Eventually there will be analysts in the dugout, of this I have no doubt.

There's a lot more to decision-making than identifying the "statistically correct" decision. And GW seems to get that to some extent, when you consider these two quotes:

The right decision is only marginally better than the wrong one

biased by his interactions with his players in getting them ready to play

"Getting players ready to play" is arguably as important to the manager as in-game decision making. The tactically correct decision for the short term may not be strategically correct in the long term for the team, and the goal of a team (and by extension its manager) isn't just to win today's game but also to build the team for the future.

Analysts in the dugout may help a team win the battle, but the manager also has to consider what it will take to win the war.

-- MWE

"Getting players ready to play" is arguably as important to the manager as in-game decision making

I'd say there's no argument for the latter. And I agree that there are many decisions that may look bad in the short-term that are, in fact, quite sound when the longer term is considered. That can get lost in the "Worst Move Ever" reactions from fans.

It is doubtful that a manager would be effective in performing what are arguably his most important functions -- dealing with the clubhouse and the media -- if he was perceived by the players and sportswriters as not really being in charge of the team. If a manager was forced to cede authority over in-game decisions to someone like MGL, I think the concern is that everyone would see the MGL figure as the real power behind the throne and dismiss the nominal manager as merely a glorified bench coach, which would compromise his ability to perform his other functions. Even if you think that MGL would be the greatest field tactician ever, it's hard to imagine him being anything other than a trainwreck when it came to managing a clubhouse full of jocks or a room full of sportswriters (suffering fools gladly is not his strong suit, to put it mildly).

If a manager was forced to cede authority over in-game decisions to someone like MGL

What if MGL was the bench coach who gave constant input, and then the manager was left to make the final decision? You need to find the right even-keeled analyst who won't care when the manager goes against him, but its possible.

Or what some teams seem to be trying - having the front office keep banging home the importance of not giving away outs, etc, to the manager behind closed doors and hoping enough of it sinks in when the game actually starts. Then just hire some intern to sit at a computer in the back and run numbers for the manager.

Which is more difficult, finding an analyst who is good at interacting with people or finding a former baseball player who is comfortable writing code and dealing with large datasets? There is no need to eliminate coaches in this hypothetical, there is only a need to delegate responsibility.

I think he thinks this is a rhetorical question with an obvious answer... That's cute.

Or what some teams seem to be trying - having the front office keep banging home the importance of not giving away outs, etc, to the manager behind closed doors and hoping enough of it sinks in when the game actually starts. Then just hire some intern to sit at a computer in the back and run numbers for the manager.

Yea, I think this all needs to be part of the overall organizational philosophy. Having an analyst in the dugout second-guessing the manager's "gut" and giving the media a great scapegoat when things don't go with the odds seems like a terrible idea to me.

I do think the NFL needs to have this kind of assistant though. NFL clubs need some sort of assistant coach whose sole job is to look at big picture strategy like time clock management, when to use challenges, when to go for it on 4th down, etc. But even then you still probably suffer from the tension between what the stats say and what old school football says.

The fact that many of the "modern stats" ignore game theory and fail to understand sometimes you should do a suboptimal (in the percentage sense) tactic leads me to believe that the idea of a stats guy in the dugout spitting out "never bunt" advice would be unhelpful at best.

This situation happens often enough in pro football, of course. Jason Garrett was recently stripped of play-calling duties for the Cowboys in favor of Bill Callahan, one of his assistants; Garrett becomes a "walk-around" head coach, in effect booted upstairs, figuratively. But of course both the complexity of the sport, and the ethos of coaching in the NFL, allow for this (not that everybody thinks it's a great move). Callahan was a small-college quarterback who got onto the coaching track right out of college; Garrett was actually an NFL quarterback, though a backup: Garrett has a resumé more typical of MLB managers, who are often benchwarmers in the majors, while Callahan is a typical "never played in the NFL" coach, a very common type.

So if baseball became as elaborate in terms of coaching as football, I can see something like this eventually happening. Baseball coaching staffs have steadily, if slowly, proliferated over the decades. The thing is, though one sees bonehead moves now and then, is it really true that managers routinely use suboptimal tactics? We saw SH and SB-attempt rates go down during the last high-offense era, and tick slightly up again over the past couple of seasons, all but unconsciously. Managers seem to adapt to changes in doctrine pretty reliably. I hate the way 21st-century use bullpens, largely because pitching changes are such a bore and involve parades of interchangeable pitchers I've never heard of, but some of these bullpens are pretty damn effective, you gotta admit.

