Strange Things in Historical Replay

Dodger1 · 04-15-2014, 10:05 PM

I was looking into a few historical replay years, and found some really inexplicable things. I choose to manage one team, and let the ai manage the others. Bt, in 1980, the ai has Paul Splitorff as a middle reliever for KC. But in real life he was in 34 games, 33 as a starter. In 1983, the ai has Scotty McGregor ad a middle reliever, but in real life he was in 38 games, ALL as a starter. In 1965 the ai has Maury Wills batting 3rd for the Dodgers, but he hit leadoff in every game I was able to locate (fitting for someone with 94 steals and zero homers, hardly 3 spot numbers.)

I tried putting myself in charge of all teams, and making the changes to match real life. But once I put the ai back in charge, it changed everything back. I really don't want to use real life lineups, I prefer using injuries. Any thougts how to fix this, and why the ai is so wildly wrong with its decisions?

David Watts · 04-15-2014, 11:16 PM

What do you have base pitcher stamina based on? I would try basing it on current season and see if that changes Splitorff and McGregor.

JohnHoward · 04-16-2014, 12:53 PM

If you are not using real lineups then you are not "replaying", you are "playing". In that regard, there is no basis for complaint against AI decisions. It has no historical memory, so it doesn't know what's "right". You are playing a fictional game with the names of real players.

BIG17EASY · 04-16-2014, 01:10 PM

Quote:

Originally Posted by JohnHoward

If you are not using real lineups then you are not "replaying", you are "playing". In that regard, there is no basis for complaint against AI decisions. It has no historical memory, so it doesn't know what's "right". You are playing a fictional game with the names of real players.

This, for the most part. I would go a bit further and say that it's actually the real players, but the AI is evaluating the real players on its own, not based on how they were evaluated in real life, so the AI sees the players the OP mentions as being most valuable in those roles.

JCarter93 · 04-16-2014, 03:01 PM

You can always use the force position function to prevent the AI from doing this.

BIG17EASY · 04-16-2014, 03:10 PM

Quote:

Originally Posted by JCarter93

You can always use the force position function to prevent the AI from doing this.

Not really. That won't tell the AI where to place guys in the lineup. Plus, that's A LOT of work to do that for every player in a historical replay.

Dodger1 · 04-16-2014, 06:28 PM

I find using actual lineups doesn't always jibe with what's going on in the game. For example, a guy who struggled in real life and was sent down might be hitting .300 in the replay, and wouldn't be sent down. To do so would be totally unrealistic. Or, a guy who was rested irl because he was in a slump might be re hot when he's removed in the game, also unrealistic.

But it is totally unrealistic to say that a team's #1 starter (mcGregor was 18-7 with a team best era) should be put in the bullpen in favor of a guy who had an era over 5? Yeah, that's a quality ai.

pstrickert · 04-16-2014, 06:38 PM

The AI usually does a decent job (though not perfect), so I'm wondering what happened with Splittorff and McGregor. How well were they pitching during the replay?

David Watts · 04-16-2014, 06:59 PM

Quote:

Originally Posted by Dodger1

I find using actual lineups doesn't always jibe with what's going on in the game. For example, a guy who struggled in real life and was sent down might be hitting .300 in the replay, and wouldn't be sent down. To do so would be totally unrealistic. Or, a guy who was rested irl because he was in a slump might be re hot when he's removed in the game, also unrealistic.

But it is totally unrealistic to say that a team's #1 starter (mcGregor was 18-7 with a team best era) should be put in the bullpen in favor of a guy who had an era over 5? Yeah, that's a quality ai.

What is your pitchers stamina based on, current season, 3 years, or entire career?

JohnHoward · 04-16-2014, 08:15 PM

Quote:

Originally Posted by Dodger1

I find using actual lineups doesn't always jibe with what's going on in the game. For example, a guy who struggled in real life and was sent down might be hitting .300 in the replay, and wouldn't be sent down. To do so would be totally unrealistic. Or, a guy who was rested irl because he was in a slump might be re hot when he's removed in the game, also unrealistic.

But it is totally unrealistic to say that a team's #1 starter (mcGregor was 18-7 with a team best era) should be put in the bullpen in favor of a guy who had an era over 5? Yeah, that's a quality ai.

Then your disagreement is with possible world metaphysics and modal logic, not artificial intelligence.

Dodger1 · 04-16-2014, 10:00 PM

Would it be so hard to have the ai select starting rotations based on, oh, say, ACTUAL NUMBER of games started? Seems pretty basic to me.

David Watts · 04-17-2014, 08:14 AM

Quote:

Originally Posted by Dodger1

Would it be so hard to have the ai select starting rotations based on, oh, say, ACTUAL NUMBER of games started? Seems pretty basic to me.

So for the 3rd time. What is your base pitcher stamina set to? Are you using recalc? If so, 1 year, 3 or 5.

Try base pitcher stamina on current season, run computer manager and see if that doesn't change Splitorff and McGregor to starters. This has worked for me in the past.

