Just My Opinion on the AI.

Bobfather · 06-19-2021, 11:20 PM

Recently there have been some threads about the Bugs/mistakes with the AI. I have limited technical knowledge. But I will try to explain what I think happens. First off for this post AI stands for any transaction/lineup/manager decision not made by a human. First, the AI does not know Mike Trout from Franchy Codero, from babe Ruth. It only sees these as a bunch of data points. Names, nationality, race, etc are superfluous to the AI. The AI will make the best decisions possible with these data points, using the parameters put in place by the devs, then tweaked by us. This includes the career paths of historical players. If you are using a rookie Mike Schmidt as a rookie, the AI does not know he is supposed to progress into a HOF, or Cal Ripken literally plays every day. So maybe Ripken gets hurt, or Schmidt never becomes more than average. Because again the AI is not prescient. It just sees data points. (maybe there is something in the background so the AI knows the career trajectory of every existing player, but I doubt it. What someone may want to try is to set up a fictional player with all the exact ratings of Mike Schmidy, Cal Ripken, or any other player you like. Then simulate a few seasons of two exact same leagues one with the real player and one with the fictional player and see what happens. So IMO the AI is not bugged or wonky. It is just trying to make the best use of the data points within set parameters. So you may see a player out of position, something I am currently doing in my save. A pitcher moving between starting a relief (again something I do all the time). Or even making trades. Just because the decision is not what one would expect, does not make it wrong. And should not result in devs being called out or called names. If you have read this far, two things, one thankyou, two get a life.

Now time to eviscerate me.
** All players named are examples only**

pgjocki · 06-19-2021, 11:34 PM

Not going to argue with your definition but rather ask the question of why doesn't the AI have the ability to "know" that some players are stars and have something in the ratings that forces the game engine to treat them differently?

Sent from my SM-G950U using Tapatalk

UltimateAverageGuy · 06-19-2021, 11:55 PM

If the AI is supposed to know what you think they should know there would be zero variety in anyone’s SIMs.

Bobfather · 06-20-2021, 12:05 AM

Quote:

Originally Posted by pgjocki

Not going to argue with your definition but rather ask the question of why doesn't the AI have the ability to "know" that some players are stars and have something in the ratings that forces the game engine to treat them differently?

Sent from my SM-G950U using Tapatalk

Again you could throw what I know about programming in a thimble and still have room for water. Even if they did there would always be some outlyers.
What I have noticed is certain players getting off to bad starts, certain pitchers having shaky first innings then settling down. etc. I started my league in 2019, I play out all of my team's games. I am currently in 2027 so my experience with great players is limited. But I recently noticed Ronald Acuna JR sitting in AA with only 2 MLB ABs.

ezpkns34 · 06-20-2021, 02:36 AM

Quote:

Originally Posted by Bobfather

Recently there have been some threads about the Bugs/mistakes with the AI. I have limited technical knowledge. But I will try to explain what I think happens. First off for this post AI stands for any transaction/lineup/manager decision not made by a human. First, the AI does not know Mike Trout from Franchy Codero, from babe Ruth. It only sees these as a bunch of data points. Names, nationality, race, etc are superfluous to the AI. The AI will make the best decisions possible with these datapoints, using the parameters put in place by the devs, then tweaked by us. This includes the career paths of historical players. If you are using a rookie Mike Schmidt as a rookie, the AI does not know he is supposed to progress into a HOF, or Cal Ripken literally plays every day. So maybe Ripken gets hurt, or Schmidt never becomes more than average. Because again the AI is not prescient. It just sees data points. (maybe there is something in the background so the AI knows the career trajectory of every existing player, but I doubt it. What someone may want to try is to set up a fictional player with all the exact ratings of Mike Schmidy, Cal Ripken, or any other player you like. Then simulate a few seasons of two exact same leagues one with the real player and one with the fictional player and see what happens. So IMO the AI is not bugged or wonky. It is just trying to make the best use of the data points within set parameters. So you may see a player out of position, something I am currently doing in my save. A pitcher moving between starting a relief (again something I do all the time). Or even making trades. Just because the decision is not what one would expect, does not make it wrong. And should not result in devs being called out or called names. If you have read this far, two things, one thankyou, two get a life.

