I think sabermetricians have to first come to grips with this concept -- that certain sporting activities trust to luck, much more than others do -- for the conversation to move forward smoothly. As it is, the operating assumption :- ) seems to be that every athletic task should inherently involve the same degree of luck.
Experience should lead us to expect that every athletic task requires a different, sometimes much different, element of luck.
Pascal's Triangle comes strongly into play when soccer and hockey players are shooting at goal -- and very strongly into play when batted ground balls leave the hitter's bat.
..................
I could be wrong. Presented strong enough evidence, I'll change my position. But experience tells me that in baseball, individual defense involves very little luck. Find a guy in cleats who will tell you that his defense involves much, if any, luck? :- )
Or, explain to me in what sense Pascal's Triangle caused Raul Ibanez' -20 UZR in Seattle and +0 UZR in Philadelphia.
..................
Certainly as mathemeticians, we understand that in a given year, x% of balls into a certain zone might be reachable. There is *some* luck in UZR, because of the way UZR is calculated (not because of the way fielders play).
I question to what extent Raul Ibanez got balls "in the fringes" of his Seattle LF zone, to the tune of a -20 runs UZR -- and then got a lot of BIP's dead-center into his LF zone once he reached Philadelphia, to the tune of an above-average UZR rating. I assert that this hypothetical "UZR luck" factor is probably not a strong one.
I assert that we should listen to Raul Ibanez' observation of the phenomenon: that in Seattle, the Mariners did a lousy job of positioning him.
In any case, the problem remains. In my view, and present company excepted, sabermetricians should keep this self-evident principle in mind. Some athletic tasks involve more Chaos Theory than others do. And there is no harm in asking the athletes which those are.
.
Voros McCracken realized, a few years ago, that the balls coming off of hitters' bats were, to a remarkable degree, producing a Pascal's Triangle of results.
Sabemetricians found this concept compelling... no, not compelling: hypnotic. The idea that you could simply apply Probability Theory to understand so much of what was going on in baseball, mesmerized analysts.
It mesmerized them so much, that it seems they didn't stop to realize that very few activities in sports can be charted along a Gaussian curve as smoothly as BIPs are. There just aren't many things in sports that are as non-skill-driven as is the direction of a baseball coming off a hitter's bat. :- )
Cheers,
Dr D
Well luck would come in the form of BABIP. All chances aren't equal for defenders and neither are the number of opportunies and difficulty levels in short samples.
There is definetly a problem with zone metrics regarding positioning or covering for an inferior defender but this is usually captured in OOZ and Dewan's plus/minus and PMR give you the extra detail to be able to evaluate this as well.
http://www.lookoutlanding.com/2010/1/26/1271512/on-the-shoulders-of-gian...
Fett posted that over at LL a while ago.
Its a nice little tutorial on some of the more recent saber discoveries. You can scroll down to defense for some more info.
Its best to know the enemy. :-)
That's got to be the best index-primer on the internet.
Why 'enemy,' though? :- ) Truth is truth.
I don't know.. It just seems like a lot of people at SSI/MC are anti-UZR.
Once you really fully understand it, then you can pick apart at its many weaknesses.
I think its a valuable data point, though I agree with you and Sandy that you'd be nuts to rely on it as your only data point to evaluate D. Thats basically the same thing as relying on ERA to evaluate pitching or on OBP with hitting.
One of the issues I was intending to emphasize in my 'luck' post, (which I think I did a lousy job of), is that BIPA and DER are (by and large) measuring exactly the same thing -- the ratio of "fell in" to "caught". You remove HR, K, BBs ... what you end up with is "fell in" or "caught", (in some form).
I think where I was headed, (but my verbosity prevented) was in noting that on a team level, the total opportunities grow to a point where DER (or BIPA allowed) become extremely difficult to be overly influenced by luck. 3,500 outs in 5,000 chances (as a theorhetical team result), makes getting lucky or unlucky hard to swallow.
But, any given fielding position divides that number by 9 (obviously influenced by position), and you're looking at 400-700 chances for an individual defender. THERE, the possibility of luck screwing up results gets a LOT more possible.
