Rethinking the AI - playing not to get fired and reputation

[Syd Thrift]

Okay, so we get a lot of stuff in here about how the AI is bad, which may be true in its way, but mostly I think we're talking about the AI as GM. I actually think everything underneath the GM needs a bit of an overhaul and this may actually even make the game quite a bit harder...

We've all watched and/or read Moneyball, right? I think the big lesson to be taken from that movie, and for that matter from a lot of baseball analysis that's been done since the 70s, is not "stats good scouts bad" or even "an economics / probability based approach to the game is ideal". I'm going to separate this out and boldify it because I think it's very important and, well, it's my central thesis:

The vast majority of coaches, scouts, and GMs do not place winning games as their highest priority. The vast majority of coaches, scouts, and GMs place not being fired and/or being able to find work after their current job as their highest priority,

The primary drama in Moneyball wasn't just "stathead overcomes all around him with stats and headedness". It was how Billy Beane had to fight against not only 29 other teams with management that was opposed to the stuff he was doing, not only a whole cadre of sportswriters and fans who had decades of training to think about baseball a certain way, but his own team's coaches and scouts, who in many cases flat-out would not do the things he told them to do. I feel like this is super-common not just in baseball and really not even just in sports but in all of life.

If you come in trying to do things differently, you're going to be faced with a great deal of skepticism. And TBF a lot of the time that skepticism is warranted because in baseball as well as in many other things, the "old ways of doing things" are very often grounded in really good reasons, at least for the time. And a lot of people who try to shake things up just wind up being crackpots: witness the disastrous managerial stint by Maury Wills with the Mariners in the early 1980s, or the Chicago Cubs "college of coaches" in the early 1960s. Even when you're fighting against stuff that seems just plain dumb on its face - I think the most on-the-nose example out there is that scene in Moneyball where the scouts say they want to draft (insert player here) because he has "a baseball face" and then refuse to take catcher Jeremy Brown seriously because he's pudgy and "doesn't look good in a pair of jeans" - you usually have to convince people with results over time, no matter how stupid you think some of the "old ways" are.

(incidentally Jeremy Brown did wash out in the minor leagues, which isn't uncommon for 7th round picks to be fair, but also to be fair, and I think you see this in basketball quite a bit, too, that 22 year old kids who have problems staying in shape turn into 27 year old men who always carry extra weight around. As a guy who was very, very fat when he was 27 I am not here to body-shame; I'm just saying, that the "he doesn't look good in jeans" criticism is actually not as risible as it appears to be on its face)

Anyway, I'm going to go in and put out some ideas specific to each role:

General Manager: I've stated before that I think the biggest problem with the "AI" for trades is that it allows you to add and subtract players until you get to it being 51% in favor of a deal. OOTP needs to figure out a way to not allow you to do that. In "don't get fired" terms, no GM wants to be either the guy who trades away a superstar in his prime for anything less than their true value (witness how Bill DeWitt is still decried by some people for trading away Frank Robinson in his prime) or for that matter trading away a top prospect who winds up being a Hall of Famer. These trades do of course happen in real life but I do think that the AI should be extra challenging when you attempt to trade for one of these pieces.

Manager: Zigging while the rest of the league zags in terms of strategy can allow you to take advantage of market inefficiencies but it can also (as was the case with Maury Wills) get you fired and never get a job again. To be fair to Wills, Whitey Herzog was a contemporary and Whitey Herzog basically tried to do the same thing in St. Louis that Wills tried to do in Seattle, which is to say he went very heavy on small ball and the stolen base. Still, I think that in game terms you should really not be able to tell the manager to do all that much compared to what his built in strategy is. Every little thing you tell the manager to do, even if it's something he already would have done, should make him less and less happy (we currently don't have personnel morale but we should!), so less and less able to control the clubhouse, less and less willing to re-sign with you, and at some point less and less willing to actually implement your strategies.

The more I think about it the more I think this should be applicable to all coaches and staff. Nobody, not in baseball, not in whatever job it is that you work at, likes to be micromanaged. Period. I don't care if you as a GM think the micromanaging you're doing makes them objectively better and that there's zero reason why the manager should be sending out a guy to steal 50 times a year if he gets caught 25 of those times. Your manager would and should dislike you for tinkering. Billy Beane famously had to go into the team's fitness center and work out during games so that he wouldn't constantly be in manager Art Howe's ear during the Moneyball era, and IIRC Howe quit the team anyway.

