quote:Originally posted by PGStaff:
"Sabermetrics, I believe, look at each AB and each game as being equal. In other words a game in April is exactly the same value as a game in a pennant race during the last week of the season. I disagree with that some what, but understand the logic. However, that is like saying what Reggie Jackson did in October was no more important or valuable than what he did in April."
For the most part, I think the first part of this is true. I don't like the analogy (it is more like saying what he does in April is the same as what he did at any point late in the regular season, but not the playoffs, which is what I associate with "October"), but I understand the point and agree with it to some degree.
There are good "sabermetric" stats that measure contributions to winning, though. I've already talked in this thread about how much I like the stats RE24 (which stands for the Run Expectancy for each of the 24 base-outs states) and WPA (which is Win Probability Added). I already talked about RE24 in this thread here, and 2013LHP did a nice description of WPA here. Both of those posts have good examples of how they work, but the main difference between RE24 and WPA is that RE24 measures the positive or negative contributions to run expectancy while WPA measures the positive or negative contribution to win expectancy, and RE24 looks at each base-out situation as the same regardless of what inning it occurs in or what the score is when that situation arises, while WPA additionally (and very intentionally) considers the "leverage" or impact of that situation. For the kind of thing PGStaff is talking about, WPA is the better and more relevant stat to emphasize as an example of how sabermetrics "gets it."
Anyway, suffice it to say that anybody who values and pushes a stat like WPA as much as as I do is at least sympathetic to the argument PGStaff makes. Some wins are "higher leverage" in terms of getting a team to the promised land - I get it - just as the results of some plate appearances have more to do with producing an individual win than others. I do think you have to work harder than to quote BA or RBI in wins, or with two outs and RISP (which is only one isolated kind of high leverage situation, of many) to make PG's point about "high leverage" wins, but you can get there - and I will. You just may be suprised at what it shows, though.
To PG's first point, I do think it is true that most (but certainly not all) "sabermetric" stats like look at things equally. This is mostly because they are designed not just to look back but to estimate future performance. If you are looking back, and trying to determine value added, then you want to know the difference between a home run in the 4th inning of a blowout and a walk-off home run. However, if you are trying to predict future performance, that difference doesn't matter as much because it has been repeatedly proven that "clutch" hitting is not so much a repeatable "skill" than it is essentially a reflection of a player's overall skill. If you look at "clutch" stat leaderboards year-to-year, players have wild variation, and over multiple years, those stats tend to look more and more like their overall stats - the best hitters are usually the best "clutch" hitters. So, when looking for predictive quality, stats that are leverage-neutral tend to be better predictors.
However, when your goal is to look backwards and look at value added, such as in an MVP or Hall of Fame discussion, those kind of "neutralized" stats (or even averages) are less helpful. You want to know who got more "clutch" base hits or extra-base hits, runs, RBI, stolen bases, etc., and when they got them - even if you know that the fact they got them last year won't tell you very well whether the same guys are likely to get them next year. That's exactly what WPA does, and why I think it is a particularly good stat to look at when considering who should be MVP.
WPA measures different actions within a game differently depending on how "high leverage" they are in terms of making it more likely that a player's team will win - whether they occurred early or late, or blow-out vs close game, giving more "credit" for things people might consider "clutch" and which contribute more to a team's probability of winning, and therefore does a good job of measuring an individual player's contribution to those wins, retrospectively. I know that only gets us part of the way to what PGStaff is talking about, and that it doesn't give added value for "high leverage wins," but I am not aware of any stats (traditional or sabermetric) that give "extra credit" for performance in wins only, or in wins only down the stretch, or in wins only down the stretch but only if your team got to the playoffs, or that weight games that are higher "leverage" in terms of a pennant race, etc. I suspect that can be done, and maybe has been done by somebody but just isn't yet widely available. If it were, I'd have NO trouble embracing that kind of stat, because I know (and respect) that most people - including me, to an extent - think that is very important.
