Wednesday, April 17, 2013
Granillo: “Rise of the Robot Umpires”
(www.baseballprospectus.com)
It troubles me that the world is just one big rip-off of Ferro Lad.
Read More...We’ve all thought it. Heck, many of us have even said it. Watching the exploits of Eric Gregg, Phil Cuzzi, Jim Joyce, and so many others, it’s hard not to believe it at least once. “We want robot umpires!”
A new project by Dan Levy is working to explore just what might happen if baseball ever heeded that call. Rise of the Robot Umpires is a graphic-style novel set in the not-too-distant future where the commissioner finally ...
Reader Comments and Retorts
Go to end of page
Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.
Page 4 of 8 pages
< 1 2 3 4 5 6 7 8 >You can sort this by oWAR. However, oWAR is by no means immune to criticism, since it includes park factors that are questionably computed and fielding/baserunning calculations that have both conceptual and computational issues. From a back-of-the-envelope examination of all that, it's very plausible to conclude that a "perfect storm" of such computational events have aligned in such a way as to boost Trout from the level of MVP candidate to all-time top 25 seasons. While only a few folk have seriously attempted to argue from that point of view, the fact that the method does that means that at the very least it needs to be re-theorized.
There's no way in hell Trout's (partial) 2012 is 25% better than Barry Bonds's 1992, when he OPS+d 204 at the peak of the last deadball era, stole 39/47 bases, and had Gold Glove CF talent that had been relegated to Gold Glove LF actuality.
They have equal oWAR, even though Bonds's OPS+ was 36 points higher and Bonds was a great baserunner and basestealer by any reasonable measurement.(*) That's absurd on its face.
(*) Bonds led the major leagues in OBP, SLG, and OPS in 1992. Trout isn't within 40 points of the AL SLG lead, and isn't within 20 points of the AL OBP lead. Yet Trout is actually at 8.3 oWAR to Bonds's 8.2 for 1992? Please.
Really? Ok. Have to disagree, Albert's defense in left was solid to good. Given a choice between Albert in the left or Holliday in left, Albert was the better fielder, and to my knowledge Holliday is considered to be at least average defensively.
Might be the most perfect replacement level season ever. He's just bad enough that when you have someone come off the DL or called up from the minors, he's the one you don't have room for on the roster anymore. But just good enough that somebody else has room for him.
Bonds only played 140 games. Just noting that. Also, according the B-R fielding numbers, 1992 is when Bonds was making his transition from excellent fielder to average fielder. There's no doubt that this transition happened, the only question is when.
Lucky for Trout that WAR stands for Wins Above Replacement and not Wins Above League Leaders. It is possible to be both a league leader and closer to the average than some other non-league leader in another place and time.
I doubt it happened at age 27.
Is it absurd that Sammy Sosa had more oWAR in 2001 than Dick Allen in 1972? Allen led the majors in OBP and OPS (missed the SLP by .003), and Sosa was over 200 points lower than the leader in OPS.
You might have a point there but we can all do better than just throwing out statements like "no way in hell" and "absurd on it's face." BB-ref has all the components right there to break down. If it's absurd, then where is it wrong?
Trout is +51 bat, +10 running, +22 fielding, -1 position and +1 GIDP.
Bonds +61 bat, +4 running, +6 fielding, -6 position and +0 GIDP. He played 140 games, which is 1 more than Trout will play (unless the Angels give him a day or two off after Martin Perez pitches the A's to a playoff spot tonight).
So: Bonds is the better hitter. Is +10 runs enough of a gap for the 204 to 168 OPS+ difference? I'm not sure if that's normal or an anomaly.
Baserunning: It's more than SB, so you'd have to look in the details on bbref, but 48-4 on basestealing is definitely better than 39-8.
Position: Bonds played the whole year in left, Trout played more CF than left. Advantage Trout.
Fielding: Bonds is +6 but averaged +28 the 3 years prior. Use multi-year fielding averages and Bonds could be +20. Make that adjustment and he's pretty much even with Trout in WAR.
