SoxProspects News
|
|
|
|
Legal
Forum Ground Rules
The views expressed by the members of this Forum do not necessarily reflect the views of SoxProspects, LLC.
© 2003-2024 SoxProspects, LLC
|
|
|
|
|
Forum Home | Search | My Profile | Messages | Members | Help |
Welcome Guest. Please Login or Register.
Vazquez Pitch Framing Log
|
Post by philsbosoxfan on Aug 7, 2014 0:54:28 GMT -5
This is from a Red Sox Stats tweet: Based on the Vazquez graph several posts above, this pitch location pattern would seem to be a perfect match for Kelly & Vazquez' talents.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 7, 2014 14:51:23 GMT -5
Once again, BP's report wasn't updated and I e-mailed them and then it was, so we have two starts combined. Let's see if they've really fixed the bug now.
The number of expected strikes over the last two games was really low; expected strikes per framing chance were .260, and the previous low was .261 in DLR's July 19th start. I'm guessing that Kelly had a lot to do with that, throwing a lot of pitches that were frameable but very tough and not many than were frameable but probable strikes. And of course it would be tougher to add run value if most of your framing chances are that iffy.
So now I'm wondering whether framing skill should be normalized to expected strikes rather than framing chances. I'll mess around with that at some point soon.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 10, 2014 2:29:28 GMT -5
Once again there was no update, for August 7, and I e-mailed BP and was told that this report requires "daily staff involvement" rather than being generated entirely automatically. Today it was updated, and the new total clearly includes a tweaking to past data, not merely what accumulated in his last two games. So all of the previous daily game logs are now in question!
However, the new totals raise him back up to 41.6 runs / 4.6 WAR per 120 games (1050 innings).
I think I'm going to combine BP's pitch-framing data with fWAR for the last 5 years or so and re-rank everyone. That should be eye-opening, and it should tell us where Vazquez will rank if the rest of his game proves to be roughly average.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 11, 2014 8:31:55 GMT -5
It looks like what I earlier thought was the 8/7 and 8/8 update plus corrections was just an incorrect 8/7 update, as later that day the report changed again with 65 further framing chances, apparently from 8/8. Data for the 19-inning game has not yet seemed to arrive.
Also, the Runs per Win factor for 2014 is now down to 9.0, and the WAR totals have been tweaked to reflect that.
Finally, I've been looking into the question of how the quality of framing chances affects a catcher's apparent performance. It's really interesting, and definitely needs to be adjusted for. I'm still working on understanding it.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 11, 2014 12:54:33 GMT -5
Ironically, for the first time this month (because of BP's problems with the report) we seem to have a reliable single-game line to consider -- and it's for a game that was more than double-header played without a break, and ended at 2:30 AM in Vazquez's brain.
Hopefully the report will be back to successful daily updates.
Oh, and Dan Butler didn't have a very good game; -0.8 strikes, -0.2 actual and -0.1 count-neutral runs. That's -1.9 WAR per season (SSSSS of course ... seriously super SSS).
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 11, 2014 17:12:14 GMT -5
There is definitely a methodological flaw of some sort in their calculations. That may helps explain why their estimations of the effect size are larger than others.
Vazquez has had a better-than-average set of framing opportunities*. Adjusting that as best I can, he's shown a 3.5 WAR skill, not 4.3. Of course, combined with his arm, that's probably enough to make him a first-division starter even if his bat is replacement level.
I'm going to look into the data pattern a bit further and then send my findings to BP. I doubt we'll get a fix until next season.
*If you plot quality of framing opportunities (Expected Strikes / Framing Chance) on the X axis and framing performance (Extra Strikes / Framing Chance) on the Y, you would expect a trendline through the data to be roughly flat, but the graph to narrow from left to right, that is, the variance to decrease. Instead, you get a steep downward trendline. Of the 20 guys who got the juiciest framing opportunities (left side of the graph), 19 score as better framers than average, and it should be maybe 12 or 13 at most, if you figure that pitchers work outside the zone more than usual if they know their catcher is a good framer. Of the 17 guys who had the least promising framing chances, 16 scored as below average framers and the exception was Joe Mauer, 2010, who was just +3 runs or so.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 11, 2014 23:26:25 GMT -5
I've come up with an adjustment for BP's Expected Strike number that really seems to work. Basically, for some reason their methodology seems to exaggerate the number of expected strikes relative to the mean number per chance. It's as if their strike zone map has too steep a gradient; pitches inside the edge of the de facto zone are being scored as likelier to be strikes than they actually are (framed more often), and pitches outside the zone are being graded as framed less often than they are. This explains why they have a greater estimate of the range of framing values.
