Missed by who? Old-school baseball people have been talking up C defense since the dawn of time. Pitchers have talked up Cs that they "like to work with" since the rules allowed them not to throw what the batter wanted. Pitch framing has been in the lexicon for at least my entire baseball life ("see the way he pulls the ball back over the plate").

What (maybe) wasn't possible until recently was accurately measuring it. It was often dismissed out of hand by analysts (and me). Catcher ERA was assigned to the random error bin. What also wasn't captured until recently was pitch-by-pitch analysis and the estimated value of a single strike/ball.

When this stuff first started, Tango and I had a short discussion, probably here, maybe offline. Anyway, given his estimates, 2 stolen strikes per game over a season was about 2 wins ... or 2 lost strikes was about -2 wins. It's not quite that simple since the stolen/lost strikes are generally borderline pitches that are called a strike, say, 50% of the time but I don't find it that hard to believe that Molina (or some C) steals 2-3 strikes a game.

That to me was the real revelation -- that the value of a couple of extra strikes per game added up to a couple of wins. What could possibly be more trivial? But it's consistent with the way teams have regularly employed guys like Ausmus.

The comments of the article ask a good question. If we start giving WAR to catchers for framing... do we take it away from pitchers?

To be annoying... Was Greg Maddux really the standout talent... or was Eddie Perez the greatest pitch framer of all time....

It would certainly explain Bill Bergen. And I definitely think this was valued before, but that "value" was/is regressed a lot because of the error bars around how much pitch framing is worth. I think in the next few years we will either 1. Realize that 5 wins is a big over estimate, or 2. Good pitch framing is going to become a bigger part of the assessment of value w when handing out contracts or promoting catchers. But to say people were missing the value of this is not quite accurate.

Catcher ERA was assigned to the random error bin.

If pitch framing is so valuable, why is doesn't it show up in CERA? There's too much other noise?

If pitch framing is so valuable, why is doesn't it show up in CERA? There's too much other noise?

I would say yes. Arguably not large enough sample size. You need the same pitchers using different catchers and there really isn't that much info to go with. You can't just assign catcher era based upon what the pitcher did, you have to assign it based upon the pitchers performance with that catcher, vs his performance with the pool of other catchers.

Only thing I personally have used Catchers era was to discredit the notion that Mike Piazza was a poor defensive catcher. It doesn't mean he was good or anything, but he never caught a(full season) team that didn't finish in the top five in era(and usually third or better) it's not actual evidence he's good, but it is definitely a knock against the argument he was poor.

When this stuff first started, Tango and I had a short discussion, probably here, maybe offline. Anyway, given his estimates, 2 stolen strikes per game over a season was about 2 wins ... or 2 lost strikes was about -2 wins. It's not quite that simple since the stolen/lost strikes are generally borderline pitches that are called a strike, say, 50% of the time but I don't find it that hard to believe that Molina (or some C) steals 2-3 strikes a game.

I don't find it hard to believe it, but unless it's the call third strike, you still have to continue the pitch sequence. I understand most of these studies look at how the average player does and awards credit based upon the difference. I.E. it's a 1-1 count, and a ball is called a strike, they look at the average performance difference throughout the league between a 1-2 count vs 2-1 count and award credit based upon that. It's nice and theoretical and all, but it doesn't really save the "run" unless the batter actually makes the out. Just changing the odds isn't the same thing as actually changing the outcome. If these studies focused strictly on ball 4 and strike 3 calls, then that would be dealing with actual value.

I think ultimately they should look at the final result of the at bat and apply their numbers as a modifier/value difference on the result. It makes no sense for the catcher to get positive 'war' on an at bat for his pitch framing ability, when the bat ultimately results in a base hit. Right now the way they are doing it, is akin to fip era. It's hypothetical, but not result based.

Bill James actually has a comment in the New Historical, written 15 years ago, which touches on this topic, although he doesn't use the term "pitch framing." It's in the Gus Mancuso comment, and also has to do with the Giants' infield defense in the Carl Hubbell era, which had a thread a few days ago here. What Bill says is that the Giants had been starting a catcher named Shanty Hogan, who was a very large man who could not give a low target. Umpires would see all those sinkers, curves and screwballs, see that Hogan was having to stab downwards with his glove, and call them balls. When the Giants switched to Mancuso, he gave a low target, caught those balls with his glove on the rise, the umpires called them strikes, and the Giant pitchers' ERAs dropped like a stone, because that staff, at the time, was all ground-ball pitchers, although probably none of them was nearly as good at it as Hubbell was. But, essentially, this is pitch framing, and you can check the Giant pitchers' ERA with Hogan and with Mancuso without too much effort, to get some idea of how much the magnitude might be for a pitcher with an extreme tendency to pitch in one direction - down, up, inside, outside.