Also though if you do demand that they do things that are against the current flow of play, those should be met with more ire than asking that a guy just, like, be normal (although there, too, if you hire the equivalent of Whitey Herzog to manage for you and then you tell him to manage like Earl Weaver, all you're going to have is an angry Whitey Herzog). In game terms, I think that means that the further in and out you set those sliders, the less a given manager is going to like it.

Also... I think those sliders should be based on a manager's current tendencies, if that makes sense. If a manager, for instance, is already +2 on Slow Hook and you goose that all the way up to +5 on the team strategy page, that shouldn't be read as "override my +2", it should be read as "take the +2 but make it the equivalent of a +7". That also means that a manager (like Whitey Herzog) who is naturally at +5 on Stealing gets instructed to be -5, that should have the effect of being average. Does that make sense?

Pitching, Hitting, and Bench Coaches: All of the micromanagey things that a manager would dislike, these coaches should dislike as well if their "station" is involved.

Also, these guys don't want to look like fools. They should just naturally get mad if a. they are a hitting coach and the team isn't hitting well (I know people hate this when players do it but it's human psychology), but also b. they should periodically inform/demand that you take players out or put players into the lineup / rotation / bullpen, and get mad when things don't line up the way they'd prefer (this should apply to managers, too, perhaps even moreso: IRL a pitching/hitting/bench coach would complain to the manager, who would then fix the issue themselves or complain to the GM - if you're playing as a manager, perhaps that's when you get these reports, but AI managers should get this feedback behind the scenes too).

Minor League Managers and Coaches: I think these guys would probably be a bit more malleable - there are often times when you do things that hurt an individual minor league team but are done for the good of the organization, like force a guy to start/train in LF when he's not rated there - but there should still be limits to that, and probably the higher up the minor league chain a manager gets, the less he's going to be OK with the micromanagement.

Base Coaches: I'm just going to say it: as micromanagement-intensive as I can get with my teams (I literally play as GM and manager of all teams in my league) I do not coach baserunning decisions, as it is too unrealistic (and micromanagey) for even myself. I think that any time you override a base coach's decision, they should get mad at you for it.

Also, in the specific instance of base coaches, the "don't look like an idiot" thing comes back into play in full swing. There are some very specific things that will get you talked about on Sportscenter or in other locker rooms in a bad way: making the first or third out at third base is a big one, advancing slow players on hits, running on guys with strong arms (think about being the guy who gave the green light to Terrance Long to move from 1st to 3rd against Ichiro on that famous throw-out; you'd never be able to live that down). Coaches should avoid these, even if you are extra aggressive in baserunning, and if you are going to call for these plays manually, you should expect your coaches to revolt if the play goes against you.

Team Trainers: I'm honestly not sure how much this may apply here. The injury system might want to be revamped at some point, maybe to incorporate something like how FM handles some injuries where you're given a decision to play someone through it or put them on the shelf, but that's beyond the scope of this and I only include this to be thorough.

Scouting Directors: This is probably the role I was thinking most about, well, that and manager, so here goes:

I've spent a lot of time talking about how I think the scouting system doesn't work right. Well, here's another way of looking at it: scouts should be kind of all over the place all the time, because that's just how scouting is. It's very inexact by its nature. Scouts should be triangulated on veterans better but only very specifically based on the statistical output of those veterans. I love, love, love TCR but if a 10 year veteran suddenly goes from having 10 HR power to 25 HR power, scouting reports should be telling you that guy has 10 HR power until the on-the-field play makes it too obvious that the player can now hit 25 HRs in a season. Likewise, if/when a player loses it, your scout should be the last person to tell you.

Think of this from the scout's perspective. If you are watching a game and you think Albert Pujols is no longer an elite hitter, and you tell your boss that, okay, he'll take your information in but in 2 months if that turned out to be just a slump, I hope you didn't try to hard to tell your boss to trade the guy because you might be fired for that. Likewise, if you're watching. like, Mitch Haniger play and you become convinced that he's going to break Mark McGwire's record, and you get your GM to pay through the nose through him and he just winds up being Mitch Haniger, it's good-bye job for you. Instead, your "scouting reports" on veteran players should mostly reflect conventional wisdom and be based almost completely on their actual statistical output.

