Getting the AI to play as shrewdly as a human?

[Dr. P.R. Park III]

Originally Posted by 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..

If it were only possible.

[Curtis]

The problem would be getting those mods to be cross-platform compatible. Most of the best Hearts of Iron mods won't work on Mac versions of the game.

Back to 1998 Yankees point, in my fictional Oriental universe I use 100% ratings, and I have never seen these waiver wire and free agent problems there. They showed up in my new MLB universe on the second day (using default evaluation percentages).

A lot of people won't like using 100% ratings for the AI because it will result in .350 hitters being pinch-hit for (or replaced altogether by) .220 hitters, but I'm fine with it.

And Syd's idea of turning down the accuracy of the scouts is a good one. I'll do that before my next session.

[Isura]

Yankee - Regarding the 5 star potential free agents, a big factor is the AI's need to keep minor league rosters around 25. That really hampers the AI's decisions. Even with ghost players it moves players around too often and plays them at the wrong level

[1998 Yankees]

Originally Posted by Isura

Yankee - Regarding the 5 star potential free agents, a big factor is the AI's need to keep minor league rosters around 25. That really hampers the AI's decisions. Even with ghost players it moves players around too often and plays them at the wrong level

Yes, I can see that. But if AI were seeing some of the same ratings that I am, with AI preference for ratings scaled up, why would not AI dump some fringe players and snap up these five-star guys? My scout cannot be that wrong on several of these players so that AI knows to stay away.

BTW, I did bump my AI preference for ratings to 70% and waited; those guys that I mentioned in a previous post went untouched (and I finally gave in to temptation and grabbed 3 or 4 of them ). So, it did not manifest any more intelligence in that regard, but I'm hoping AI will be smarter in general with both of us on a more level playing field.

[OldFatGuy]

Originally Posted by 1998 Yankees

but I'm hoping AI will be smarter in general

Good luck with that.

[1998 Yankees]

Originally Posted by OldFatGuy

Good luck with that.

I hear ya!

[koohead]

I'm liking the heavier weight on stats. One thing I noticed in my universe is a return of some older players from over seas. I had about a few vets leave the MLB for Japan a couple seasons ago since they were no longer the superstars that they once were. After adjusting the AI to rely more on stats (I'll have to look but I think I'm at 45 for ratings, 35 current 15 last, 5 2 years ago), I'm seeing a handful of these vets return from Japan and get contracts from MLB teams and contribute. yeah, they aren't the superstars they once were, nor are they even all stars...but they still make the highlight reel and get Player of the Week awards every now and then. Its good to see the AI bring them back. I haven't found that sweet spot yet though...I think I may flip flop ratings and current year and see how the AI plays...but I do think the AI is smarter when stats are weighed more heavily.

[koohead]

Originally Posted by StyxNCa

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.

There's an interesting thought...it would be nice to have 2 different AI talent evaluation settings...one for players with MLB experience (say at least 2 years of stats) where you can have the scout look at stats more, and another for "prospects", where the scout/AI will rely on potential ratings a bit more heavily when making evaluations.

Has this been posted in the suggestions forum? if not, I'll do so.

[The Wolf]

Originally Posted by koohead

...but I do think the AI is smarter when stats are weighed more heavily.

I do too. You see more decisions that makes sense and less that make you go "Huh?"

[Syd Thrift]

In the complete absence of stats, the game will rely upon ratings. ISTR Markus saying something about there being something of a sliding scale within the engine regarding basing things on stats vs. ratings. In any case, my current dynasty weighs ratings as 5%, this year's stats at 65%, last year's at 25%, and 2 years ago at 5%. For the most part, it seems to be working really well. I see former stars stay on teams into their 30s despite not being very useful because the team isn't aware of any better option, but at the same time I see spectacular flame-outs, particularly at pitcher, followed by a few years of the team that owns the flame-out trying to coax some value out of him. OTOH, I'm really not seeing a lot of instances of young kids getting cut and not latching on because they had one bad season, which is what I was afraid was going to happen with the big emphasis on stats.

[The Wolf]

Originally Posted by Syd Thrift

I see former stars stay on teams into their 30s despite not being very useful because the team isn't aware of any better option

That's called realism.

[Syd Thrift]

Oh yeah. I'm not complaining.

[myasu]

Very very interesting ideas in this thread!
I will try the "Heavier weight on Stats" way of playing in my next universe.

[StyxNCa]

Originally Posted by koohead

There's an interesting thought...it would be nice to have 2 different AI talent evaluation settings...one for players with MLB experience (say at least 2 years of stats) where you can have the scout look at stats more, and another for "prospects", where the scout/AI will rely on potential ratings a bit more heavily when making evaluations.

Has this been posted in the suggestions forum? if not, I'll do so.

No.I hadn't mentioned it in the suggestions area. This thought just came to me as I was typing and I didn't think about suggesting it.

