full seasons at LF (282 g) and Ernie Banks started 158 games at SS.

Now if you think it's weird that Banks doesn't have a career rating at SS, join the club. There's a quirk in how career ratings are calculated and it treats Ernie as a 1B, so he becomes incapable of playing SS.

The problem with Clendenon is that he played 17 games at LF in 1962 so the sim front-loaded him with 200 fielding experience at that position. Even though he doesn't have career outfield ratings, that experience at LF was retained and allowed the AI to play him at LF... which it did for two seasons.

Otherwise, just having players with some playabilty at more positions (with the freakish exception of Banks) is the main reason this problem almost completely disappeared with career ratings.

Here are the bad starts I logged for this sim. You can see how quickly this problem becomes a non-issue if Banks and Clendenon are addressed.

1962: 31 GS
1B 4 Bob Cerv, NYY
1B 1 Jim Lemon, MIN
1B 6 Haywood Sullivan, KCA
3B 20 Chuck Hinton, WAS

1963: 141 GS
LF 141 Donn Clendenon, PIT

1964: 54 GS
SS 54 Ernie Banks, CHC

1965: 212 GS
2B 2 Chuck Harrison, BOS
SS 69 Ernie Banks, CHC
LF 141 Donn Clendenon, PIT

1966: 25 GS
SS 15 Ernie Banks, CHC
SS 10 Dalton Jones, PIT

1967: 21 GS
SS 20 Ernie Banks, CHC
CF 1 Len Gabrielson, KCA

1968: 4 GS
2B 1 Jim Beauchamp, SFG
2B 1 Jack Hiatt, NYM
3B 1 Skip Lockwood, LAD
3B 1 Manny Mota, PIT

Brad K · 07-13-2024, 07:19 PM

The 1B is interesting. For years I've seen the computer manager cover 1B well, often having two players who can only play that position. Were there other players on the roster who could play 1B?

uruguru · 07-13-2024, 07:28 PM

I think it's sort of a bug. All three of those 1B (Cerv, Lemon and Sullivan) did not play enough 1B to have career infield ratings at the position, but the game gave them 200 fielding experience anyway. Fortunately they were not good enough to be starters (unlike Clendenon in LF) so the AI just started them there presumably as substitutions.

If the game did not grant fielding experience for unrated positions, I suspect there would almost be zero bad starts when using "Entire Career" for fielding ratings.

Brad K · 07-13-2024, 07:34 PM

What's strange about this is any IF can be trained to play a passable 1B. It's crazy for the computer manager to pick OFs and Cs.

uruguru · 07-13-2024, 07:56 PM

Quote:

Originally Posted by Brad K

What's strange about this is any IF can be trained to play a passable 1B. It's crazy for the computer manager to pick OFs and Cs.

The most important thing it looks at (I suspect) is the position rating.

A SS (great infield ratings) with no fielding experience at 1B will have a position rating of 0.

An OF or C (terrible or no infield ratings) with the max 200 fielding experience at 1B will have a position rating of 1. In fact, having the max 200 fielding experience at a position guarantees a rating of at least 1 at that position... even at catcher.

That means, from the perspective of the AI, a catcher or outfielder who has 200 fielding experience at 1B is a better option than an actual infielder who would conceivably be a much better option because he is actually rated for the infield.

So the game handing out 200 fielding experience during the recalc to players who are completely unrated at the position helps break the logic the AI uses to put players at positions.

Brad K · 07-13-2024, 08:12 PM

Seems during spring training the computer manager would train someone to play 1B.