Gosh, you know how to hurt a guy.
That's pretty absurd.
And the "where is it wrong?" inquiry misses the fact that each of the components could be "reasonable" yet simply wrong in a perfect storm sort of way, such that when you add them up you get nonsense. That may not be the case all the time, but it's certainly the case some of the time, and "some of the time" - especially when a lot of players are involved such as in a sorting of 2B over the past 50 years - can be a big mess.
This is why it's problematic for a stat to have layers upon layers of questionable entries, at which point "Oh? Where is it wrong?" becomes a flawed inquiry. The flaw is a structural one: there are too many questionable layers to an uber stat.
Bonds only played 140 games. Just noting that.
He actually had fewer plate appearances than Trout does this year.
Let's just ignore the fielding component:
Bonds comes out 10 runs ahead by wRAA. This is basically wOBA with a park and league factor adjustment. Fine if you don't trust park factors, but they are what they are.
Bonds was an excellent baserunner that season - add 4 runs for him in baserunning. Trout is an exceptional baserunner, 10 runs better than average. Trout gains six runs there.
Now, replacement value. Higher for a LF than a CF. Trout gets another 4 runs there.
So, ignoring double plays, this puts them even.
If you want to tell us why you don't trust wRAA or wOBA, please go ahead. Maybe there's something up there for these two guys or these seasons. "Absurd on its face" is not a valid argument.
PEACE
Perfect
Evaluator (of)
Actual
Comprehensive
Excellence.
This should get the old 1960s protest crowd on board.
I was focusing on offense primarily. Trout has more oWAR this year than Bonds did in 1992.
Maybe 10 runs for pure batting isn't enough to capture the secular difference between a 205 OPS+ season and a 168. Those are fundamentally different accomplishments; one is excellent, the other is rare and ultra elite. Bonds's 1992 is one of the five or so best post-integration, pre-Selig Era offensive seasons (*). Trout's 2012 is nothing like that.
(*) Arguably, the very best since he blew away #2 by 29 OPS+ points.
That's because the defense component of WAR is treated with precisely as much confidence as the offense component.
And it simply is not the case that we are anywhere near as sure of defense as we are of offense. Nobody looks at Adam Jones's EqA and is like, ".299? WTF?"
he OPS+d 204 at the peak of the last deadball eraThat's addressed in the OPS+, and Bonds shows a 10 run (1 win) advantage in hitting. However, Bonds was a LF (where you expect better hitting than CF); thier oWAR are the same (8.2 vs. 8.3).
stole 39/47 bases In '92 NL play, the average team attempted 192 SB. Trout has stolen 9 more bases, at a much better rate, in a league where the average team is attempting only 141 SB. That's significant.
and had Gold Glove CF talent that had been relegated to Gold Glove LF actualityUZR shows Bonds as pretty pedestrian in '02, and he had only one more "Gold Glove talent" year after; BIS shows a 3-year "valley" in his defensive numbers from '90-92, followed by a 4-year return to "Gold Glove talent", followed by being average or worse - and that's compared to LF, not CF. Trout, on the other hand, is +13 runs in UZR and +21 runs by BIS - compared to CF.
But I think this brings another issue up - why, when offensive stats disagree with our eyes we believe the stats but when defensive stats disagree we dismiss the stats out of hand? Defensive stats may not be perfect (or even good enough) yet, but isn't it also likely that sometimes players just aren't as good defensively as we think they are?
Maybe the 204 OPS+ is inflated by 32 intentional walks. Take those out (and do the same for the NL batting line as a whole) and Bonds's OPS+ goes down to 197 - not a huge drop, but a drop. Do the same for Trout and the 2012 AL, and his OPS+ stays the same.
Really? I've never heard of it and I'm a native who has lived almost his entire life here.
(*) In both years he played CF more than 3 games, Bonds beat the CF league range factors by more than Trout is beating them by this year. Frankly, I'm not seeing what's so special in Trout as a CFer in the range factors (understanding that range factors aren't everything.) In 1987, Bonds played 46 games in center and beat the league RF/9 by 30 points. Trout's advantage this year is 14 points.