One important note, or at least I think so, is that a catcher's tendencies, viewed in this way, may produce mixed results. Suppose the Giants had had a staff with Carl Hubbell and a bunch of guys who pitched high heat, like Robin Roberts. Mancuso's contribution to Hubbell would be countered by his catching high pitches by stabbing upwards, so the guys other than Hubbell would get hurt, not helped. In other words, if you're going to talk about the magnitude of pitch framing, you need to know the tendencies of the pitchers involved, as well as the catchers. There may be catchers, and Yadier Molina and IRod are certainly places to start looking, who may be able to leverage glove position in any direction, changing from pitcher to pitcher, or maybe even pitch to pitch. But that's asking a lot from anyone worse than Molina. Not adjusting for the tendencies of the pitchers involved may be one reason why the results of pitch-framing analysis seem unreasonably large. I don't know, but I do think that pitcher tendency needs to be factored into any pitch-framing analysis. - Brock Hanke

Another thought that I had right after I hit Submit: This business of pitch framing may be why some pitchers, including superstars like Steve Carlton, had favorite "personal catchers", like Tim McCarver for Carlton. It may be that McCarver's normal positioning led to his glove moving in the direction that most benefitted Carlton's stuff. Don't know; thought it was worth throwing into the mix, especially since you can probably get pitchers who have or had personal catchers to tell you why they liked that catcher so much. - Brock

I've wondered for years if Piazza's real reason for staying a catcher was being great a framing pitches. He got a terrible defensive rep due to a weak arm, while I-Rod had a wonderful rep due to a great arm. But I-Rod's teams consistently (until late in his career) finished near the bottom in ERA while Piazza's finished near the top. I-Rod was known early on for skipping pitcher/catcher meetings and being focused purely on his arm (high risk throws to first or third to try to pick off runners for example) while Piazza was known for spending extra time with his pitchers and working hard on being a good catcher. Seems to me that their reps would hint at Piazza being better at the 'little things' which stats back then wouldn't pick up on like pitch framing. It would be very, very interesting to have had those stats for the mid-90's in order to see if that hypothesis is correct. Imagine if Piazza's defense was actually a positive back then and I-Rod's a negative despite the arms...boy would that turn a lot of value equations from that era off kilter.

I remember there was a stat called Zone rating, which had Piazza as one of the few players to ever score a perfect score in a season (I think it's since been modified and that is no longer the case) what it basically said was every foul ball possible that he could have caught, he caught. I've always argued that his defense was overstated, never went so far to say it was good, but it just seemed like a big stretch to think that a guy who caught as many games as he did, on a team that was as good as his teams were at preventing runs, was somehow a record liability defensively.

If these studies focused strictly on ball 4 and strike 3 calls, then that would be dealing with actual value.

This is the equivalent of rating hitter value on RBIs. Changing a count into a pitcher's count has value. It won't pan out every time, but it averages out over a season to some positive value added. It's the same line of argument as why we use ERA+ or WAR or wOBA (or whatever) to measure value instead of RBI or wins.

This is the equivalent of rating hitter value on RBIs. Changing a count into a pitcher's count has value. It won't pan out every time, but it averages out over a season to some positive value added. It's the same line of argument as why we use ERA+ or WAR or wOBA (or whatever) to measure value instead of RBI or wins.

I understand that. And am not proposing they stop on the overall scorekeeping, but at the same time, when they list a WAR based upon the value of changes in the at bat, and then the at bat results in the opposite of the value they are given, effectively their contribution has been fully nullified and has no bearing on the real world.

Again it's like the fangraphs travesty of pitcher war. It's theoretical on what could have been in an alternate universe, but has no bearing on the real world, when you see something like "Molina created 2+ wins because of his pitch framing ability." and you go back and look at some of those at bats and see them still resulting in hits.....it's not "True" wins that he contributed, he helped the odds some but ultimately they beat the odds.

No catcher should get positive value from his defense handling in an at bat that result in the batter reaching base. And the opposite is true. At least on a war perspective. You can use some other silly nonsense like in a wpa perspective and that is fine, but generally people (except weirdo fangraph groupies) consider war to be a measurement of actual tangible value.

Now of course this could be a tool to work and help determine how much percentage of each plate appearance can be assigned to a catchers value, pitchers value and maybe even defender's value. (Say we discover that babip with 1-2 count is lower than babip with a 2-1 count, then when an out is made with a 0-2 count in which the catcher helped get the count there, he would get an increased share of the defensive credit.)

This is the equivalent of rating hitter value on RBIs.

Not really, it's equivalent to rating of a hitter value based upon obp.

Not really, it's equivalent to rating of a hitter value based upon obp.

No - By the logic proposed, getting on base doesn't do anything except improve the odds that you'll score. If you don't score, you might as well not have gotten on base, Scoring is what matters. Except that scoring doesn't matter - it just improves your odds to win, so nothing matters except wins. If you don't win, you might as well have not scored. Because wins matter. Except that wins don't matter, they just improve your odds to make the playoffs, so making the playoffs is all that matters. So if you don't make the playoffs, you might as well have not won. Making the playoffs matters. Except that making the playoffs just gives you a chance at winning the Series. Because COUNT DA RINGZZZZZZ.

Or, we can accept that balls and strikes matter.

No - By the logic proposed, getting on base doesn't do anything except improve the odds that you'll score. If you don't score, you might as well not have gotten on base, Scoring is what matters. Except that scoring doesn't matter - it just improves your odds to win, so nothing matters except wins. If you don't win, you might as well have not scored. Because wins matter. Except that wins don't matter, they just improve your odds to make the playoffs, so making the playoffs is all that matters. So if you don't make the playoffs, you might as well have not won. Making the playoffs matters. Except that making the playoffs just gives you a chance at winning the Series. Because COUNT DA RINGZZZZZZ.