07-13-2024, 12:39 AM	#1
uruguru All Star Starter Join Date: May 2022 Posts: 1,300	With regards to the 'starting out of position' issue, here are some hard numbers I decided to run a 1962-1968 sim and collect some actual metrics on this problem to see if it is really a big issue or just some hyperventilation about some odd cases. Taking Garlon's advice, I am testing without minor leagues, although honestly it's possible that an mlb-only sim could be worse. After all, once a team has to start someone out of position, they have fewer options to solve the problem (i.e. no minor leaguers to call up). A quick summary of the problem: As historical sims progress through seasons, players gain or lose the ability to play certain positions based on changes in the real world. A good example would be Craig Biggio coming up as C, moving to 2B, and playing a season in CF late in his career. But there is also the more common case of older players (3B, OF and C) moving to 1B in their later career. The problem arises in that OOTP players "remember" their playing experience at those forgotten positions, and the AI may not only occasionally play them in those positions, but make them regular starters. This is more common with stars who the AI wants to keep in the lineup. However, these guys are atrocious fielders and are clearly playing ahistorical positions. What constitutes 'out of position'? On a scale of 1-10, if a player's maximum rating in a position is '1', I am calling it out of position. 2 and 3 are also really bad, but I want these to be clear cases. When a player is playing out of position, he becomes a defensive liability along the lines of 'all-time bad' liability. The sim starts in 1962, no minors, fielding is set to 3 year averages, trading volume is average, and rookies enter the league as free agents. For each season, I tabulated how many games were started out of position (and the amt for each player), and how many total innings were played out of position (starters and emergency subs) As you would expect, the first season was almost problem-free. The problems start when recalcs start changing the fielding ratings of players. I am not drawing any conclusions. This is just data so that everyone can see what is happening in a routine sim. You can decide for yourself if you think this is an issue or not. I suspect it would almost disappear completely if you set fielding to 'entire career', but then you'd end up with a lot of players in ahistorical positions (e.g. Hank Aaron at 2B) for most of their careers. Not sure if that is better or worse. From 1962-1968: 1962 - 1 GS, 12 IP 3B 1 Lou Klimchock, ML1 1963 - 34 GS, 350.1 IP 3B 32 Deron Johnson, KCA 3B 2 Bobo Osborne, DET 1964 - 162 GS, 1445.2 IP 3B 15 Manny Mota, PIT SS 1 Tom Tresh, NYY SS 13 Felix Mantilla, NYM SS 6 Julio Gotay, STL CF 123 Rick Reichardt, PHI CF 4 Ted Savage, PHI 1965 - 203 GS, 1633.2 IP 3B 2 Dom Demeter, PHI 3B 41 Deron Johnson, KCA SS 2 Bob Johnson, BOS SS 2 Tony Larussa, MIL LF 156 Charley Smith, CHW 1966 - 554 GS, 6262.2 IP C 12 Ed Kirkpatrick, BAL 1B 36 Bill Freehan, HOU 3B 134 Harmon Killebrew, MIN 3B 125 Tommy Harper, CIN 3B 2 Jim Hickman, NYM 3B 2 Lou Klimchock, ATL SS 3 Tony Larussa, ATL LF 9 Harmon Killebrew, MIN LF 135 Orlando Cepeda, SFG RF 96 Ed Kranepool, NYM 1967 - 491 GS, 4295.1 IP C 6 Ed Kirkpatrick, BAL 2B 35 Chuck Hinton, PHI 2B 7 Frank Kostro, DET 3B 150 Harmon Killebrew, MIN 3B 37 Deron Johnson, KCA 3B 96 Tommy Harper, CIN 3B 9 Rick Joseph, HOU 3B 6 Frank Kostro, DET SS 1 Cap Peterson, SFG SS 19 Joey Amalfitano, HOU LF 5 Willie McCovey, SFG LF 120 Orlando Cepeda, SFG 1968 - 401 GS, 3599.1 IP 1B 11 Ron Swoboda, BAL 2B 26 Frank Kostro, DET 3B 62 Harmon Killebrew, MIN 3B 5 Frank Kostro, CHW SS 8 Ernie Banks, CHC LF 150 Willie McCovey, SFG LF 86 Harmon Killebrew, MIN LF 40 Cookie Rojas, CIN CF 11 Ron Stone, ATL CF 2 Howie Bedell, ATL Last edited by uruguru; 07-13-2024 at 10:10 AM.

