player eval AI settings

[oomm]

Some use settings like:
5
55
25
15


in which the emphasis is on the current year statistical performance. But how is this accurate? Seems to me that player eval IRL is always based on past rather than present performance. Shouldn't it be more like


5
30
40
25?


(of course this is presuming nothing about how OOTP uses such settings nor how they play out in the course of a OOTP season)

[PSUColonel]

Ahh…the age old debate…lately I’ve been using 25/25/25/25 and even though I have traditionally supported a high ratings weight, it seems to be working very well.

[Reed]

The setting depends entirely on what you are trying to accomplish.

For veterans I would agree that past performance may be more important (unless they are starting to enter their declining years then current year is very important) but for youngsters then more emphasis is on present performance.

[Sweed]

Originally Posted by PSUColonel

Ahh…the age old debate…lately I’ve been using 25/25/25/25 and even though I have traditionally supported a high ratings weight, it seems to be working very well.

This is where I've settled in at too. Can't remember if this came up in the yearly evaluation thread last year or the year before. In any case with my slow style of playing 'em all out it has worked well.

[Pdubya64]

It makes me wonder what happens to how the game evaluates prospects if the Ratings number gets below a certain threshold... say, 35% or so?

They have essentially no "history", so wouldn't lowering the Ratings # affect how they end up being evaluated? Does it even matter?

[oomm]

Ya I know this has likely been kickda round and back again. Ratings weight refers to ratings, and ratings would reflect performance data. So stats from previous seasons is reflected in the ratings, no? iow, ratings, if used logically, would reflect a player's 'reputation' garnered from previous seasons, nay?

[Brad K]

What is the default on these settings?


EDIT: OK, I figured it out. for historical its 30 50 15 5

[Pdubya64]

65/20/10/5 for Ratings/Current Yr/1 Yr/2Yr

Personally, as of late I am using 55/25/15/5. Haven't run enough time by so far to get feedback on it though.

[itsmb8]

I remember someone vouching for either 0/50/25/25 or 0/67/22/11 and said it worked VERY well. I cant remember which one it was though.

[oomm]

Ok. but zero ratings means zero accumulated reputation. So the game ONLY knows the past 2 years, plus current season. Wouldn't ratings cover more time that that? My point is how this mirrors RL. No way does player value get truncated to basically last year.



0/67/22/11 would likely show abrupt changes, nada? So a bad current season and the AI makes sudden moves. It's my guess.


25/25/25/25 With no emphasis, this basically turns it off?



55/25/15/5 So (again if my guess is right) reputation is offset more by current season than previous 2 combined (though previous two would be reflected in the ratings, nein?)

[Brad K]

Originally Posted by Pdubya64

65/20/10/5 for Ratings/Current Yr/1 Yr/2Yr

Personally, as of late I am using 55/25/15/5. Haven't run enough time by so far to get feedback on it though.

If you start a new game does if give you default or does it use the settings from the last game? I have looked at my current game and 5 old saves. Two oldest are 40 40 15 5 and the others are 50 30 15 5. This is on OOTP 21.

[SirMichaelJordan]

Originally Posted by PSUColonel

Ahh…the age old debate…lately I’ve been using 25/25/25/25 and even though I have traditionally supported a high ratings weight, it seems to be working very well.

I’ve been using this for nearly 10 versions now.

I’ve tried others throughout the years but I always find myself back here.

[Reed]

I do not know if these settings have any effect on a player's "ratings". The manual does not indicate that it does. This is used by the AI to "evaluate" players for lineup selection, for trade purposes, and for signing/releasing players.
For example if I use a setting of 50 for ratings and 50 for current stats and a player ratings indicates he should be a .300 hitter but he is currently hitting .200; by rating he might be a leadoff hitter and by his current stats he should be on the bench but the AI will see him as a .250 hitter and place him maybe 7th in the lineup. As the season go on his stats will toward his ratings. At least that is how it works in my historical league. The development settings will affect the players ratings, not this setting IMO.
This can lead to very unbalanced trades IMO. I think it might be fun to use in a fictional game while hiding the ratings.