The fact that many of the "modern stats" ignore game theory and fail to understand sometimes you should do a suboptimal (in the percentage sense) tactic leads me to believe that the idea of a stats guy in the dugout spitting out "never bunt" advice would be unhelpful at best.

Sure, you want to keep the defense guessing, but the majority of bunts, just to stick with your example, seem to come right into the teeth of the defense when they're expecting it. There's no game theory to take advantage of when the pitcher is throwing high fastballs and the 3B is already in on the grass.

Starters are left in too long... closers are left sitting on their asses during high-leverage moments

Many years ago I maintained for a couple of weeks a bullpen forecast for the Red Sox, at a time when it seemed to the crowd here that their relief pitcher usage was unfathomable. Each day I would post to GC, well before game time, the recent bullpen usage for each reliever and my forecast of who was "available" for that night and who wasn't. IIRC, I also considered who the game's starter was, and the next day's schedule/starter.

Most of the time, the bullpen usage followed sensible/sabermetric principles given all that other info. So, if a high-leverage opportunity arose, but their closer had thrown 30 pitches each of the prior two nights, they went with the best available reliever other than the closer.

Naturally, since then I've generally assumed that there's a lot more to reliever usage than just the momentary matchup and who is on the roster, and that any judgment of a manager just on those items is insufficient. Sabermetrics is about finding useful information and throwing out useless information, not about finding low-hanging fruit then ridiculing people who are carrying around ladders.

The fact that many of the "modern stats" ignore game theory and fail to understand sometimes you should do a suboptimal (in the percentage sense) tactic leads me to believe that the idea of a stats guy in the dugout spitting out "never bunt" advice would be unhelpful at best.

Clearly that isn't the type of guy you'd want in the dugout. There are good times to bunt and bad times. If your analyst can't quickly figure out the difference between these two he's not very good and not the person this author is talking about. I'm fairly certain that a good analyst understands that even bunting in suboptimal situation from time to time can lead to better results down the road. If we're talking bunting, it brings the infield in, keeps them honest and opens up holes. So again, if the analyst can't figure this out he's not very good.

I'm fairly certain that a good analyst understands that even bunting in suboptimal situation from time to time can lead to better results down the road.

So does a fairly good manager, so what is the analyst bringing to the dugout?

I am a huge stats guy in general, but I see no point in having it in the dugout. I don't think it add much to the day to day running of the club/game. It is most effective at talent evaluation and possibly teaching/helping instruction over the long haul and not in game tactics.

Most of the time, the bullpen usage followed sensible/sabermetric principles given all that other info

My big problem is that way too often managers will go with a fairly strict 7th, 8th and 9th inning guy, where everyone gets to start their own inning. Or will see L-L-R hitters coming up, and start the inning with their lefty reliever, leave him in for the third batter, and bring out the RH setup guy to start the next inning. I know, lacking specific examples and whatnot, but I think there is something to be gained by breaking away from traditional roles without doing anything more tramautic than asking your best reliever to get 4 outs instead of 3 more often.

I think the best solution would be to start with the "biggest" decision a manager can make - substitution - and at least offer a manager a "menu" of options about possible substitutions and their (super simulated) consequences.

You can easily imagine a simulation engine (something akin to villageidiom's #10) saying

1) Here are all possible substitutions available to you and your opponent right now.
2) Taking into account park factors, rest days, injuries, positioning, and of course all the sabermetric data your heart could desire, here are the expected outcomes of all these substitutions.
3) Manager gets the "substitution matrix" and determines if now's the time to pinch hit / bring in a new reliever / etc.

You could even have expected outcomes beyond the game to account for some of the long-term personnel management issues. The point is, if it has an impact, measure it, account for it, and make decisions accordingly.

My big problem is that way too often managers will go with a fairly strict 7th, 8th and 9th inning guy, where everyone gets to start their own inning.

All other things being equal, I think relievers are more successful when they enter a game at the start of an inning than in the middle. I also think teams are more successful when they align relievers with leverage. These will run into conflict sometimes, and where the dividing line is between the two isn't clear.