If that was the logic proposed, then the statement I would have made would have said "we shouldn't count any positive war for a catchers performance unless it resulted no runs(or runs if it was a negative performance)." That was not what I proposed, and it's ridiculous to think that was what I proposed.

We accept each at bat as an individual piece of the puzzle working to scoring a run. But we have never given credit to pitchers who get ahead of the count and then allow a base hit. This is what this system does. We are giving credit for a portion of the result without looking at the actual result. I mean by the logic that they are using for catchers rating, a batter who gets ahead on a count 2-0 and then grounds out would probably still be credited with a positive performance. (mind you, I would have no problem with a system that does any of that as an analytical tool, just don't see why it matters in a war like stat)


There are lots of ways to go with this, but I think any system that assigns catchers defense as a bonus to war, should only be based upon how it affected the actual results that happened and not the theoretical of what is possible. Example if a guy grounds out on that 1-2 pitch that should have been 2-1, the catcher gets some of the defensive credit for the improvement in the odds on that play happening. If he smacks a homerun.... well no credit is given as it didn't ultimately make a difference.

If you don't think ball and strike calls other than ball 4 and strike 3 matter, then go take a look at the difference between overall hitting lines after an AB with say a 1-2 and a 2-1 count. For all of MLB in 2013, the average hitting line after 1-2 was .179/.228/.271 and after a 2-1 count it was .255/.387/.412. So from a 1-1 count, that next pitch going one way or the other ends up swinging the hitter's OBP by 159 points and SLG by 141. That is a MASSIVE difference.

When people say saving runs, they aren't talking about theoretical runs. They are talking actual runs. You can't call changing a count to a positive chance for the team that ultimately results in a base hit as saving actual runs.

If you don't think ball and strike calls other than ball 4 and strike 3 matter, then go take a look at the difference between overall hitting lines after an AB with say a 1-2 and a 2-1 count. For all of MLB in 2013, the average hitting line after 1-2 was .179/.228/.271 and after a 2-1 count it was .255/.387/.412. So from a 1-1 count, that next pitch going one way or the other ends up swinging the hitter's OBP by 159 points and SLG by 141. That is a MASSIVE difference.

I fully understand the changes to the chances. But again, when talking about a value stat like war or runs saved, you can't be focusing on the theoretical. This is the same silly nonsense that make people list fangraph war for pitchers. And again, track that information, just like fip, it serves a purpose. It's a predictive stat as it stands, and it helps gauge how a catcher could help a team, but it is not actual runs unless it actually changes the outcome of the at bat to a positive(for the team) until then it's just potential sitting in the ether.

Another thought that I had right after I hit Submit: This business of pitch framing may be why some pitchers, including superstars like Steve Carlton, had favorite "personal catchers", like Tim McCarver for Carlton. It may be that McCarver's normal positioning led to his glove moving in the direction that most benefitted Carlton's stuff. Don't know; thought it was worth throwing into the mix, especially since you can probably get pitchers who have or had personal catchers to tell you why they liked that catcher so much.

I think this was a big part of the Negro Leagues' Biz Mackey's defensive rep - that he could smoothly & naturally make a "strike" out of nearly anything the pitcher threw up there. Apparently everybody liked pitching to Biz Mackey.

I don't find it hard to believe it, but unless it's the call third strike, you still have to continue the pitch sequence. I understand most of these studies look at how the average player does and awards credit based upon the difference. I.E. it's a 1-1 count, and a ball is called a strike, they look at the average performance difference throughout the league between a 1-2 count vs 2-1 count and award credit based upon that. It's nice and theoretical and all, but it doesn't really save the "run" unless the batter actually makes the out. Just changing the odds isn't the same thing as actually changing the outcome. If these studies focused strictly on ball 4 and strike 3 calls, then that would be dealing with actual value.

I have to disagree. It seems that the best you are ever going to do is to use statistical methods that measure the incremental difference of batting say "1 ball, 2 strikes" vs "2 ball, 1 strike".

Because, you never know if that batter hit a weak grounder on 1-2 because the previous pitch was a strike not a ball. Or the batter was able to lay off that 1-2 pitch because the previous pitch was a ball.

YOu seem to think that the only outcomes that matter are outcomes that end in walks or strikes (see the part where you say "strictly on ball 4 and strike 3 calls'"). It's not just about ball 4 and strike 3; it can never be.

There are going to be plenty of times when someone hits a weak grounder because behind in the count; or lays off an unhittable pitch early in the count then later gets a single, as well as the vice versa of those.

The only way you can do that, seems to me, is to try to measure these very small, % based, incremental differences. Your suggestion that only strike 3 and ball 4 matter cant possibly be true there are all sorts of outcomes that are happening BEFORE you get to ball 4, and maybe POSSIBLY based on getting the strike call.

YOu seem to think that the only outcomes that matter are outcomes that end in walks or strikes (see the part where you say "strictly on ball 4 and strike 3 calls'"). It's not just about ball 4 and strike 3; it can never be.

I just don't see how you can assign a run value to something that ultimately failed. Again, I understand why we want this data and fully agree with tracking it. I just do not agree with this data, as currently being tracked, as being useful in a stat like war or a concept like runs saved, which is at least tangentially related to the real world.

Same with fip. They are both fantastic evaluation tools, fantastic for understanding the game, but not pertinent to a value stat like war or runs created etc.