[SirMichaelJordan]

Originally Posted by Reed

I do not know if these settings have any effect on a player's "ratings". The manual does not indicate that it does. This is used by the AI to "evaluate" players for lineup selection, for trade purposes, and for signing/releasing players.
For example if I use a setting of 50 for ratings and 50 for current stats and a player ratings indicates he should be a .300 hitter but he is currently hitting .200; by rating he might be a leadoff hitter and by his current stats he should be on the bench but the AI will see him as a .250 hitter and place him maybe 7th in the lineup. As the season go on his stats will toward his ratings. At least that is how it works in my historical league. The development settings will affect the players ratings, not this setting IMO.
This can lead to very unbalanced trades IMO. I think it might be fun to use in a fictional game while hiding the ratings.

Individual ratings decide lineup order, OVR rating decides who’s in that lineup.

The option to use stats in scouting (located in the global settings) can potentially change the individual ratings during the season based on performance.

[oomm]

25/25/25/25 basically turns it off though.



Yes, the player eval is about how the game evaluates player worth effecting how the player is used relative to other players. Were the settings 0/0/0/100 then the game evaluates player worth according to his stats generated 2 seasons ago. Not last season, nor this season.


Still though, his ratings would reflect past years performance, so I wonder how that factors in..................................

[Sweed]

Originally Posted by oomm

25/25/25/25 basically turns it off though.



Yes, the player eval is about how the game evaluates player worth effecting how the player is used relative to other players. Were the settings 0/0/0/100 then the game evaluates player worth according to his stats generated 2 seasons ago. Not last season, nor this season.


Still though, his ratings would reflect past years performance, so I wonder how that factors in..................................

You may find this thread from last year an interesting read..
https://forums.ootpdevelopments.com/...ght=Evaluation

It is the thread I alluded to in my earlier post in this thread. It is a good detailed discussion on this subject. SMJ goes into how and why he went to the 25/25/25/25 setting and how it all works within the context of individual GM tendencies.

It is the setting I use now and would not consider it "turning off" the feature.

As to your 0/0/0/100 example? Have no idea other than extreme settings like that might lead to unrealistic outcomes? I myself would not expect the developer to waste time trying to cover how the AI would react to an extreme setting like this. If someone wants to evaluate based only on two years ago, have at it. But please don't complain about the result

[oomm]

Originally Posted by Sweed

You may find this thread from last year an interesting read..
https://forums.ootpdevelopments.com/...ght=Evaluation

It is the thread I alluded to in my earlier post in this thread. It is a good detailed discussion on this subject. SMJ goes into how and why he went to the 25/25/25/25 setting and how it all works within the context of individual GM tendencies.

It is the setting I use now and would not consider it "turning off" the feature.

As to your 0/0/0/100 example? Have no idea other than extreme settings like that might lead to unrealistic outcomes? I myself would not expect the developer to waste time trying to cover how the AI would react to an extreme setting like this. If someone wants to evaluate based only on two years ago, have at it. But please don't complain about the result

Thanks for the link. Am finding many links (as in virtually all) are broken now.



However, though I haven't reached the end of the discussion there, there hasn't been any direct discussion of ratings percent representing a basic culmination of past performance (particularly performance beyond 2 seasons ago). And this remains my question: does not the ratings percent reflect the broader, general career performance? Because, otherwise, the game would be blind to performance prior to 2 seasons ago. Which wouldn't seem to work too well.


Therefore, again, following the above premise-- 25/25/25/25 appears to flatten everything. By 'flatten' I mean it makes everything equal to the other. For example, say it's a 10 season career in question: 2 years ago would be given same weight to this year and last year and the culmination of 10 years.


If that doesn't 'turn it off' what am I not understanding?

[PSUColonel]

Originally Posted by oomm

Thanks for the link. Am finding many links (as in virtually all) are broken now.