I agree that if a high-leverage situation arises with 2 outs in the 8th, ideally you want your relief ace pitching. But for your relief ace to be pitching with 2 outs in the 8th, it's possible (and somewhat likely) you would have had to start warming him up with no outs in the 8th, in a lower leverage situation, when it's unclear a high-leverage situation is on the horizon. Unless you have your best reliever always warming up at the start of the 8th, just in case a high-leverage situation arises, there will be some high-lev situations in the 8th for which he will be unprepared. And if you always have him warming in the 8th, that will also take its toll on his performance.

I don't disagree with your premise, that managers should break away from tradition where warranted. I'm just saying game-state isn't enough context to say when a manager should break away, yet a lot of would-be sabermetricians judge it that way.

The right decision is only marginally better than the wrong one

Yes, but it's like gambling. The house only has a marginally better chance of winning than the players have, but over the long run, the house will always clean out the players.

If a manager always makes the right decision, over the course of a season it will pay off.

To paraphrase Mr. Spock: models and simulations consider only those conditions that they were designed to consider. A thoughtful manager might still have a great many more factors to take into account than the model, thus leading to a different outcome.

The fact that many of the "modern stats" ignore game theory and fail to understand sometimes you should do a suboptimal (in the percentage sense) tactic leads me to believe that the idea of a stats guy in the dugout spitting out "never bunt" advice would be unhelpful at best.

That would be a very bad analyst. I would hope that if MLB teams hired 30 bench analysts, they could find better hiring choices than that. Start with guys who have read and can grok MGL's chapter on bunting in The Book.

I think that all of the stuff that an analyst can do to help a manager is best done outside of the dugout before the game. During the game, let the manager make the decisions and react to situations as they occur. You don't need an analyst in the dugout to tell a manager that a bunt is a nonoptimal tactical decision; the manager should know the percentages before he ever gets to the bench.

-- MWE

15/village -Agree with pretty much everything you said. But I still see it in situations you can be somewhat prepared for, say a big bopper is coming up 3rd in the 8th, or you can get your ace reliever up as soon as a guy gets on. And if a high leverage situation arises that you aren't prepared for, well, stall as long as you can. And if I were a manager, I think I'd have my closer standing/stretching/throwing light tosses as soon as the 8th starts. Nothing strenuous, but cut as much time as possible between sitting with his legs up holding a coffee and ready to pitch.

And 19 - absolutely agree. I had a similar conversation elsewhere about when NFL teams should go for two. Someone thought good game management was when coach should know exactly when to go for two according to the chart. I feel good game management is having some intern ready with the chart, and the coach being able to have an innate ability to combine that number with time left, performance of his offense or defense, and the flow of the game.

i think in baseball there is honestly not enough strategy to justify said analyst. the player tries to hit the ball. when the pitcher cannot throw any more, you replace him. in some situations, you want to replace a hitter that is a terrible fielder at the end of a game with a weaker fielder that is not as good at hitting. most of the decisions are too obvious, and the vast majority of analysis just needs further data over some sort of situation, which any manager can decipher. good coaches are made in the offseason and rely on soft science. they make sure their players are working hard, they build the best possible roster with the possible help of sabermetrics, and they try to lead the team emotionally and manage public relations.

i enjoy trying to rate historical players across different periods in baseball history using sabermetrics, but if i were a coach i would never consult a statistician for a situational call. i would discuss roster moves in the offseason with him, use it in contract negotiations, and possibly ask him to research various split stats for me. i would review them before the game and keep a couple of the key values in my head. but i would never ask him to tell me what the numbers say on whether i should do such and such in this situation. data can be gathered on the various relievers, etc. but any probabilistic analysis? i don't see the point in using it for the paltry amount of strategy there is in baseball. maybe i am making a semantic point and people on here are calling data gathering "analysis."

now football has a heck of a lot more in-game strategy, but the statistical measures for football are not nearly as reliable as those for baseball. baseball is pretty astoundingly good for statistics being an actual reflection of skill: it's iterative, uniform, often given in terms of percentages, and has a large sample size. in football, you can have the quarterback with the most yards in the league be worst, and vice versa, and the sample size is small. completion percentage and yards per completion are often misleading because teams pass more and do so more effectively when behind as the winning team plays a prevent defense. in fact, the defensive back with the most tackles often IS the worst and vice versa, etc. i guess i'm responding to the earlier post by saying that you couldn't even have this analyst in a high-strategy game like football, because the statistics are not nearly as reliable apropos a chance of winning. and there is a correlation because the good statistics are going to be iterative and uniform, i.e., antithetical to complex strategy.