I fully understand the changes to the chances. But again, when talking about a value stat like war or runs saved, you can't be focusing on the theoretical. This is the same silly nonsense that make people list fangraph war for pitchers.

YOu seem to equate "theoretical" with "Statistical based" and then dismiss it or something.

It's theoretical in the sense, that we cannot prove that outcome X was in fact directly based on factor Y (where factor Y is Molinas ability to get that borderline call on strike 1).

That's merely because just about any outcome in baseball, any statistic say, is always going to be based on a myriad of factors.

Here's an example...

CLemente faces Gibson CLemente is a career .317 hitter; but this year he's a .350 hitter, but this month he's a .280 and today??? Does it matter what he is today?? or not??... Meanwhile Gibson is a 150 ERA pitcher career, but this year he's 190, and in night games he's 130. And in this year's night game's he's 140? Does it matter? Do you only count the career? Or only count the last 10 AB? or is it simply a continuum? Plus there's runner on base, in hundred years of baseball, the runners on base increase b.a. by 5 pts say, in this decade it's only 4 pts; and in Forbes field at night this year it's 6 pts.... Plus Gibson is a fastball pitcher and Clemente is only .250 against him; PLus Clemente hurt his back last thursday, plus in the second inning Gibson felt a twinge, plus in the last several minutes the sun is casting a giant shadow in RF.


Now CFB; how do you repond:

a) only some of this stuff matters, the career stuff Ok, the rest is irrelevant. or
b) it's all relevant, but it's real hard to measure.


Of course it's "ALL RELEVANT" is my answer. it's all relevant it's just hard to work out the numbers becase:

a) there are so many factors; and
b) many of these factors are going up and down, IN REAL TIME.

Now, we see Clemente pops out to SS. Now what do you say:

a) Clemente was a .000 hitter in that situation because he was out;
b) CLemente was a .31744555 hitter because of factors : a, b, c, d....

I would go with B. What do you say?

HEre's a better suggestion


"Not everything that matters can be counted, and not everything that be counted matters." Einstein.

Get it?

CFB: lets put it this way: do you believe that baseball outcomes can in fact be predicted statistically or not? Given that: a) we are never going to be fully informed, as the states of molecules in front of homeplate is constantly changing, as well as lots of other stuff; and b) we cant' state FOR CERTAIN what will happen.

What other position is there? That everything is deterministic? And no amount of statistical crunching will change Clemente's pop out to a hit?

I confess I dont see another possibility. What DO YOU THINK?

******

This is the same silly nonsense that make people list fangraph war for pitchers.

I confess I dont know what that controversy was about, but if you could provide some detail I would be interested.

Because you are putting into a giant bin called a "data set" and you are using that to do draw STATISTICAL conclusions, and not pure DETERMINISTIC conclusions.

You cant say factor A, caused Clemente to get a hit. THe best you can say is: after 10,000 data pts we know that fastballs are hit with a .300 BABIP, and after 100,000 data pts. we know that night games increase b.a. by 5 pts.

You cant say that because it was a night game, that caused Clemente to get a hit.

That just how reality is. Unless you subscribe to something like determinism. WHich I guess I do in a way to. I think ultimately everything is pre ordained.

But given I dont have all the factors I need to predict reality, then I am left with percentages, and the limitation that I cant predict things perfectly.

There are lots of ways to go with this, but I think any system that assigns catchers defense as a bonus to war, should only be based upon how it affected the actual results that happened and not the theoretical of what is possible. Example if a guy grounds out on that 1-2 pitch that should have been 2-1, the catcher gets some of the defensive credit for the improvement in the odds on that play happening. If he smacks a homerun.... well no credit is given as it didn't ultimately make a difference.

This is sort of the crux, so I will attempt to address this.

Preface: it's basically the difference between what some people refer to as Performance versus Value. People have different terms for these two things, but basically the difference between what actually happened versus someone true talent level.


Let us suppose we have two outfielders. We put them through rigourous testing, we find Outfielder A; can run very fast, he can cover 50 yards in 3 seconds, he only drops 1/100 balls. WE put him through stress tests so there is no reason to think that he will flub this in a real game. Player B covers half as much ground and drops 1/10 balls.

We assign ratings based on Value, A=99; B=40.

THey then play an actual season of baseball. Player is forced to play LF in Fenway. Fenway gets a reconstruction so that the Monster is moved into 100 feet. And it's 100 feet high. Player A protests that there is no reason to play there but no one listens. Player B plays in Puerto Rico where it's 500 feet to RF and everything there gets caught in a breeze.

At the end of the season:

Player A, 0 put outs, 0 assists, RField: -50.
Player B, 700 Put outs, 50 assists, RField: 1469.


WHich player performed better? Which player has more talent?

I confess I dont know what that controversy was about, but if you could provide some detail I would be interested.

Fangraphs war for pitchers use fip data instead of actual results. So if in theory the pitcher should have allowed 20 hits but in actuality allowed 25,they go by 20 hits allowed in determining what the era should have been. bWar is based upon actual results.

CFB: lets put it this way: do you believe that baseball outcomes can in fact be predicted statistically or not? Given that: a) we are never going to be fully informed, as the states of molecules in front of homeplate is constantly changing, as well as lots of other stuff; and b) we cant' state FOR CERTAIN what will happen.

