OOTP Developments Forums - View Single Post

PSUColonel · 08-24-2021, 02:50 PM

Here are some observations:

One thing I have learned is I think the bulk of the weight needs to go to ratings. Something else I feel we all may have overlooked, is the "two years ago" slot doesn't get enough love. I'm not saying it should be on par with "this year", or "last year", but I think the idea of it only getting 5% all of this time in most equations is likely too low.

I have been running tests with the three mentioned above...and really like them all. There isn't a lot of difference between them but there are a few.

I think which I will ultimately end up using will depend on what I might want.

The first thing I notice is that trading is tighter (in terms of a tougher AI) with 55 as the ratings weight. Not between AI teams though. (as some commonly mistake) I am talking about negotiations between the player and the AI.

Now, moving from here, I find the more stats you implement, the better the AI is at contracts, lineups, pitching rotations (bullpen) and other transactions. The one exception can be the waiver wire. I find 55 ratings weight not only helps with human vs AI trades, but also helps the AI with choosing what players go on waivers.

Another observation is that last year also needs some more love. Now, when I say this, you have to take in context. What I mean by this is:

let's say I choose 55 as my ratings weight. Now, everything else becomes relative.

if I were to choose 55/25/15/5...2 years ago and last year jest aren't getting enough representation. ratings in this case are 11x more important than 2 years ago. Last Year is 3x more important than two years ago, and ratings are 3.6x more important than last year. This year is 1.6x more important than last year. (which isn't bad) Ratings is 2.2x more important than this year. (not bad either). To me the issue is that ratings are 3.6x more important than last year, and 11x more important than two years ago.

if I shift the ratings around a little, to say 50/20/20/10 it makes a bit of a difference. In this case, the first and most obvious thing we all see is a 50/50 split between ratings and stats. That's nice, but it's not the reason I like it. The reason I like it, is the weights. I believe "This Year" & "Last Year" need to be very close in valuation. In this case they are obviously identical. Ideally I might like to have "This Year" slightly higher, but I don't think this is too bad. Clearly those two combined account for 40% of the valuation...but they are both only 2x more important than 2 years ago. Ratings are now 5x (as opposed to 11)more important than two years ago, and 2.5x more important than "This Year, and "Last Year".

If you take the example of 55/20/15/10...you get some pretty good results also. Like I said, with this setting, Human vs AI trading and waivers are helped out a bit more.

Last year is weighted 1.5x more than "Two Years Ago". "This year is weighted double two year ago (2x) and 1.3x more important last year.

Ratings are now 5.5x greater than two years, and" 3.6x. greater than last year. They are also weighted 2.75 times more than this year.

So, in conclusion, The ideal setup to me is either 55/20/15/10 or 50/20/20/10.

Both work well. The only difference is. little minute difference in trading and waivers AI. 55 just seems to be that threshold for some reason.

08-24-2021, 02:50 PM	#79
PSUColonel Hall Of Famer Join Date: Feb 2002 Posts: 13,011 Blog Entries: 37	Here are some observations: One thing I have learned is I think the bulk of the weight needs to go to ratings. Something else I feel we all may have overlooked, is the "two years ago" slot doesn't get enough love. I'm not saying it should be on par with "this year", or "last year", but I think the idea of it only getting 5% all of this time in most equations is likely too low. I have been running tests with the three mentioned above...and really like them all. There isn't a lot of difference between them but there are a few. I think which I will ultimately end up using will depend on what I might want. The first thing I notice is that trading is tighter (in terms of a tougher AI) with 55 as the ratings weight. Not between AI teams though. (as some commonly mistake) I am talking about negotiations between the player and the AI. Now, moving from here, I find the more stats you implement, the better the AI is at contracts, lineups, pitching rotations (bullpen) and other transactions. The one exception can be the waiver wire. I find 55 ratings weight not only helps with human vs AI trades, but also helps the AI with choosing what players go on waivers. Another observation is that last year also needs some more love. Now, when I say this, you have to take in context. What I mean by this is: let's say I choose 55 as my ratings weight. Now, everything else becomes relative. if I were to choose 55/25/15/5...2 years ago and last year jest aren't getting enough representation. ratings in this case are 11x more important than 2 years ago. Last Year is 3x more important than two years ago, and ratings are 3.6x more important than last year. This year is 1.6x more important than last year. (which isn't bad) Ratings is 2.2x more important than this year. (not bad either). To me the issue is that ratings are 3.6x more important than last year, and 11x more important than two years ago. if I shift the ratings around a little, to say 50/20/20/10 it makes a bit of a difference. In this case, the first and most obvious thing we all see is a 50/50 split between ratings and stats. That's nice, but it's not the reason I like it. The reason I like it, is the weights. I believe "This Year" & "Last Year" need to be very close in valuation. In this case they are obviously identical. Ideally I might like to have "This Year" slightly higher, but I don't think this is too bad. Clearly those two combined account for 40% of the valuation...but they are both only 2x more important than 2 years ago. Ratings are now 5x (as opposed to 11)more important than two years ago, and 2.5x more important than "This Year, and "Last Year". If you take the example of 55/20/15/10...you get some pretty good results also. Like I said, with this setting, Human vs AI trading and waivers are helped out a bit more. Last year is weighted 1.5x more than "Two Years Ago". "This year is weighted double two year ago (2x) and 1.3x more important last year. Ratings are now 5.5x greater than two years, and" 3.6x. greater than last year. They are also weighted 2.75 times more than this year. So, in conclusion, The ideal setup to me is either 55/20/15/10 or 50/20/20/10. Both work well. The only difference is. little minute difference in trading and waivers AI. 55 just seems to be that threshold for some reason.