Been testing those AI evaluation numbers again

PSUColonel · 09-20-2013, 06:52 PM

I have been using 40/30/20/10 with hard/neutral trade settings for some time now. I feel there is still a small weakness when it comes to players with a lot of upside who maybe didn't have a good year somewhere...especially the past year. I have been testing extensively and come up with a new set that I feel are just about perfect. They mimic the 40/30/20/10 numbers, but are just tweaked a little to not allow the AI to give away or pass on a player that is scouted highly, but as not to ignore the stats either.

My new numbers are : 45/27/20/08

See what you guys think.

pgjocki · 09-20-2013, 08:18 PM

I haven't had time to run the numbers in my saved league but how does this affect the variance between established players and prospects?

I am always concerned about tilting to far in one direction and skewing the results which will cause the AI to overvalue and not build their teams for sustained future success.

PSUColonel · 09-20-2013, 09:00 PM

I feel it helps the AI realize (on some level) the difference between a highly touted prospect who has not been performing well and makes it less likely to easily trade away a potentially good player for less.

PSUColonel · 09-20-2013, 09:02 PM

Conversely however, I am wondering if something like 30/40/20/10 would act as a setting between average and neutral? (something I know many have been clamoring for)

EDIT: that is with difficulty set at "hard"

PSUColonel · 09-20-2013, 09:17 PM

In fact, after just doing some testing, that's just what it does. I am still testing further though.

For the purpose of making hard a little less hard, I think I might suggest 34/39/19/8

PSUColonel · 09-20-2013, 09:31 PM

OK, for those of you who feel the hard setting at 40/30/20/10 is still to hard, and that average is way too easy, I suggest using the hard setting with neutral preference at 34/39/19/8

PSUColonel · 09-20-2013, 10:08 PM

So, if you want to round it off...to me 35/35/20/10

not much different from the original 40/30/20/10 but I think it give a little more weight to a player who is having a good year, but also has good ratings.

Might not seem like a big difference, but I do think it has a nice effect.

RchW · 09-20-2013, 10:36 PM

I appreciate this thread because I'm struggling with poorly rated players getting mega contracts. I don't care what the evaluation parameters are OOTP should be able to determine bad players from good players and not give the dregs 7yr/$70 million contracts.

See this post.

http://www.ootpdevelopments.com/boar...ml#post3569330

VanillaGorilla · 09-20-2013, 10:46 PM

Since we are talking eval settings, a question:

Does a player's past performance affect his development?

In other words: does a player who played well two years ago, all other things being equal, have a better chance of playing well next year than the player who did not play well 2 years ago?

PSUColonel · 09-20-2013, 11:21 PM

sir MJ seems to think more than 10 is needed for two years ago. Maybe tomorrow I will fool around with 35/30/20/15

PSUColonel · 09-21-2013, 01:22 AM

Quote:

Originally Posted by RchW

I appreciate this thread because I'm struggling with poorly rated players getting mega contracts. I don't care what the evaluation parameters are OOTP should be able to determine bad players from good players and not give the dregs 7yr/$70 million contracts.

See this post.

http://www.ootpdevelopments.com/boar...ml#post3569330

This argument would suggest a high ratings percentage is in fact a good thing.

PSUColonel · 09-21-2013, 01:23 AM

Also going to play around with 45/20/20/15

TGH-Adfabre · 09-21-2013, 10:02 AM

When using AI Eval, does Ratings include Potentials or just current ratings. I am assuming potential but just checking.

Thx

PSUColonel · 09-21-2013, 02:08 PM

yes it includes potentials

SirMichaelJordan · 09-21-2013, 03:35 PM

Quote:

Originally Posted by PSUColonel

sir MJ seems to think more than 10 is needed for two years ago. Maybe tomorrow I will fool around with 35/30/20/15

Yup I use 25 across the board. I settle with that last year after tweaking because I had players who were bad getting long term contracts, etc...

Also I believe the % are prorated until a certain amount of AB's are reached and ratings are used if there isn't enough stats even if you have 0% ratings.

With 25 across the board, A prospect would need 3 bad season. 1 bad season would not over power because hypothetically speaking, he would have two good seasons which are weighted the same as his 1 bad year and not to mention his ratings being ok. Lower rating for 2 years ago will make that good year not mean anything because its not weighted much.

During the current season, his last year stats will over power (which lets say he's coming off a terrible year) because even though he's playing well, his current year stats aren't being fully weighted until he reach a certain amount of AB's. During this time more weight are going into his last year stats until he reach that magic number of AB's.

PSUColonel · 09-21-2013, 03:53 PM

Quote:

Originally Posted by SirMichaelJordan

Yup I use 25 across the board. I settle with that last year after tweaking because I had players who were bad getting long term contracts, etc...

Also I believe the % are prorated until a certain amount of AB's are reached and ratings are used if there isn't enough stats even if you have 0% ratings.

but a lot of your trades must be whacked with numbers like those.

SirMichaelJordan · 09-21-2013, 04:00 PM

Quote:

Originally Posted by PSUColonel

but a lot of your trades must be whacked with numbers like those.

Not at all. I am anal when it comes to realism so I wouldn't have settle with those numbers if that was the case. Main reason I was testing in the first place because trades were unrealistic to me.

Also if you dont check that AI eval setting then you are not seeing what the AI see and that may make trades not make any sense.

nyy26wc · 09-21-2013, 04:27 PM

Quote:

Originally Posted by SirMichaelJordan

Not at all. I am anal when it comes to realism so I wouldn't have settle with those numbers if that was the case. Main reason I was testing in the first place because trades were unrealistic to me.

Also if you dont check that AI eval setting then you are not seeing what the AI see and that may make trades not make any sense.

I think I might try the 25 across the board.

