Sooner or later it was bound to happen.

What was bound to happen?  As far as I can tell the only thing that has happened is that two electrical engineers wrote a paper for a research paper competition in which they claimed to be able to create a computer model that will out-perform human coaches in deciding when to yank a starting pitcher.

Maybe I  missed the part where an MLB club elected to install the RoboCoach x2012 in their dugout?  Or to use this paper in any way at all?

That said, I do appreciate your bringing this to attention of the board.  There are a bunch of interesting topics addressed.  I'm looking forward to showing them to my 2 HS students and hoping it may spark some interest.  I think my youngest will be fascinated by the pick-and-roll research, and might even learn some playing strategy from  the rebounding study.   I suspect my eldest will be most interested in the study   by the two Stanford guys about what makes baseball umpires call balls and strikes. If this research can be validated I think MLB should use it in their training.

JCG,

Is it necessary that some ML club use any particular scientific study to validate it? No. Its science, not guesses, opinion, or smoke and mirrors.

Ah, I'm not a scientist so I didn't realize that one paper was all it took to establish scientific fact.  So thanks for that.  Well, gotta go.  Long drive ahead. But you won't believe the mileage I'm getting from my Honda Cold Fusion (TM) hybrid.

It is interesting, but I don't know how to operationalize it.  

Since it takes 80% of the MLB season to build the pitcher-specific profiles, it's unlikely a manager in a pennant race will be willing to switch to this tool on August 1st.

However, it is very likely any manager presented with this tool would believe he has additional relevant information (e.g., how good the pitcher's stuff is, how he is sustaining velocity, how confident he is in the pitcher's competitiveness, how his mechanics look, how the pitcher says he feels, where the pitcher is due to bat in the next inning, heat, humidity, whether his team is ahead or behind and by how much, how rested the bullpen is, etc.) that would overrule the predictive value of the model.  

Hard to see anyone using it as more than one more piece of info to consider along with all the other factors a major league manager must continuously and subjectively weigh.  

Real hard to see any other levels of ball having the data to employ it because of the shorter season, absence of data on opposing hitters, less consistency among pitchers.

Another impediment to deploying it is it doesn't take into account what happens after the decision to pull the starter.  It's one thing to have a model that predicts more accurately whether the starter will give up a run in the next inning; it's quite another thing to predict the outcome of all the remaining innings in the game based on going to the bullpen sooner.  It's like when your broker suggests it's time to sell a stock.  Taking the advice forces a subsequent decision, whose consequences are unknowable.

Wonder why they didn't just come up with a specific % drop in velocity to determine when to take the pitcher out?  

Seems a little confusing to have to take many different factors into consideration. 

I only read about four sentences and came across 12 words I didn't understand so I just stopped reading.  To quote Josh Stern - "I have a keen sense of the oblivious".

Like many of you, I've watched a lot of baseball games at all levels.  As the father of 3 pitchers, you kind of get a sense for when a pitcher needs to taken out or when he is cruising.  I can see it in my sons as well as others.   There is a finite number of pitches to be thrown in a day.  Some days they have it and other days they don't.  A statistical model make sense to me.

On the flip side.....What absolutely intrigues me is when a pitcher is pulled because the position players fail to make a routine play.  For example:  the ground ball doubleplay ball muffed by the 2nd baseman as the result of the RHPs change-up to a LH hitter.   The result of the failed execution of the play is to pull the pitcher?  Really?  It drives me bat-sh*t crazy!  You can do all the modeling and analytics you want but I'd like to see a documented study on why position players aren't pulled.  

This shows part of the problem with that study. 

Compare this...from the study:

5.1 Game 2 of 2013 ALCS

Game 2 of 2013 ALCS between Detroit Tigers and Boston Red Sox was a pivotal game in the series. A highly debated moment occurred when the Detroit starter, Max Scherzer, was pulled after the 7th inning.  Our model would have indicated that Scherzer should not have been removed. Detroit relievers (Detroit used 4 relievers in the bottom of the 8th inning) ended up giving 4 runs.

With this:

“I just knew I was at my limit,” said Scherzer, who took a no-hitter into the sixth inning before giving up a one-out single to Shane Victorino, followed by an RBI double to Dustin Pedroia."

“You gotta be smart,” said Scherzer, who allowed two hits, two walks and one earned run over seven innings with 13 strikeouts. “It’s still early in the series. You gotta gauge your health. My health is important to the team. I had reached my pitch count limit.”    "I told them I was done."

"Scherzer said there was no way he was going to be able to pitch the eighth." 

http://www.freep.com/article/2...nings-told-them-done-

There are many more things a manager has to take into account rather than just focusing on whether or not a run is going to be scored in the next inning, including the pitching strategy for other games in the season or series (whether that pitcher can come back and pitch another game), as well as the long term and short term health of the pitcher.   What if Scherzer, after he said he was done, was convinced to pitch another inning due to some statistical model, and then tore his labrum/ended his career?   

HAHA....I stopped when I hit the hieroglyphics on page 3

Stats...sending you a dialog

There are so many human and environmental factors in baseball. I believe greatly in statistical data, but wonder if this study is too broad to fit all situations. Rain, humidity, temperature, altitude, ball park dimensions and umpires all play roles worth considering in baseball because they muddle the science.

Specific to pitching and offensive strategy, wouldn't the technology already being used by some teams and PG be more reliable in these decisions. I can't find the name of it, but it's the device that measures the rotation and trajectory of pitches. Originally used to track missiles I believe. There was a thread here before but I can't come up with the right key words to get a successful search.

I would think those measurements would be more consistent for tracking pitching as well as which batters hit which type pitchers better. There's probably already history and maybe teams wouldn't need 80% of a season to create models.

here's a couple of links to a new study currently underway.  The results will not be ready for a couple of years but will be an interesting read when its done.  I just hope they use plain English, I am with coach2709, to big of words for us common folk.  Maybe we should have sniffed some paint fumes 1st Coach?

https://www.mercyhurst.edu/mu-...february/pitch-count

http://www.erietvnews.com/stor...tioned-at-mercyhurst

Well not going to lie mmm1531 but after I started reading it I went and put my forehead on a bat knob and spun around a few times.  Got really dizzy but didn't help to understand any better on the second read.