AI Evaluation Settings

[r0nster]

Originally Posted by Rain King

I play mostly "God Mode" and have settled on around 35/35/20/10.

You don't see veteran players just get dumped like you do with the higher ratings weight...they do more of a fade out as they stop producing which is what I was looking for.

Interesting you handled it that way what i did was to change development level and speed of loss and that seemed to help at least to what i was looking for. Longer player life especially for those doing really well. I didnt mess with injury setting. ONLY development speed. It seemed to make players end of prime years around 33 and 34 sometimes RARELY 36.

[koohead]

Originally Posted by r0nster

Interesting you handled it that way what i did was to change development level and speed of loss and that seemed to help at least to what i was looking for. Longer player life especially for those doing really well. I didnt mess with injury setting. ONLY development speed. It seemed to make players end of prime years around 33 and 34 sometimes RARELY 36.

The thing with adjusting the AI Eval values is even if a player were to start the decline and drop off from their prime, they wouldn't be dumped as quickly because more 'value' is placed on their performance from the past 2 years. So regardless of how slowly or quickly a player falls off the cliff, giving more emphasis to stats, and especially stats from the previous year and 2 years ago, keeps the player signed and in the lineup longer.

[PSUColonel]

Originally Posted by koohead

The thing with adjusting the AI Eval values is even if a player were to start the decline and drop off from their prime, they wouldn't be dumped as quickly because more 'value' is placed on their performance from the past 2 years. So regardless of how slowly or quickly a player falls off the cliff, giving more emphasis to stats, and especially stats from the previous year and 2 years ago, keeps the player signed and in the lineup longer.

Yes, I do think it's important to give last year and 2 years ago more weight than what the MLB Default settings are. That's what I was trying to convey a bit in my post above. 5% for two years ago is just way too low IMO. So it 10% for "last year". I myself would like to give more to stats, but I ademently believe you need to have more than 50% weighted for ratings, which leaves you with 45%. I also figure there is a reason the developers seem to believe in a declining system of distributing the weights, so to me, 55/20/15/10 just seems to be the best I can do without sacrificing a very very challenging AI. (at least for trades and waivers)

[MizzouRah]

Originally Posted by PSUColonel

Yes, I do think it's important to give last year and 2 years ago more weight than what the MLB Default settings are. That's what I was trying to convey a bit in my post above. 5% for two years ago is just way too low IMO. So it 10% for "last year". I myself would like to give more to stats, but I ademently believe you need to have more than 50% weighted for ratings, which leaves you with 45%. I also figure there is a reason the developers seem to believe in a declining system of distributing the weights, so to me, 55/20/15/10 just seems to be the best I can do without sacrificing a very very challenging AI. (at least for trades and waivers)

I think am going to use the same, 55/20/15/10 or 55/25/15/5 in my Reds franchise I'll start up later today. I loved 40/30/20/10 for several versions but I agree the ratings weight needs to be at or above 50.


I was thinking 50/30/15/5 or 50/25/20/5 but I like yours as well. This might be my go to for now though 55/25/15/5. (to me, I don't care as much what a player did 2 years ago, I'm looking at ratings, current season if I'm at least to the halfway point, and last season)

[Pelican]

I had been trying to absorb the varying wisdom in the 22 thread discussion on this. I was at 50/25/15/10 ("recency bias") but this year realized that any reference to 2020 stats was probably confounding, for so many reasons. So for 2022 "two years ago" is set to zero. I'd just rather forget the whole year.

I have some questions on how my choice gets applied.

Does "this year's stats" mean they will be updated as games are played in 2022 Season IRL (if I so choose), or is this a reference to "this year's stats" in my 2022 sim?

How does the AI adjust, when "last year's stats" are in the minor leagues? And/or when "two years ago" was minor leagues - maybe lower minors?

[PSUColonel]

I am still advocating 55/20/15/10

[Anteater]

Do you recommend that for fictional leagues also?

[PSUColonel]

Originally Posted by Anteater

Do you recommend that for fictional leagues also?

I do

[progen]

It's very simple my fine, feathered, friends. Stats only, and 2-67-11-20. I always enjoy this discussion, and have tried different settings as mentioned by previous posters, but overall I'm fine with this setting for the way I play. To each his own though.

And admittedly, I don't drill down to stats, trades, etc., as some of you do, but I haven't seen anything outrageous as far as trades, or AI roster moves. In fact on some AI trades, I was asking myself, why I didn't go after that player. The AI really did know what it was doing!

