June 14, 2007

Forecasting Doom

One of the things I wrote about yesterday was that Justin Verlander's first pitch strike % has been really low - below or at 50% in his most recent outings. I thought this was interesting, so I went through some more numbers this morning: Over his last five starts, Verlander's FPS% has been 48.9%, a very bad number. (Remember, this includes his no-hitter.) For the sake of comparision, I calculated the numbers of two other young phenoms - Philly's Cole Hamels has a 56.6% over his last five starts, and SF's Matt Cain owns a 52.4%. So it seems Verlander has been having some control issues lately.

But I went farther back into his
game log, and found some more interesting results. In Verlander's five starts before his five previous ones, he did much, much better - 60.3%. So what is the cause for this? The number one factor I would look at is fatigue. I don't know for a fact that 1st pitch strike % is directly related to fatigue, but it seems to make sense. Remember, Verlander threw over 200 IP last season - almost twice his previous season high - and pitched deep into October as the Tigers reached the World Series. As the season went into September and October, Verlander's numbers started to decline last year. This year? With the exception of his last two starts, Verlander's ERA has climbed consistantly throughout the season. He gave up just 2 HR in his first seven starts; he's given up five in his last six outings. His walk rate was pretty high for the first couple starts, and then dropped in the middle of the year, but it's climbing back up again; meanwhile, except for his 12 K outing two days ago, his strikeout rate has stayed pretty much the same.

So, my point is: The Tigers had better be careful with how they use Verlander. Here's what Baseball Prospectus 2007 said about Verlander:
...Postseason included, Verlander threw nearly 60 percent more innings in 2006 than in his pro degbut in 2005 and looked tired toward the end of the year, particularly in the playoffs. There are arguments for doing something creative with him this year, such as using him out of the bullpen for a month at the start of the season. At the very least, the Tigers need to err on the side of caution, because his arm is much too valuable to put in any long-term jeopardy.
So, why do I bring this up? The Tigers are not erring on the side of caution. Verlander has thrown 97+ pitches in all but one start, and he's been over 110 three times. And there's no reason to think this is going to stop. The Tigers have lost a lot of players from their bullpen and rotation, including Joel Zumaya, Kenny Rogers, Nate Robertson, and Zach Miner. Because of this, Jim Leyland is going to try to get a lot of innings out of his healthy starters, which could be very dangerous for Verlander this year and in the future.

Anyone remember
Felix Hernandez's one-hitter against the Red Sox, back in April? After that game, everybody in the media jumped all over him, anointing him the next Johan Santana. Well, during his next start, he got injured and hasn't been nearly the same since coming back. Well, now the media's giving Justin Verlander more than enough love, and I'm seeing some parallels to the King Felix situation. I'm not saying he's going to get hurt on Sunday, but I will predict that he either goes on the DL or sees a dramatic rise in his ERA before the season is over.

June 13, 2007

Verlander's No-no

In case you have secluded yourself from all forms of media except this blog, Justin Verlander threw a no-hitter last night. Verlander rode the benefit of 12 strikeouts and a couple really good defensive plays behind him to record the second no-hitter of the season (the first coming from Mark Buerhle). This was a complete domination of the Brewers' lineup (who, by the way, recorded 22 hits the previous game) - Verlander struck out 12, and the ones who did make contact didn't get it very far; only three balls were hit to the outfield.

Much has been made of Verlander throwing 102 mph into the ninth inning; his velocity climbed as the game continued, going from an average of 96 in the first few innings to 98 around the sixth to 100 over the final inning. Now, the radar gun at Comerica is famous for not always being accurate - last year in the playoffs, Zumaya was repeatedly clocked at 103 when he was really throwing 100, and Verlander was consistantly hitting triple digits when all season long he'd been clocked at 97. But this one seems fairly legit -
ESPN's Inside Edge (probably insider-only) had Verlander reaching 102 also, and I trust their numbers.

Verlander mixed in a fair amount of changeups and sliders as well, finishing the game with 112 pitches - actually a bit high, considering he faced only 30 hitters. One of the reasons why was that he only threw 50% first-pitch strikes, which is not a terrible number but is less than ideal. Looking through some of his
game logs, this doesn't look like a new thing for him - by my count, in his previous three games his first pitch strike ratios have been 12/26 (46%) against Texas last week, 13/26 (50%) vs Cleveland a few weeks ago, and an especially atrocious 11/29 (38%) against Cleveland on May 31. I can't find FPS data anywhere except by going through game logs, but I know that, when I'm pitching, I usually shoot for a number in the mid-60's. I'd guess that the major league average is somewhere from 55-60%, and a star like Verlander should be in the sixties.