The fix eliminates the very strong tendency for good framers to have gotten much better framing opportunities (I left in a slight effect just because it made the graph look better), and it even produces the expected result of the variance of the guys who got better than average opportunities (more than a standard deviation better) to have more variance (21%) than the guys who got subpar chances. (I haven't adjusted for that, yet.)
Do the following past the smell test?
Runs saved per 1050 innings, career, raw:
Ryan Lavarnway 3.5 (yes, the sample is just big enough to be meaningful) Victor Martinez -9.4 Jason Varitek -11.4
Adjusted:
Jason Varitek -3.5 Ryan Lavarnway -10.1 Victor Martinez -12.5
Raw 13 Best Framers, Career (minimum 1 full season)
Jose Molina Greg Zaun David Ross Jonathan Lucroy Chris Stewart Paul Bako Russell Martin Carlos Corporan Ryan Hanigan Eli Whiteside Brian McCann Hank Conger Martin Maldonado
Adjusted 13 Best Framers, Career
Jose Molina Chris Stewart Jonathan Lucroy David Ross Ryan Hanigan Russell Martin Paul Bako Hank Conger Jan Gomes Francisco Cervelli Yadier Molina Greg Zaun Brian McCann
Raw 10 Worst Framers, Career
Ryan Doumit Carlos Santana Rob Johnson Gerald Laird Lou Marson Jon Jaso Koyie Hill Kenji Johjima Chris Iannetta Jorge Posada
Adjusted 10 Worst Framers, Career
Ryan Doumit Koyie Hill Kenji Johjima Gerald Laird Carlos Santana Lou Marson Wellington Castillo Rob Johnson Chris Iannetta Jorge Posada
5 Best Seasons, Raw
J Molina, 2010 J Molina, 2008 J Molina, 2013 Zaun, 2008 Stewart, 2011
5 Best Seasons, Adjusted
J Molina, 2008 J Molina, 2012 J Molina, 2013 J Molina, 2010 Stewart, 2011
5 Worst, Raw
Doumit, 2008 Doumit, 2012 Johnson, 2010 Doumit, 2009 Santana, 2012
5 Worst, Adjusted
Doumit, 2008 Doumit, 2009 Doumit, 2012 Doumit, 2011 Johjima, 2009
|
|
|
Post by Oregon Norm on Aug 12, 2014 0:54:32 GMT -5
This looks to me like a list I recognize, though only from watching many of these guys... and I can smell the bottom group in each of those lists! Thanks for all the effort
|
|
|
Post by jimed14 on Aug 12, 2014 7:34:44 GMT -5
I'm not clear Eric. Is your conclusion that catcher framing has a greater or lesser effect on games than previously thought?
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 13, 2014 2:41:26 GMT -5
I'm not clear Eric. Is your conclusion that catcher framing has a greater or lesser effect on games than previously thought? It's less than BP has it, but not hugely so. Their methodology exaggerates the effect somewhat. The reduction of Vazquez from their 4.3 to my 3.5 WAR per season is probably typical of the size of their error.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 15, 2014 18:10:40 GMT -5
Last two games are about to be added.
As I suspected, BP has tweaked all their data: stats for past years are now slightly different than the data I captured on June 27th, and have been analyzing ever since.
Fortunately, it should be easy to capture the data again and do the same effort of correcting it. In the meantime, though, my estimate of the correction on Vazquez is just a guess, so I'm doing a worst-case version (I think).
|
|
dd
Veteran
Posts: 979
|
Post by dd on Aug 16, 2014 9:18:34 GMT -5
Warning: I'm about to show my baseball ignorance once again. I have a fee questions about pitch framing in general:
1) Is the work you're doing, Eric, based on the the camera / computerized strike zones like the ones we see in the graphics on TV? I ask that because I've often wondered how accurate those are. Are they different from park to park due to different camera angles? I read somewhere a while back that park differences can be a problem. Also I understand they need to be recalibrated from time to time and that doesn't always happen.
Don't mean to get unto a technology discussion. Just wondering if that's what you use and how good is it.
