Do GM's have too much power now?

[T.G.]

How do you guys feel about the developing trend of GM's really making "all" the decisions now?  The influence and direction of today's manager seems to be disappearing to a certain degree.  Is that a good thing or a bad thing?

Obviously, the best scenario is a truly collaborative effort.  That being said, when we look at the Red Sox and the Dodgers - it seems (because we don't actually know the details) that Roberts takes orders and Cora still gives them.  Both are influenced by the data - but one uses it as a guide where the other appears to use the data as a directive.

 

[Dtwncbad]

Self-policing.

A very successful team is never going to be accused of having a GM that has too much power.

A team that underachieves might, and then they get replaced.

I don"t think this is something that can (or should) be answered in generality. 

[T.G.]

Good points, but I still think it could be an interesting conversation.  I wonder how this plays out for the Angels. 

[Angel Oracle]

Has MLB become too Stratomatic-ized, with the 3 outcomes push?

[ettin]

I would go a step further actually and say that computer simulations could end up running the calls to some degree. All of that data is in the system and it would not be entirely shocking if teams are running real-time simulations of batter-hitter match-ups in-game.

If you had a powerful enough computer you could probably run a short series of Monte Carlo simulations based on defensive alignments and batter-hitter match-ups to spit out a probability number to determine if you should let the hitter hit or bring in a pinch-hitter and/or a new pitcher.

Teams want to win and that is why Eppler specifically spoke about hiring a manager with a probability-based mindset. Data drives the decisions now so that teams maximize their probability of having a good outcome on the field of play.

I would be surprised if GM's are driving all decisions during  a game, it is probably computers that pull all of the emotion out of the decisions.

[T.G.]

20 minutes ago, ettin said:

I would be surprised if GM's are driving all decisions during  a game, it is probably computers that pull all of the emotion out of the decisions.

That's a good point.  Is this a good thing?  I'd really like to know more about how this really works.  What's the actual process?  How is it communicated? I think it's fascinating.

[ettin]

 

Just now, True Grich said:

That's a good point.  Is this a good thing?  I'd really like to know more about how this really works.  What's the actual process?  How is it communicated? I think it's fascinating.

Well the inputs to the computer simulations will only be as good as the data you have available.

For example it generally takes years of defensive data on a particular player to understand how good of a defender they are in terms of assigning percentages of how often they field the ball successfully (or unsuccessfully). Hitters on the other hand, for example, usually firm up their strikeout numbers in only 50 plate appearances (meaning that you have a decently good idea of their K% in a small sample size).

First you have to identify what data means something (good correlation) and then how that data interacts with other data in the game. It is not an easy process but if you have all of this data assembled and put together in the right way it can be a powerful tool. So if you know generally that Kershaw has averaged over the last three years a 32% K% against left-handed hitters and he is pitching to Cody Bellinger who has averaged over the last two seasons a 25% K% against LHPs, you can begin to develop a BASIC model that you can use to determine how often that match-up will turn out in either the pitcher's or hitter's favor.

I use the word BASIC because there is a lot more that should go into that computation including base runner states (how many other hitters are already on-base and at which bases), day/night conditions, atmospheric conditions (high winds), ballpark configurations and factors, pitcher pitch types and how well the hitter handles those pitch types, previous match-up history (if the sample size is good enough which it usually is not), more recent pitcher/batter history (meaning for instance Kershaw is getting older so his "stuff" may be losing velocity, etc.).

This is why having really good, conditioned data (and why teams are investing so much in it) that you have confidence in is so important. That data can tell you a lot and if you put it into complex simulators it can give you relatively good probability outcomes which you can then apply to in-game use in some form, shape, or capacity.

I don't pretend to know all of the factors here Grich. I do know that in simulating rocket launches we use Monte Carlo runs (thousands of simulations) to understand how we can break the rocket. Ideally most of those runs should be along a "nominal' path but when you purposely throw wrenches into the equations you begin to see a handful of "bad flights" that violate the boundaries of your planned flights. In baseball the simulations are strictly looking to maximize good match-ups (a hitter having an advantage over the other team or a pitcher having an advantage against a hitter) and minimize bad ones. Teams want that greater than 50% probability all the time. This is probably a good reason why the Dodgers have focused so much on building complete team depth on their roster because any injury can potentially decrease your odds at any one position and having good replacements mitigates that a LOT.