That said, I think it is also important to add that in this year's MVP discussion, I doubt that some kind of "Playoff Probability Added" sort of stat - something that would consider "high leverage" wins in addition to how a player performed in "high leverage" situations within individual games - would add that much to the debate, because both Mike Trout and Miguel Cabrera were playing very competitive games and competing for a playoff spot right until the very end (or very close to it).
If you were to measure something like "Playoff Probability Added, you have to be *really* careful to separate individual from team contributions. Right now, we have WPA to do that on an individual win basis, and since ultimately, what gets a team to the playoffs is wins (all of 'em), that's where you need to start. Until something like "Playoff Probability Added" is invented, you have to look at individual game WPA during whatever you consider to be the "more important" stretches of the season, in order to measure what PGStaff is talking about when he talks about those "higher leverage" or "more important" wins.
So that's what I did. I arbitrarily decided to look at August 15 on as a proxy for "the stretch" because I had to choose something and that actually proved to be a pretty good starting point. Both the Angels and the Tigers had played 117 games to that point, with 45 games left for each. Their records were almost identical at that point (the Angels were 61-56, and the Tigers were 62-55). Neither team had their divisional lead (the Tigers were 2 games back of the White Sox, and the Angels were 7 games back of Texas), but both the Tigers and Angels were right in the thick of the Wild Card race (1.5 and 2.5 games out of a playoff spot, respectively). So starting with the August 15th games, I look at Cabrera's and Trout's WPA overall for the rest of the season, as well as in each player's team's wins, and in each player's team's losses. I figured that, especially if viewed together with each player's overall seasonal WPA, that kind of analysis would do a complete job of measuring each player's offensive "value" down the stretch.
I'll get to the results in a sec...first, a detour to address PGStaff's next point.
quote:"Getting back to whether one thinks games near the end are more important than games played earlier. For those that feel the last 2 months are more important than the first two... They have a legitimate case for Cabrera being the MVP. Also, here is a stat that really boggled my mind. In the Tigers 86 wins, Cabrera hit .368 with 34 HRs and 95 RBI.
"And how about this stat. Cabrera hit .420 with 2 outs and runners in scoring position with a 1.211 OPS. Trout hit .286 with 2 outs and runners in scoring position with a .782 OPS. No matter how one looks at stats, that shows some real 'clutch' hitting by Cabrera."
Actually, as I said before, whenever you slice and dice stats from a single season, you inevitably create a sample size that is small enough that it doesn't mean as much. Obviously it happened, and reflects value actually produced in the smaller samples you're looking at, but it doesn't mean that value produced is a reflection of real differences in talent because the smaller the sample, the greater the chance that any differences observed can be the result of simple chance and random variation rather than talent. And it doesn't really break out "high leverage" wins or ABs from lower leverage ones, anyway. Also, quoting BA and OPS (especially OPS without correcting for park factors and competition, because this was a HUGE difference in any comparison of stats between Cabrera and Trout this year) in 2 outs / RISP only (a relatively small subset of high-leverage situations we might describe as "clutch") is particularly vulnerable to this criticism. It is interesting, but not really all that meaningful.
Beyond that, with two great players having MVP-quality years, I would fully expect that you'd see terrifically better numbers in their team's wins, because they are likely to have been a big part of that. But if you ONLY focus on wins, you'll not only get a smaller sample (and less meaningful results), you will also ignore positive contributions that these kinds of players make - plays that made their team's chances of winning more likely, sometimes by a lot - in games their team eventually lost (as well as ignoring their negative contributions to their team's losses as well as wins). That's crazy, IMO - that kind of slicing and dicing isn't helpful to understanding the complete picture.
So, I think the methodology I came up with is much better, and more meaningful. Here's what it showed:
First of all, Mike Trout had a higher WPA for the entire season than Miguel Cabrera did. Trout led the AL (and MLB) with 5.32 WPA, while Cabrera was third in the AL with 4.82 WPA (behind Trout, and his teammate Prince Fielder's 4.93 WPA) and fifth in MLB (behind Andrew McCutcheon's 4.85 WPA as well). [Trout also led Cabrera in the more leverage-neutral measure of individual players' contribution to run scoring, RE24.]
