fWAR leader: Time to adjust the contributions of the defense

[BTS]

Did Rasmus put on weight too? Hutch? It's pretty tough to believe a few pounds slowed Trout down so much that he went from the best defender around to replacement level. You stick to that though.

 

I haven't commented about the magnitude of his defensive decline. What I've said is that his base-running has seen a very sharp decline, and given that he plays CF, it wouldn't exactly be shocking if his defense were also impacted to a significant degree.

[Nox]

Well if it's not predictive then how are you supposed to use it if not retrospectively

 

WAR itself isn't predictive but other models are.

 

Said models output a set of events that you can then use the WAR framework on to value them properly.

 

(measured past events + random variation) -> WAR -> noisy, confusing stuff

 

(projected event set based on our best estimation of a player's true talent) -> WAR -> Something more useful

[theblujay]

WAR itself isn't predictive but other models are.

 

Said models output a set of events that you can then use the WAR framework on to value them properly.

 

(measured past events + random variation) -> WAR -> noisy, confusing stuff

 

(projected event set based on our best estimation of a player's true talent) -> WAR -> Something more useful

 

Is this projection model like the Oilver projections? And I agree with what you're saying but I don't see how it relates to this thread exactly.

[eastcoastjaysfan]

And, yes, Nox is still saying that WAR itself is neither retrospective nor predictive. Essentially, as a stand-alone end-game measure, it's poo.

 

So, what they do, is take this thing they believe is poo, and input a variety of variables and constants in different computations, and then characterize and scale the answers to mimic that poo.

 

It's brilliant, really. Statisticians hard at work ruining the world of logic and reason.

 

Think about it this way. Let's say you wanted to measure how fast someone/something is and you didn't have all these "newfangled" measures of distance and time that we have in today's modern world. Some drunken guy, who is drooling on himself and wears a special protective helmet made of rock, and tends to shout out random silliness like "clap clap" and "troglodyte" over and over again, suggests using "fingers in the bum" as a benchmark. How long it takes to get them in there, how many you can get in there in that amount of time, how long the fingers are ... you know, the normal stuff. Everyone looks at him as if he's mad. Yet, some adopt this measure. We'll call them the trollhunters. They go around spotting trolls and yelling "fingers in the bum!" Some bright minds realize this isn't the best way to measure velo, so they construct other ways, but then label their findings still in the context of "fingers in the bum!" without it being based on fingers in the bum.

 

And everyone is happy and confused.

 

You and Nox. Yonge and Bloor. 11pm tonight. Spanky is the special guest referee.

[Nox]

Is this projection model like the Oilver projections? And I agree with what you're saying but I don't see how it relates to this thread exactly.

 

Oliver sure but preferably something better. Zips/Steamer publicly, something proprietary for teams (incorporating scouting data, hitfx data etc).

 

As for how this all applies to this thread: We're saying that retrospective 1 yr defensive metrics are only loose measures of past defensive ability. Nobody disagrees with that. So when we use that sketchy output as an input to the WAR framework, we should hardly be surprised the resultant numbers look weird. It's more an issue of the way we measure D being imperfect (it's a hard problem, scouts are demonstrably no better at it) than WAR being broken.

[HERPDERP]

You and Nox. Yonge and Bloor. 11pm tonight. Spanky is the special guest referee.

 

On the subway? I think there's enough room for Spanky to push both on the tracks.

[Angrioter]

So Let’s Talk About Alex Gordon

by Dave Cameron

 

For most of the last few years, if you clicked on the Leaderboards tab here on FanGraphs, you’d find Mike Trout‘s name at the very top. Today, that is not the case, as Trout has been surpassed in 2014-to-date WAR, slipping to #2 for the first time since late April. That isn’t necessarily controversial in and of itself, as it’s not that unusual for the best overall player in the game to not rate at the top of the WAR leaderboards every season, but what is somewhat controversial is the name of the player who has usurped Trout at the top of the list at this moment.

 

Alex Gordon, you see, is not exactly what most people think of as a superstar. He’s a corner outfielder who is hitting .286 with 13 home runs. Among 153 qualified Major League hitters this season, he’s ranked 36th in batting average, 32nd in on-base percentage, and 53rd in slugging percentage. Even using wOBA as a better evaluator of overall offensive performance, his .357 wOBA puts him in a tie for 33rd with Neil Walker and Jayson Werth. Add in park effects, and his wRC+ of 128 falls to 39th. As a hitter, he’s basically having the same season as Matt Kemp. This is the batting profile of the guy who currently leads all position players in WAR, and for many, that simply highlights the limitations of the model.

 

Even sabermetrically-inclined writers who live in Kanas City think this is weird.