[Angel Oracle]

As much as computerization is coming into MLB, there is still the clubhouse element.   

It seems that managers to be successful will always need to be solid enough at managing clubhouses as well?   

[T.G.]

A lot of the focus this post season has been about Alex Cora and how he drives the process.  I like what I've read and heard.  He's the guy asking for the data and pushing his front office to give him data he can use. He still manages with his eyes to a certain extent too.  I hope at some point - someone sits down with him to really learn about the process he goes through.  It's interesting stuff.

[Blarg]

Using data driven outcomes, in 1988 Gibson remains in the clubhouse in street clothes. This game is becoming more boring every year. 

[ettin]

3 minutes ago, Blarg said:

Using data driven outcomes, in 1988 Gibson remains in the clubhouse in street clothes. This game is becoming more boring every year. 

Whenever money enters into the equation, it tends to drive the business model sadly.

[Angel Oracle]

38 minutes ago, True Grich said:

A lot of the focus this post season has been about Joey Cora and how he drives the process.  I like what I've read and heard.  He's the guy asking for the data and pushing his front office to give him data he can use. He still manages with his eyes to a certain extent too.  I hope at some point - someone sits down with him to really learn about the process he goes through.  It's interesting stuff.

Agreed, about Alex

[Angel Oracle]

7 minutes ago, Blarg said:

Using data driven outcomes, in 1988 Gibson remains in the clubhouse in street clothes. This game is becoming more boring every year. 

And things like this are exactly why actual managing of people still needs to be an important element.  

We saw what happens with a new manager, who's only skill set seemed to be numbers crunching (Kapler).   Phillies imploded impressively.

[stormngt]

4 hours ago, True Grich said:

How do you guys feel about the developing trend of GM's really making "all" the decisions now?  The influence and direction of today's manager seems to be disappearing to a certain degree.  Is that a good thing or a bad thing?

Obviously, the best scenario is a truly collaborative effort.  That being said, when we look at the Red Sox and the Dodgers - it seems (because we don't actually know the details) that Roberts takes orders and Cora still gives them.  Both are influenced by the data - but one uses it as a guide where the other appears to use the data as a directive.

 

Data should be used as a guide not a directive. Remember data only explains things in the past.  It dies not explain the future.

[T.G.]

4 minutes ago, stormngt said:

Data should be used as a guide not a directive. Remember data only explains things in the past.  It dies not explain the future.

I agree with you... but I think baseball is trying to use it to predict outcomes... or as Eppler would say... " “We’re looking for somebody who can think with a probability-based mindset."

[stormngt]

4 minutes ago, True Grich said:

I agree with you... but I think baseball is trying to use it to predict outcomes... or as Eppler would say... " “We’re looking for somebody who can think with a probability-based mindset."

I agree that stats are useful for prediction.  However I am fearful decisions are becoming too dependent of sabre metrics 

[BackUpTheTruck]

There is something to be said about Moneyball (Statistics), and there is something to be said about old-school and new-school baseball wisdom that a former major league player would know well about. You cannot rely on just one and neglect the other.

[ten ocho recon scout]

2 hours ago, Angel Oracle said:

As much as computerization is coming into MLB, there is still the clubhouse element.   

It seems that managers to be successful will always need to be solid enough at managing clubhouses as well?   

Youd think. And hope. But with the way athletes are kind of babied these days (and the money invested in them) you wonder how much sway a manager has over a clubhouse malcontent. 

[ettin]

1 hour ago, True Grich said:

I agree with you... but I think baseball is trying to use it to predict outcomes... or as Eppler would say... " “We’re looking for somebody who can think with a probability-based mindset."

I don't think they are trying to predict I think they are trying to create potential outcomes that are more likely to fall their way. You put Kershaw as your Game 1 starter because he gives you the best probability of winning a game but the actual outcome can vary obviously. Clayton likely gives the Dodgers their best chance to win.

If you ran a bunch of simulations with Kershaw as a starter against most teams in baseball it would probably spit out something like a greater than 50% probability (maybe something like 51%-54% give or take).

[T.G.]

29 minutes ago, ettin said:

I don't think they are trying to predict I think they are trying to create potential outcomes that are more likely to fall their way. You put Kershaw as your Game 1 starter because he gives you the best probability of winning a game but the actual outcome can vary obviously. Clayton likely gives the Dodgers their best chance to win.