For young players, especially potential draftees, it really should be a crapshoot. Like, I'd be okay if OSA was like 90% random chance and scouts were like 70%. The fact is that in real-life baseball, high draft picks do not come through. Even first round picks don't make it to the majors something like half the time, and MLB is filled with superstars who were straight up not 1st or 2nd overall picks. Mike Trout, for instance, was drafted 25th. Of the top 5 picks in that draft (2011), 2 never made it to the major leagues at all, a 3rd got a very brief cup of coffee, a 4th (Dustin Ackley) made it but was never very good, and only 1 guy (Stephen Strasburg) actually turned into a long-term major league player. I know that the NBA and NFL have occasional, similar stories, but this is the norm in baseball.

So you should have this almost-useless OSA report and the slightly-less-useless scouting report and from there... well, from there I think your scouts should just be triangulating their initial reports with a player's on-field play, and that's all. If a scout thinks a player has 70 potential and then regrades them to a 25 based on nothing more than being wrong the first time, how does that apply to the real world? IRL scouts do just plain get youngsters wrong a *lot* - they'll call the next Mike Trout a bum or hype up some guy with good velo but little actual potential as the next Roger Clemens - but a scout who changes his mind 2 months in is a scout who, again, gets fired. How can you trust a guy like that?

This is basically the foxes vs hedgehog debate - fox-like thinking that allows you to completely throw out your previous analysis might be the optimally best way to proceed, but hedgehog-like thinking where you stick to your priors until you're overwhelmed by contrary evidence makes people remember your hits more, and in an environment where literally everybody misses more than half the time, nobody remembers them anyway except as kind of funny footnotes. Literally everywhere you go, people take hedgehogs more seriously than foxes. Like, the whole entire point of sites like 538 and books like Freakonomics is to encourage more fox-like thinking in a world where everyone pursues the hedgehog.

Final note for all roles: Forget about all this "LEGENDARY" reports and stuff. Like, sure, behind the scenes some scouts (also coaches, managers, etc.) are better than others but as presented to you the player, all you should know about is their reputation, and their reputation should be 100% based on what's actually happened in the game. You want to hire a new scout? Their rep is based on how many stars they've recommended and that their GM drafted. Likewise, a manager has to be based on the success of their team, the pitching coach on the success of their pitching staff, and so on. A trainer is judged based solely on how many major injuries their team has suffered. These guys can maybe have tendencies that can clue you in to if they prefer power pitchers or veterans or what have you but in terms of actual ability? Nah, that's got to be blocked off from the player.

If you think a guy is underperforming, you fire him and try to find someone else. Even when it comes to things like personality or the sorts of things you'd learn about in an interview... in the real world as much if not moreso than in athletics, there are people who aren't very good at what they do who interview really well and people who interview poorly who are pretty awesome. A new, untested coach/scout/etc. should be just flat-out unknown; not bad, necessarily, but unknown, and the only way to get data on them is to put them into a given environment and see how they do.

I will say that for scouts and maybe trainers, I think this way of thinking means you ought to have players working underneath the main director/trainer, just as a way of seeing if they're any good. We don't have to go back to the way the one edition of OOTP had of micromanaging scouts; you can just, for instance, have a North American guy, a South American guy, and so on. Trainers, maybe you'd have someone who deals with pitchers vs. one that deal with position players (you'd definitely need to make any reputation account for that) or else, maybe, your sub-trainers get randomly assigned to deal with each injury along with your main guy.

[Syd Thrift]

I didn't want to include this in the first post but some of these things can apply to players, too. Nobody wants to look like an idiot out there. If you're force-starting Barry Bonds at shortstop, Barry Bonds is going to quit the team until you fire the guy who's force-starting him there and replace him with someone who's not trying to make him look like a sucker. There are some more subtle things going on too: I can't remember the name of the guy now but Foolish Baseball did a report on a catcher who couldn't throw anybody out for like a year. He concluded that this was actually mostly because his team (the Pirates I think) had their pitching staff not pitch from the stretch with runners on base and essentially hung this guy out to dry. If I was the catcher put in this position, I would be pissed, like this alone, if I thought I was good enough to play in the major leagues, would make me demand a trade (IIRC Felipe Alou did something similar with the Expos in the mid-90s).

