Getting the AI to play as shrewdly as a human?

[The Wolf]

You can turn stars on just for the draft if you want to.

[1998 Yankees]

I greatly respect the opinions of The Wolf and others who recommend a stats-only style of play. However, I find myself agreeing with OFG when he says:

Originally Posted by OldFatGuy

. . . given that one of its mottos is "It's your game, play it your way" I would like to play it with ratings, with no house rules, and have it challenging.

I'm wondering about these two quotes . . .

Originally Posted by jmknpk2

I use the default of 30/50/15/5. I kind of use that same evalution scale when I evaluate players.

Originally Posted by Isura

That could be the issue. Increase the weighing of ratings. That tends to make the AI smarter

and whether more people who have stayed with ratings find better AI results by giving ratings a much higher weight in AI Player Evaluation.

For example, I have always played, up until now, with the default AI Player Evaluation weights. In eight years, I have assembled a powerhouse franchise that won 120 games last year and is on pace to win even more this season.

Take a look at the player below. After reading in this thread about finding five-star potential minor-league players being neglected by the AI in the free agent pool, I went to look and sure enough there were several guys like him hanging around.

This player was released in June after a poor performance in SSA last year and not playing at all this season. With these potentials (my scout is highly talented, so they are probably accurate), why has nobody else signed him?

I agree with OP; this is one of the reasons why I can be so successful after only a few seasons. Seasons gone by, I would have snapped up this player but now I'm going to leave him and the others alone. I want to see what happens if I make a change.

What I am going to change is the AI Player Evaluation to give much higher weight to ratings. If the AI uses only 30% of its "brain" to the ratings that I am seeing, leaving 70% (really just 15% in this case) to focus on only one partial year of year-old stats, no wonder I have an advantage. The AI and I have to be on a level playing field if we are both going to be looking at ratings in our player evaluations.

We've heard convincing arguments for no-ratings gameplay. Anybody care to speak up for AI "ratings evaluation enhancement" style of play?

[Afino]

Yeah - I've completely ignored the waiver wire and don't sign 5-star prospects like that. I see them, but leave them for the other teams.

I just don't think it's fair to be scooping them up. Guess it's a part of the "house" rules.

[The Wolf]

There's no right answer here. There's only what works best for you. For some of us, playing stats only gives us something we really like. It may not do the same thing for you. Whatever, it's your game, play it your way.

[StyxNCa]

Originally Posted by 1998 Yankees

I greatly respect the opinions of The Wolf and others who recommend a stats-only style of play. However, I find myself agreeing with OFG when he says:

I'm wondering about these two quotes . . .



and whether more people who have stayed with ratings find better AI results by giving ratings a much higher weight in AI Player Evaluation.

For example, I have always played, up until now, with the default AI Player Evaluation weights. In eight years, I have assembled a powerhouse franchise that won 120 games last year and is on pace to win even more this season.

Take a look at the player below. After reading in this thread about finding five-star potential minor-league players being neglected by the AI in the free agent pool, I went to look and sure enough there were several guys like him hanging around.

This player was released in June after a poor performance in SSA last year and not playing at all this season. With these potentials (my scout is highly talented, so they are probably accurate), why has nobody else signed him?

I agree with OP; this is one of the reasons why I can be so successful after only a few seasons. Seasons gone by, I would have snapped up this player but now I'm going to leave him and the others alone. I want to see what happens if I make a change.

What I am going to change is the AI Player Evaluation to give much higher weight to ratings. If the AI uses only 30% of its "brain" to the ratings that I am seeing, leaving 70% (really just 15% in this case) to focus on only one partial year of year-old stats, no wonder I have an advantage. The AI and I have to be on a level playing field if we are both going to be looking at ratings in our player evaluations.

We've heard convincing arguments for no-ratings gameplay. Anybody care to speak up for AI "ratings evaluation enhancement" style of play?

