AI Thread - Substitions and other AI Issues..

[david limbaugh]

I started this thread in hopes we can give exact situations where we see the AI could be improved.

Instead of anecdotal comments perhaps if we give Markus and the gang exact game situations they can walk through the code and make improvements....

Who wants to go first?

[SirMichaelJordan]

Figured I should post this here.

I've been paying extra attention to the Closers getting way more decisions than they should compared to real life problem (its been a thread before) and I believe the problem is that in some situations, the closers are getting put in the game despite the game being tied. The gameplay strategy settings will not tame any of this. Its as if the closer is program to come in the game during the 9th+ innings (if you are using "use closer" at the "very often" setting) as long as they are not losing.

I haven't seen a situation where the closer comes in for a losing team but I wouldn't rule that out with the small sample size.


Never seen any of this in the previous games but I am not sure if its a bug.


Edit- I am seeing closer come in at the 9th despite losing but I should mention that I am in the playoffs right now. It should be interesting to see what happens during the season.

In my last game, my team was up 9-1 but the AI brung in their closer for the top of the 9th.

Checking the AI games there was a situation where the AI used its closer in a tied game in the 9th but replaced him with the other setup man during the 10th inning.

Maybe this should go in that AI thread lol.

I should mention that this caught my attention when I seen one my closer (in a sim save) record 17 wins along with 40 saves. He also had 3 loses and never recorded a GS.

[ScottM816]

As you can see CF James Jones is starting the game at SS, where he has no experience at.



Also not sure why they are carrying 3 1Bs and 1 backup in the infield (Rickie Weeks)



Keep in mind during this game, Cano got hurt so Weeks replaced him

[david limbaugh]

Rescouting monthly.

First of the month. Lineup - Manager sets lineups/depth charts.

David Wills HOT - 10-22 in last 6 games.

AI / Manager - bench him for Davey Crockett

Benches a 1-16-.351 in 27 games for a 0-2-333 in 7 games

I wish AI / Managers would not do this....

[SirMichaelJordan]

Team focus needs to be more dynamic. I am 99% sure it only changes after a season. It would be cool to see it change around July (or whatever the half point in the season is in the league in question) because there are times where a rebuilding team end up winning as much as 90 or so games but end up trading one of their top players for prospects near the deadline.

I am not suggesting that a team should jump from Rebuilding to Win now in July but I think in this scenario it would be more realistic (and benefit the AI) for a overachieving team go from rebuilding to neutral. Same for underachievers as they should go from Win now to neutral during the season. I would also like to see teams that are neutral jump to win now or rebuild depending on the situation.

[Orcin]

Team focus changes for several teams in early June.

[SirMichaelJordan]

Originally Posted by Orcin

Team focus changes for several teams in early June.

I know it did in previous versions but this year I think its more tied to the owner and his personality. Its normal to see a team who was set to rebuild at the start of the year, do great and still has a focus to rebuild. This would not be a problem if these teams didn't start trading away their best players for minor league prospects and most of the time their best players are barely 26. (Up for Arb)

Have you tracked this?

[RchW]

Originally Posted by david limbaugh

Rescouting monthly.

First of the month. Lineup - Manager sets lineups/depth charts.

David Wills HOT - 10-22 in last 6 games.

AI / Manager - bench him for Davey Crockett

Benches a 1-16-.351 in 27 games for a 0-2-333 in 7 games

I wish AI / Managers would not do this....

Can you explain the bold. Indecipherable to me. I must be missing something obvious.

I've seen a player get hot for a month. If my AI manager sat a solid starter (especially if he is not cold) for more than a week because his backup was hot, I'd consider that unrealistic first and a firing offense second.

I'd be totally against any AI adjustment that favored streakiness for more than 5-8 games. Even then I'd favor platoon splits over a streak.

[david limbaugh]

Originally Posted by rchw

can you explain the bold. Indecipherable to me. I must be missing something obvious.

I've seen a player get hot for a month. If my ai manager sat a solid starter (especially if he is not cold) for more than a week because his backup was hot, i'd consider that unrealistic first and a firing offense second.

I'd be totally against any ai adjustment that favored streakiness for more than 5-8 games. Even then i'd favor platoon splits over a streak.

hr-rbi-avg :-)

[madJ]

Originally Posted by david limbaugh

hr-rbi-avg :-)

Driving in runs is not a skill. You need runners on base in front of you to be able to collect RBI. Thus a .333 hitter isn't that much worse than a .351 hitter. That one home run (in 20 more games played) isn't that big of a difference. So I'd argue that the AI had to decide which one of the two hot hitters to bench and you just didn't like the decision. How did the 1-16-.351 hitter do in his last 7 games (not his last 27)?

[david limbaugh]

Originally Posted by madJ

Driving in runs is not a skill. You need runners on base in front of you to be able to collect RBI. Thus a .333 hitter isn't that much worse than a .351 hitter. That one home run (in 20 more games played) isn't that big of a difference. So I'd argue that the AI had to decide which one of the two hot hitters to bench and you just didn't like the decision. How did the 1-16-.351 hitter do in his last 7 games (not his last 27)?

:-) The 1-16-.351 guy is the 'hot' guy. My argument with the AI is that what did he do to get benched? Not run out a ground ball? Stand and look at a pop up? Take the managers daughter out on a date? ????

[madJ]

Originally Posted by david limbaugh

:-) The 1-16-.351 guy is the 'hot' guy. My argument with the AI is that what did he do to get benched? Not run out a ground ball? Stand and look at a pop up? Take the managers daughter out on a date? ????

Maybe the AI doesn't like his face or thinks he doesn't have the kind of body scouts like at that position.

