OOTP Developments Forums - View Single Post

jdw · 09-14-2025, 09:49 PM

Furthering luckymann's response:

The AI GM kinda-sorta sees "future", but also kinda-sorta doesn't. Best example I can give would be the same 1970 Import mentioned above.

At one extreme, the HOU AI GM can't properly calculate the future value of that Morgan+Wynn+Rader+Marshall+Geranimo fleecing mentioned above.

In the other extreme, it see the future value of Bobby Grich enough that it won't even bite at a self-fleece offer of Reggie Smith + Carlton Fisk + Carl Yastrzemski for him.

The only way it could "rate" Grich that high is based on some level of Future Analysis given Grich isn't highly rated for the 1970 season in which he barely played. But it values him so much that it won't take a combo package containing:

1. a Top 10 position player in the 1970s who is very good for the balance of the decade

2. a prospect of no less than equal value at an equally scarce position with a longer career ahead of him

3. a 1970 MVP level player who is an above average-to-good hitter the balance of the decade

So it's pretty obvious the engine see immense value in Grich, which on some level is a very good thing. You don't want the Human GM to fleece the O's for Grich.

But it also values him so extremely high that it doesn't let the Human GM to build what would be a reasonable package for him, which would be Fisk and at most some very nominal player that you often see the AI GM ask for in one of those "You're very close, just a little more" deals.

* * * * *

I will admit that it probably is a very complicated thing to program.

If one is using Remaining Peak as the driver for Potential Ratings, that having an increased significance on AI GM's rating players sounds like a good idea. That would address the Morgan issue. In a 1970 Import in the configuration, no position player in theory should have their Potential Ratings crunch a higher value. There simply is no position player in a 1970 Import that can match 1975 Joe Morgan.

But...

Bobby Murcer and Reggie Jackson in a same scenario in a 1970 Import would have a very close crunch to their Potential Ratings. We as a Human GM know how their careers turn out, and a straight up Bobby-Reggie deal in 1970 or 1971 or 1972 would end up terribly one sided.

We know in Real Life bad deal along those lines happen. It feels like "equal talent" is moving, but over time turns into a wipe out.

How to balance this out to make the GM less likely to make incredibly bad deals... I really don't know.

But as much as I point to stuff like the valuation of elite players (Morgan, Grich, Wynn, etc in 1970), we also need to keep in mind those bad valuations of players like Giusti & Rooker in the examples given above. Also the Watson one. If you're a Human GM who doesn't care, it's no big deal. If you care to a degree, it's painful to see how regularly it happens.

09-14-2025, 09:49 PM	#7
jdw All Star Starter Join Date: Nov 2003 Posts: 1,161	Furthering luckymann's response: The AI GM kinda-sorta sees "future", but also kinda-sorta doesn't. Best example I can give would be the same 1970 Import mentioned above. At one extreme, the HOU AI GM can't properly calculate the future value of that Morgan+Wynn+Rader+Marshall+Geranimo fleecing mentioned above. In the other extreme, it see the future value of Bobby Grich enough that it won't even bite at a self-fleece offer of Reggie Smith + Carlton Fisk + Carl Yastrzemski for him. The only way it could "rate" Grich that high is based on some level of Future Analysis given Grich isn't highly rated for the 1970 season in which he barely played. But it values him so much that it won't take a combo package containing: 1. a Top 10 position player in the 1970s who is very good for the balance of the decade 2. a prospect of no less than equal value at an equally scarce position with a longer career ahead of him 3. a 1970 MVP level player who is an above average-to-good hitter the balance of the decade So it's pretty obvious the engine see immense value in Grich, which on some level is a very good thing. You don't want the Human GM to fleece the O's for Grich. But it also values him so extremely high that it doesn't let the Human GM to build what would be a reasonable package for him, which would be Fisk and at most some very nominal player that you often see the AI GM ask for in one of those "You're very close, just a little more" deals. * * * * * I will admit that it probably is a very complicated thing to program. If one is using Remaining Peak as the driver for Potential Ratings, that having an increased significance on AI GM's rating players sounds like a good idea. That would address the Morgan issue. In a 1970 Import in the configuration, no position player in theory should have their Potential Ratings crunch a higher value. There simply is no position player in a 1970 Import that can match 1975 Joe Morgan. But... Bobby Murcer and Reggie Jackson in a same scenario in a 1970 Import would have a very close crunch to their Potential Ratings. We as a Human GM know how their careers turn out, and a straight up Bobby-Reggie deal in 1970 or 1971 or 1972 would end up terribly one sided. We know in Real Life bad deal along those lines happen. It feels like "equal talent" is moving, but over time turns into a wipe out. How to balance this out to make the GM less likely to make incredibly bad deals... I really don't know. But as much as I point to stuff like the valuation of elite players (Morgan, Grich, Wynn, etc in 1970), we also need to keep in mind those bad valuations of players like Giusti & Rooker in the examples given above. Also the Watson one. If you're a Human GM who doesn't care, it's no big deal. If you care to a degree, it's painful to see how regularly it happens.