Now time to eviscerate me.

*cries out as the wall of text smushes me*

For what it's worth, if Schmidt is rated with 80 potential, or 5* or whatever method you use, then hopefully the AI would have some idea that at least he should be good

AdamLindsBeard · 06-20-2021, 01:28 PM

Among the players who were taken before Schmidt in the 1971 draft:

Tommy Bianco
Roy Branch
Roger Quiroga
Ed Kurpiel
David Sloan

... you get the idea.

(Then again, so were Jim Rice, George Brett, Frank Tanana and Rick Rhoden.)

Point: Scouts are often dead wrong. And Mike Piazza went in Round 62 as the first-ever 1390th pick. The AI probably gets it wrong just about as often as the real scouts do, though I've never done the actual research.

06-19-2021, 11:20 PM	#1
Bobfather All Star Starter Join Date: Jun 2016 Location: Boston Ma. Posts: 1,559	Just My Opinion on the AI. Recently there have been some threads about the Bugs/mistakes with the AI. I have limited technical knowledge. But I will try to explain what I think happens. First off for this post AI stands for any transaction/lineup/manager decision not made by a human. First, the AI does not know Mike Trout from Franchy Codero, from babe Ruth. It only sees these as a bunch of data points. Names, nationality, race, etc are superfluous to the AI. The AI will make the best decisions possible with these data points, using the parameters put in place by the devs, then tweaked by us. This includes the career paths of historical players. If you are using a rookie Mike Schmidt as a rookie, the AI does not know he is supposed to progress into a HOF, or Cal Ripken literally plays every day. So maybe Ripken gets hurt, or Schmidt never becomes more than average. Because again the AI is not prescient. It just sees data points. (maybe there is something in the background so the AI knows the career trajectory of every existing player, but I doubt it. What someone may want to try is to set up a fictional player with all the exact ratings of Mike Schmidy, Cal Ripken, or any other player you like. Then simulate a few seasons of two exact same leagues one with the real player and one with the fictional player and see what happens. So IMO the AI is not bugged or wonky. It is just trying to make the best use of the data points within set parameters. So you may see a player out of position, something I am currently doing in my save. A pitcher moving between starting a relief (again something I do all the time). Or even making trades. Just because the decision is not what one would expect, does not make it wrong. And should not result in devs being called out or called names. If you have read this far, two things, one thankyou, two get a life. Now time to eviscerate me. All players named are examples only Last edited by Bobfather; 06-20-2021 at 03:42 PM. Reason: Added Note

06-19-2021, 11:34 PM	#2
pgjocki All Star Starter Join Date: Jun 2013 Location: Maryland - just outside DC Posts: 1,585	Not going to argue with your definition but rather ask the question of why doesn't the AI have the ability to "know" that some players are stars and have something in the ratings that forces the game engine to treat them differently? Sent from my SM-G950U using Tapatalk __________________ - - - World Series championships: 1926, 1931, 1934, 1942, 1944, 1946, 1964, 1967, 1982, 2006, 2011

06-19-2021, 11:55 PM	#3
UltimateAverageGuy All Star Reserve Join Date: Jan 2017 Posts: 587	If the AI is supposed to know what you think they should know there would be zero variety in anyone’s SIMs.

06-20-2021, 01:28 PM	#6
AdamLindsBeard Minors (Rookie Ball) Join Date: Apr 2014 Posts: 46	Among the players who were taken before Schmidt in the 1971 draft: Tommy Bianco Roy Branch Roger Quiroga Ed Kurpiel David Sloan ... you get the idea. (Then again, so were Jim Rice, George Brett, Frank Tanana and Rick Rhoden.) Point: Scouts are often dead wrong. And Mike Piazza went in Round 62 as the first-ever 1390th pick. The AI probably gets it wrong just about as often as the real scouts do, though I've never done the actual research.