2) If the automated systems are really more accurate than the home plate umps, isn't that really an indictment of MLB? If they have a way to do it better, shouldn't they be using it?
|
|
|
Post by ethanbein on Aug 16, 2014 22:57:25 GMT -5
Warning: I'm about to show my baseball ignorance once again. I have a fee questions about pitch framing in general: 1) Is the work you're doing, Eric, based on the the camera / computerized strike zones like the ones we see in the graphics on TV? I ask that because I've often wondered how accurate those are. Are they different from park to park due to different camera angles? I read somewhere a while back that park differences can be a problem. Also I understand they need to be recalibrated from time to time and that doesn't always happen. Don't mean to get unto a technology discussion. Just wondering if that's what you use and how good is it. 2) If the automated systems are really more accurate than the home plate umps, isn't that really an indictment of MLB? If they have a way to do it better, shouldn't they be using it? Eric can give a more complete answer I'm sure, but this is all based on pitch f/x, which is the same technology that drives those strike zones on TV. It is optical camera based, but it's an advanced system with special cameras, not just TV cameras. I know some people have found park differences, but they're generally small and I think they are usually corrected for. I don't know if BP does, though. And yes, we should be on computerized strike zones by now, but that's not happening anytime soon.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 17, 2014 0:11:03 GMT -5
Warning: I'm about to show my baseball ignorance once again. I have a fee questions about pitch framing in general: 1) Is the work you're doing, Eric, based on the the camera / computerized strike zones like the ones we see in the graphics on TV? I ask that because I've often wondered how accurate those are. Are they different from park to park due to different camera angles? I read somewhere a while back that park differences can be a problem. Also I understand they need to be recalibrated from time to time and that doesn't always happen. Don't mean to get unto a technology discussion. Just wondering if that's what you use and how good is it. 2) If the automated systems are really more accurate than the home plate umps, isn't that really an indictment of MLB? If they have a way to do it better, shouldn't they be using it? Indeed, sometimes the pitch/fx data gets skewed because the cameras get out of position. Once in a great while you can notice this. BrooksBaseball.net, the best source of pitch/fx data, park-corrects the data (and the pitch-type identification). Since BP has links to them, I think they're using that database rather than the raw data. I'd be very surprised if they weren't. I have been an outspoken advocate for automating the calling of balls and strikes for many, many years. I have altered my position a little bit: I would like to see this implemented by MLB about the time Chrisstian Vazquez's career is winding down.
|
|
|
Post by ethanbein on Aug 17, 2014 18:40:48 GMT -5
I was at Saberseminar in Boston this weekend, and today Dan Brooks gave a talk on his pitch framing model, which is the one that's on BP. As has been said earlier, it starts with a probability that a pitch will be called a strike at each point in the strike zone and gives credit from there. A few other things that he mentioned:
The model controls for the pitcher throwing, so while Koji might be easier to frame than Rubby, that shouldn't show up in the data since it's already accounted for.
The probability of the pitch being called a strike is adjusted for count, batter and pitcher handedness, and pitch type.
The results become reliable very quickly, in just a few games, so we can have good confidence that Vazquez is really, really good at this.
I'm sure there are a few things he mentioned that I'll rember later, but I left feeling very impressed with the model and pretty confident in its results.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 18, 2014 6:43:22 GMT -5
I was at Saberseminar in Boston this weekend, and today Dan Brooks gave a talk on his pitch framing model, which is the one that's on BP. As has been said earlier, it starts with a probability that a pitch will be called a strike at each point in the strike zone and gives credit from there. A few other things that he mentioned: The model controls for the pitcher throwing, so while Koji might be easier to frame than Rubby, that shouldn't show up in the data since it's already accounted for. The probability of the pitch being called a strike is adjusted for count, batter and pitcher handedness, and pitch type. The results become reliable very quickly, in just a few games, so we can have good confidence that Vazquez is really, really good at this. I'm sure there are a few things he mentioned that I'll rember later, but I left feeling very impressed with the model and pretty confident in its results. Thanks for this! Did he say anything about controlling for umpire? That would seem to be more basic than controlling for pitcher. What I've found is that a very slight error in the calibration of expected strikes (in my last analysis, 1.7%) gets amplified and leads to much larger errors in the final data, more on the 10% - 20% side, and it especially leads to errors with guys who had very low or high numbers of expected strikes, so that all of the former show up as good framers and all of the latter show up as bad ones. I think there may be other ways to correct the data than the one I've already messed with. I hope to re-grab the data (which they subsequently tweaked) and re-do the analysis, and then send it on BP, sometime this week.
|
|
|
Post by ethanbein on Aug 18, 2014 11:34:00 GMT -5
I don't remember anything about controlling for umpire, although I know BIS talked a little bit about their proprietary system that used a similar technique to divide credit between the pitcher, catcher, batter, and umpire. I bet if you emailed him directly at brooksbaseball he'd get back to you, and I bet he would be interested to see any other research you have on it.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 19, 2014 18:43:25 GMT -5
Either Vazquez had his best game ever yesterday, or they tweaked the algorithm again. In any case, he's not looking any worse.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,911
|
Post by ericmvan on Aug 26, 2014 9:50:58 GMT -5
After a week too busy with real life, I've updated the main table. The data strongly suggests that BP is continuing to tweak their algorithm. I caught one supposed game's worth of data that clearly looked like it included a correction, then missed a bunch of updates whose aggregate seems more likely to be explained by more tweaks rather than a strange extended slump.