I have used the same settings you have and I have seen some players with real high potential sitting in the free agent list as well. I have sort of chalked it up to my scout being one of the best and his opinion is different than other scouts in the league. However, I also find it hard to believe that none of the other 23 scouts see what mine does.

I don't sign free agents. I build my team strictly through the draft and I allow ghost players in the minors so I have no need to fill every slot. When need arises I may make a trade. I have trade set on hard so I can't succumb to the occassional craving to cheat the other team.

I had not thought of putting a lot more emphasis on ratings. That would make sense to me. I was thinking of making it 100% rating evaluation since when I look at ratings and stats of some players, one named Drew Butler for example, the two don't jive meaning their stats have never lived up to the ratings. So this would make it seem going strictly by ratings may elevate their evaluation, but that doesn't mean they will live up to those. This seems most realistic to me when it comes to going after players. If you're going for a free agent with no major league experience, in real life scouts base their opinions more on what they see of the player than their stats from the past, even though that would play a role to a point because what someone does in high school or college is a lot different than what they may do on a professional level where they are playing against good players every day. Looking through the history of the major league draft shows what I am saying. How many early round draft picks actually had any kind of career?

The only way for stats to matter heavily would be if they had a history in the majors. Then you could combine ratings, stats and age of player to figure out where they are going a lot more accurately. Then if you use coaches, you would have to look at the quality of your coaches to see if you could get that player to have more success than in the past. These are things I look at, even more than ratings.

Another thing I look at is intelligence and work ethic. I refuse to go after anyone with super potential who has super low work ethic because I don't feel that player will ever reach his potential since he doesn't seem willing to work to get there. On the other hand, I will go after someone with so-so potential with extremely high work ethic and good intelligence because I have had success with those type of players. Combined with my coaches those players seem to overachieve a lot more often than fail.

Anyway, these are just my thoughts. I guess using just stats would results in the same ends.

[deadringer]

1998 Yankees, for me it's all about suspension of disbelief. With ratings on I can see all the little things that bug me and I can't get over them and enjoy the ride. I find the game way too easy with ratings on without choking the hell of myself with house rules. In an ideal world, I'd much prefer to have ratings on and be challenged but it just isn't happening.

Personally I'd love to see a whole new mode of the game offered in the future where instead of ratings we get something like "tools". IRL most of the scouting reports I've got to see talk more about a players physical abilities or characteristics and what the scouts belief those tools will translate into numbers. You can kind of get that in the game using potentials only but it's not quite the same.

By the way, on the waiver wire issue I still have to use house rules to keep from cherry picking all the high end talent. I only allow myself to claim players who have major league experience and are out of options. Then I only claim players who would go on my major league roster right now.

[1998 Yankees]

Originally Posted by Afino

Yeah - I've completely ignored the waiver wire and don't sign 5-star prospects like that. I see them, but leave them for the other teams.

I just don't think it's fair to be scooping them up. Guess it's a part of the "house" rules.

I checked - they are still out there with AI "ratings evaluation enhancement" settings of 65/25/10/5 although it's been only "two weeks" in game time since my last post. I'm with you, though, now that I've read through this thread. It just is too easy to grab those guys, even if only one of the six or seven or so actually turns out to be a superstar. Yet, if I don't, it will bother me that the AI continues to ignore them also. A quandary.

Originally Posted by The Wolf

There's no right answer here. There's only what works best for you. For some of us, playing stats only gives us something we really like. It may not do the same thing for you. Whatever, it's your game, play it your way.

Well said. I may be joining you ultimately.

Originally Posted by StyxNCa

I have used the same settings you have and I have seen some players with real high potential sitting in the free agent list as well. I have sort of chalked it up to my scout being one of the best and his opinion is different than other scouts in the league. However, I also find it hard to believe that none of the other 23 scouts see what mine does.
Yeah, but you are having the same feeling as me; if I look at those ratings and drool, at least one of the 29 AI GM's should be as well. Moving my AI ratings evaluation weight up should have worked in that regard, but not yet.