What does this mean? Well, it means Verlander throws too many pitches. His 3.88 pitches per plate appearance are 31st out of 145 pitchers with at least 40 IP this year (incidentally, the lowest number belongs to none other than Greg Maddux, with 3.15), and he's thrown 97 pitches or more in all but one start this year, topping 110 three times. Why does this matter? Verlander is only 24 years old. Last year, he threw over 200 innings (including the postseason); his previous high was 129. And he throws 102 mph. My point is, the Tigers should be careful about Verlander's workload; he could easily end up on the DL for at least a couple weeks before this season is over.

The good news is, he's striking out more hitters. He somehow posted a low K rate last year despite having almost the same stuff; this year, he's upped his rate almost a full K/9 IP. As Rob Neyer wrote last night in
his blog: (insider-only, of course)

Verlander has been "pitching to contact" for a while now, and he's good enough to make that work. But tonight we saw what can happen when he pitches to non-contact. No, he might never throw another no-hitter (Roger Clemens hasn't thrown even one). If Verlander boosts his strikeout rate to eight or nine per nine innings, though, someday he'll add a Cy Young Award to the hardware he's already got.

Verlander's a great pitcher witH Cy Young-caliber stuff. And I like how he was throwing harder late in the game, proving that he has the stamina to go deep into the game and that he isn't throwing every pitch as hard as he can, saving some for when it matters. But the Tigers had better be careful with how they use him, at least until he gets into his prime years.

Ripping ESPN.com, continued...

This probably isn't the best way to start this blog - it's probably more interesting when I blog about, you know, actual baseball. But I thought I'd do a follow-up on yesterday's "discussion" of ESPN's new player rating system. Later on Monday, the creator of the ratings - Jeff Bennett - did a chat on ESPN's SportsNation explaining the rankings some more. Here's the link (I think it's Insider-only).

Here are some of the more interesting questions and responses...


Scot (TX): Some questions on formulas for you Jeff: 1. Why team win percentage? Is Teixeira really a "better" player if he played for Detroit instead of Texas? 2. Why not differentiate CF from the corner OF spots in the defensive standings? A player who can play CF seems to have more value. 3. Did you ever try to work in quality of defense? Right now an excellent fielder and a poor fielder get the same amount of defensive points as long as they both play SS.

Jeff Bennett: A multi-part question off the top....

1) I think there is some value to player on a wining ball club. This is evident with the BBWAA voting. However, it make up only 5% of a score and the difference between Detroit and Texas currently ranslates to about 0.4 rating points.

2. Currently this is a function of the way our stats are coded. Looks for this to be adjusted for in the future. the defensive spectrum would go obviously CF, RF, LF.

3. Yes, much like VORP, quality of defense is not accounted for.

1. Yes, there is some value to a player on a winning ball club. There is also some value to a player on a losing ball club. Typically, the players on losing ball clubs will, on average, have worse stats than players on winning ball clubs, so you shouldn't have to make an extra adjustment for this. By the way, I absolutely love that he thinks the fact that the MVP voting is based largely on team performance is a valid reason to adjust for this in his system.

2. This has been changed, according to the explanations. The difference isn't enough yet, but it's a step in the right direction.

3. So, you're defending a new system by pointing out an earlier, better system that has the same flaw. Remember, this was the stat that was labeled "cutting-edge" on the front page of ESPN.com. This wasn't what I was looking for.

Otter (Boston MA): What do you do for players that have a significant split in where they play defensively? Weighted average of difficulties? Is there any bonus given for this type of versatility?

Jeff Bennett: The position the player plays the majority of his games at is designated his primary position. There is no bonus of versatility, but it is something I considered.
So, if Victor Martinez plays 75 bad defensive games at catcher next year and 70 games at first, and Joe Mauer plays 145 good games at catcher next year...they get exactly the same points for their defense? VORP at least accounts for different positions played during the year (I think it does, anyways)...

Paul (Wildomar, CA): Anybody who has read Bill James knows saves are overrated. Why 40% allocated to saves? Seems to me the other 4 categories, maybe especially inherited runners scored, should be higher. We all know Pat Neshek is better than David Weathers.