[SirMichaelJordan]

My main gripe with OOTP have been with the way fielding and positions are handled. Changing the setting to have players rated compared to each other rather than filtered by positions helps (player listed as a 2B and rated against other 2B despite playing SS for the entire year as well as giving the AI a false sense of weakness to their depth for example) as does changing AI eval settings, but it will not correct something that is flawed which is the AI not being smart enough to change a player's position as he switches on the depth chart. Maybe the simple fix is to not give so many generated players good to great position ratings at multiple positions which fills the universe up with HOF utility players. I understand players can play multiple positions but do we really need recently drafted college players being rated 4's 5's & 6's (on a 2-8 scale) at more than one position? This causes things like the AI DFAing the wrong people. For example, recently I've seen a team that had a weakness at 3B, their best 3B should have easily been DFA'd and not been on the opening day roster. Instead of making a move to upgrade the position they smartly placed their best 1B at 3B who have been playing 3B for them last season at a All star level (because of course he can play there very well) but failed to list him as a 3B for some reason.

The second best 1B is their top FA signing this past offseason who was bought in to fill a hole at RF (they currently have a terrible RF as if now and the guy they brought in played most of his career at RF) but for some reason he's listed as a 1B.

Now, here's the kicker...The 1B who started all last season and led the team in HR's is now DFA'd because he's the 3rd 1B on the roster..The team now have a terrible starting RF while giving up one of their top players in the process all in favor of one minor league player because he's listed as their best 3B at the moment. I am sure there are more examples of this going on if I keep looking for it and the problem might correct itself as the season move along when the AI finally decides to make the position change but this might be what other users are seeing when they posted about the AI waivering good players despite taking advice of using a stat heavy AI Eval.

I am sure option years are affected by this as well.

Note that this problem may not exist with feeder leagues as the generated players for those leagues do not tend to have godly ratings at multiple positions.

[David Watts]

Originally Posted by SirMichaelJordan

My main gripe with OOTP have been with the way fielding and positions are handled. Changing the setting to have players rated compared to each other rather than filtered by positions helps (player listed as a 2B and rated against other 2B despite playing SS for the entire year as well as giving the AI a false sense of weakness to their depth for example) as does changing AI eval settings, but it will not correct something that is flawed which is the AI not being smart enough to change a player's position as he switches on the depth chart. Maybe the simple fix is to not give so many generated players good to great position ratings at multiple positions which fills the universe up with HOF utility players. I understand players can play multiple positions but do we really need recently drafted college players being rated 4's 5's & 6's (on a 2-8 scale) at more than one position? This causes things like the AI DFAing the wrong people. For example, recently I've seen a team that had a weakness at 3B, their best 3B should have easily been DFA'd and not been on the opening day roster. Instead of making a move to upgrade the position they smartly placed their best 1B at 3B who have been playing 3B for them last season at a All star level (because of course he can play there very well) but failed to list him as a 3B for some reason.

The second best 1B is their top FA signing this past offseason who was bought in to fill a hole at RF (they currently have a terrible RF as if now and the guy they brought in played most of his career at RF) but for some reason he's listed as a 1B.

Now, here's the kicker...The 1B who started all last season and led the team in HR's is now DFA'd because he's the 3rd 1B on the roster..The team now have a terrible starting RF while giving up one of their top players in the process all in favor of one minor league player because he's listed as their best 3B at the moment. I am sure there are more examples of this going on if I keep looking for it and the problem might correct itself as the season move along when the AI finally decides to make the position change but this might be what other users are seeing when they posted about the AI waivering good players despite taking advice of using a stat heavy AI Eval.

I am sure option years are affected by this as well.

I hope everyday that more and more people start thinking this way. Rchw has been preaching this for quite some time now. Great post.

[SirMichaelJordan]

Originally Posted by David Watts

I hope everyday that more and more people start thinking this way. Rchw has been preaching this for quite some time now. Great post.

Yes, its been getting better especially for this version as players are more likely to be placed at their best rated position but this doesn't fix the problem of the AI listing players at one position but decides to play them at another.

For starters there should be a bigger gap in ratings between a player's best position and his secondary positions.

A typical generated player at the age of 22 after the draft would have position ratings of 7, 6, 6, 5 (2-8 scale) at four different positions. Which translates to Good, above average, above average, average on any scale.

[markmcghee]

I'm seeing terrible pinch-hit-for-pitcher-late-in-close-games logic. Even with the Pinch Hit for Pitch setting set to Very Often, the AI will usually allow his pitcher to bat in the 7th, 8th, and 9th innings when trailing by 2 or more runs and often even when two are out. I've seen this in old time leagues as well as in leagues utilizing the 2015 rosters. Can't count how many games the AI has lost due to poor pinch hitting logic. If the AI is trailing in the 7th, 8th, or 9th and the pitcher is due up with runners on-board, the AI should pinch hit 99% of the time to at least give himself a shot at getting back into the game.

[MizzouRah]

The issue I'm seeing is the AI playing say, a 3/10 rated RF but then put a 9/10 rated RF at DH. Should be the other way around.

[ScottM816]

If only there was a way to use No Force Start for certain positions...Say I have Josh Hamilton, I want him only use in RF/LF, maybe something to say Do Not Start at CF.

[david limbaugh]

Originally Posted by ScottM816

If only there was a way to use No Force Start for certain positions...Say I have Josh Hamilton, I want him only use in RF/LF, maybe something to say Do Not Start at CF.

I am sorta asking for that in asking for a settings option to ONLY play players at positions they have a certain rating at, with the option of setting that rating threshold, on either a league wide basis, or by manager. Some managers may not want Josh in CF. Some might...

[IsaacR]

Not a huge deal, but the AI is starting Howe in left in Marroquín in right, despite Howe being superior to Marroquín as a right fielder, while Marroquín is more experienced in left.