However, though I haven't reached the end of the discussion there, there hasn't been any direct discussion of ratings percent representing a basic culmination of past performance (particularly performance beyond 2 seasons ago). And this remains my question: does not the ratings percent reflect the broader, general career performance? Because, otherwise, the game would be blind to performance prior to 2 seasons ago. Which wouldn't seem to work too well.


Therefore, again, following the above premise-- 25/25/25/25 appears to flatten everything. By 'flatten' I mean it makes everything equal to the other. For example, say it's a 10 season career in question: 2 years ago would be given same weight to this year and last year and the culmination of 10 years.



If that doesn't 'turn it off' what am I not understanding?

This is not true because different staff members value things differently also. What it does do is take out the artificial emphasis of one area over another.

[Sweed]

Originally Posted by oomm

Thanks for the link. Am finding many links (as in virtually all) are broken now.



However, though I haven't reached the end of the discussion there, there hasn't been any direct discussion of ratings percent representing a basic culmination of past performance (particularly performance beyond 2 seasons ago). And this remains my question: does not the ratings percent reflect the broader, general career performance? Because, otherwise, the game would be blind to performance prior to 2 seasons ago. Which wouldn't seem to work too well.


Therefore, again, following the above premise-- 25/25/25/25 appears to flatten everything. By 'flatten' I mean it makes everything equal to the other. For example, say it's a 10 season career in question: 2 years ago would be given same weight to this year and last year and the culmination of 10 years.


If that doesn't 'turn it off' what am I not understanding?

SMJ has done more to look into this than anyone who's work I have read. The purpose of the 4-25's is to get the AI to make the best decisions. His experience is this rating works best. Like all things OOTP that is subjective and the best I can add is it works for me. Could be my very slow style since I play out every inning of every game?

Level out to me does not mean turns off. What it does IMHO is to not let the AI overreact to any one category and make poor decisions because of that overreaction. A player could hit into bad luck this year with career low babip and, if an average player, be sent down or cut. With level evaluation that will hopefully not happen as the other two years plus ratings can save the AI from itself. Reverse that and he has career high babip and now instead of being cut the AI gives him a long term deal as it's his FA year. Again the other two years of stats and ratings can put this all in context for the AI.

Another way it helps the AI is when the AI did not take stats into account then, of course, it only reacted to ratings. You could have a reigning MVP take a ratings hit the following season half way through and be released even though he was still putting up good numbers. Some would argue that is a good thing as the AI needs all the help it can get. Others would argue IRL as long as that player was putting up numbers he would never be benched or released. Imagine an aging Reggie Jackson with 20 HRs at the All Star break being released the following week because his ratings went down. I think this is the type of thing these settings are trying to avoid.

Also consider the AI prorates some of these evaluation numbers when the sample size is still small. IE a hot first 2 weeks of the season does not get the "benefit" of the full 25% evaluation.

Too me I don't know why one would want to go more than 2 years back? If the AI is evaluating an aging Harmon Killebrew in 1973 I would argue it should not care how he hit in 1963 or 1970 for that matter. The idea, I think, is to get the AI to look at the player as he is today and recently so it can make a good decision on extensions, trades, non-tender, cut etc. I'm not a statistician but IIRC, when these weights were added to the game and talked about in the forums, 3 seasons was deemed a good sample size to get a look at where a player has been and is going.

Then at the end also consider that each AI GM/scout has their own interpretation on how to evaluate players. This then gets mixed into the 25/25/25/25 to vary things even more. At least that is the theory from SMJ as I understand it. He can correct me I'm not understanding him.

IMHO SMJ's 4-25's works well in helping the AI with handling it's rosters. This means I get a more realistic world in regards to the transactions I see.
You might try it and not agree and that is fine. There is no right answer

[Sweed]

Originally Posted by PSUColonel

This is not true because different staff members value things differently also. What it does do is take out the artificial emphasis of one area over another.

Thanks. While I was typing my response you posted this. You said in one sentence what I tried to say in a few paragraphs