[PhillieFever]

Another thing to take into consideration when trying to get the AI to make smart decisions is by adjusting the AI's roster strategy.While playing out a game earlier today,I noticed that the Dodgers were using Jim Lefebvre at SS rather than the much better(and much older) Maury Wills,well come to find out that their manager's AI setting was biased pretty heavily towards playing prospects over veterans,so I changed all of the roster settings to neutral and voila,Wills was in the next game.

[southmonaco]

Based on all the positive feedback on the 'stats only' approach, I'd like to create a new solo league and give it a shot, but I can see an obvious problem at league creation and wonder if anyone has a realistic way to deal with it.

If I create this game using the "New 2008 MLB Game" feature, I already have a pretty good idea what the ratings are for many of the players, so turning the ratings off won't much change the game experience, at least not for a while. Randomizing names will help some but I can probably still figure out who most of the players are based on age, position, team, and stats.

If I create a new custom league modeled after MLB, I'm given a team with no stats (at least initially) and no ratings, and as a GM I'm pretty much paralyzed until the season is well under way and I have some stats to give me some idea what I'm working with. Is that the best I can do, or is there some way to create a new league with unknown players and fictional historical stats that can be used from day 1?

[PhillieFever]

Originally Posted by southmonaco

Based on all the positive feedback on the 'stats only' approach, I'd like to create a new solo league and give it a shot, but I can see an obvious problem at league creation and wonder if anyone has a realistic way to deal with it.

If I create this game using the "New 2008 MLB Game" feature, I already have a pretty good idea what the ratings are for many of the players, so turning the ratings off won't much change the game experience, at least not for a while. Randomizing names will help some but I can probably still figure out who most of the players are based on age, position, team, and stats.

If I create a new custom league modeled after MLB, I'm given a team with no stats (at least initially) and no ratings, and as a GM I'm pretty much paralyzed until the season is well under way and I have some stats to give me some idea what I'm working with. Is that the best I can do, or is there some way to create a new league with unknown players and fictional historical stats that can be used from day 1?

If you want to play with real players,you can just go back to an earlier era,start with no ratings and keep player development on at the default 100.With that,they'll be enough talent movement to keep you guessing.

[Afino]

Originally Posted by southmonaco

Based on all the positive feedback on the 'stats only' approach, I'd like to create a new solo league and give it a shot, but I can see an obvious problem at league creation and wonder if anyone has a realistic way to deal with it.

If I create this game using the "New 2008 MLB Game" feature, I already have a pretty good idea what the ratings are for many of the players, so turning the ratings off won't much change the game experience, at least not for a while. Randomizing names will help some but I can probably still figure out who most of the players are based on age, position, team, and stats.

If I create a new custom league modeled after MLB, I'm given a team with no stats (at least initially) and no ratings, and as a GM I'm pretty much paralyzed until the season is well under way and I have some stats to give me some idea what I'm working with. Is that the best I can do, or is there some way to create a new league with unknown players and fictional historical stats that can be used from day 1?

Well here's three options:

a) Sim a few years before you start so you have a history with fictional players and stats to base your decisions on.

b) Play with ratings on for the first year or two, then turn them off.

c) Play with ratings on for the draft/initial setup only, then turn them off.

Oh, and I'd keep talent randomness far below 100 if you're playing stats only. IMO, 50 is good for my tastes, but some people in this thread advocate even lower.

[southmonaco]

Originally Posted by PhillieFever

If you want to play with real players,you can just go back to an earlier era,start with no ratings and keep player development on at the default 100.With that,they'll be enough talent movement to keep you guessing.

That's not a bad idea, but I'd like to play using the current MLB division and team structures, so I can only go back as far as the last expansion (1998). This helps some because I don't know the players as well, but I'll still be able to guesstimate what the ratings are for a lot of the good players.

What I really would like is a modern MLB structure populated with completely fictional players that come with an appropriate level of fictional stats (i.e., if one of my new players has 4 yrs experience, the computer shows his 4 yrs of stats, whether in the majors or minors).

If this isn't possible to set up, I'm curious what others using the 'stats only' approach have done when they create new games.

[Afino]

Originally Posted by southmonaco

That's not a bad idea, but I'd like to play using the current MLB division and team structures, so I can only go back as far as the last expansion (1998). This helps some because I don't know the players as well, but I'll still be able to guesstimate what the ratings are for a lot of the good players.

What I really would like is a modern MLB structure populated with completely fictional players that come with an appropriate level of fictional stats (i.e., if one of my new players has 4 yrs experience, the computer shows his 4 yrs of stats, whether in the majors or minors).

If this isn't possible to set up, I'm curious what others using the 'stats only' approach have done when they create new games.

Does it matter if you start in 2008???

I mean, you could start in 2008 with all fictional players, sim out 20-30 years to clear out all the initial draftees, and then begin.

That's probably all you could do.