I don't sign free agents. I build my team strictly through the draft and I allow ghost players in the minors so I have no need to fill every slot. When need arises I may make a trade. I have trade set on hard so I can't succumb to the occassional craving to cheat the other team.
Yeah, but not signing free agents is a rather harsh and unrealistic house rule. I also have my trading set on hard, which is much better than Average for sure.

I had not thought of putting a lot more emphasis on ratings. That would make sense to me. I was thinking of making it 100% rating evaluation since when I look at ratings and stats of some players, one named Drew Butler for example, the two don't jive meaning their stats have never lived up to the ratings. So this would make it seem going strictly by ratings may elevate their evaluation, but that doesn't mean they will live up to those. This seems most realistic to me when it comes to going after players. If you're going for a free agent with no major league experience, in real life scouts base their opinions more on what they see of the player than their stats from the past, even though that would play a role to a point because what someone does in high school or college is a lot different than what they may do on a professional level where they are playing against good players every day. Looking through the history of the major league draft shows what I am saying. How many early round draft picks actually had any kind of career?
I'm thinking of bumping ratings weight up to 75% and leaving the stats weights to total 25%. That about matches my style of play anyway; I depend heavily upon scouting reports.

The only way for stats to matter heavily would be if they had a history in the majors. Then you could combine ratings, stats and age of player to figure out where they are going a lot more accurately. Then if you use coaches, you would have to look at the quality of your coaches to see if you could get that player to have more success than in the past. These are things I look at, even more than ratings.
I think that, for human and AI to be somewhat closely matched in ability in this game, that the AI has to be tuned to look at things the way the human does. Hence my prior post about making AI use ratings as much as I do. See the example that I talked about again where, if I am correct, my entire brain is focused on five stars while poor AI only has 45% of his working on the problem.

Another thing I look at is intelligence and work ethic. I refuse to go after anyone with super potential who has super low work ethic because I don't feel that player will ever reach his potential since he doesn't seem willing to work to get there. On the other hand, I will go after someone with so-so potential with extremely high work ethic and good intelligence because I have had success with those type of players. Combined with my scouts those players seem to overachieve a lot more often than fail.
Big ditto on these comments about intelligence and work ethic. I was fixated on work ethic for a long time as a determinant of development but read just recently here that intelligence also plays a role in that. I had always thought intelligence was just for better ballplay.

Anyway, these are just my thoughts. I guess using just stats would results in the same ends.
Good thoughts. Thanks for sharing them.

Originally Posted by deadringer

1998 Yankees, for me it's all about suspension of disbelief. With ratings on I can see all the little things that bug me and I can't get over them and enjoy the ride. I find the game way too easy with ratings on without choking the hell of myself with house rules. In an ideal world, I'd much prefer to have ratings on and be challenged but it just isn't happening.
Well, like I said to The Wolf, if I cannot resolve this dilemma of wanting to use ratings and scouting reports without either dominating the league year after year (I'm not that rabid a Yankee fan, just a devoted one) or needing to adhere to unrealistic house rules such as never signing a free agent, then I may be joining you guys soon.

Personally I'd love to see a whole new mode of the game offered in the future where instead of ratings we get something like "tools". IRL most of the scouting reports I've got to see talk more about a players physical abilities or characteristics and what the scouts belief those tools will translate into numbers. You can kind of get that in the game using potentials only but it's not quite the same.
Yes, Wolf talked a bit about enhancements just for stats-only play. See some of his earlier posts in this thread.

By the way, on the waiver wire issue I still have to use house rules to keep from cherry picking all the high end talent. I only allow myself to claim players who have major league experience and are out of options. Then I only claim players who would go on my major league roster right now.
Ratings off and still need house rules? Remind me never to join an online league with deadringer.

[The Wolf]

There's really no argument here. Play the game the way you like it.

