AI too many strange waivers

clamel · 06-23-2008, 07:55 AM

Started a dynasty in 1987

Tom Glavine got waived from Atlanta
Dave Stieb got waived and claimed (Toronto to Milwaukee)
Blue Jays decided to have 68 game reliver Jeff Mussleman in the rotation with Tom Filer/Jim Clancy/Jimmy Key/Joe Nunez and Musselman

OK strange things can happen but these are just after looking over the startup after 4-5 days. I´m sure numerous other very strange things happens.

The Musselman thing might be because he's got 15 starts in 1988, but Dave Stieb should more suitable IMHO. John Cerutti also got waived btw

If this is the top of the mountain I see I would like to see perhaps a more traditional AI, that is more holding on to players they got when you start.
The 80´s and 90´s didn´t see that much transaction as today.

One thing I could try is to as a commish look over every roster and make certain long serving players very popular with their teams. Perhaps it solves some troubles.

Any more have any view on this, or should I drop the problem to be unresolveable ???

Another stuff is that the active rosters seams to "tilt" on either 4-5 RF or LF then just 1 on the other. Strange with the AI setting up who´s on right and who´s on left.

The Wolf · 06-23-2008, 09:18 AM

Change the settings on how your AI evaluates talent.

I keep seeing "the game does X" threads, where the settings can be changed to stop it from happening. ("The game's way too easy!" "Okay, turn ratings off so you have to judge by stats only, set trades to 'very hard' and 'prefer rookies' and come back in a week and say that.")

ashantewarrier · 06-23-2008, 10:21 AM

Quote:

Originally Posted by The Wolf

Change the settings on how your AI evaluates talent.

I keep seeing "the game does X" threads, where the settings can be changed to stop it from happening. ("The game's way too easy!" "Okay, turn ratings off so you have to judge by stats only, set trades to 'very hard' and 'prefer rookies' and come back in a week and say that.")

Agreed.

The AI does not evaluate talent based on name..it would not distinguish Tom Glavine from Joe Nobody if the ratings were similar. That's the beauty of OOTP..customizable options.

Raidergoo · 06-23-2008, 10:49 AM

Tom Glavine was not good in 1987 statistically, and was not that much better in 1988. He walked more batters than he struck out, and he does not strike many batters out.

If there would have been a pressing need, the AI would have ditched him.

The 1987 release of Steib is troubling. He's got some flaws, but he's I think that he's in the top 33% of SP for that season.

clamel · 06-23-2008, 05:36 PM

I have AI very hard but not to scrap ratings ....

How will the AI judge without ratings.

Of course the AI doesn´t know Tom Glavine or what he becomes but somehow his ratings should indicate that he could be at least a valuable pitcher in the 2-3 years.

The Glavine issue of course gets bigger looking at Stieb. You got a prime starter ditched for a good reliever Musselman. he had 68 games no starts and still have a 5 in endurance letting him be judge as a starter in 1987.
I had hoped that he had lower endurance for 1987, then in 1988 add a little endurance to be good enough to start.

I tried out Double values on the present year, which I hope is perhaps one key to this Stieb problem. Hardly to Glavine (even the opposite).
Of course the waiver never happend.
So I move on and consider it a fluke happening, like some real life trades have been during history.
How could Cubs get rid of Maddux (kind of crazy AI)

What will happend if one in commish mode make future player very popular in the editor. Will the AI change it back or might the player stay popular until he really comes of age ??

SittingDuck · 06-23-2008, 06:18 PM

Really, the best thing to do is create your league, back it up and then test with various settings. You have to find what works best for you in your settings.

For me, I use this for historicals:

Ratings = 20
This years stats = 45
Last year's stats = 25
2 year's ago = 10

There are all sorts of variables, so try them out and see how that affects things.

ALSO, give your teams a bigger cash allowance in the League Setup/Rules section. You don't want them dumping them for salary purposes. The whole financial aspect of ootp9 is still be analyzed and tweaked.

andydct · 06-24-2008, 04:27 AM

Quote:

Originally Posted by The Wolf

Change the settings on how your AI evaluates talent.

I keep seeing "the game does X" threads, where the settings can be changed to stop it from happening. ("The game's way too easy!" "Okay, turn ratings off so you have to judge by stats only, set trades to 'very hard' and 'prefer rookies' and come back in a week and say that.")

This sounds like a bit of a copout for the AI to me, basically we're saying that the AI can be competitive with a human who has limited information. Basically the AI will continue to make mistakes but its human opponent will not notice. I guess that's "better" but it still seems like it could use work.

That being said, I may give this a try personally, because I love a challenge.

rasnell · 06-24-2008, 05:34 AM

Quote:

Originally Posted by andydct

This sounds like a bit of a copout for the AI to me, basically we're saying that the AI can be competitive with a human who has limited information. Basically the AI will continue to make mistakes but its human opponent will not notice. I guess that's "better" but it still seems like it could use work.

That being said, I may give this a try personally, because I love a challenge.

Some players master the game sooner than others. They are merely looking for ways to increase the challenge. The game has so many ways to customize that you can challenge yourself. I would not call this an AI flaw.

Hador · 06-24-2008, 06:18 AM

Quote:

Originally Posted by andydct

This sounds like a bit of a copout for the AI to me, basically we're saying that the AI can be competitive with a human who has limited information. Basically the AI will continue to make mistakes but its human opponent will not notice. I guess that's "better" but it still seems like it could use work.

