Suggestions: Artificial Intelligence

[battists]

This suggestion thread is designed to collect suggestions for the next version of OOTP, on the following topic:

Artificial Intelligence

Includes
- Suggestions specifically related to the game's artificial intelligence, including situational AI in the in-game engine, and roster descision-making

Excludes
- Probably some overlap with other threads, but that's OK!

Instructions
We will use this thread to collect suggestions on the above topic. Later on, we hope to present this information back to the community in a way that would allow you to vote and prioritize certain features.

If you would like to participate, here are the rules!

1. Only 1 suggestion per post.
2. Suggestions must be on-topic for the thread.
3. Constructive, non-inflammatory suggestions only.
4. No comments on others' suggestions are necessary. That means, we don't need 15 posts in the thread saying "Great idea!" or "I agree!"
5. Please don't link to other forum threads as your way of making suggestions. ("My suggestion can be found HERE.") If it's important enough to you that you want it in this thread, then it should be important enough to put it in the format I've asked for here!
6. You can certainly post suggestions that conflict with or build upon earlier suggestions.
7. Mockups are welcomed, they really help Markus see what you're interested in accomplishing.

Format for Suggestions

Suggestion
What is your suggestion? Be as specific as possible; include the screen you're talking about, if applicable.

Reason
Why do you think this is important? If this is to correct what you perceive as a problem in OOTP 2007, what is that problem? Be as specific as possible.

Priority
In your opinion, how important is this suggestion?
High (MUST have), Medium (should have), Low (nice to have))


Handy Cut-and-Paste Suggestion Form:

Suggestion


Reason


Priority


Thanks for participating!

Steve

[JerryArr]

Suggestion
Different individual AI GM's have varying levels of intelligence/preferences for prospects/etc.

Reason
Adds variety to the game.

Priority
Low

[jmolony3]

Suggestion
Allow the major league club to direct the development of minor league talent. ie if a player is a SS but the major league club has that filled. Then push the minor league club to use the player in other positions that might be needed, like 3b or 2b. this would be a check box and allow up to 3 alternate positions to be trained. this could be used to convert a started to reliever or the reverse. This could be limited by initiating a designation of ML prospect.
this is something beyond the 40 man rosters where you believe the player may make it to the major leagues. maybe 40 players per team but only players with minor league contracts would be eligible.

Reason
This can be done manually but with 6 minor league teams it is tedious to try and accomplish. i also believe that the Major league teams have this influence today to direct their prospects development.

Priority
Low but it would be nice.

[Comedian2004]

Suggestion
Teach the AI to look ahead at his team's schedule and make changes to his rotation based on the upcoming schedule.

Reason
Makes AI Smarter. Example, it sees that the team it is playing next series is a HR hitting team. Pitch around in the rotation, when it can, to now allow starters to start that are prone to allow more home runs.

PriorityMedium

[Stu]

Suggestion
Make the AI avoid using their backup C unless necessary.

Reason
I've seen my backup C substituted into all sorts of weird positions. He should only be brought in to replace my starting C unless injuries make it absolutely necessary.

Priority
Medium

[Stu]

Suggestion
Change depth chart substitution priorities. This is difficult to explain, but here goes. Currently the AI doesn't handle multiple substitutions using the depth chart very well at all. As an example if I have:

1B Starter: Kevin Youkilis
1B Depth: David Ortiz
DH Starter: David Ortiz
DH Depth: Wily Mo Pena

If I sim 7 days Kevin Youkilis got hurt on the 2nd day of the sim, the AI would start some random bench player at 1B (even if they aren't rated at 1B) and keep David Ortiz at DH. It should fill out the fielding positions first, starting Ortiz at 1B and Wily Mo at DH. The AI should do whatever it can to make sure players aren't playing out of position.

Reason
Smarter AI

Priority
HIGH

[Stu]

Suggestion
Have the AI retaliate to multiple beanings in a series/game. Also have umpire warnings and suspensions for beanball wars.

This could also be tied into the rivalry idea I posted in the Grab bag thread. Teams might be more likely to throw at a rival team.

Reason
Realism/Entertainment

Priority
Medium

[KGAGE]

Suggestion
Create umpires with different ratings. Some ratings could be strike zone size, temperment, accuracy. Would tie in to things such as ejections/suspensions. Also pitchers might have a better day with umpires who have a large strize zone.

Reason
Adds more unpredicatability to the game, immersion.

Priority
Low

[Comedian2004]

SuggestionThe AI has to be made as smart as possible, and then some.