[PhillieFever]

Originally Posted by 1998 Yankees

I checked - they are still out there with AI "ratings evaluation enhancement" settings of 65/25/10/5 although it's been only "two weeks" in game time since my last post. I'm with you, though, now that I've read through this thread. It just is too easy to grab those guys, even if only one of the six or seven or so actually turns out to be a superstar. Yet, if I don't, it will bother me that the AI continues to ignore them also. A quandary.

Well said. I may be joining you ultimately.

I'm definitely interested in hearing about your results with the ratings eval set at 75,keep us posted.

[StyxNCa]

1998 Yankees....

I was kind of vague on the free agent statement. It isn't really a "house rule". It's more that I have never been a fan of free agency so I don't build my teams that way. However, I guess some do use it as a "house rule". I like to keep the money to re-sign my own people.

[StyxNCa]

Originally Posted by The Wolf

There's really no argument here. Play the game the way you like it.

I wasn't arguing with you. I agree with you when you say play it your way. I reckon no two of us plays it exactly alike.

[Afino]

I guess it's somewhat different for me since I play in the pre-free agency era.

I've built a dynasty that pretty much can only be stopped by players aging beyond their effectiveness or my own doing.

In this past offseason, I traded away the league's MVP, benched another MVP-winning, consistent .300/25/120 guy since his age was eroding his skills VERY quickly at age 38 (I do use Skydog's settings for aging/dev), traded away a 80/80/50 (1-100 scale) scout find RP because my bullpen was loaded, and traded away my stud 24 y/o RF who had just won the ROY for a #2 power lefty starter that had given me fits over the last few years. that player was the ONLY ML player I got back in any of those deals.

You'd think this would hurt me, right?


I've started the season 18-5.

I was actually HOPING for a year or two where I'd struggled....

[Syd Thrift]

Originally Posted by StyxNCa

I have used the same settings you have and I have seen some players with real high potential sitting in the free agent list as well. I have sort of chalked it up to my scout being one of the best and his opinion is different than other scouts in the league. However, I also find it hard to believe that none of the other 23 scouts see what mine does.

Just curious: have you tried playing with ratings on and scouting accuracy set to low? I haven't done that myself but it seems like it would be worth a shot.

IMO the problem that a lot of people are experiencing isn't so much that the game AI is stupid (although it does have its issues and always will, given that it's a computer game and not a human being) but that the scouting model is still waaaaaaaay too accurate. My ideal for the scouting model would be to bring back multiple scouts (perhaps the head could hire them himself and save you the hassle of micromanaging all that stuff) and have them actually base their reports on "watching" a guy play over a series. If a stud prospect went 1-12 with 5 strikeouts, he'd come back saying that the guy is terrible at avoiding Ks. Perhaps you'd want to temper this assessment with actual current ratings (the equivalent of seeing a lot of loud outs or quiet hits), but I think you'd still want the games witnessed themselves to be at the forefront. A scouting department might have the following personnel:

Minor league scout (for your minors, of course)

Several regional scouts (for scouting amateurs; I don't think computers are at a point where you can reasonably expect the game to simulate 10,000+ high school and college players a year, so you'd need to abstract this into having the game generate the draft-worthy players and having each regional guy look at 20% or 25% depending on how many regional scouts you had, which in turn would be based on your team scouting budget. The more guys a scout would have to look at, the less accurate his reports would be. If you had one
guy scouting the entire country for a 25-round draft by himself, his results might be no better than OSA's and similarly incomplete)

National cross-checker (to re-evaluate the best of the players the original scout found; this could be an optional position depending on a team's scouting budget)

Advance scout (to give you a heads-up on the next opponent)

Director of Player Development, (other country) (each DoPD would cost quite a bit of money themselves, but this would be the guy who would find hidden players and/or provide feedback on guys that can be had from Japan, Cuba, South Korea, the Domican Republic, and so on)

Assistant GM(provides info on the other teams in your league; also, he could be the person to interface with when you wanted to know about trade rumors throughout the league)

fireyournamehere.com(not paid for by your scouting department, of course; this would be a place you could go to read a blog written by some kid living in his mom's basement who has nothing better to do than to call out your bad GM moves)