 

 

 

Passan, it should be noted, is arguing against a strawman, since I haven’t seen a single person argue that Alex Gordon is “the best player in baseball this year.” For one, even if you used WAR as the sole basis for determining “best player in baseball” — and you shouldn’t do that — then the answer would be Felix Hernandez (+6.2 WAR, a half-win ahead of Gordon), so the most aggressive argument you could make is that WAR has Gordon as the best position player so far.

 

But really, even that is a far too aggressive interpretation, since no one has ever rationally argued that WAR is precise to the decimal point. The reality is that WAR has always been best used for grouping players of similar levels of contribution, not for arguing that a 0.1 WAR difference means that Player X is having a better year than Player Y. No one actually argues for using WAR as a precise tool to measure minuscule differences. I’d suggest that what WAR is actually saying is that Alex Gordon, so far, is having one of the best seasons of any position player in baseball this year, and I don’t think that statement is at all absurd.

 

First, let’s start with just the less controversial offensive component, since we already went through his hitting numbers. We noted that by wRC+, Gordon ranks just 39th among MLB hitters this season, but then again, hitting isn’t the only way to produce offensive value. We know that players convert their number of times on base into runs at different rates, and that having Jarrod Dyson reach is more likely to lead to a run for the Royals than if Billy Butler reaches.

 

Gordon isn’t Dyson, but he is a very good baserunner, having taken 11 extra bases and only making two outs in the process this year. Once you combine baserunning value with hitting value, we find Gordon ranks 22nd in Offensive Runs Above Average this year, better than his raw hitting marks would suggest. He’s still more of a good offensive player than a great one, but simply looking at his BA/OBP/SLG marks will undersell his contributions to run scoring.

 

But, of course, that isn’t why Gordon ranks #1 in position player WAR at the moment. He ranks at the top because he’s #3 in MLB in Defensive Runs Above Average, coming in at +17 runs relative to a neutral defensive player. People are rightfully more skeptical of defensive metrics than they are of offensive metrics, and we absolutely have more uncertainty surrounding Gordon’s defensive performance this year than we do his offensive performance. But rather than saying that the defensive component rating Gordon as an elite player is absurd, we should instead ask what the magnitude of the measurement error might actually be, and how that should affect our view of his performance overall.

 

After all, Gordon isn’t exactly a defensive schlub. He moved to left field full time in 2011, and has won a Gold Glove in every season since. While there are all kinds of problems with the Gold Glove voting, Major League managers very quickly adapted to Gordon as a terrific defensive outfielder, so this isn’t just a trust-the-numbers-over-the-eyes situation. By the Fans Scouting Report, Gordon rated as an 82 last year, the fourth best mark of any player in the entire league. People who watch Alex Gordon play defense regularly think Alex Gordon is really good at defense, so we shouldn’t be too terribly surprised that Alex Gordon ranks as a very good defender.

 

But this is absolutely an outlier season for him in terms of UZR, which is the defensive component of WAR. Since moving to left field full time in 2011, here are Gordon’s UZR/150 numbers:

 

2011: +12

2012: +12

2013: +7

2014: +27

 

Over the last three years, Gordon has rated as a very good defensive left fielder, rating about 10 runs per season better than the average left fielder. This year, he’s pushing close to 30 runs better than the average left fielder, which is why he ranks #1 in WAR right now and he never has before. Alex Gordon is clearly not a true talent +27 defender in left field, and some skepticism about that number is entirely justified.

 

But again, let’s keep in mind that even a normal Alex Gordon defensive rating would still rank him as one of the best players in baseball this year. Since moving to left field full time in 2011, he’s ranked as +30 runs above an average defensive player, which includes the positional adjustment. Even if we put no weight on more recent data and simply use a straight average of the total, we’d expect him to have a defensive rating around +6 or +7 right now; instead, he’s at +17, meaning that the bump in defensive rating this year has given him credit for about 10 extra runs, or about one win.

 

Subtract a win off Gordon’s total, and instead of ranking #1 in seasonal WAR to date, he falls all the way to a tie for #9. That’s the magnitude of the difference. Using an overly regressed defensive assumption, Gordon is still a top 10 position player in Major League Baseball this year. If you were a bit more reasonable in your regression and weighted recent performance more heavily than past performance, you’d end up with a weighted average defensive value of closer to +9, and Gordon would again find himself in the top five among position players in Major League Baseball.

 

And here’s the thing; there is absolutely no reason to assume that defensive performance is more static than offensive performance. In fact, there are all kinds of reasons to believe the exact opposite, and to expect fluctuations in defensive performance of a greater degree than we find in offensive performance.