Stuff like:

(Based on a hitters intellegence and skills), a hitter try to hit the ground ball to the right side to bring a run in, or just hit a deep fly to right field; pitchers being aware of hitters that have hurt them in the past and pitching with more care to them; managers using the DH to 'semi-rest' older players and catchers; managers aware of the upcoming schedule, like not pitching a strong left hander in the last game of a series because they know the next series is vs. a strong left hitting team.

Reason
Realism - The game is full of a lot of great features, it is going to be hard to add major features and IMHO, 75 percent of the work for the new game should improve the AI and fine tuning the features in the game.

Priority
HIGH

Additional Notes:

Fix ALL bugs of course is #1 priority, no matter how small the bug is. 2nd would be the AI and 3rd would be fine tuning the features introduced in 2007. 1-3 major features would be needed also, but I would like to see this being the #1 complaint for 2008:

The AI is too smart, I can't beat it.

[lt2cents]

Suggestion

The new 2006/2007 OOTP platform is good, and now provides a soliid basis to build this game into a tremendous sim and role play experience. That is not to say it certainly needs tweaking in many places. But it is undoubtedly into the indefinite future that improvements to the AI can be made.

Here is a basic one-- line up selection and ordering.

Unrealistically often very slow players and catchers are at the top of the lineup, when comparable OBP players with more speed are lower in the lineup. For various reasons RL catchers often hit down the lineup.

There are other batting order shortcomings. Often the human eye can immediately spot reasonable naturals for 3rd, cleanup, and other line up spots. Unfortunately the AI misses the big picture too often.

Also player positions often make little sense, as when two players are correctly chosen for the line up, and could play each others position better, but are placed in their weaker defensive spots.

Line up making needs better heuristics with some use of fuzzy tolerances to improve the resulting line up overall strength. The AI needs to be able to step back and see the bigger picture of the resulting line up.

Reason

For those who play out games as manager, to see the AI team with a more sophisticated, realistic feeling line up.


Priority

[ms2002]

Suggestion
Make the player evaluation more realistic, or at least more customizable. Differentiate managers/GM's as being either "Sabremetric-based" or "Traditional." Sabremetric GM/Managers will seek out and prefer players based on their VORP and other secondary statistics, while the Traditional GM/Managers will prefer players with experience, raw numbers, and just plain luck.


Reason
It's much more realistic. The main gripe I have with OOTP2007 is that there is an extreme over-emphasis on VORP. It's good for statheads, but in real life, very few GMs actually follow this model. Teams like Oakland and Toronto are more sabremetric-based, while teams like Seattle, San Francisco and Chicago are much more traditional. This should definitely play a factor in the game. Also, there should be far more GMs following the traditional model than the sabremetrics. Although, it'd be nice if this was customizable.

An example would be the many cases I've seen in OOTP2007 where a 40hr, .270 hitting 1B is rated 1 star and disregarded by the AI just because his VORP is not high. As we know, 1B is an extremely offensive position and they should be penalized for this. They should get high-paying contracts as well (this does not happen in the current version). A change in the way the players are evaluated would fix this.

Priority
Very high. What good is a perfect player development model when the AI evaluation is flawed? I would love to see this being a top priority for OOTP2008. I'd be more than happy to clarify anything as well.

[injury log]

This may sound like an 'Improve the AI!' suggestion, but here goes:

Suggestion

Improve AI player evaluation, particularly of younger players, and make AI GMs give much greater consideration to long-term implications of roster decisions:

-the AI should take greater account of Potential ratings, and AI GMs should be very reluctant to waive or release a young player with high Potential, regardless of current ratings;
-AI GMs should consider both short-term and long-term consequences of roster decisions (trades, waiver moves, etc)- the AI seems too focused on the short-term;
-a simplified version of MORP could perhaps be incorporated into AI decisionmaking, as a way for AI GMs to judge the long-term value of prospects (versus the more expensive established players).

Reason

Even with player evaluation set to 'Heavily Favor Prospects', AI GM's don't seem to value prospects nearly as highly as real-life teams. With trading set to 'Hard', I can always trade a good upcoming free agent at the trade deadline for a top prospect, even when the players have similar Overall ratings. The game doesn't seem to value the six cheap years a MLB-ready prospect can give a team, as opposed to the three month rental the impending free agent will provide, and instead seems to focus on the small difference in their current ratings. So the AI seems focused on the short-term. If MORP (or something similar) were incorporated into AI decisionmaking, AI teams would be better able to make decisions that reflect the long-range forecasting of real-life GMs.