They now have Vazquez at 3.6 WAR per 120 games, which is pretty much what I thought their 4.5-ish figure should be when corrected. I think I'll wait till the season is over to re-test the data for the miscalibration I detected earlier. In the meantime, assume that any day's update might include a tweak to the algorithm!
|
|
|
Post by jmei on Aug 26, 2014 22:46:52 GMT -5
|
|
|
Post by ethanbein on Aug 27, 2014 10:40:00 GMT -5
I thought a bunch about this, and here's what I came up with: 1) It wouldn't matter that the pitcher can adjust and throw more borderline strikes to a good framer if every pitcher adjusts the same. If every pitcher optimized his pitch location based on the probability that each pitch would be called a strike, the only differences would be between *catchers*, and therefore all the credit should go to the catcher. Now, obviously that's a little extreme, and it stands to reason that some pitchers adjust better to good framers than others do, so the credit that pitchers get should be based on the spread of this skill among pitchers. How to measure that, I have no idea, but I'm sure someone much smarter than me could do some sort of pitch f/x study. But you can't have some sort of generic "70/30" kind of split that Dave has suggested. 2) If we're going to believe that pitchers are going to pitch more outside the zone when they have good framers, doesn't that mean framing is even more important than we've acknowledged before? Everything is based on called strike percentage, but everything else is better for pitchers outside the zone, like swinging strike %, weak contact, power, etc. Think about it in the case of a weak framer. If a pitcher has to pitch in the zone more because he feels he can't get the call on the corner, isn't it the catcher's fault if he gets hit a little harder? Therefore, shouldn't a guy like Lucroy also get a little credit for the swinging strikes and weak contact his pitchers produce? Obviously that's hard to sort out and the effect is probably small, but if you believe Dave, it seems like you have to take that step logically. This thread over at Tango's blog about this article is very good. MGL explains my thoughts a lot better than I do.
|
|
|
Post by okin15 on Aug 27, 2014 14:17:53 GMT -5
1) It wouldn't matter that the pitcher can adjust and throw more borderline strikes to a good framer if every pitcher adjusts the same. If every pitcher optimized his pitch location based on the probability that each pitch would be called a strike, the only differences would be between *catchers*, and therefore all the credit should go to the catcher. Now, obviously that's a little extreme, and it stands to reason that some pitchers adjust better to good framers than others do, so the credit that pitchers get should be based on the spread of this skill among pitchers. How to measure that, I have no idea, but I'm sure someone much smarter than me could do some sort of pitch f/x study. But you can't have some sort of generic "70/30" kind of split that Dave has suggested. 2) If we're going to believe that pitchers are going to pitch more outside the zone when they have good framers, doesn't that mean framing is even more important than we've acknowledged before? Those are exactly the thoughts I came up with too. It's one of the few positions where the player's value becomes exponential once the pitchers know about it. Does the statistic subtract points for guys who miss strikes in areas where there is some chance of a called strike (eg. .3 points in the 30% mentioned by the blog)? I assume so, but am just trying to think about where the crazy numbers might be coming from.
|
|
|
Post by jimed14 on Aug 27, 2014 14:44:03 GMT -5
One could easily suggest that Lester purposefully took advantage of superior pitch framing this year and probably a lot of last year. It really seems to help a guy with a great cutter like his.
The other exponential factor could be batters widening the strike zone because they think there will be more borderline strikes called.
|
|
|
Post by mgoetze on Aug 27, 2014 14:44:21 GMT -5
Didn't make any sense to me on reading it, made even less sense after reading MGL's comments. Dave Cameron's got this wrong.
|
|
|
Post by burythehammer on Sept 1, 2014 10:46:59 GMT -5
Just wanted to point out that Clay Buccholz has a sub 2 BB/9 while being caught by Christian Vazquez. Kid is so good that I might actually buy that Shields/Clay/Kelly can be our top of the rotation next year.
|
|
|