No, but that when you are assigning run value to a player, which is what this is trying to do, that it should be based upon something that actually happened. Generally speaking I think that the lowest individual action we should award a run value to is the entire plate appearance. This is cutting it down to awarding a run value to each individual pitch. If they are going to do something like that at one level for one player, you have to do it at that level for every player if you want to assign a consistent value. As it stands, for pitchers we are only evaluating them on the ball in play. For batters same thing, for fielders same thing(mind you, not sure how to cut them down to a per pitch basis) , but we are now expecting to put individual pitches into the pool, but only for catchers? Why not assign individual pitch values to batters who manage to work the count to 2-1 instead of hacking on a borderline pitch that would have made it 1-2? Same with pitchers?

Dave Ortiz comes up, gets ball one...borderline pitch that he takes and which the data says that 70% of the time the rest of the league swings and misses it or it gets called a strike on them, and instead he gets a ball call on him...now he has worked his way into a favorable hitters count, shouldn't he get the same type of bonus the catcher who worked the count in the opposite way did? Of course he then pops up to the second baseman, but by this system, he gets more "runs" for working the count to 2-1 instead of 1-2 even if he ultimately has the same out result.

I believe that there are several ways to evaluate everything, and when you come up with a backwards looking stat that is assigning a run value to a player, it should logically be based upon actual events that happened.

AGAIN. I have no problem with tracking this information. I have no problem with using this as an evaluation tool. I just don't see how this actually works in a war like framework and produces reasonable results. In fact I think this data is EXTREMELY valuable. I just don't remotely think it's an accurate representation of the runs saved when people are using that as a measurement of this type of analysis.

WHich player performed better? Which player has more talent?

Not really understanding the question. Too extreme of an example to make any sense of the scenario. But when it comes to evaluating which player had the better year in that one regards, obviously the guy who actually played and made the out had the better year. Not really understanding how this relates.


This article was about what is in store for the future of this type of analysis. And the truth is that this is still in it's infancy and needs lots of feedback on what it does. Again. I like the analysis. I like the information. It's impressive that they are able to gather all of this information, create a good framework for measuring what is going on, and come up with pretty definitive answers that a player affects the call of the pitch. And it's pretty conclusive on who is good at it, and who is bad at it and that it's a repeatable skill. I'm just pointing out that when they assign a run value to something like this it's more removed from "real" run values that we have associated for years in the stat community, in that it's not really based upon the same "definition" of event that we have used forever. And that they are still a ways away from incorporating this type of analysis into a stat like war and having it maintain the integrity of the stat.

We accept each at bat as an individual piece of the puzzle working to scoring a run. But we have never given credit to pitchers who get ahead of the count and then allow a base hit. This is what this system does. We are giving credit for a portion of the result without looking at the actual result. I mean by the logic that they are using for catchers rating, a batter who gets ahead on a count 2-0 and then grounds out would probably still be credited with a positive performance. (mind you, I would have no problem with a system that does any of that as an analytical tool, just don't see why it matters in a war like stat)

OK I see what you are saying here and none of us has really responded to it. THis is valid point and it's different than what I was talking about. I apologize.

What if you did this:

Batter gets ahead 2-0; he ultimately grounds out. Instead of being charged say -0.45 (example only) for the out. WE charge him at -0.65 because he was ahead in the count and had a better chance to get a hit.

Conversely. The pitcher gets ahead 2-0, he ultimately gives up a hit, instead of charging him for 1 single, we charge him 1.2; same reasoning.

ANd similar for Molina.

Why would you have a problem with this system? IT would not change Willie Mays from HoF to average. It would hardly change anything at all in fact.

Nor would Molinas pitch framing ability hardly change him that much. It might slightly change the value of Cardinal pitchers.

OK so what? Isnt that what the system is telling us? YOu cant just ignore what it is telling us.

It turns out that Wainwright is not as good as Verlander, in fact Wainwright is only as good as Gerrit Cole. And it turns out that Molina is better than Johnny Bench. He's incredible.

BUt the numbers still balance. You add up Molina and you add up the Cardinal pitching staff and it still comes out to 95 wins. Very close to actual. What is wrong with a system like that? IT all balances out.

EDITED: for some substantive changes

BUt the numbers still balance. You add up Molina and you add up the Cardinal pitching staff and it still comes out to 95 wins. Very close to actual. What is wrong with a system like that? IT all balances out.

They aren't there yet. Again, they have produced incredible information, but they haven't reached the point that they can include it in a war like stat yet, and maintain the integrity of the stat. Sure they could incorporate it into a p.o.s. stat like fangraphs war, after all, it already is broken, much in the same way that incorporating this into it would be.

Not everything has to be incorporated into the stat, and again, the only real way to maintain the same consistency is to only evaluate the value of the pitch handling skill on the ball in play of that particular at bat. Mind you, that isn't the only way to use that stat, again, the system they have now is GREAT. it's just not being measured on the same scale/event that an all inclusive stat like war is, or even a generic stat like "runs created/saved" is done.

Dave Ortiz comes up, gets ball one...borderline pitch that he takes and which the data says that 70% of the time the rest of the league swings and misses it or it gets called a strike on them, and instead he gets a ball call on him...now he has worked his way into a favorable hitters count, shouldn't he get the same type of bonus the catcher who worked the count in the opposite way did? Of course he then pops up to the second baseman, but by this system, he gets more "runs" for working the count to 2-1 instead of 1-2 even if he ultimately has the same out result.