[AlpineSK]

Originally Posted by progen

It's very simple my fine, feathered, friends. Stats only, and 2-67-11-20. I always enjoy this discussion, and have tried different settings as mentioned by previous posters, but overall I'm fine with this setting for the way I play. To each his own though.

And admittedly, I don't drill down to stats, trades, etc., as some of you do, but I haven't seen anything outrageous as far as trades, or AI roster moves. In fact on some AI trades, I was asking myself, why I didn't go after that player. The AI really did know what it was doing!

What's the quality of your scouting like with something like this? I've played this game since OOTP4 and I've never tried Stats Only. This might be the year though.. I'm just wondering how it incorporates in.

[wodi]

The problem I generally have is that the AI is slow to call up top prospects, what setting do you recommend curb this?

Is ratings just current ratings, or does potential factor in?

[progen]

Originally Posted by AlpineSK

What's the quality of your scouting like with something like this? I've played this game since OOTP4 and I've never tried Stats Only. This might be the year though.. I'm just wondering how it incorporates in.

I have scouting on normal. With international signings, because you don't have any stats, you have to trust your scouts opinion, which I like. It makes the game more challenging. I also have found that I really slow down my simming, and spend a lot of time breaking down the stats of the various players. Also, when the email comes around about the top draft prospects, I delete it without looking, AND I never look at the mock draft. I want to go on my own scouting reports for the draft, and who I believe will help the team down the road.

And I never use stars. Lazy persons way of playing the game, and they are so annoying. Hey, not for everyone, and even though I will skip buying OOTP23, which breaks my record of buying the game every year since OOTP4, playing this way does keep my interest. Just not "wowed" by what OOTP23 has to offer, and OOTP22 still scratches my itch for the baseball genre. Eventually, when it goes on sale for $9.99 as it always does, I may pick it up, but probably will wait on OOTP24 when it comes out next March.

[BarneyRubble]

I use 25/25/25/25, but it really depends on what your goal is. I play as the GM and I don't micromanage. That said, I haven't tested this in 2 or 3 versions and may not have a clue what I'm talking about.

I used to get disheartened when a 34 year old all-star suddenly becomes a scrub in the off season (while he's celebrating the holidays with his family). I had this issue even when using 40/30/20/10. So I changed to 25/25/25/25 about the time OOTP implemented scouting changes and GM tendencies. What 25/25/25/25 does is minimize the impact of scouting when dealing with non-prospects (players with history). They can be a crap shoot, especially veterans. Depending on the Managers tendencies, a veteran will stay in the line-up longer. Track record is important.

I assume that, when scouting younger, international and draft players, the scout uses ratings because there are no stats (or very little). I also assume that OOTP prorates it's evaluation based on quantity of the stats (playing time). I also assume 'current year' is in fact 'current year' and if it's opening day, then '2 years ago' should be relevant and current year gets very little weight (no stats). I assume that, if an aging player has a monster season, followed by an injury filled season, then '2 years ago' matters. I assume that, at the end of a season, '2 years ago' has less weight than it had on opening day because current year is 'fulfilled'.

These are all assumptions on my part, as OOTP does not explain how the mechanics work. As for results, I haven't used anything other than 25/25/25/25 in quite a while. One reason for this is because it's very realistic. I use default trade settings. I used to get hung up on whether trades were realistic, then some extremely bad real life trades came into mind and I stopped worrying about it. I dominated the game when scouting was mostly ratings based, but I find it much more challenging with these settings. I once traded away a 26 year old, left handed pitcher and I thought I was fleecing the AI. He became a Cy Young winner. It was my Jake Arrieta moment. Sometimes one can rip off the AI, but most of the time, it's beneficial to both teams. I always find a reason why a team will part with a player. I've traded for players that flourished and I've had them bomb out. I've traded players because of money, chemistry or I have a better option. I take what I can get because I have no use for them. An expansion team's best player came in a trade with NY. Guess what, NY didn't need him. While I don't ignore them, I try not to get hung up on 'ratings' when trading. Specifically, OVR/POT.

What I've 'noticed' (I don't know this for sure) is that 'ratings' is the key. If you want a game where your scout can report to you that he saw your starting SS in the grocery store the other day and he can no longer field or hit, then crank up ratings as high as it will go (100/0/0/0), hire the best scouts and set you scouting budget as high as you can afford.

If I had two questions, I would ask...

1) When does 'current season' start?

2) The day 'current season' starts (using 25/25/25/25), is AI evaluation effectively 50/0/25/25, 34/0/33/33 or something different?

The answers to these questions really do matter.

[Rain King]

The devs have sprinkled enough knowledge here and there about this stuff that I feel pretty certain your assumptions are correct.

