Need Input With AI Ratings/Stats Weight

PSUColonel · 12-12-2021, 03:30 AM

I am undecided about what ratings and stats weight I want to use and cannot makeup my mind. I have it narrowed down to a few candidates, but I just cannot seem to make a decision without constantly second guessing. Here are my candidates...help me to decide which one you think helps create the best challenge and realism as far as AI behavior goes. I do realize most of you will think it probably doesn't make that much of a difference which one I use, but like I said, I can't makeup my mind, so I am looking for input in terms of which one you think is best of these. OK thanks...

PSUColonel · 12-12-2021, 01:48 PM

I realize we as a forum community have debated and discussed this topic ad nauseam for years...but the reasoning for some of the seemingly slight differences here, is IMO adding even just a little more weight to last year, and 2 years ago helps the AI to not overreact to a good or bad single year. At the same time, I still feel the weights should be mostly ratings driven, and I also believe in a a "slope" effect where the weight values regress a bit as you go back in time.

I have always been a big advocate of 55/25/15/5....but I recently feel this mix may very well be a problem for last year, and two years. That's why I am leaning a bit with 55/20/15/10. While the 15 for last year is the same, but the 25 becoming a 20, and the 5 becoming a 10 (in theory) should help the AI not overvalue one year over another too greatly.

koohead · 12-12-2021, 05:16 PM

I too have used 55/25/15/5 for many versions. Every time I read one of these threads I second guess myself but there is never consensus so I just leave well enough alone.

PSUColonel · 12-12-2021, 05:50 PM

Quote:

Originally Posted by koohead

I too have used 55/25/15/5 for many versions. Every time I read one of these threads I second guess myself but there is never consensus so I just leave well enough alone.

yup...I am pretty set with ratings at 55 I think....for me, it's really about 55/25/15/5 & 55/20/15/10. They seem very similar (and they are) but I am starting to think that little extra in the 2 years ago category helps the AI make some more realistic evaluations, while still not sacrificing challenge (55 ratings)

PSUColonel · 12-13-2021, 01:19 AM

If you think about it....when using 55/25/15/5....ratings are weighted 11x more than 2 years ago. When using one of the other ones, ratings are weighted either 5 or 5.5x more than 2 years ago. The later seems more reasonable to me.

Syd Thrift · 12-14-2021, 09:55 AM

I use 0/67/22/11 but ymmv.

progen · 12-14-2021, 10:48 AM

Quote:

Originally Posted by Syd Thrift

I use 0/67/22/11 but ymmv.

Same here

View Poll Results: Which of these AI Evaluations Would You Use?
55/25/15/5	3	33.33%
55/20/15/10	2	22.22%
50/25/15/10	4	44.44%
50/20/20/10	0	0%
Voters: 9. You may not vote on this poll

12-12-2021, 03:30 AM	#1
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,107	Need Input With AI Ratings/Stats Weight I am undecided about what ratings and stats weight I want to use and cannot makeup my mind. I have it narrowed down to a few candidates, but I just cannot seem to make a decision without constantly second guessing. Here are my candidates...help me to decide which one you think helps create the best challenge and realism as far as AI behavior goes. I do realize most of you will think it probably doesn't make that much of a difference which one I use, but like I said, I can't makeup my mind, so I am looking for input in terms of which one you think is best of these. OK thanks...

12-12-2021, 05:16 PM	#3
koohead Hall Of Famer Join Date: Dec 2001 Location: Phoenix, AZ Posts: 3,211	I too have used 55/25/15/5 for many versions. Every time I read one of these threads I second guess myself but there is never consensus so I just leave well enough alone. __________________ GM - New Jersey Bears of the NPBL;

12-12-2021, 01:48 PM	#2
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,107	I realize we as a forum community have debated and discussed this topic ad nauseam for years...but the reasoning for some of the seemingly slight differences here, is IMO adding even just a little more weight to last year, and 2 years ago helps the AI to not overreact to a good or bad single year. At the same time, I still feel the weights should be mostly ratings driven, and I also believe in a a "slope" effect where the weight values regress a bit as you go back in time. I have always been a big advocate of 55/25/15/5....but I recently feel this mix may very well be a problem for last year, and two years. That's why I am leaning a bit with 55/20/15/10. While the 15 for last year is the same, but the 25 becoming a 20, and the 5 becoming a 10 (in theory) should help the AI not overvalue one year over another too greatly.

12-13-2021, 01:19 AM	#5
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,107	If you think about it....when using 55/25/15/5....ratings are weighted 11x more than 2 years ago. When using one of the other ones, ratings are weighted either 5 or 5.5x more than 2 years ago. The later seems more reasonable to me.