I think we agree on this. THe system just has to be balanced out for the Pitcher, the catcher and also the batter. It would be hard to do that, doesnt make the system invalid.

SO then what is your concern with pitch by pitch evaluations? doesnt this mean you cant defend that argument now?

Generally speaking I think that the lowest individual action we should award a run value to is the entire plate appearance. This is cutting it down to awarding a run value to each individual pitch.

It seems you are taking an arbitrary line in the sand and making this some sort of sacred point.

LInear runs, takes a single and awards it a portion of a run. Do you have a problem with that? Why not? It's the same reasoning isnt it? It usually takes 3 or 4 at bats to go from Runner on base to scoring a run. By your reasoning we cannot use a single plate appearance to award a portion of a run.

Same thing with going from Runs above average to Wins above average. They say 10 runs is about equal to a win. I have a problem with that because of leveraged situations. But the concept seems OK to a point. You dont need to use an actual game w/l outcome to prove that 5 runs is worth 0.5 wins do you ?

It's the same reasoning isnt it? A point was made above, that your reasoning would lead to you to not count runs because COUNT DA RINGZ! YOu seem to dismiss that argument, but I think it's the same.

You say we cant deconstruct AB into pitch by pitch. But we deconstruct wins into runs, we deconstruct runs into AB. Whats the difference? WHy are you stuck on this point?

What if you did this:
Batter gets ahead 2-0; he ultimately grounds out. Instead of being charged say -0.45 (example only) for the out. WE charge him at -0.65 because he was ahead in the count and had a better chance to get a hit.

Conversely. The pitcher gets ahead 2-0, he ultimately gives up a hit, instead of charging him for 1 single, we charge him 1.2; same reasoning.

ANd similar for Molina.

And that is kinda my point. I wouldn't have a problem with that for the most part. But personally what I would do, would to look only at bats in which the system claimed that the catcher did a positive or negative thing in regards to pitch framing, and if the result of the at bat correlates, then look at sharing "defensive" value between the catcher and the fielder.

This wouldn't be rewriting the entire system, and you would still be dependent on the same definition of event that we have been using for decades.

(I wouldn't exclusively look at those at bats as an evaluation tool, everything is valuable, so keep all the data that they have and make it available, but for a war or runs saved frame work, not everything is necessary)

I'm just pointing out that when they assign a run value to something like this it's more removed from "real" run values that we have associated for years in the stat community, in that it's not really based upon the same "definition" of event that we have used forever

yes, I see this part.

But personally what I would do, would to look only at bats in which the system claimed that the catcher did a positive or negative thing in regards to pitch framing, and if the result of the at bat correlates, then look at sharing "defensive" value between the catcher and the fielder.

But then you're skewing data, that does not seem to be scientific at all, to have this great big data set and then to only take outcomes that you want to take. I realize you're not being biased for St Lou, or to Molina, but you are still taking only certain outcomes.

Same example as before. what if you were only to use, batter data from games where player A's team won? And only those hits in games where the Pirates won would be used for Clemente's Wins Above Average.

Probably it wouldnt not change things much, maybe Pete Rose still gets say 4 WAA in CIN in '69 (example only)


But you cannot say that if we take 300 Clemente AB and 300 Rose AB that is better than taking 600 AB for both those guys. Can you?

Again your argument against deconstructing stats/value. What if for example Rose batted .400 in games CIN lost? and batted .250 in games they won. He still gets a b.a. of .300 but your Win Shares method Rose only gets 1.0 WAA instead of say 3.0.

Would that bother you? Isnt this the same as your argument against deconstructing?

EDIT: massively, read it again.

If you have all that data, then you have to use it all to be as accurate as possible. Your argument seems arbitrary.

SO then what is your concern with pitch by pitch evaluations? doesnt this mean you cant defend that argument now?

Again, I like the information. It's very useful and is pretty much a lock in being relatively accurate way of studying it. It's not an accurate accounting of how many runs a catcher saves/costs his team though. It's a misleading way of labeling the catchers performance on this. It might be the best way to label it, but it's misleading based upon the "industries" standards of definition of a run, which is based upon a result from a specific, finalized event(plate appearance usually). This is assigning a run value to a portion of an event, which ultimately may not have the same positive results we have associated with the individual plate appearancs.

As I pointed out in post 26. My comments are about what is in store for the future, and what is right or wrong with the stat as is now.

It seems you are taking an arbitrary line in the sand and making this some sort of sacred point.

Not really, a single plate appearance or defined event such as a stolen base attempt has been the standard for evaluating runs among baseball stat analysis. Fip doesn't deal the pitch count(although it probably could be enlightened a little more if it did) linear weights doesn't deal with the pitch count, war etc...nothing in baseball that is used as an backwards value looking tool, is based upon anything other than the individual plate appearances(and other separate distinct plays)

LInear runs, takes a single and awards it a portion of a run. Do you have a problem with that? Why not? It's the same reasoning isnt it? It usually takes 3 or 4 at bats to go from Runner on base to scoring a run. By your reasoning we cannot use a single plate appearance to award a portion of a run.

No. By my logic we can't use a 3-0 count and argue it's worth .1 run on an at bat that the batter eventually grounded out on. Especially when combined with a system that uses one plate appearance as the definition of event.
Portions of runs is still being based upon the single event of a plate appearance.