This is significant, because WPA is a cumulative, "counting" stat, not a rate stat - you can't look better by producing at a higher rate for less than a full season. Basically, WPA looks at the probability a team has of winning in every conceivable score, inning, and base-out state. You accumulate positive and negative WPA with each plate appearance depending on whether the result of each plate appearance either advanced or diminished your team's chance of winning, and higher leverage situations count more (again, the example of the difference in value of a 3-run home run in a 4th inning blow-out versus a walk-off 3-run blast). I mention this because it rebuts the idea that just because Cabrera played all six months of the season, and Trout only played five months, Cabrera was automatically "more valuable." He wasn't, and this proves it. [I got at the same point in another post by saying that, although Cabrera had 58 more plate appearances than Trout did, he also made 56 more outs - almost completely negating the opportunity advantage he had over Trout to provide extra value to his team.]
But what about "the stretch," August 15th to the end of the season? Well, the Angels went 28-17 the rest of the way, and Mike Trout accumulated 1.386 of his season's WPA total in that period. Meanwhile, the Tigers went 26-19, and Miguel Cabrera accumulated 1.088 of his season's WPA total in that stretch. On a purely cumulative basis, Trout contributed more, to more wins, than Cabrera did in the same "stretch" period.
Does that change if you break it down by negative and positive contributions to individual wins and losses? No. First, Trout had negative contributions in fewer of those 45 games (18 of 45) than Cabrera did (20 of 45). That was true not only overall, but when you break it down by contributions to wins versus losses:
* Trout was a negative contributor in 8 of 28 Angels wins in those 45 games (with a cumulative negative contribution of -0.298 WPA in those 8 games). Conversely, Trout was a positive contributor in 5 of 17 Angels losses (with a cumulative positive contribution of +0.493 WPA in those 5 games).
* On the other hand, Cabrera was a negative contributor in 9 of 26 Tigers wins (with a cumulative negative contribution of -0.591 WPA in those 9 games). Significantly, too, 5 of the 9 Tigers' wins down the stretch where Cabrera was a negative contributor came in in the Tigers' last 8 wins (all within the last 10 games) - hardly "closer" material. Cabrera was a positive contributor in 8 of 17 Tigers losses (with a cumulative positive contribution of +1.246 WPA).
* Overall, in the 45 "stretch run" games from August 15th to the end of the season, Trout contributed +1.700 WPA to 28 Angels wins, and contributed -0.314 WPA to 17 Angels losses, while Cabrera contributed +0.729 WPA to 26 Tigers wins, and +0.359 to 19 Tigers losses.
What this means in a nutshell is that, during the part of the season and in that part of each game "when it really counts," Mike Trout contributed more to more wins than Miguel Cabrera did, and Miguel Cabrera derived far more of his WPA "value" in games the Tigers ultimately lost and contributed less to wins down the stretch than Mike Trout did for the Angels down the stretch Whatever the sliced and diced, neutralized splits, cherry-picked numbers, and averages tell you about the stretch run, this completely refutes the idea that Cabrera contributed more to winning down the stretch that Trout did. That idea is demonstrably untrue.
Even though his team didn't make the playoffs (and Cabrera's did), Mike Trout contributed more, to more wins in a tougher division and in a tougher hitting environment, than Miguel Cabrera did. And remember, WPA either doesn't measure or doesn't fully measure those things than make Mike Trout special, like going first-to-third (or more) on a single, or Gold Glove quality defense at a premium position. Even ignoring THAT, Mike Trout was simply better. And saying that does nothing to diminish Cabrera's great season.
quote:"...To me the MVP should simply be the guy who is most valuable. That does not mean the guy with the best stats was automatically the most valuable. The triple crown was a great accomplishment, but doesn't mean Cabrera was Most Valuable. The stats compiled by Mike Trout were great, but it doesn't mean he was most valuable. Best stats = Best stats but not necessarily Most Valuable Player."