2. If we're going with range factor, Bonds's RF in 1992 is better than the average left fielder's, but not by so much that he's clearly a "Gold Glove CF talent." He beats the average LF in '92 by less than Trout's margin over the average CF this year.
Using the simplest form of runs created possible, the difference in their season is about 14 runs. Park and era adjustments would easily make it 10.
Now, Trout has more PAs/G (due to hitting leadoff), but the 10 runs is not an absurd number.
The purported difference in Mike Trout's fielding and a 27-year-old Barry Bonds's is (a) enough to wipe out the difference between a 205 and 168 OPS+; plus (b) another 60%.
That simply can't be right. The scales and magnitudes and impacts of the various components are fundamentally mismatched.
Because defensive stats can be messed up by positioning of the defense. A guy who is playing out of position due to bad coaching, but who has great range, is still going to be hurt by the numbers. It's one of the knocks against Zobrist is that he is being helped by his defensive positioning/coaching. Mind you, on a strictly value method, the player who is getting that advantage is earning the value.
Park and era adjustments would easily make it 10.
Again, the scales are off here. There's no way the differences in those stadiums and 2012/1992 shave 30% off that differential -- particularly when park/league/era are already accounted for in OPS+.
You might be right. 2012 is actually a higher run-environment than 1992 (5% across MLB) and that would counter more than half of the park effect. Maybe the IBBs are worth 3-4 runs that are being dropped.
gets.
It doesn't mean the outfielders are taking the CF's chances, or anything close. CFs typically take chances from corner outfielders, since they take essentially everything they get to.
http://sports.yahoo.com/news/10-degrees--miguel-cabrera-s-triple-crown-push-overshadowing-mike-trout-s-mvp-season.html
We've tread over this ground for 10 years, but, obviously it's because the offensive stats count REAL things (singles, homeruns, etc) and we're not sure if the defensive ones do or not. You cannot deny that Neifi Perez just hit a homrun - but was that catch that Pete Incaviglia made a tough one? Who knows?
There's a conceptual mismatch there, too -- we aren't crediting hitters only with "tough" singles or other hits, but we are with fielders. So we're combining the measurements of two things that aren't really the same. Maybe that's why the scales of measurement don't really work in combination.
Then you need to re-watch your Three's Company re-runs. They dealt with this subject in a 1978 episode entitled Days of Beer and Weeds:
Roper: Good, good, then you'll clear out the garden tomorrow. Thank you.
Jack: (dismayed) Yeah.
Janet: (sighs) Well, goodbye weekend. That garden is right out of tarzan of the apes.
Chrissy: It is a little overgrown.
Jack: A little? Chrissy, there are pockets of Japanese in there who don't know the war is over.
Defensive stats count real things, too, just like RBI. And just like RBI, we don't trust them to stand alone as a measure of quality. But unlike RBI, we don't have anything else to put around them.
If Albert Pujols makes a game-ending play on a hard grounder up the first base line, it keeps one of his teammates from having to make another game-ending play later in the inning. If Prince Fielder doesn't make that play, another Tiger will have a chance to make another play later on. Whether the chance is being directly stolen from another player on the specific play in question isn't what I'm talking about.
How often does a fielder get a chance to do something better than a replacement-level fielder? If it's 100, which seems extremely high (*), then fielding should have roughtly 1/6 the magnitude of hitting. (You'd have to tweak the formula based on the type of hits saved by a measured play -- bad OF's "hit" triples by turning outs to triples; IFs really don't. Their screwups are typically single-like).
(*) Mike Trout has had 265 chances total in CF this year.
(1) Invariable in the sense that it's going to be somewhere between 24 and 27, and usually exactly one of those two, in any given team's game, with few enough outliers from extra-inning games not to make very much difference in the long run. A team can score as many runs as it can get, and the difference between a 4-run league and a 5-run league is very large, percentagewise; but the amount of outs teams make is all but inflexible, and has been since forever.