GM meddling is one of these things that really should be looked on very similarly to owner meddling: tolerable when it works, the source of player and personnel revolts when it does not. Whereas a GM that does nothing but sign guys and then allows his underlings to make all the decisions, and who takes recommendations from his scouts on who to draft, and who doesn't piss off other GMs by constantly pestering them and trying to fleece them... that guy probably gets way more leeway, even if - especially if - his team is not winning. A micromanagey GM whose team finishes with 100 losses is probably out of a job after 1 or 2 seasons.

[PSUColonel]

Yes, I have always advocated for a "dynamic" reputation system within OOTP for staff. I brought this up years ago....I am glad you also pointed this out so perhaps we can re-visit the issue. We need to have chats with the developers however, and it seems they drop by less and less these days to discuss potential concepts with us. We used to have some pretty good brain storming sessions as I recall.

[Syd Thrift]

Originally Posted by PSUColonel

Yes, I have always advocated for a "dynamic" reputation system within OOTP for staff. I brought this up years ago....I am glad you also pointed this out so perhaps we can re-visit the issue. We need to have chats with the developers however, and it seems they drop by less and less these days to discuss potential concepts with us. We used to have some pretty good brain storming sessions as I recall.

I feel like they do come in every now and then, particularly when they aren't busy writing code - which, ironically, makes what I'm talking about right now maybe the worst time to get them to respond (on the other hand, this could also give a lead dev an idea of something to add into this version while it's still being worked on!), given that they're all going to be heads-down on 23 right now.

But yeah, knowing exactly how good your personnel are, even if you don't know exactly how good your players are due to the randomness that's currently added with scouting, makes the game way too easy for the human player.

[OrangeP47]

I mean, I don't need to read your post (though I did): As a HUMAN PLAYER I'm playing to not get fired, not necessarily to win, most of the time

That's more because lately I've been GM only-ing small market teams with owners who are decreasing the budget year after year, in a universe that has inflation, so it's a double whammy, and I feel as if I can sometimes only complete certain owner goals or win, not both.

Recently I jumped ship though from that specific team after 12 years, so we'll see what happens next.

[oldfatbaldguy]

I like what you're saying about making development more of a crapshoot. Specifically, what I think I don't see enough of in the game is minds changing inside an organization. I've traded young players out of the low minors who my staff considered to be no-hopers and seen them develop into good major-league players for other teams. When it happens I just think, "Well, if I'd kept him we'd still have him in Double-A." Because what I almost never see is the scouting staff doing a pivot on a guy and saying whoa, we really underestimated this kid, he might have a chance.

On my fictional team I've got a pitcher who was a seventh-round pick out of the University of Florida who has developed well beyond his supposed potential and spent time in the majors two seasons after being drafted. A seventh-round Gator isn't exactly coming out of nowhere, and it feels miraculous. I think it's too rare in the game. But this is based on my gut feelings about my OOTP experiences, nothing more, and everybody else's mileage may vary.

[PSUColonel]

This is a good post..I just want to give it a bump so it can be discussed a bit more.

[Syd Thrift]

I was also thinking that I got sidetracked into the “don’t micro” thing, which is important, but wanted to hit the “don’t embarrass yourself” bit more. A manager won’t want to…

- give a lot of PT to a guy who’s not hitting for average. Yes, Chris Davis, but there are limits.

- give PT to guys who commit a lot of errors. This isn’t just bad fielding; it’s embarrassing to have a guy out there who’s fielding under .900. Butch Hobson, but Hobson played 40 years ago and did wind up losing his job.

- At least some managers should completely melt down when players walk guys too much. Maybe that one isn’t universal but I mean I watched Lou Piniella manage…

A lot of this stuff probably gets mitigated with better records, but that’s part of why I think chemistry would work for personnel.