You say we cant deconstruct AB into pitch by pitch. But we deconstruct wins into runs, we deconstruct runs into AB. Whats the difference? WHy are you stuck on this point?

Never said we can't deconstruct ab into pitch by pitch. Said that we shouldn't be using pitch by pitch data in a system that is based upon plate appearance by plate appearance data. If you want to rewrite the entire system in the way you discussed in post 27, then the way they are currently assigning value to catchers defense makes total sense. But in the systems that we have currently, it's incorrect to include the two. You cannot include pitch by pitch data into a data set that was created using per plate appearance data. If you do, you need to smooth out the pitch by pitch data based upon the result of each of those particular at bats where the pitch by pitch data figures into the play.

But then you're skewing data, that does not seem to be scientific at all, to have this great big data set and then to only take outcomes that you want to take. I realize you're not being biased for St Lou, or to Molina, but you are still taking only certain outcomes.

You do realize the Molina they are talking about isn't Yadier Molina but his brother.

If you have all that data, then you have to use it all to be as accurate as possible. Your argument seems arbitrary.

And you can if you want. But you have to use all the data, in the way you pointed out in post 27. You want to include catchers ability on a pitch by pitch basis, so if even though the catcher created a beneficial situation, the base hit happened, that is fine, you can assign the blame to the fielder, hitter or pitcher, just ultimately make it a zero sum event.

You do realize the Molina they are talking about isn't Yadier Molina but his brother.

I thought it was Benji.

nothing in baseball that is used as an backwards value looking tool, is based upon anything other than the individual plate appearances..

what about ball and strike data? I guess I need to know what you mean by "backwards looking tool"? to understand what your concern is here. I think I understand the difference between outcome based stuff and peripheral stuff that might lead to better predictions. I think I need more info to understand your pt here.

I realize pitch framing is quite a bit different than other stats; I dont see that that concept, per se, makes it invalid or not useful.

I still think you could successfully combine a pitch by pitch metric, and use it in combo with an AB based metric. I dont see any fundamental reason why that would be impossible. I guess that is where we differ.

Good discussion. thank you.

what about ball and strike data? I guess I need to know what you mean by "backwards looking tool"?

Backwards tool is looking towards the past event and assigning a value based upon that event to evaluate the past performance. (War-not fangraphs war--, runs created, ops, etc) are backwards looking tools. Forwards looking tools are tools that is useful for evaluating going forward/predictive tools would be something like fip. You aren't stuck with actual events for a predictive tool. You can deal with theoretical concepts as much as possible.

Ball and strike data isn't currently used in any popular value tools that is out there. Strikeouts and walks and pitch counts are used, but individual strike and ball data isn't really tracked in any meaningful way in the way that we are talking about here.

I realize pitch framing is quite a bit different than other stats; I dont see that that concept, per se, makes it invalid or not useful.

Agree...Never said it wasn't useful. I think it's an incredible tool and has been as groundbreaking in it's own way as Fip was for it's time. Just stating that I don't think that they are currently at the level to incorporate it into a "one size covers all" value tool like War and was giving the reasons why I think that.

I still think you could successfully combine a pitch by pitch metric, and use it in combo with an AB based metric. I dont see any fundamental reason why that would be impossible. I guess that is where we differ.

Again, never said it would be impossible. I've only been debating/arguing with taking the current structure of pitch framing ability and incorporating it into a current war (or other value) structure and pointing out the flaws with just dropping it into those systems(in my opinion).

Enjoying this conversation quite a bit. Wondering about a slightly different aspect. Is there any way that framing affects DIPS theory? If hitters fare far worse after a count is flipped toward the pitcher due to framing, then should the framing effects be adjusted for in a DIPS study? Because it may be that catchers are exerting an ability to influence quality of contact and therefore skewing pitcher DIPS data. A corollary to my question might be to ask whether pitchers exert greater or lesser control over BIP outcomes depending on the count. Since some pitchers will work into more hitters counts than others (and the opposite), there could be skewing of the data toward contact during hitter's or pitcher's counts.

As I was participating in this thread, a lot of those questions came up to me. Not just about catchers affecting Dips theory, but what about a pitcher that has the ability to get ahead of the count 0-1 consistently. Is there any correlation between babip and the count? (mind you a lot of the reason the numbers are better or worse for a hitter based upon the count, is how close it makes it to them striking out or walking...which dips already accounts for effectively)

I imagine that 10-20 years ago, the count figured more heavily in babip than it does in todays take and rake style of play. I'm sure someone has researched dips by count, it seems somewhat too obvious not to have been put under a microscope somewhere along the line.

OK, I dont think there is much we differ about. Do you believe that data can only be useful if it only uses final outcomes (strike out, walks, hit) in the data or do you think that it can work if it uses ordinary pitch by pitch data?

I think this is where we parted company earlier now I am not so sure.

I think all data can be useful, even the atrocities such as WPA and productive outs have some value.

OK, I dont think there is much we differ about. Do you believe that data can only be useful if it only uses final outcomes (strike out, walks, hit) in the data or do you think that it can work if it uses ordinary pitch by pitch data?

I've only been discussing catchers pitch framing ability in regards to "war" and value stats. Mostly because of this from the article.