2012 Trout has 132 RC and 52 Adjusted batting runs. Roughly the same as Cabrera, Roughly the same as 2nd place in the 1992 NL.
So, what's interesting is that the difference is only 16 RC and 23 ABR. So "percent of the difference" has no merit as the baselines are note equivalent. Ditto with any replacment level factor. OPS+ is not linear. Runs are linear. The difference in offense between Bonds 1992 and Trough 2012 is 15-21 runs. I don't think either of those are park adjusted, but let's call it 20 - 2 wins (also linear except for extremely low values of runs allowed). This is actually a vast margin. The difference between a no-brainer MVP and a close race.
So, can you get 20 runs back on position, defense, and baserunning?? CF vs. LF is probably 10 right there.
Range factor doesn't show it. To appreciate Trout's defense you have to A) trust Dewan's metric or B) watch the games and see him jumping over the walls to save homers, and running in and diving to rob singles.
True that you can't do a direct comparison for B, 20 years apart. I'm sure Bonds made some great defensive plays in 1992 but I can't remember a single one.
Or you can do a little more work and look at the other OFs on the teams. If you do, you'll see that the Pirates' other CF in 1987 (*) -- the year I'm talking about --beat the league RF by less than Bonds did when Bonds played CF. And if you look at the teams' primary right fielder, RJ Reynolds, you'll see that his range factors trail the league's badly in both right field and left field.
So the balls were hit to the Pirate OF, some players got to a whole bunch of them, and other players didn't.
You'd likely see the opposite effect if you compared Derek Jeter to other Yankee shortstops and Derek Jeter vs. league compared to Other Yankee IFs vs. league.
(*) Andy Van Slyke, highly regarded defensively. Five straight CF gold gloves, 1988-1992.
How often does a fielder get a chance to do something better than a replacement-level fielder? If it's 100, which seems extremely high (*), then fielding should have roughtly 1/6 the magnitude of hitting.
That's not really how it works, though. Even the best hitters make outs in at least half of their PA, Bonds excepted, and replacement level hitters don't make outs all the time. The difference comes in (1) the number of plate appearances that have positive outcomes, and (2) the magnitude of those positive outcomes.
Comparing, at random, Jim Thome and Neifi Perez in 2002. Thome reached base 269 times, Neifi 151 (Thome had 28 more PA). Adjusting for the PA difference, that's only about 111 extra good outcomes for Thome.
Of course, 49 of those were home runs, so there's that.
Range factor would be a greatly improved, as a simple and basic metric, if it were displayed as "player outs per 30 balls in play" or something like that instead of "player outs per 27 team outs".
How many "robbed" hits do you figure Trout has this year?
How about adding robbed hits to offensive production (*), subtracting "failed" robberies, and calcluating a slash line? That would make some sense.
(*) If a SS robs a single, he's deemed to have hit a single.
I'm confused - are we talking about Bonds's 1992 season, or his 1987 season?
Edit: FWIW, TotalZone thinks Bonds was awesome in '87, +24. It does not think he was awesome in '92.
And even the worst defenders make putouts (or IF assists).
So you either throw everything in with fielding, as we do with hitting, or we have to find some way to effectively scale the fact that fundamentally different things are being measured in hitting and fielding.
The idea of 'replacement level' doesn't really work for defence, except to indicate the player who will be taken out in the late innings of a close game.
A replacement-level hitter might well field his position better than about a third of a league regulars.
That's at least -- at least -- double the number of extra good outcomes for comparably capable fielders.(*) Moreover, Thome's good outcomes included 49 homeruns and a bunch of extra base hits. Other than the occasional robbed home runs, fielders can't "hit" homeruns (by aaving them). Middle infielders really can't even "hit" extra base hits, only singles.
(*) Probably more like quadruple.
Page 4 of 8 pages
< 1 2 3 4 5 6 7 8 >You must be Registered and Logged In to post comments.