If you ran a bunch of simulations with Kershaw as a starter against most teams in baseball it would probably spit out something like a greater than 50% probability (maybe something like 51%-54% give or take).

I think we're saying the same thing - you just articulated it better than I did. 

[Inside Pitch]

4 hours ago, Blarg said:

Using data driven outcomes, in 1988 Gibson remains in the clubhouse in street clothes. This game is becoming more boring every year. 

https://www.fangraphs.com/blogs/the-recipe-for-the-red-sox-secret-sauce/

The Red Sox walk like an average team, but they do much better than league average with strikeouts. It’s a contact-oriented team. The only AL teams with a lower strikeout rate than Boston were Cleveland and Houston. The Red Sox have a higher batting average — likely the result of making more contact — but they also hit for more power, as seen by the high ISO. Overall, the Red Sox were one of the five best offensive teams in baseball. As for the strikeouts, the club has a lot of players with good track records in that regard.>>

So....   The three AL division winners all have low K rates... or you know, make actual contact...    But three outcomes.....

[Stradling]

I heard ARod on George Sedano show yesterday.  He said that his understanding is that Roberts almost entirely goes by the script of the front office and that Cora, is allowed much more freedom to manage what he sees.   I think he said because Dombrowski is an old school type of guy.  An example he made is if the pitcher was pitching lights out that Roberts would still pull him when the script said to versus Cora would keep throwing him longer.   

[hangin n wangin]

1 hour ago, Stradling said:

I heard ARod on George Sedano show yesterday.  He said that his understanding is that Roberts almost entirely goes by the script of the front office and that Cora, is allowed much more freedom to manage what he sees.   I think he said because Dombrowski is an old school type of guy.  An example he made is if the pitcher was pitching lights out that Roberts would still pull him when the script said to versus Cora would keep throwing him longer.   

If Roberts is really going by a script and every move is followed by the order of the analytics guys, he shouldn’t be fired, one would think....

[Docwaukee]

the assumption is that the front office controls the metrics and that the manager is being neutered from their natural inclination of 'going by feel'.  I think that is a misconception.  

I think it's more of a collective where certain scenarios are anticipated and pre-planned.  

Prior to all of this data being available, why were certain managers successful over others?   It's most likely that their ability to observe and assimilate from those observations was far superior to those that were less successful.  Of course it's all relative as the primary factor in determining the success of a manager is based on the performance of the players.  

There's actually be more of a responsibility shift to the front office in terms of performance whereas the manager's primary responsibility is now to maintain the clubhouse and manage personalities.  

To me, that feels like a more appropriate allocation of responsibility.   The GM picks the players.  Why aren't they the most responsible for said player's performance? If there is information available to optimize the potential of a player then that info should be disseminated but if that info doesn't result in that optimized performance, then it becomes about how it was executed.  

The manager and front office should be on the same page with regard to execution. Yes, there is room to disagree as nothing is ever 100%.  The GM should choose the manager just as they choose the players.  It's extremely rare these days that a gm doesn't get to choose their own manager.  We're just not used to it because our manager survived four GMs.  That dynamic is an outlier and likely at odds with the vision of the person tasked with the overall success of the team.  It's why Dipoto quit.  He wasn't allowed to fully execute his plan.  

Here's something to consider.  If it wasn't for Mike Scioscia, Jerry Dipoto might still be our GM.      

[Ace-Of-Diamonds]

21 hours ago, ettin said:

If you had a powerful enough computer you could probably run a short series of Monte Carlo simulations based on defensive alignments and batter-hitter match-ups to spit out a probability number to determine if you should let the hitter hit or bring in a pinch-hitter and/or a new pitcher.

Teams want to win and that is why Eppler specifically spoke about hiring a manager with a probability-based mindset. Data drives the decisions now so that teams maximize their probability of having a good outcome on the field of play.

Maybe to some extent but you won't need a computer to tell you to let Ohtani hit over Pujols or don't PH for Trout. Some decisions are too obvious, such as walking Trout with 1B open and game on the line with RISP. Which is why we need a great hitter right behind Trout in the line up.

I wish they would make a shifts illegal (to some extent), and get baseball back to the way it was when I was growing up.