I often discuss pitch framing with my colleagues. The most common source of doubt I hear: The numbers don’t pass the sniff test. The infamous Jose Molina has too many smart people crinkling their brows. This is a determination each of us has to make. Can there be a possible 5-win data inefficiency that existed for 100-plus years of baseball history? Can it be possible such a big deal was missed for so long?

And my point is that the reason it's not passing the sniff test, is that it's dealing with two different connotations of what a win equals. When someone says a player is 3 wins, they instinctively are thinking of, in regards to war or another all encompassing stat that ranks players by wins. And most of these systems out there are based upon event generated numbers that are generated on a per plate appearance basis. Now we are looking at a stat that is generating wins, not based upon the single event of plate appearances that we are used to going by, but based upon varying states in a particular plate appearance. We are dealing with a different scale.

There is NOTHING wrong with the way that they are coming up with their data. There is nothing wrong with their research. It's very valuable data and research. I just am not comfortable with using the same terms for wins/runs that we have used based upon plate appearance events with per pitch events.

There is NOTHING wrong with the way that they are coming up with their data. There is nothing wrong with their research. It's very valuable data and research. I just am not comfortable with using the same terms for wins/runs that we have used based upon plate appearance events with per pitch events.

Put another way, it's not that pitch-framing wins are too high. It's that with pitch-framing wins catchers are the only players being credited with influencing the count, and consequently pitchers and hitters might not be getting enough credit in tbeir win-based metrics.

Intuitively, a seventh of a run per pitch (0.16 for you sticklers) is much too high and may be injecting confusion into the discussion. Put things that way and a catcher who frames two pitches in an inning has a bigger impact than a shortstop who steals a single up the middle. If that's any kind of repeatable skill, you could have several years where the "real" MVP is the best pitch framing catcher.

I dont think you're doing the math right, Zach. Or to be more fair, the math your are quoting is wrong. But I gather you already know that.

Look at the breakdown posted in post 15. That's a 14% differential in OBP. I dont how much to call getting on base, a single is .45, but this includes Extra Bases as well as walks.

Let's call it 0.4 as a starting pt.

Multiply by 0.14 and we get what? about 0.06 of a run.

if you multiply by two pitches a game for 160 games, that might work out to 2 games per season.

Which is close to what BJ said earlier upthread. The numbers are not really that hard to believe. I'd like to see better ways to measure CS and stuff like that for catchers..

But it remains to be seen whether anyone is really stealing this many strikes every game. Stealing two strikes in a game does not seem like a feat. But doing it every day, for 162 games? Maybe a stretch....

This thread wandered off into other issues, but the more I think about it, the more I think I have figured out something serious. It may very well be that the vast majority of catchers are good at framing a subset of pitches (high heat, low curves, sliders to the 3B side, or the 1B side, etc.). This strikes me as a very important context if you're looking at pitch framing. I don't know who is doing all the detail work of watching film to compile these numbers, but if someone has the info available as to the whole context of what pitches got thrown to a catcher, instead of just a count of Strikes Stolen, and Strikes Lost, I think it might be VERY useful to figure out which catchers are good at framing which pitches, and then putting that into a context of what their pitching staffs actually threw. When that gets done, you may have more sniffable (that's a word?) evaluation. Mancuso might not have been any better a pitch framer than Hogan. He was just on a team with the right pitching staff. If the staff had been the Robin Roberts Phillies of the early 1950s, Mancuso might grade out as a poor framer, while Hogan might look sensational. And most pitching staffs are not going to be monolithic. There may well be many catchers who helped some of their pitchers, but hurt the rest, because they were only good at framing certain types of pitch. Pitch framing analysis is in its infancy. There could be a lot of factors that would reduce some of the wilder results by putting them in context. The stuff of the staff is certainly one of them. - Brock Hanke

RE post 47:

I've seen breakdowns of which catchers are gaining/losing calls in specific parts of the strike zone. It was basically showing rendered heat maps to show the size of the called strike zone each catcher was getting to compare between catchers, and there were absolutely conclusions you could take away about high/low or inside/outside framing for specific catchers. I honestly don't remember the specific article but I think it was on Baseball Prospectus late last year. IIRC Lucroy gains a lot of his strike calls on low pitches while Jose Molina tends to steal outside pitches more.

another thing to think about: once this aspect becomes well known it might go away or change. Similar to the Heisenberg uncertainty principle in that the observation itself may change the outcome. In this case, once umpires are made aware of this they may operate to lessen the effect. Which will only further provide more grist for the mill here as we argue about strikes that were stolen in April but now it's may.

Then we can have peripheral trends like: monthly differential in strikes lost by umpire or something.

Thank goodness! This is the discussion on pitch framing I have been seeking for some time. If I were to summarize my concerns about all of the ground-breaking studies that have been over the past few years, then I would say that the measurement of the value of catcher's ability to frame pitches is being computed on a pitch-by-pitch basis for catcher-framing only while the rest of the linear weights based calculations are being done on a plate appearance/base-outs basis. It seems to me that all the parties involved need to be measured on the basis for it too work. This makes me skeptical of the measurement of the value of an individual catcher's pitch framing performance. To me, the numbers coming out of the pitch-framing studies seem too high. I personally would feel more comfortable with the results of these studies if they were not measuring the effect on a different scale (pitch outcome vs. PA/base-out outcome) from the other components of WAR/linear weights.

Great discussion!