One of my problems is I find that the AI tends to move starters into the bullpen in unrealistic situations. So, maybe this setting might cure that problem too.

TGH-Adfabre · 09-21-2013, 05:08 PM

I am assuming you guys have used Ratings only. It is what I use but I do not play much solo so my online experience has shaped most of my OOTP perspective.

What does this provide that Ratings only does not. I am here almost daily, don't post much, I do not remember any threads contrasting the two options.

Sorry if I missed them.

Thanks,
Tom

RchW · 09-21-2013, 05:43 PM

Quote:

Originally Posted by PSUColonel

This argument would suggest a high ratings percentage is in fact a good thing.

When evaluation is high on ratings then highly rated players that have poor stats get the big contracts. Not sure what's worse.

Quote:

Originally Posted by PSUColonel

yes it includes potentials

Are you sure about this?

09-20-2013, 06:52 PM	#1
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,089 Blog Entries: 37	Been testing those AI evaluation numbers again I have been using 40/30/20/10 with hard/neutral trade settings for some time now. I feel there is still a small weakness when it comes to players with a lot of upside who maybe didn't have a good year somewhere...especially the past year. I have been testing extensively and come up with a new set that I feel are just about perfect. They mimic the 40/30/20/10 numbers, but are just tweaked a little to not allow the AI to give away or pass on a player that is scouted highly, but as not to ignore the stats either. My new numbers are : 45/27/20/08 See what you guys think.

09-20-2013, 08:18 PM	#2
pgjocki All Star Starter Join Date: Jun 2013 Location: Maryland - just outside DC Posts: 1,575 Blog Entries: 1	I haven't had time to run the numbers in my saved league but how does this affect the variance between established players and prospects? I am always concerned about tilting to far in one direction and skewing the results which will cause the AI to overvalue and not build their teams for sustained future success. __________________ - - - World Series championships: 1926, 1931, 1934, 1942, 1944, 1946, 1964, 1967, 1982, 2006, 2011

09-20-2013, 09:02 PM	#4
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,089 Blog Entries: 37	Conversely however, I am wondering if something like 30/40/20/10 would act as a setting between average and neutral? (something I know many have been clamoring for) EDIT: that is with difficulty set at "hard" Last edited by PSUColonel; 09-20-2013 at 09:06 PM.

09-20-2013, 09:17 PM	#5
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,089 Blog Entries: 37	In fact, after just doing some testing, that's just what it does. I am still testing further though. For the purpose of making hard a little less hard, I think I might suggest 34/39/19/8 Last edited by PSUColonel; 09-20-2013 at 09:27 PM.

09-20-2013, 10:08 PM	#7
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,089 Blog Entries: 37	So, if you want to round it off...to me 35/35/20/10 not much different from the original 40/30/20/10 but I think it give a little more weight to a player who is having a good year, but also has good ratings. Might not seem like a big difference, but I do think it has a nice effect. Last edited by PSUColonel; 09-20-2013 at 10:09 PM.

09-20-2013, 09:00 PM	#3
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,089 Blog Entries: 37	I feel it helps the AI realize (on some level) the difference between a highly touted prospect who has not been performing well and makes it less likely to easily trade away a potentially good player for less.

09-20-2013, 09:31 PM	#6
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,089 Blog Entries: 37	OK, for those of you who feel the hard setting at 40/30/20/10 is still to hard, and that average is way too easy, I suggest using the hard setting with neutral preference at 34/39/19/8

09-20-2013, 10:36 PM	#8
RchW Hall Of Famer Join Date: Jul 2004 Location: Toronto ON by way of Glasgow UK Posts: 15,629	I appreciate this thread because I'm struggling with poorly rated players getting mega contracts. I don't care what the evaluation parameters are OOTP should be able to determine bad players from good players and not give the dregs 7yr/$70 million contracts. See this post. http://www.ootpdevelopments.com/boar...ml#post3569330 __________________ Cheers RichW If you’re looking for a good cause to donate money to please consider a Donation to Parkinson’s Canada. It may help me have a better future and if not me, someone else. Thanks. “Conservatism consists of exactly one proposition …There must be in-groups whom the law protects but does not bind, alongside out-groups whom the law binds but does not protect.” Frank Wilhoit

09-20-2013, 10:46 PM	#9
VanillaGorilla All Star Starter Join Date: Dec 2011 Posts: 1,371 Blog Entries: 2	Since we are talking eval settings, a question: Does a player's past performance affect his development? In other words: does a player who played well two years ago, all other things being equal, have a better chance of playing well next year than the player who did not play well 2 years ago?

09-20-2013, 11:21 PM	#10
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,089 Blog Entries: 37	sir MJ seems to think more than 10 is needed for two years ago. Maybe tomorrow I will fool around with 35/30/20/15

09-21-2013, 01:23 AM	#12
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,089 Blog Entries: 37	Also going to play around with 45/20/20/15

09-21-2013, 10:02 AM	#13
TGH-Adfabre Hall Of Famer Join Date: Sep 2007 Posts: 2,599	When using AI Eval, does Ratings include Potentials or just current ratings. I am assuming potential but just checking. Thx __________________ You mock me, therefore I am My wife

09-21-2013, 02:08 PM	#14
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,089 Blog Entries: 37	yes it includes potentials

09-21-2013, 05:08 PM	#19
TGH-Adfabre Hall Of Famer Join Date: Sep 2007 Posts: 2,599	I am assuming you guys have used Ratings only. It is what I use but I do not play much solo so my online experience has shaped most of my OOTP perspective. What does this provide that Ratings only does not. I am here almost daily, don't post much, I do not remember any threads contrasting the two options. Sorry if I missed them. Thanks, Tom __________________ You mock me, therefore I am My wife