Carlton · 04-17-2014, 09:02 AM

That issue will come up alot, and I will get flack for this, but in a true baseball sense, the AI is average when it comes to anything historical. Fictional league? It's the bees knees as it uses the same algorithms to create the players as it does to set the lineup

For the P's what database are you using? And like David stated, what are you importing them as?

The easiest fix for P are to go through them and up their stamina until it says projected role as Starter in the Player editor

For the lineup
It WILL put a leadoff hitter in the #3 hole if all the players it has in the lineup (very power and OBP heavy no matter what the era) have lower BABIP and Avoid Ks, and those catchers that got free passes and rarely K'd because the P let them hit the ball rather than waste pitches? They will bat 2nd WAY too often.

The best way to fix this is to have managerial tendencies and roster management actually hold some heavy weight in the programming, that way you can make the manager love AVG>OBP a bit and a good contact hitter will be the starter instead of the .230 30 HR guy, but it doesn't have that weight and I do not think 15 will address that

The workaround for the leadoff hitter would be to lower the speed and steal of the leadoff hitter on that team, but that opens up a new can of worms

David Watts · 04-17-2014, 09:41 AM

I've asked before if it might be time that OOTP installs a "use according to real life" feature to the game. I think it would be a cool feature for people that play using real transactions/retire according to history. That being said, how many people would use it? This game doesn't market itself as a replay game in the same sense Diamond Mind, Action Pc or Strat do. It's a career sim. Once fictional transactions, fictional injuries, fictional expansions etc etc get added to the mix, how is the AI supposed to field a team if it's hogtied by real usage?

pstrickert · 04-17-2014, 10:56 AM

Notice the "Base roles/positions on" option under Historical Settings. This is supposed to help the AI (assuming it is working properly). Out of the Park Baseball Manual

David Watts · 04-17-2014, 11:19 AM

Quote:

Originally Posted by pstrickert

Notice the "Base roles/positions on" option under Historical Settings. This is supposed to help the AI (assuming it is working properly). Out of the Park Baseball Manual

Exactly. This does have a huge effect on AI usage. Maybe your screenshot will get the point across better than my posts have

JohnHoward · 04-17-2014, 01:08 PM

I always regret responding to these, so here is my final word: the AI will properly reflect historical usage if you give it the chance. That means setting up your league properly and doing the extra work to maintain it. It also means using actual as-played schedules, transactions, roster moves, rotations, and lineups. But you have to do some work.

On the other hand, if you don't do these things, and you do have player development on, and injuries on, and the AI can make roster moves at will, then all bets are off. If you introduce noise into the signal and then expect the AI to filter out all the noise that YOU introduced, then you are working against the AI, not with it. You get what you deserve.

The Game · 04-18-2014, 12:59 AM

Quote:

Originally Posted by Dodger1

I was looking into a few historical replay years, and found some really inexplicable things. I choose to manage one team, and let the ai manage the others. Bt, in 1980, the ai has Paul Splitorff as a middle reliever for KC. But in real life he was in 34 games, 33 as a starter. In 1983, the ai has Scotty McGregor ad a middle reliever, but in real life he was in 38 games, ALL as a starter. In 1965 the ai has Maury Wills batting 3rd for the Dodgers, but he hit leadoff in every game I was able to locate (fitting for someone with 94 steals and zero homers, hardly 3 spot numbers.)

I tried putting myself in charge of all teams, and making the changes to match real life. But once I put the ai back in charge, it changed everything back. I really don't want to use real life lineups, I prefer using injuries. Any thougts how to fix this, and why the ai is so wildly wrong with its decisions?

Default ratings scale thing (sorry forgot what it is called) uses a 3 yr scale for stats. In 1981 Split started 15 games and appeared in 21. I am fairly confident that is affecting his 1980 stats and the way he is being used. I find that a 1 yr scale is best for MLB replays. Also if KC is stacked then Spit may be forced into the pen.

Carlton · 04-18-2014, 08:42 AM

Quote:

Originally Posted by pstrickert

Notice the "Base roles/positions on" option under Historical Settings. This is supposed to help the AI (assuming it is working properly). Out of the Park Baseball Manual

and that does work

but NOT for the initial year...weird I know

in my 1995 replay, it has many SP with relief stamina (but accurate pitch ratings) so I needed to go through all of them and increase the stamina by 40

When my first draft came along...the stamina was fine.

I've seen that happen with the generic DB and Spritze's. To not place fault entirely on the AI, most of the players imported will be guys who haven't played in 4 or 5 years but are still 30-35 yrs old

04-16-2014, 06:28 PM	#7
Dodger1 Minors (Double A) Join Date: Aug 2012 Posts: 137	Disagree I find using actual lineups doesn't always jibe with what's going on in the game. For example, a guy who struggled in real life and was sent down might be hitting .300 in the replay, and wouldn't be sent down. To do so would be totally unrealistic. Or, a guy who was rested irl because he was in a slump might be re hot when he's removed in the game, also unrealistic. But it is totally unrealistic to say that a team's #1 starter (mcGregor was 18-7 with a team best era) should be put in the bullpen in favor of a guy who had an era over 5? Yeah, that's a quality ai.