Well, that is the case in EVERY game. No matter what game you play the AI will always make mistakes. If you'd ever come up against a perfect AI you would lose every single game. How much fun would that be?

Amongst the most difficult things when designing a game is to program an AI that is challenging while still leaving the average player a chance to win the game. That is why games have various difficulty settings and that is why all those settings you can change in OOTP make this game so much fun for so many different people.

Hador · 06-24-2008, 06:22 AM

One other thing:

If you look at other games (Civ 4 comes to mind) then you will see what a real cop out looks like. In those games the game designers don't spend enough time on the AI and if you play the game at higher levels what will happen is simply that the AI starts to cheat. For instance if building a unit takes you 10 rounds the AI will be able to build it in 5 and other stuff like that.

In OOTP as far as I can tell that has never been the case. Rather the game uses a very sophisticated AI that, on top of all, can be customized in just about every way you can think of.

JWay · 06-24-2008, 08:38 AM

also in historicals the AI doesn't know who these players are, and you know how their careers finish, so its quite an advantage to you. Every league is just like a fictional league to the AI.

06-23-2008, 07:55 AM	#1
clamel All Star Starter Join Date: Apr 2008 Location: Across the Pond Posts: 1,037	AI too many strange waivers Started a dynasty in 1987 Tom Glavine got waived from Atlanta Dave Stieb got waived and claimed (Toronto to Milwaukee) Blue Jays decided to have 68 game reliver Jeff Mussleman in the rotation with Tom Filer/Jim Clancy/Jimmy Key/Joe Nunez and Musselman OK strange things can happen but these are just after looking over the startup after 4-5 days. I´m sure numerous other very strange things happens. The Musselman thing might be because he's got 15 starts in 1988, but Dave Stieb should more suitable IMHO. John Cerutti also got waived btw If this is the top of the mountain I see I would like to see perhaps a more traditional AI, that is more holding on to players they got when you start. The 80´s and 90´s didn´t see that much transaction as today. One thing I could try is to as a commish look over every roster and make certain long serving players very popular with their teams. Perhaps it solves some troubles. Any more have any view on this, or should I drop the problem to be unresolveable ??? Another stuff is that the active rosters seams to "tilt" on either 4-5 RF or LF then just 1 on the other. Strange with the AI setting up who´s on right and who´s on left.

06-23-2008, 05:36 PM	#5
clamel All Star Starter Join Date: Apr 2008 Location: Across the Pond Posts: 1,037	I have AI very hard but not to scrap ratings .... How will the AI judge without ratings. Of course the AI doesn´t know Tom Glavine or what he becomes but somehow his ratings should indicate that he could be at least a valuable pitcher in the 2-3 years. The Glavine issue of course gets bigger looking at Stieb. You got a prime starter ditched for a good reliever Musselman. he had 68 games no starts and still have a 5 in endurance letting him be judge as a starter in 1987. I had hoped that he had lower endurance for 1987, then in 1988 add a little endurance to be good enough to start. I tried out Double values on the present year, which I hope is perhaps one key to this Stieb problem. Hardly to Glavine (even the opposite). Of course the waiver never happend. So I move on and consider it a fluke happening, like some real life trades have been during history. How could Cubs get rid of Maddux (kind of crazy AI) What will happend if one in commish mode make future player very popular in the editor. Will the AI change it back or might the player stay popular until he really comes of age ?? Last edited by clamel; 06-23-2008 at 05:40 PM.

06-23-2008, 10:49 AM	#4
Raidergoo Hall Of Famer Join Date: Mar 2003 Posts: 9,005	Tom Glavine was not good in 1987 statistically, and was not that much better in 1988. He walked more batters than he struck out, and he does not strike many batters out. If there would have been a pressing need, the AI would have ditched him. The 1987 release of Steib is troubling. He's got some flaws, but he's I think that he's in the top 33% of SP for that season.

06-23-2008, 06:18 PM	#6
SittingDuck All Star Starter Join Date: Apr 2007 Location: Richmond, VA Posts: 1,834	Really, the best thing to do is create your league, back it up and then test with various settings. You have to find what works best for you in your settings. For me, I use this for historicals: Ratings = 20 This years stats = 45 Last year's stats = 25 2 year's ago = 10 There are all sorts of variables, so try them out and see how that affects things. ALSO, give your teams a bigger cash allowance in the League Setup/Rules section. You don't want them dumping them for salary purposes. The whole financial aspect of ootp9 is still be analyzed and tweaked.

06-24-2008, 06:22 AM	#10
Hador Minors (Single A) Join Date: May 2005 Location: Hannover, Germany Posts: 65	One other thing: If you look at other games (Civ 4 comes to mind) then you will see what a real cop out looks like. In those games the game designers don't spend enough time on the AI and if you play the game at higher levels what will happen is simply that the AI starts to cheat. For instance if building a unit takes you 10 rounds the AI will be able to build it in 5 and other stuff like that. In OOTP as far as I can tell that has never been the case. Rather the game uses a very sophisticated AI that, on top of all, can be customized in just about every way you can think of.

06-24-2008, 08:38 AM	#11
JWay Hall Of Famer Join Date: Dec 2001 Location: Sioux Falls, SD Posts: 5,021	also in historicals the AI doesn't know who these players are, and you know how their careers finish, so its quite an advantage to you. Every league is just like a fictional league to the AI.