Well, the number of fielding chances (for individual fielder), and the number of BIP chances for an individual hitter are essentially identical. K-rates and spot in the batting order impact chances for hitters, while whether you're playing short or RF change things for fielders. But the scale is roughly equivalent.
The conclusion is: when discussing TEAMS, (or leagues), DER/BIPA should each be treated as largely luck-free.
When looking at individual players ... luck "can" be a significant factor in individual cases, but it's not a good idea to immediately assume that every oddity is either luck or skill. The reality isn't that simple.
My gripe is that many of the same people will look at a BIP result for an individual player and say - "that's luck" - we can expect him to return to normalcy in the future ... but, when discussing UZR, the result is treated as "luck-independent" -- where, no matter how severe the skew from normalcy, the instinct is to assume a change in skill level.
Both assumptions are incorrect. The truth is, because the individual results for BIP and UZR are volatile, when there is volatility, you CANNOT know how much luck/skill is involved.
There are those that view Junior's .186 road BIPA as evidence he's due to bounce back. Yet, we know with Sexson that his off-the-charts bad BABIP was an indication that he had actually swooned BELOW the minimum sustainable athletic level to remain a major leaguer. Junior is a career .288 BIPA guy. He went .269 in '08 and .220 in '09. But, because his '09 results were skewed so bad on the road, the "analysis" suggests that he's due to bounce back.
Mind you ... the statistical analysis cannot take into account his off-season workouts, and losing X pounds, etc., etc. But, for any other player, whose BABIP had dropped so drastically for two seasons, (and turning 40), would any analyst be so willing to play the luck card?
IMO, too many are too quick to play the luck card when viewing BIPA for hitters, and not willing enough to consider the luck card when viewing UZR.
Team DER, and individual UZR, those are two completely different questions.
Yes, we all realize that % of balls caught are caught by several different metrics: BABIP, H%, DER, UZR, etc.
DER is definetly going to be more accurate year to year since theres a lot less noise. Its important not to treat it as 100% skill though.
You do get situations even offensively where you have a team-wide lucky BABIP or unlucky Rs scored. The '09 Angels come to mind as the most extremely lucky BABIP team last season. The '09 Mariners were unlucky in Rs scored.
Saying "runs scored was below expected runs generated, given a team's linear weights" does not equate directly to "they were unlucky".
There is an ASSUMPTION that the Angels had a 'lucky' BABIP in 2009, and that the Ms were unlucky in Runs scored in '09 -- based on standard analytical model. But, "luck" and "we cannot account for this (today)" aren't equal.
Part of the problem I see is that too many "standard" traits get to be ingrained and every outlier becomes pegged as luck. Well, how many times do you have to be "lucky", before the analysts will pause, scratch their chins and say, "well, maybe there is something going on here that we aren't accounting for."
The Angels in 2009 were a nearly unique club (offensively), in that they had TEN (10) players all hitting over 100 OPS+, but lacked the 2 or 3 150 OPS+ mashers typically found on high scoring teams. It is HARD to find clubs in history with offense as evenly distributed up and down the lineup as the Angels. If your "comp" set is 4 or 5 teams in the past 100 years, (I'm saying as an example - not as gospel), then you don't really have a reasonable statistical comparison.
Current methodology typically throws all the numbers in one hopper, and gets statistical saugage out the other side. But, in a game where ORDER matters, having two clubs with identical .350 OBP and .441 slugging "might" have very different runs scored, if one had a trio of 150 OPS+ hitters, while the other had a much flatter distribution of hitters from 1-9. But, as a general rule, teams with similar aggregate stats "tend" to have similar talent distribution, so they "tend" to behave in a particular way.
Were the Ms "unlucky" in runs scored --- or was there run scored shortage related to the fact that for the season, their 7th best hitter was batting in the #2 slot in the order? They had a .710 from the 6th spot, and a .673 in the 2nd. What if you reverse just those two bats? What if the best bat, (.837 in the 5th slot), had been hitting 3rd instead?
Most teams put their best hitters higher in the order. More chances. The Ms got their best production out of the #5 slot, (actually tied with #1). Should that be classified as "luck"? Or should it be classified as a fixable tactical error? If they score runs in line with their linear weights in 2010, it will be "perceived" as luck. But, if the club has its top 5 hitters batting in the top 5 slots in the order, then there might be a specific reason for the change - not only identifiable, but also very much in control of the club.