[oldfatbaldguy]

Originally Posted by Syd Thrift

- At least some managers should completely melt down when players walk guys too much.

Well, there's one in my game who does, but it's the one who isn't controlled by anything the programmers set up.

[Pdubya64]

I love the discussion Sid, so thanks for starting this thread. I want to think on it some more before writing a decent response, but based on my experiences in other computer game genres I will say this:

I would be worried that so many moving parts in a OOTP AI with all the various contributors is going to make it near impossible to stay clear of bugs and unintended AI gameplay, etc.
Yeah, I am pessimistic. But AI is REALLY hard stuff to do. I sort of compare it to those robots that are almost human looking, but end up just creeping humans out... we are very good at detecting any anomaly in "artificial" stuff.
Perhaps we will get to the point where things get pretty close, maybe with quantum computers? I AM seriously impressed with EPIC's new Unreal 5 work. It's being used for movie-making already.

[Syd Thrift]

I don't think any of this is quantum-level AI stuff. The heart of it is basically adding chemistry for your personnel and then making them get madder and madder when you micro them. And then on the flip side the AI is already setting all types of business rules; I'm just point out where said business rules could/should be changed.

[Syd Thrift]

Originally Posted by oldfatbaldguy

Well, there's one in my game who does, but it's the one who isn't controlled by anything the programmers set up.

Admittedly I was reminded of that by a recent game where I was like "ANOTHER LEADOFF WALK %@#&%&(@ THIS GAME".

[oldfatbaldguy]

Originally Posted by Syd Thrift

Admittedly I was reminded of that by a recent game where I was like "ANOTHER LEADOFF WALK %@#&%&(@ THIS GAME".

Oh, the replies I make in my head when I hit that mound visit button and the pitcher says it's all good.

[timmermac]

Originally Posted by oldfatbaldguy

Oh, the replies I make in my head when I hit that mound visit button and the pitcher says it's all good.

Honestly, I never hit the Mound Visit button. It appears to be completely useless in my experience.

[Pelican]

Please don’t say that. I at least use it to give my relievers more time to get ready, when the guy on the mound has suddenly imploded. The AI gives broad hints a SP is done, such as giving up rockets on the first pitch. In my head I imagine what I would say to the pitcher who is obviously out of gas. I’m infuriated when I get the generic “plenty of gas left in the tank” when they are pooping the bed big time. I wish just once we could have a productive conversation. Occasionally, I admit, the SP says he is out of gas and running on fumes. I appreciate honesty.

[oldfatbaldguy]

I THINK based on experience that a good pitching coach can settle a pitcher down if you hit the mound visit button.

I KNOW that a pitcher in the bullpen who is listed as "warming up" will usually be "ready" after a mound visit.

[Sweed]

Originally Posted by timmermac

Honestly, I never hit the Mound Visit button. It appears to be completely useless in my experience.

Your loss

[beepollen]

Originally Posted by oldfatbaldguy

Oh, the replies I make in my head when I hit that mound visit button and the pitcher says it's all good.

"I have yet to even break a sweat."


Yah, dummy, because it's only the first inning and you've already let the first six guys reach base...

[pgjocki]

I both love and hate this post so much.

I'm currently struggling and am ready to fire my coach and burn the team to the ground; but of course I'm the dummy who signed most of these guys and I'm fairly certain it's my inability to manage pitchers that is causing most of my problems.

I love the concept of interaction with the coaching staff and then a real "Cause and Effect" (or is it Cause and Affect?) from your actions. That might take the game a little away from emulation to RPG land and that is probably another debate to be had.

My only hill I'd attempt to die in is that any of the consequences need to be transparent - I'm not interested in another aspect of the game where the AI is doing things and I can't trace it back to why.

[Pelican]

Originally Posted by beepollen

"I have yet to even break a sweat."


Yah, dummy, because it's only the first inning and you've already let the first six guys reach base...

Or, "great, then you won't even have to shower. Just go into the clubhouse and change into your street clothes and catch a flight to AAA."

Or, "maybe if you put a bit more effort into pitching, you might see better results..."

Or, "it's possible you have a serious medical condition that is preventing your body from perspiring, so we are going to shift you to the IL for sixty days to have a specialist run a bunch of tests on you and your rag arm".