As for your questions.

I am pretty sure that "Current Season" flips as soon as you move to the Off-Season.

I think a pretty good assumption would be that Current Season stats just hold 0 weight until they begin to accumulate and the value of the other categories gets pro-rated to compensate, which would effectively be your 34/0/33/33 example if your settings are at 25/25/25/25

[PSUColonel]

The stats get prorated if there isn't enough information for the current year. While I agree with everything you have said above (regarding veteran players suddenly falling off a cliff and 25/25/25/25 helping to mitigate this) I must also point out when it comes to trade AI, this setting will have a negative affect on the human player when negotiating deals. The AI will allow "sweeteners" (worthless young players) in deals that are completely worthless. When ratings are at least 55...this ceases to happen. So I suggest keeping ratings at 55, and then deciding how you want to arrange the rest. You could go 55/15/15/15, or you could go 55/20/15/10....IMO these are two of the best bets.

[oomm]

Originally Posted by Rain King

The devs have sprinkled enough knowledge here and there about this stuff that I feel pretty certain your assumptions are correct.

As for your questions.

I am pretty sure that "Current Season" flips as soon as you move to the Off-Season.

I think a pretty good assumption would be that Current Season stats just hold 0 weight until they begin to accumulate and the value of the other categories gets pro-rated to compensate, which would effectively be your 34/0/33/33 example if your settings are at 25/25/25/25

This introduces a cardinal point re analyses: all discussion seems to center on peoples impressions, which is fine. But a systematic analysis of how the software and database interact would be perhaps more definitive. I am not the man for that job.

[Sweed]

Originally Posted by BarneyRubble

I use 25/25/25/25, but it really depends on what your goal is. I play as the GM and I don't micromanage. That said, I haven't tested this in 2 or 3 versions and may not have a clue what I'm talking about.

I used to get disheartened when a 34 year old all-star suddenly becomes a scrub in the off season (while he's celebrating the holidays with his family). I had this issue even when using 40/30/20/10. So I changed to 25/25/25/25 about the time OOTP implemented scouting changes and GM tendencies. What 25/25/25/25 does is minimize the impact of scouting when dealing with non-prospects (players with history). They can be a crap shoot, especially veterans. Depending on the Managers tendencies, a veteran will stay in the line-up longer. Track record is important.

I assume that, when scouting younger, international and draft players, the scout uses ratings because there are no stats (or very little). I also assume that OOTP prorates it's evaluation based on quantity of the stats (playing time). I also assume 'current year' is in fact 'current year' and if it's opening day, then '2 years ago' should be relevant and current year gets very little weight (no stats). I assume that, if an aging player has a monster season, followed by an injury filled season, then '2 years ago' matters. I assume that, at the end of a season, '2 years ago' has less weight than it had on opening day because current year is 'fulfilled'.

These are all assumptions on my part, as OOTP does not explain how the mechanics work. As for results, I haven't used anything other than 25/25/25/25 in quite a while. One reason for this is because it's very realistic. I use default trade settings. I used to get hung up on whether trades were realistic, then some extremely bad real life trades came into mind and I stopped worrying about it. I dominated the game when scouting was mostly ratings based, but I find it much more challenging with these settings. I once traded away a 26 year old, left handed pitcher and I thought I was fleecing the AI. He became a Cy Young winner. It was my Jake Arrieta moment. Sometimes one can rip off the AI, but most of the time, it's beneficial to both teams. I always find a reason why a team will part with a player. I've traded for players that flourished and I've had them bomb out. I've traded players because of money, chemistry or I have a better option. I take what I can get because I have no use for them. An expansion team's best player came in a trade with NY. Guess what, NY didn't need him. While I don't ignore them, I try not to get hung up on 'ratings' when trading. Specifically, OVR/POT.

What I've 'noticed' (I don't know this for sure) is that 'ratings' is the key. If you want a game where your scout can report to you that he saw your starting SS in the grocery store the other day and he can no longer field or hit, then crank up ratings as high as it will go (100/0/0/0), hire the best scouts and set you scouting budget as high as you can afford.

If I had two questions, I would ask...

1) When does 'current season' start?

2) The day 'current season' starts (using 25/25/25/25), is AI evaluation effectively 50/0/25/25, 34/0/33/33 or something different?

The answers to these questions really do matter.

This pretty much describes my experience with 25/25/25/25.
Yes, when there is a small sample size Matt and Markus have said the scout takes that into account and prorates until there is a suitable sample size in place.