Many of us have also seen AI teams draft a monster prospect in the inaugural draft who isn't ML-ready; the AI then invariably waives the player because he isn't soon going to play in the Majors. To me, this would be similar to the Devil Rays waiving Delmon Young (who was signed to a Major League contract) the year after he was drafted, clearly absurd. In this case, the AI team again seems focused on the short-term- the small value an extra 40-man roster spot could provide for the current season- and not the long term value of the prospect.

Priority

Very high.

[t-bone shuffle]

Suggestion:

The in-game AI must be made smarter, or perhaps more realistic. I'm not really sure which is the issue but it's certainly one or the other or both.


Reason:

The two most disappointing examples I can give are:

1) The opposing AI Manager consistently intentionally walks the tying run to the plate, late in games. Usually it's with a 3 run lead and a runner on second, but I've seen it in other situations as well. Regardless, a big league manager would NEVER, EVER, intentionally bring the tying run to the plate with a 3 run lead in the 7th inning or later. And unfortunately, this seems to be done without regard to the personnel involved. This must stop. (Please, I can hear those of you who will say you've seen this and that, but as a rule, you'd never intentionally bring the tying run to the plate with a 3 run lead late in the game).

2) I'm not completely sure if this isn't also a roster AI issue, but way too often I see the opposing AI managed team lose a game late because it has no rested relievers and leaves a reliever in to finish, and almost invariably lose a game. My opinion is that in reality, a manager would make a change to another tired reliever at some point just to make a change. But in ootp, when the AI has nothing but unrested relievers left, it usually just leaves the current pitcher in to absorb a beating, regardless of the game situation. I've won too many games in this fashion and it feels completely unrealistic, and truthfully leaves a pretty bad taste in my mouth. Again, I'm sure that some of this is roster AI, but the in-game AI too often sacrifices a game because of player fatigue.


Priority:

For me it's very high. Winning games when the AI makes bonehead moves such as the above is no fun and leaves me questioning the credibility of the game itself. It's gotta change.

[cbbl]

Suggestion

Another AI improvement suggestion:

In non-DH leagues, the AI should recognize that the pitcher is usually the worst hitter in the lineup and seek to pinch hit for him whenever practical. I don't know how the "Pinch Hit for Pitchers" option affects this, but the specific situation is as follows:

Late inning game, the pitcher is getting tired and is not likely to return for the next inning. The pitcher comes up to bat. Unless it is a strategic bunting situation (and the pitcher is a good bunter) or, in rare cases, the pitcher is also a decent hitter, the pitcher should ALWAYS be pinch hit for. It makes no sense, usually, to allow a pitcher to bat if he is not returning the next inning to pitch. The AI must employ some amount of "future" vision about what it will do in the next inning.

Reason

Improved strategy. Pitchers are often easy outs. If the AI allows a pitcher to bat when he will be pulled from the mound before the next inning, it's basically giving the other team a free out -- and fails to mimic RL.

Priority

I'll say medium.

[Raidergoo]

Processed

[Ktulu]

Suggestion
Before features are made optional, there should be appropriate AI to handle all the available options.

Reason
The OOTP community loves options and with good reason, customizability is arguably OOTP's best feature. However, at times I think there are too many things turned into "options" just because of the reasoning "more options are better, so why not make it optional?" Well the main reason is, what good is an option if the there is no AI behind it?

It's great that we can select X number of minor league levels, but it kinda sucks if the AI teams don't know how to use them or player development is adversely affected because a player is playing above/below his appropriate level even if that level doesn't exist in your universe.

It's also cool to be able to choose the active roster and "40" man roster sizes, but does it do much good if the AI GMs have no idea how to manage your custom roster settings?

* not saying these things actually happen (i honestly don't know), just using them as examples

So yeah, basically when things are made optional there should be programming in place to handle the available options

Priority:
Medium

[swampdragon]

Suggestion: The AI should only play players at an unrated position as a last resort.

Reason: I hate watching zero rated fielders kick the ball around. This isn't a soccer sim.

Priority: High

[pat]

Originally Posted by Raidergoo

Processed

What do you mean of processed? All the suggestions will be made or it is just noted?

[Raidergoo]

Originally Posted by pat

What do you mean of processed? All the suggestions will be made or it is just noted?

It means I have collected all suggestions to that point onto the master spreadsheet.

[pat]

Originally Posted by Raidergoo

It means I have collected all suggestions to that point onto the master spreadsheet.

Ok. Thanks. Pardon my ignorance.