Again, I mostly agree with you here, though I responded more directly to your "best stats doesn't necessarily equal "best player" or MVP" comment in this response to another one of your posts. I guess where we disagree boils down my feeling that there are some stats that do a better and more comprehensive job of measuring complete value than others, and generally speaking, the player with the best numbers in THOSE stats is almost always going to be the more valuable player.
The exceptions might be when there is a large field of relatively mediocre offensive candidates but one who stands out as a fielder (think Zollo Versalles in 1965) or as a baserunner (think Maury Wills in 1962). Or sometimes because there is a great player who has been overlooked when he had better years, especially when a great leader on a playoff team (think Willie Stargell in 1979).*
[*The point about being previously overlooked is a big intangible that worked in Miguel Cabrera's favor this year, and IMO it may have carried as much or more weight than the Triple Crown.]
Obviously there is a bit of a continuum here, but generally speaking, the player who stands out in RE24, WPA, WAR, oWAR, and OPS+ (or wRC+, which is slightly better than OPS+ but harder to understand) is almost always going to be the most valuable player, and not coincidentally, almost always DOES win the MVP. When they don't and you look back at it, it looks very strange and usually like an obvious mistake in overlooking somebody who becomes a Hall of Famer (and often several Hall of Famers). That is the kind of historic season Mike Trout had, and I predict right now that is what it will look like looking back - at least (knock on wood) if Mike Trout stays healthy and has a career of 6000-8000 PA or more. NOT THAT MIGUEL CABRERA IS UNDESERVING - he had a great and historic season - but the kind of season Mike Trout had is greater still, just as historic or more so (though in a less traditionally-recognized way), and very rarely has failed to be recognized with the MVP.
quote:"Granted Mike is the better all around player. But the truth is, when the Angels needed him the most he tailed off a lot at the end and Cabrera turned it up a notch for the Tigers. That has to be considered 'valuable'."
It is valuable, but I hope I've convinced you that this statement ("when the Angels needed [Trout] the most he tailed off a lot at the end and Cabrera turned it up a notch for the Tigers") isn't so, or at least that there is another (and I'd argue better) way to look at this and conclude that it isn't so.
quote:"Anyway, no matter how you look at it, there's always another way to look at it. Both deserve the MVP. Mike was the better player this year, Cabrera was the better hitter this year. I can see reasons why people would vote either way. I'm a Mike Trout fan!"
You're right - there's almost as many ways to look at these things are there are opinions. Look, I'm a fan of BOTH guys, and last year, I beat the drum for Cabrera. This year, it's different, and I don't really believe in the "career reward" or "make-up" MVP.
I just think that when you finally come down to two players who played such critical roles on teams that played meaningful games right up until the end, "the better player" should win (and I don't concede that Cabrera was "the better hitter" though I would concede that it was close, and an argument can be made that he may have been).
I know that many of these stats are new to many people, but they are not really that complex (especially RE24 and WPA), and the idea behind them is very much rooted in REAL BASEBALL and real things that almost everybody says is "valuable." When you come right down to it, the main reason Cabrera won even though Trout had demonstrably more value comes down to undervaluing how important it is not to make outs, and overvaluing (even romaniticizing) the importance of Triple Crown stats as a measure of all-around value over not just better traditional stats (OBP, SLG, and their combination, OPS) but newer stats and ideas (like park factors, OPS+, WAR, RE24 & WPA) as well. It is too bad that the debate got turned into a referendum on whether "geeks" are really baseball fans, and rather than opening their minds too many traditionalists just dug in their heels and refused to consider that there might be a better way of looking at things. Because there's always a better way of looking at things; what was once new and considered "state of the art" will someday be supplanted or even discredited by a better way of looking at the same thing.