Originally Posted by PSUColonel

The stats get prorated if there isn't enough information for the current year. While I agree with everything you have said above (regarding veteran players suddenly falling off a cliff and 25/25/25/25 helping to mitigate this) I must also point out when it comes to trade AI, this setting will have a negative affect on the human player when negotiating deals. The AI will allow "sweeteners" (worthless young players) in deals that are completely worthless. When ratings are at least 55...this ceases to happen. So I suggest keeping ratings at 55, and then deciding how you want to arrange the rest. You could go 55/15/15/15, or you could go 55/20/15/10....IMO these are two of the best bets.

I'm sure this is all true but...

When playing a solo league one can easily control how they go about trades.
I rarely use "make this work now" but when I do I can't say I've seen the AI tell me it will make the deal by using "sweeteners". They normally try to "take me to the cleaners" wanting top talent for their second tier player.

With regard to player movement, AI trades, both with me and between AI teams I see reason. The trades, for the most part, make sense. The few that look funny? I attribute to differences in how my scout views the players involved along with how they may have different preferences when making those evaluations.

I have used 25/25/25/25 for 3(?) versions now and have seen no reason to change. I don't use house rules and don't really see the need. No, I don't pile on "sweeteners" when trying to make trades but, to me, that isn't a house rule as it's nothing I would do anyway. I put players in a deal that I believe I would put in if I were a real GM. Too me that is the point of the game and what makes it fun. I appreciate all Markus and Matt have done with the trade AI over the years but I don't need Markus or Matt to make an AI that polices me in a solo league to stop me from being tempted to cheese the AI.

For a solo league, real world experience I have found the four 25's as good as it gets. Just my 2 cents FWIW

[MathBandit]

I know I've said this in other threads, but I still don't see the logic behind weighting each of the last three years equally.

Entering 2022, did any of us think Vladdy and Yelich were equally-good hitters and likely to put up the same hitting numbers in 2022? They each had one ridiculous season, one good season, and one average season in their last three. Do you truly think that a player who was worth 8 WAR, then 4 WAR, then 2 WAR is just as good as one who was worth 2 WAR, 4 WAR, 8 WAR assuming all else is equal? Telling the AI to treat the most recent numbers as being worth no more than the numbers from two years ago is absolutely a huge departure from the way (just about?) every human evaluates players.

[Sweed]

Originally Posted by MathBandit

I know I've said this in other threads, but I still don't see the logic behind weighting each of the last three years equally.

Entering 2022, did any of us think Vladdy and Yelich were equally-good hitters and likely to put up the same hitting numbers in 2022? They each had one ridiculous season, one good season, and one average season in their last three. Do you truly think that a player who was worth 8 WAR, then 4 WAR, then 2 WAR is just as good as one who was worth 2 WAR, 4 WAR, 8 WAR assuming all else is equal? Telling the AI to treat the most recent numbers as being worth no more than the numbers from two years ago is absolutely a huge departure from the way (just about?) every human evaluates players.

It's about getting the game to behave, with regard to transactions, in the most realistic fashion possible. Each of us has a standard as to what that is. It has nothing to do with real life.

I could write many paragraphs over the how and why but that is easy enough to be found with a search or simply reviewing this thread from past versions.

[PSUColonel]

Originally Posted by MathBandit

I know I've said this in other threads, but I still don't see the logic behind weighting each of the last three years equally.

Entering 2022, did any of us think Vladdy and Yelich were equally-good hitters and likely to put up the same hitting numbers in 2022? They each had one ridiculous season, one good season, and one average season in their last three. Do you truly think that a player who was worth 8 WAR, then 4 WAR, then 2 WAR is just as good as one who was worth 2 WAR, 4 WAR, 8 WAR assuming all else is equal? Telling the AI to treat the most recent numbers as being worth no more than the numbers from two years ago is absolutely a huge departure from the way (just about?) every human evaluates players.

I can see both sides of the coin here. What Bandit wants to do is to ensure this behavior amongst every staff member in the OOTP universe. What Sweet is saying, is the 25/25/25/25 will allow staff member to act as their personalities truly are...and just so we are clear...staff members will never value 2 years ago more than this year or last year...it could be equal, but even that is more rare.

As I have said so many times now...to me, I think it's just important to have the majority of weight go to scouting grades. If you think about it...it's the ratings which produce the stats, not the other way around. I just hate the "Mae this work now" button (but admittedly use it simply because it's there) and this helps curtail the AI from asking for those worthless sweeteners...instead with 55 ratings, it will ask for a legit player/prospect. It also helps to ensure there isn't any odd waiver behavior.