04-16-2014, 10:00 PM	#11
Dodger1 Minors (Double A) Join Date: Aug 2012 Posts: 137	Simple Question Would it be so hard to have the ai select starting rotations based on, oh, say, ACTUAL NUMBER of games started? Seems pretty basic to me.

04-15-2014, 10:05 PM	#1
Dodger1 Minors (Double A) Join Date: Aug 2012 Posts: 137	Strange Things in Historical Replay I was looking into a few historical replay years, and found some really inexplicable things. I choose to manage one team, and let the ai manage the others. Bt, in 1980, the ai has Paul Splitorff as a middle reliever for KC. But in real life he was in 34 games, 33 as a starter. In 1983, the ai has Scotty McGregor ad a middle reliever, but in real life he was in 38 games, ALL as a starter. In 1965 the ai has Maury Wills batting 3rd for the Dodgers, but he hit leadoff in every game I was able to locate (fitting for someone with 94 steals and zero homers, hardly 3 spot numbers.) I tried putting myself in charge of all teams, and making the changes to match real life. But once I put the ai back in charge, it changed everything back. I really don't want to use real life lineups, I prefer using injuries. Any thougts how to fix this, and why the ai is so wildly wrong with its decisions?

04-15-2014, 11:16 PM	#2
David Watts Hall Of Famer Join Date: Apr 2002 Location: Looking for a place called Leehofooks Posts: 8,869	What do you have base pitcher stamina based on? I would try basing it on current season and see if that changes Splitorff and McGregor.

04-16-2014, 12:53 PM	#3
JohnHoward All Star Reserve Join Date: Jun 2005 Posts: 624	If you are not using real lineups then you are not "replaying", you are "playing". In that regard, there is no basis for complaint against AI decisions. It has no historical memory, so it doesn't know what's "right". You are playing a fictional game with the names of real players.

04-16-2014, 03:01 PM	#5
JCarter93 Minors (Double A) Join Date: Apr 2014 Posts: 157	You can always use the force position function to prevent the AI from doing this.

04-16-2014, 06:38 PM	#8
pstrickert Hall Of Famer Join Date: Dec 2005 Posts: 15,738	The AI usually does a decent job (though not perfect), so I'm wondering what happened with Splittorff and McGregor. How well were they pitching during the replay?

04-17-2014, 09:02 AM	#13
Carlton Hall Of Famer Join Date: Mar 2002 Posts: 3,765	That issue will come up alot, and I will get flack for this, but in a true baseball sense, the AI is average when it comes to anything historical. Fictional league? It's the bees knees as it uses the same algorithms to create the players as it does to set the lineup For the P's what database are you using? And like David stated, what are you importing them as? The easiest fix for P are to go through them and up their stamina until it says projected role as Starter in the Player editor For the lineup It WILL put a leadoff hitter in the #3 hole if all the players it has in the lineup (very power and OBP heavy no matter what the era) have lower BABIP and Avoid Ks, and those catchers that got free passes and rarely K'd because the P let them hit the ball rather than waste pitches? They will bat 2nd WAY too often. The best way to fix this is to have managerial tendencies and roster management actually hold some heavy weight in the programming, that way you can make the manager love AVG>OBP a bit and a good contact hitter will be the starter instead of the .230 30 HR guy, but it doesn't have that weight and I do not think 15 will address that The workaround for the leadoff hitter would be to lower the speed and steal of the leadoff hitter on that team, but that opens up a new can of worms

04-17-2014, 09:41 AM	#14
David Watts Hall Of Famer Join Date: Apr 2002 Location: Looking for a place called Leehofooks Posts: 8,869	I've asked before if it might be time that OOTP installs a "use according to real life" feature to the game. I think it would be a cool feature for people that play using real transactions/retire according to history. That being said, how many people would use it? This game doesn't market itself as a replay game in the same sense Diamond Mind, Action Pc or Strat do. It's a career sim. Once fictional transactions, fictional injuries, fictional expansions etc etc get added to the mix, how is the AI supposed to field a team if it's hogtied by real usage?

04-17-2014, 10:56 AM	#15
pstrickert Hall Of Famer Join Date: Dec 2005 Posts: 15,738	Notice the "Base roles/positions on" option under Historical Settings. This is supposed to help the AI (assuming it is working properly). Out of the Park Baseball Manual

04-17-2014, 01:08 PM	#17
JohnHoward All Star Reserve Join Date: Jun 2005 Posts: 624	I always regret responding to these, so here is my final word: the AI will properly reflect historical usage if you give it the chance. That means setting up your league properly and doing the extra work to maintain it. It also means using actual as-played schedules, transactions, roster moves, rotations, and lineups. But you have to do some work. On the other hand, if you don't do these things, and you do have player development on, and injuries on, and the AI can make roster moves at will, then all bets are off. If you introduce noise into the signal and then expect the AI to filter out all the noise that YOU introduced, then you are working against the AI, not with it. You get what you deserve.