 

We all generally understand that performance variance decreases as sample sizes get larger, and not just on defense. Batting average over a full season is more credible than batting average over a month’s worth of games played. We don’t freak out when Josh Harrison leads the NL in wRC+ over the last 30 days, as he does now, as we know that stuff like that happens, even with metrics with very minimal measurement error.

 

And the reality is that one of the primary reasons why offensive statistics are more reliable is simply because the samples are larger. Over the course of a season, an everyday player will bat 600 to 700 times, allowing much of the small sample variance to wash out in the end. On the other hand, even a very good left fielder like Gordon averages about 300 putouts per year, and most of those are routine plays that any ambulatory Major Leaguer could have made, so they have no real effect on his defensive rating.

 

According to the Inside Edge data here on the site, 71% of the balls hit in Gordon’s direction this year have been “Routine”, meaning that they are converted into outs 90-100% of the time. Gordon has converted 99.6% of those plays, so even that range is likely too large. These are routine pop flies that basically everyone catches. In addition, another 42 plays have been labeled “Impossible”, meaning that no one ever catches balls hit at that location, angle, and velocity. That leaves just 53 plays this season ranging somewhere between “Remote” (1-10% chance of conversion) and “Likely” (60-90%), and those are the plays that determine 100% of defensive rating.

 

We’re really talking about evaluating a player based on his performance on something like 50 marginal plays throughout the course of the season. It would be ludicrous to expect performance over an N of 50 to be the same every single trial, especially when the result of the play made or not made has such a large swing in run value. Making a catch on one of these marginal plays in the outfield is often times the difference between saving an extra base hit or making an out, and the gap in run value between a double or a triple and an out is worth more than a full run. With just a few 50/50 balls going one way or another, a player’s defensive performance in runs saved can vary dramatically.

 

Think of defensive performance as similar to home run rate. If a player hits a ball one foot shy of the wall, he makes an out; if he hits one foot beyond the wall, he creates at least one and often times more than one run for his team. A very small difference in a player’s swing can have a very big difference in the outcome, and we’re not that surprised when we see things like Chris Davis‘ HR total going from 33 to 53 to 21 over the course of the last three years. When dealing with samples this size and high magnitudes of difference for an out versus a non-out, we should rationally expect fluctuations. Single year blips may be correct, or they may not be — we don’t know for sure — but deciding that a player’s defensive rating is incorrect because it fluctuates from his prior history is a very flawed way of thinking. The cliche is wrong; defense can slump just like anything.

 

There is absolutely an argument to be made that Gordon’s UZR may be incorrect — though interestingly, people only ever seem to assume that numbers are too extreme, ignoring the possibility that the measurement error could also mean that his defensive rating might be too low — and if you were trying to answer the question of who “the best player in baseball” is, you’d definitely want to use multi-season regressed defensive numbers. But even using those kinds of calculations, there’s no way to get Alex Gordon out of the top 5-10 position players in MLB this year. The only “absurd” argument would be that Gordon hasn’t been one of the best players in baseball this year.

 

The very best? WAR can’t tell you that. But the good news is that it’s not trying to. What WAR is trying to tell you, though, is that Alex Gordon is having a great season, and you should accept that conclusion even without putting as much faith in defensive metrics as you do in offensive ones.

 

http://www.fangraphs.com/blogs/so-lets-talk-about-alex-gordon/

[KingKat]

Wow! That's a really well written post. Half-way through I assumed it was written by someone other than Cameron and then I doubled back and to my surprise saw his name.

[BTS]

Wow! That's a really well written post. Half-way through I assumed it was written by someone other than Cameron and then I doubled back and to my surprise saw his name.

 

Cameron gets a bad rap. His articles are generally pretty good. He's not a statistician like Tango or Bill James, but he's a pretty great baseball blogger IMO.

[KingKat]

Cameron gets a bad rap. His articles are generally pretty good. He's not a statistician like Tango or Bill James, but he's a pretty great baseball blogger IMO.

 

I don't read him or Fangraphs regularly. Whenever his stuff is brought to my attention it's usually because he wrote something really stupid. Nox has brought up examples in the past and seems to really dislike him generally.

[BTS]

I don't read him or Fangraphs regularly. Whenever his stuff is brought to my attention it's usually because he wrote something really stupid. Nox has brought up examples in the past and seems to really dislike him generally.

 

Yeah, Nox dislikes him, but I don't think people like Nox are really his intended audience.

[KingKat]

Yeah, Nox dislikes him, but I don't think people like Nox are really his intended audience.

 

Yeah but thank kind of misses the point. People like Nox can tell you exactly when and how he's full of s***.