i am into poker, where doing mathematical analysis for individual "in-game" strategies is way more justified than in any sport. even in a purely logical situation like that, while you consider the math, you eventually rely on your perception of the strategy of your opponent. your playing environment, i.e. your opponents and how you react to their strategy, completely overwhelms any probabilistic analysis you may develop, no matter how refined. because your opponent can flout your system, or, in the case of multiple opponents, the system can simply become chaotic and what you do is literally irrelevant (such as when you have 4 or more opponents in a poker hand). if you have aces, you should fold before the flop if you don't think you can get less than 4 people to the flop. In early position, A-7 you should fold, in late position, you should raise. even in the same position with the same action before you your strategy often changes. if you're playing elky grospellier and he raises over your 3-3 in the small, you should shove on him because he is a notoriously aggressive bluffer. if it's liv boeree, you should fold because she is known as an extremely conservative player. your opponent defines your strategy much more than the numbers.

you should know the numbers, yes, but often times just to discount them.

#7 Despite the number of people who say it, "never bunt" is flat wrong -- even before considering game theory aspects.

It is true that base for an out is rarely a good trade with a position player at the plate, but it's plain wrong to do the analysis on that basis. It ignores a moderately common outcome -- runner reaches (either by beating out the bunt, reaching on an error or reaching when the defense tries for the lead runner and fails to get him).

And yes, it also ignores other bad outcomes. pop up into a DP, defense gets the lead runner or batter fails to get the bunt down and has to hit away with two strikes.

It is accurate to say that you should never sac with somebody who isn't a good bunter. It's also accurate to say that bunting will always be a reasonable play for a good bunter with reasonable speed (and thus a reasonable chance of reaching) -- particularly if the opposing 3rd-baseman is not an excellent defender.

The fact that many of the "modern stats" ignore game theory and fail to understand sometimes you should do a suboptimal (in the percentage sense) tactic leads me to believe that the idea of a stats guy in the dugout spitting out "never bunt" advice would be unhelpful at best.

Um excuse? Have you ever read anything MGL's ever written? Like seriously are you ####### kidding me?

http://www.fangraphs.com/blogs/index.php/were-the-yankee-sac-bunts-in-the-8th-inning-correct/

#21 Chess is more interesting. There is often a provably correct option for a position. And yet some very strong players often opted for something other than the optimum moves.

Reti wrote that Lasker played objectively bad moves to take his opponent out of his comfort zone. Larsen once wrote (playing for a win against a lesser grandmaster -- Pomar) I don't care if it's sound, just that it was unfamiliar. And Tal would often accept the inferior side of a very complex position -- expecting to out-play his opponent in the complications.

IOW some very successful players opted to play the man, not the board, in a perfect information game.

You don't need an analyst in the dugout to tell a manager that a bunt is a nonoptimal tactical decision; the manager should know the percentages before he ever gets to the bench.

But whether or not a bunt is correct changes based on the situation.

Note no one's saying Dave Cameron (who is not a sabermetrician) should be hired to help make in game decisions. But MGL, Tango, David Gassko? Hell yeah. Those guys obviously know waaaay more about proper strategy than most managers and there is surely a way to utilize their expertise without damaging the manager/player relationship.

if you have aces, you should fold before the flop if you don't think you can get less than 4 people to the flop.

What do you mean by this?

if you have aces, you should fold before the flop if you don't think you can get less than 4 people to the flop.

Can I play poker with you?

I hear tell they have these things called computers these days.

I see no point or need for an analyst in the dugout (and especially not MGL! :-) but an analyst's model providing input for the manager's decision-making. Sure.

Also if you put an analyst in the dugout then every manager is gonna turn into Philip Seymour Hoffman as Art Howe. What a horrible world!