I tend to distrust the word "luck" in regards to analysis, because I believe it assumes a state where not only is the outcome unexplainable - but that it will ALWAYS be unexplainable, (and completely out of the control of the people involved).
If the wind blowing out/in ratio in Wrigley is 60/21 in 2009 and 21/60 in 2010, you've got a variable that's out of the control of the people. But, it would be relatively easy to quantify the impact of said wind, and adjust stats accordingly. Or, you can just say in 2010, the offense was "unlucky".
Sandy...you might be right about the offense not being "unlucky" last year...there might have been a problem with line-up balance that prevented the Mariners from scoring as many runs as their linear weight RC or their OPS would suggest. However, I still factor those 50 runs into my offensive projection, because the character and composition of the line-up has shifted in such a way that there probably won't be a horrible black hole anywhere offensively except SS and it's unlikely that the Ms will have anyone batting 4th who does as poorly as their clean-up spot did collectively last year. So they're not just gaining in LWRC by adding players...they're gaining that much plus the 50 runs they're likely to get back by having a more balanced, patient offense this year.
And this is precisely why "assuming" luck can be problematic. I think you're likely correct ... that the 2010 lineup balance will be more typical, and as a result, the 50 "missing" runs will reappear. But, if you understand that this is case, then management can work to prevent this from happening, (like by not bothering to sign the slugger who balks at hitting cleanup -- or by saying 'no' when some future slugger says he doesn't want to hit cleanup and you don't have a better option).
... as we know, the M's two luck factors Pythag (+) and runs/OPS (-) essentially canceled last year, and I'm not sure either one was exclusively due to random fluctuation...
My hero Griff had a saaaaaaad BABIP last year, but am not sure that either was exclusively due to random fluctuation.
A player gets into a slump, he's taking (relatively) defensive swings -- at the very least, dialing back his ambitions a little -- and that will affect what saberdudes consider 100%-luck metrics.
Couldn't agree more.
Is quite a bit different. Figgins 2 and Bradley 3 really attack a pitcher, in the sense of going up there with bad intentions, getting their pitches and doing something with them.
Last April's lineup went up there hoping for mistake pitches. Here comes a pitcher's pitch, and bang, there's a ball weakly put in play.
Would tend to suspect that whatever abuse last April's lineup suffered, it was well deserved, including 'soft' runs per OPS. But fortunately, the lineup is evolving...
I did a line on Griffey showing that while he was unlucky he deserved a sub .250 BABIP last year.
The Angels' BABIP was legitimately lucky. They had players across the board who were outperforming their career BABIPs. They also performed extremely well in the clutch.
The Ms naturally were unlucky as well considering they didn't hit well "in the clutch". You could argue that this is partially skill too though.
Hey Doc,
Go check out the splits for these HoFers as they were winding down toward retirement. "Signs" typically begin showing up around age 36, followed by adjustments (usually seen in fluctuating patience), and eventually turning into unsustainable BABIPs. The precise timing isn't set in stone - but the pattern repeats with a LOT of high grade players, (and I suspect that the same is true for some tier two guys).
Though I do remember Sullivan posting some GB/LD/FB breakdown that gave Griffey credit for more actual bad luck than I thought he'd had.
Griff did hit some weak balls last year, no doubts there.
Am I going to see declining BABIP's with those three guys? (it's 4:17 a.m., LOL)...
Probably not as much as Griffey got, right? Although Willie's last year or two was quite a sight... :- / were you watching baseball during Mays' last gasp?
Okay ... here's the BIP swoons for FR, HA and WM. Career BIP listed first, then age-BIP for final few seasons
Robinson: Career = .295;
34 - .312
35 - .268
36 - .265
37 - .271
38 - .260
39 - .200
40 - .226
Hank Aaron: Career = .291
37 - .291
38 - .235
39 - .256
40 - .242
41 - .238
Willie Mays: Career = .300
38 - .313
39 - .303
40 - .339
41 - .282
42 - .242
Griffey - Career = .288
35 - .305
36 - .248
37 - .284
38 - .269
39 - .220