07-13-2024, 06:48 PM	#5
uruguru All Star Starter Join Date: May 2022 Posts: 1,300	ok, good news! I re-ran the 1962-68 sim with fielding ratings set to "Entire Career". As a result, only 488 games were started with players out of position, as opposed to 1846 games when using 3 year averages. That's a 74% reduction from one simple change. (settings: fielding is career, 5-year recalc, rookies enter as FA, trading is average, no TCR) Even better, 440 of those 488 "bad" starts came from just two players. Donn Clendenon started two full seasons at LF (282 g) and Ernie Banks started 158 games at SS. Now if you think it's weird that Banks doesn't have a career rating at SS, join the club. There's a quirk in how career ratings are calculated and it treats Ernie as a 1B, so he becomes incapable of playing SS. The problem with Clendenon is that he played 17 games at LF in 1962 so the sim front-loaded him with 200 fielding experience at that position. Even though he doesn't have career outfield ratings, that experience at LF was retained and allowed the AI to play him at LF... which it did for two seasons. Otherwise, just having players with some playabilty at more positions (with the freakish exception of Banks) is the main reason this problem almost completely disappeared with career ratings. Here are the bad starts I logged for this sim. You can see how quickly this problem becomes a non-issue if Banks and Clendenon are addressed. 1962: 31 GS 1B 4 Bob Cerv, NYY 1B 1 Jim Lemon, MIN 1B 6 Haywood Sullivan, KCA 3B 20 Chuck Hinton, WAS 1963: 141 GS LF 141 Donn Clendenon, PIT 1964: 54 GS SS 54 Ernie Banks, CHC 1965: 212 GS 2B 2 Chuck Harrison, BOS SS 69 Ernie Banks, CHC LF 141 Donn Clendenon, PIT 1966: 25 GS SS 15 Ernie Banks, CHC SS 10 Dalton Jones, PIT 1967: 21 GS SS 20 Ernie Banks, CHC CF 1 Len Gabrielson, KCA 1968: 4 GS 2B 1 Jim Beauchamp, SFG 2B 1 Jack Hiatt, NYM 3B 1 Skip Lockwood, LAD 3B 1 Manny Mota, PIT Last edited by uruguru; 07-13-2024 at 06:56 PM.

07-13-2024, 10:35 AM	#2
Brad K Banned Join Date: May 2016 Location: St Petersburg Florida USA Posts: 6,693 Infractions: 0/2 (4)	Takes a lot of offense to make up for a 1 rating at a position. Thanks for posting.

07-13-2024, 05:43 PM	#4
fredbeene All Star Starter Join Date: Aug 2016 Posts: 1,633	thank you for posting. i love this type of analysis to make game better. It would probably help if you posted your exact settings. Recalc, Dev on of, Talent change randomness. That said, i have always found it frustrating to have to run so many test sims to see why the AI always does the same thing. ie in the 70s AI want to use RICO CARTY as a starting C. OK kinda cool for a random try. However everytime the AI does the same thing. Not really adressing your issues, but there should be some RANDOMNESS to AI and it should be different for different team, not just one overall AI (not really AI)

07-13-2024, 07:19 PM	#6
Brad K Banned Join Date: May 2016 Location: St Petersburg Florida USA Posts: 6,693 Infractions: 0/2 (4)	The 1B is interesting. For years I've seen the computer manager cover 1B well, often having two players who can only play that position. Were there other players on the roster who could play 1B?

07-13-2024, 07:28 PM	#7
uruguru All Star Starter Join Date: May 2022 Posts: 1,300	I think it's sort of a bug. All three of those 1B (Cerv, Lemon and Sullivan) did not play enough 1B to have career infield ratings at the position, but the game gave them 200 fielding experience anyway. Fortunately they were not good enough to be starters (unlike Clendenon in LF) so the AI just started them there presumably as substitutions. If the game did not grant fielding experience for unrated positions, I suspect there would almost be zero bad starts when using "Entire Career" for fielding ratings.

07-13-2024, 07:34 PM	#8
Brad K Banned Join Date: May 2016 Location: St Petersburg Florida USA Posts: 6,693 Infractions: 0/2 (4)	What's strange about this is any IF can be trained to play a passable 1B. It's crazy for the computer manager to pick OFs and Cs.

07-13-2024, 08:12 PM	#10
Brad K Banned Join Date: May 2016 Location: St Petersburg Florida USA Posts: 6,693 Infractions: 0/2 (4)	Seems during spring training the computer manager would train someone to play 1B.