One of my favorite scenes in the movie is the "why did you use Venafro, I want Bradford there" "it was a lefty batter Billy" "I don't care about lefty/righty." Chad Bradford with the lifetime 856 OPS lefty split and more BB than K. You'd be hard-pressed to find a worse pitcher for a high-leverage situation against a LHB. Is there a single point in that movie when Howe was actually wrong? I'll probably never decide if that was an inside joke or not.

if you have aces, you should fold before the flop if you don't think you can get less than 4 people to the flop.

I think he means the more players stay in the more likely the flop will put your aces at a disadvantage to another player. It does you no good if you can beat almost everyone in that hand.

But, still... aces.

I think he means the more players stay in the more likely the flop will put your aces at a disadvantage to another player.

Sure, but you shouldn't fold them pre-flop except in super-rare tournament end-game situations, I believe.

But MGL, Tango, David Gassko? Hell yeah. Those guys obviously know waaaay more about proper strategy than most managers and there is surely a way to utilize their expertise without damaging the manager/player relationship.


Someone like Tango knows next to nothing about humans and what makes them tick and knows next to nothing about what makes talented jocks tick. He'd have been the butt of ridicule as a coach of my 7th grade basketball team and would last less than 30 seconds addressing a major league baseball team before being laughed out of the room.

There's nothing about him that indicates any ability to lead, command, or motivate and a whole bunch of things indicating that he has no capacity whatever to do these things -- the main one perhaps being that he sits by himself chewing through data as a means of proving himself "right" about things. I wouldn't hire him to manage a lemonade stand.

Sure, but you shouldn't fold them pre-flop except in super-rare tournament end-game situations, I believe.

Right. You can concoct scenarios where you should fold AA, like if you have the chip lead with 11 players left when the top ten qualify for the World Series or whatever, but you should essentially never fold them in real-life. Its essentially impossible to set up a situation where you should fold them in a cash game. For instance, you have positive equity in playing AA against two other players, no matter what the second guy has, even if you somehow know the third guy also has AA!

Someone like Tango knows next to nothing about humans and what makes them tick and knows next to nothing about what makes talented jocks tick. He'd have been the butt of ridicule as a coach of my 7th grade basketball team and would last less than 30 seconds addressing a major league baseball team before being laughed out of the room.

There's nothing about him that indicates any ability to lead, command, or motivate and a whole bunch of things indicating that he has no capacity whatever to do these things -- the main one perhaps being that he sits by himself chewing through data as a means of proving himself "right" about things. I wouldn't hire him to manage a lemonade stand.

I think you might be my least favorite poster on any website ever.

A) How the #### do you know anything personal about Tango?

B) Who the #### said anything about Tango knowing more about managing players or about hiring him to manage a team? I said he knows more about strategy and that expertise could be used to help the managers make better strategic decisions.

C) You are a ####### moron.

closers are left sitting on their asses during high-leverage moments

This one's been done to death. The Red Sox tried to abandon the "closer" role 8 or 9 years ago in favor of a more saber-approved pen structure, the pitchers revolted, and they changed back in the span of about a month.

It's the players' game. If they don't buy in, things aren't going to happen. And they aren't going to buy in to Geeks Bearing Spreadsheets.

Note no one's saying Dave Cameron (who is not a sabermetrician) should be hired to help make in game decisions. But MGL, Tango, David Gassko? Hell yeah. Those guys obviously know waaaay more about proper strategy than most managers and there is surely a way to utilize their expertise without damaging the manager/player relationship.

I think every team has someone in such a role at this point. The catch is that humans are imperfect. Some of the stat guys are obviously better than others and of course some managers are more willing to accept the input than others.

The fact is we've seen many of these guys employed. Hell, there are people posting here who are or have been employed by MLB teams.

The fact is we've seen many of these guys employed. Hell, there are people posting here who are or have been employed by MLB teams.

This is true, but as far as I know these guys are being employed at the front office level and have zero input on managerial decisions. I will admit I'm completely ignorant about this, but given the abundance of intentional walks still going on, I'm guessing managers remain completely on their own.

This one's been done to death. The Red Sox tried to abandon the "closer" role 8 or 9 years ago in favor of a more saber-approved pen structure, the pitchers revolted, and they changed back in the span of about a month.

That's not really what happened.

But there is room for having a 9th inning designated closer and also using that closer in the 9th of a tie game on the road - which many managers refuse to do.