[BTS]

Yeah but thank kind of misses the point. People like Nox can tell you exactly when and how he's full of s***.

 

IIRC Nox thinks he's kind of a charlatan, which maybe he is. He definitely doesn't advance knowledge like some of the more mathy types, but I still enjoy most of his articles. He's a decent writer and creates interesting content, but not groundbreaking content. I think Nox is of the opinion that he tries to insert himself in conversations with the types who post on Tango's blog, but that isn't really the nature of his contributions.

 

That's what fangraphs is now, for the most part. Interesting content, but not groundbreaking, and for the most part, not quality research.

[KingKat]

IIRC Nox thinks he's kind of a charlatan, which maybe he is. He definitely doesn't advance knowledge like some of the more mathy types, but I still enjoy most of his articles. He's a decent writer and creates interesting content, but not groundbreaking content. I think Nox is of the opinion that he tries to insert himself in conversations with the types who post on Tango's blog, but that isn't really the nature of his contributions.

 

That's what fangraphs is now, for the most part. Interesting content, but not groundbreaking, and for the most part, not quality research.

 

"I think Nox is of the opinion that he tries to insert himself in conversations with the types who post on Tango's blog, but that isn't really the nature of his contributions."

 

That's part of it but there's also the part where he gets concepts flat out wrong in his own writing. Down is up, up is down, kind of stuff.

[Angrioter]

Yeah, Nox dislikes him, but I don't think people like Nox are really his intended audience.

Nox is pretty smart, but very demanding; he's always looking for the third eye.

[BTS]

"I think Nox is of the opinion that he tries to insert himself in conversations with the types who post on Tango's blog, but that isn't really the nature of his contributions."

 

That's part of it but there's also the part where he gets concepts flat out wrong in his own writing. Down is up, up is down, kind of stuff.

 

Yeah, I could see that. IMO a big part of the criticism is that people tend to view Cameron as a SABR-genius and that grinds the gears of people who know that he isn't, and leads to people pointing out his flaws. I'll read Cameron over 95% of bloggers, and definitely over any Jays-dedicated types like Tao or Stoeten. He's an entertaining writer and usually has something relatively interesting to say.

[KingKat]

Yeah, I could see that. IMO a big part of the criticism is that people tend to view Cameron as a SABR-genius and that grinds the gears of people who know that he isn't, and leads to people pointing out his flaws. I'll read Cameron over 95% of bloggers, and definitely over any Jays-dedicated types like Tao or Stoeten. He's an entertaining writer and usually has something relatively interesting to say.

 

Yeah well I did really enjoy this particular piece and it's the first time, I've really seen his qualities on display instead of his flaws so it gives me a more balanced perspective. If he were still writing for S.S. Mariner or whatever his original blog was, it would be easier to forgive sloppy back of the napkin math but he really should have someone review that stuff for him at Fangraphs. An argument could be made that the Fangraphs editor should be someone who has some statistical background and not just someone who can write. So maybe the real issue is that he's a bit miscast as something more than just a contributor.

[Nox]

Meh, the article is ok. Certainly better than all the "LOL Alex Gordon is not the best player in baseball, WAR is broken" types we've seen over the past week.

 

But when he says things like: "What WAR is trying to tell you, though, is that Alex Gordon is having a great season",

 

I'm not sure he actually understands WAR either. WAR isn't telling you Alex Gordon is having a good season. WAR is telling you that the measurements that have been assigned to Alex Gordon thus far would indicate an elite player IF we take those measurements to be accurate measures of true talent. A few months of UZR and a number of offensive numbers aren't really that.

 

It's not an instance where he truly just embarrasses himself. I mean, up until MGL shamed him publicly for the 10th or so time, he didn't really understand how to regress platoon splits properly. That lack of knowledge is obviously fine for the average poster here (who might read the Valencia threads here from last month and pick something up) but for the lead writer for what's turned into the top analytics site, it's pretty awful.

[GD]

I haven't commented about the magnitude of his defensive decline. What I've said is that his base-running has seen a very sharp decline, and given that he plays CF, it wouldn't exactly be shocking if his defense were also impacted to a significant degree.

 

http://www.fangraphs.com/blogs/has-mike-trout-gotten-slower/

 

I haven't read this but it seems relevant to this discussion

[BTS]

http://www.fangraphs.com/blogs/has-mike-trout-gotten-slower/

 

I haven't read this but it seems relevant to this discussion

 

Well that's good timing.

[Angrioter]

http://www.fangraphs.com/blogs/has-mike-trout-gotten-slower/

 

I haven't read this but it seems relevant to this discussion

 

Great read, Gracias hermano!