My side issue is something that's been raised before: scouts are useful for judging potential ability, but why use them at all to tell you something the stats should already be revealing (current ability)? If a guy is blocked at 2nd base in the majors and therefore isn't getting at bats, asking the scouts is an okay way of abstracting watching him yourself in BP, but I'd like to see that more, well, realistic. Specifically in that case, I'd like to see a potentially wildly inaccurate assessment of his hitting abilities. Could you imagine if you had, to bring in a name from the past, Rob Ducey on your team and all you had to go by was how he did in practice? You'd want to start him for sure. Rob Ducey looked like Mickey Mantle in BP. Of course, in actual games he looked like Rob Ducey, but that's the point. A real-life manager might look at Ducey's past record (in game terms, his stats) and scouting reports from seasons past to determine if the guy is worth it. IRL there is a sense that once a guy gets to a certain age he's not going to suddenly change what kind of ballplayer he is. If you have younger guys who you think might have a chance to be big-league players, you put them somewhere where they get regular PT. This is for their development's sake, sure, but it's also so that you as a GM have a record you can base your moves on.

[StyxNCa]

Originally Posted by Syd Thrift

Just curious: have you tried playing with ratings on and scouting accuracy set to low? I haven't done that myself but it seems like it would be worth a shot.

No I haven't. Didn't even think of it. I may try it.

Originally Posted by Syd Thrift

My side issue is something that's been raised before: scouts are useful for judging potential ability, but why use them at all to tell you something the stats should already be revealing (current ability)?

You know, this is a great point! I wouldn't mind seeing the "current ability" ratings taken out. Thinking about it, even though I do see the current ratings with my eyes, I really don't pay much attention to them. It does seem more realistic just to use potential and current stats combined.

[The Wolf]

Originally Posted by Syd Thrift

Just curious: have you tried playing with ratings on and scouting accuracy set to low?

That's an idea worth trying.

[TonyJ]

Originally Posted by deadringer

By the way, on the waiver wire issue I still have to use house rules to keep from cherry picking all the high end talent. I only allow myself to claim players who have major league experience and are out of options. Then I only claim players who would go on my major league roster right now.

I've played stats only since v5 at least and my house rule has always been that I don't claim players on waivers. Period. As The Wolf likes to say, "it's your game, play it your way" and it sucks that I can't use waivers, but the AI has always been really stupid when it comes to waivers. More challenge, in any case.

[TonyJ]

Originally Posted by Syd Thrift

Just curious: have you tried playing with ratings on and scouting accuracy set to low? I haven't done that myself but it seems like it would be worth a shot.

IMO the problem that a lot of people are experiencing isn't so much that the game AI is stupid (although it does have its issues and always will, given that it's a computer game and not a human being) but that the scouting model is still waaaaaaaay too accurate. My ideal for the scouting model would be to bring back multiple scouts (perhaps the head could hire them himself and save you the hassle of micromanaging all that stuff) and have them actually base their reports on "watching" a guy play over a series. If a stud prospect went 1-12 with 5 strikeouts, he'd come back saying that the guy is terrible at avoiding Ks. Perhaps you'd want to temper this assessment with actual current ratings (the equivalent of seeing a lot of loud outs or quiet hits), but I think you'd still want the games witnessed themselves to be at the forefront. A scouting department might have the following personnel:

Minor league scout (for your minors, of course)

Several regional scouts (for scouting amateurs; I don't think computers are at a point where you can reasonably expect the game to simulate 10,000+ high school and college players a year, so you'd need to abstract this into having the game generate the draft-worthy players and having each regional guy look at 20% or 25% depending on how many regional scouts you had, which in turn would be based on your team scouting budget. The more guys a scout would have to look at, the less accurate his reports would be. If you had one
guy scouting the entire country for a 25-round draft by himself, his results might be no better than OSA's and similarly incomplete)

National cross-checker (to re-evaluate the best of the players the original scout found; this could be an optional position depending on a team's scouting budget)