I wouldn't hire him to manage a lemonade stand.

Imagine his dismay. He'll have to keep doing this for a living rather than running a lemonade stand owned by an egomaniac.

I don't know Tango in real life, but I've interacted with him online over the years. I've rarely dealt with anyone more courteous, which says something, given that I've frequently seen him dealing with people who are raising objections or arguments that he must have considered and rejected. I dunno what he's like as a leader of men but people who are willing to hear others out and take the time to explain why they've considered their point of view and rejected it or change their way of thinking to reflect what the other person thinks seem to me like the sort of people that can do quite well as leaders.

Imagine his dismay. He'll have to keep doing this for a living rather than running a lemonade stand owned by an egomaniac.

Heh. Well done.

Indeed – Tango introduced me to Primer by taking an interest in things I'd written, and the few times I've talked with him online he has been one of the nicest (virtual) guys I've ever (not) met. I see no reason to believe that sabermetricians in general would have any lower people skills on average than baseball managers or accountants or English professors.

he sits by himself chewing through data as a means of proving himself "right" about things.

Maybe he does it because he loves baseball.

Tommy is good people, of this he convinced me a long, long time ago. Are his skills well suited to managing a baseball team as the job currently entails? Probably not. If only for no other reason than they are not what is usually expected of guys who are normally MLB managers. Through no fault of his own there's a good chance that would sink him.

But then there's a lot of different ways a dugout can be organized and I think there's at least some value to trying to work out a functional analytical framework for managerial decisions and tactics. So far in the NBA, the experiment hasn't been a failure, though maybe you can't call it wildly successful just yet.

For what one of us geeks costs, the number of wins you have to add a year to justify it is ridiculously low.

I think that all of the stuff that an analyst can do to help a manager is best done outside of the dugout before the game. During the game, let the manager make the decisions and react to situations as they occur. You don't need an analyst in the dugout to tell a manager that a bunt is a nonoptimal tactical decision; the manager should know the percentages before he ever gets to the bench.

I agree with this. Jeff Luhnow had an interesting idea with the Astros. He said he wants to put Bo Porter's office in the top floor of the Front Office, so that he can have interaction with the analysts, scouting executives, GM, etc., with the concept that they all work as a team.

if you have aces, you should fold before the flop if you don't think you can get less than 4 people to the flop.

Obviously you should see a flop for cheap. But I think this adage holds if you CANNOT (psychologically) get away from the hand post flop. Especially in a No-Limit game. Also it applies stronger if the other 3 players are playing tight and likely to be holding an ace. But really, pairs don't play well multi-way unless you hit a set.

The fact that many of the "modern stats" ignore game theory and fail to understand sometimes you should do a suboptimal (in the percentage sense) tactic leads me to believe that the idea of a stats guy in the dugout spitting out "never bunt" advice would be unhelpful at best.

True, but hasn't there already been a manager in the game, in Davey Johnson, who would happily incorporate all the modern stats into his in-game decision making? I'd be surprised, too, if Earl Weaver had contemporary stats at his disposal and wouldn't use them to best advantage, overruling them as it made sense to him.

Then, you also have the newer manager with a head for numbers, who is hired based in part on his willingness to absorb and use sabermetric principles. I think that kind of manager could have and use a stats-oriented bench coach without running into second-guessing. The bench coach would be seen, I think, as an adjunct; not as someone with a polar, opposed point of view, but a supplement to the manager.

This one's been done to death. The Red Sox tried to abandon the "closer" role 8 or 9 years ago in favor of a more saber-approved pen structure, the pitchers revolted, and they changed back in the span of about a month.

The question of making moves based on tactics vs making moves based on personnel is interesting and not really resolvable. On the one hand, you'd always like to have your best pitcher on the mound at the crucial moment. On the other hand, players love routine and being able to mentally prepare for their role. So you might have your best pitcher, but he's out of his comfort zone a little.

Personally, I'd prefer to see more of a return to the fireman role. But if you're only getting 80 innings out of Mo Rivera this year, wouldn't you prefer to see him at his best?

Second issue: doesn't a fireman role make more sense in a lower scoring environment? If the team's 3-4-5 hitters are the only power threat, there's a clear benefit to having your closer face the heart of the order. But if the #7 hitter can put one out of the park, why burn the closer early?