Advance scout (to give you a heads-up on the next opponent)

Director of Player Development, (other country) (each DoPD would cost quite a bit of money themselves, but this would be the guy who would find hidden players and/or provide feedback on guys that can be had from Japan, Cuba, South Korea, the Domican Republic, and so on)

Assistant GM(provides info on the other teams in your league; also, he could be the person to interface with when you wanted to know about trade rumors throughout the league)

fireyournamehere.com(not paid for by your scouting department, of course; this would be a place you could go to read a blog written by some kid living in his mom's basement who has nothing better to do than to call out your bad GM moves)

My side issue is something that's been raised before: scouts are useful for judging potential ability, but why use them at all to tell you something the stats should already be revealing (current ability)? If a guy is blocked at 2nd base in the majors and therefore isn't getting at bats, asking the scouts is an okay way of abstracting watching him yourself in BP, but I'd like to see that more, well, realistic. Specifically in that case, I'd like to see a potentially wildly inaccurate assessment of his hitting abilities. Could you imagine if you had, to bring in a name from the past, Rob Ducey on your team and all you had to go by was how he did in practice? You'd want to start him for sure. Rob Ducey looked like Mickey Mantle in BP. Of course, in actual games he looked like Rob Ducey, but that's the point. A real-life manager might look at Ducey's past record (in game terms, his stats) and scouting reports from seasons past to determine if the guy is worth it. IRL there is a sense that once a guy gets to a certain age he's not going to suddenly change what kind of ballplayer he is. If you have younger guys who you think might have a chance to be big-league players, you put them somewhere where they get regular PT. This is for their development's sake, sure, but it's also so that you as a GM have a record you can base your moves on.

Your ideas intrigue me and I wish to subscribe to your newsletter.

[bababui]

Perhaps its time to develop OOTP Mods in a similar way as Hearts of Iron mods: That change the game, sometimes in radical ways to give users very different gaming experience..

[Eckstein 4 Prez]

Originally Posted by Syd Thrift

IMO the problem that a lot of people are experiencing isn't so much that the game AI is stupid (although it does have its issues and always will, given that it's a computer game and not a human being)

I agree with a lot of what you've said here, but I wanted to highlight this because it's a total copout. Yes, for the foreseeable future the in-game AI will always be less intelligent than a human. But the current in-game AI makes mistakes that a ******ed 8-year-old shouldn't be making. I could go on and on for days about this, since I play out all the games in my solo league, but here's my favorite example:

The AI, with the game tied in the bottom of the ninth, will get men on first and third, one out, and then bunt the runner from first to second. I have seen this happen maybe five times.

That kind of mistake is inexcusable. There are way, way, way, WAY too many examples of such "buggy thinking" that should be getting caught. Given the fact that attention to detail is not exactly OOTP's strong suit (to put it nicely), I'm not optimistic about this getting fixed. I'm even less optimistic when you say things like what you posted and give the game a pass on some truly wretched moves.

[PhillieFever]

Originally Posted by Eckstein 4 Prez

I agree with a lot of what you've said here, but I wanted to highlight this because it's a total copout. Yes, for the foreseeable future the in-game AI will always be less intelligent than a human. But the current in-game AI makes mistakes that a ******ed 8-year-old shouldn't be making. I could go on and on for days about this, since I play out all the games in my solo league, but here's my favorite example:

The AI, with the game tied in the bottom of the ninth, will get men on first and third, one out, and then bunt the runner from first to second. I have seen this happen maybe five times.

That kind of mistake is inexcusable. There are way, way, way, WAY too many examples of such "buggy thinking" that should be getting caught. Given the fact that attention to detail is not exactly OOTP's strong suit (to put it nicely), I'm not optimistic about this getting fixed. I'm even less optimistic when you say things like what you posted and give the game a pass on some truly wretched moves.

I just posted a thread about these types of things earlier today(or yesterday) the title is In Game Strategy Lacking? check it out!