Jeff Bennett: Saves are overrated. That is what the actual stat used here has a stiff penalty for blown saves. Neshek does quite well for himself here. He is #116 today overall. Not bad and ahead of many, many closers.
Saves are overrated. Therefore, there is a penalty for when you pitch poorly in a save situation. Does that make sense? And, although there is a penalty for blown saves, getting actual saves is the way relievers get 40% of their points. Pointing out how the best middle reliever in baseball is at #116 doesn't really help your case...

Taylor (Escondido CA): Did you take park effects into account at all? Of course the Padres hitters are going to rank poorly and the Padres pitchers awesome because of Petco. Or, for example, Ian Kinsler (Texas launching pad) v. Jose Lopez (Safeco death to hitters), have the same OPS, yet your rankings have Kinsler higher?

Jeff Bennett: Excellent question. Park effects are not directly applied to adjust the ratings. BIg parks doesn't seem to hurt the Comeria hitters this year or say Dmitri Young. I have a hard time de-valuing Peavy. He is 3-0 with 1.06 ERA on road this year. I see he is ranked 31st in Win shares today behind many pitchers. That seems a little low.
These rankings are flawed in many ways, but the lack of park factors is the biggest one. Saying "Big parks doesn't seem to hurt the Comerica hitters this year or say Dimitri Young" not only is very gramatically flawed, it is logically flawed as well. If a player has good stats in a pitcher's park, he would have even better stats in a hitter's park. That's simple logic. Hitters are always "hurt" by big parks, whether or not their stats are good. And Comerica has been a hitter's park this year, by the way, with a park factor of 1.071 (above 1.000 is hitter-friendly).

Bryan (Madison, WI): Don't you think that your formula for relievers makes it too difficult for actual relievers to be rated highly? You've said that saves are overrated, yet they comprise 40% of the reliever rating whereas holds are ignored completely. It seems to me a reliever could have a excellent season and unless he picks up a lot of wins (which is generally determined by the events earlier in the game and by the actions of his offense, not the pitcher himself) it would be very difficult for him to crack the top 100 or 150 players.

Jeff Bennett: Fair question, but set-up men can do well here by having a low ERA, Opp. BA, K-to BB-ratio. Last year Cla Merideth was one of the majors to 20 pitchers when applying retroactively. I think that would be news for most of mainstrream baseball fans. He has unhittbable in 2006 and didn't register many wins (5).
Yes, but those numbers are only 60% of a reliever's value. The other 40% is wins and saves. Basically, from this and some of the other questions, what Bennet seems to be saying is that great middle relievers can move themselves up to the level of a solid-to-good closer if they have a great year. When, in reality, a great middle reliever is often worth just as much as a great closer.

Ian, NYC: I don't understand what your point is behind this list. There are people who have put a heck of a lot of science and research into coming up with formulas like this (Win Shares, VORP, etc), while much of what you have selected here is totally arbitrary (why exactly %10 for BA for eaxmple?), and by pretending this is somehow scientific degrades the whole field of work on this subject. Much of what you are including here has been proven to be no reflection on individual player quality (like saves, wins and RBI to a large extent), not to mention penalizing someone because they play on a bad team. Why should anyone take this list seriously?

Jeff Bennett: Ian, I think you hit on something. There is no sucjh thing as the perfect way to evaluate a baseball player. Win Shares and VORP are great, but you can ask the same types of questions about their lists. This system is very fluid and puts players in perspective based on where they rank in the majors vs their peers. Nothing more scientific than that.
Actually, there are things more scientific. It's great that you compare relievers to relievers, hitters to hitters, etc....but Win Shares compares all players to all other players. Wouldn't that be "more scientific"? And I can't ask the same question about WS and VORP that I've been asking about this one: What the hell do the final numbers mean? WS and VORP numbers have meaning. This system just gives you numbers.

There's some more, but it's mostly just about individual players. Still no answer as to why there's no park factor adjustment, still no answer as to what the numbers mean, still no actual reason there's no defense adjustment. Hopefully ESPN gives up on this quickly.

I'll try to get to some actual on-the-field baseball stuff later today, namely Verlander's no-hitter...

June 12, 2007

A Bad Idea...

Okay, now that school's basically over, I'm launching this new private blog. I'll explain more about who I am later, but since there's no high school sports to write about in the summer, I'm going to spend my summer blogging about my true love - baseball. But average fans beware - if you don't know what OPS is, you may want to look somewhere else...

What better way to start off a blog than by ripping ESPN? Let me start with a prelude - I really do like ESPN.com's website. I like some of the colorful stuff like
DJ Gallo, and I love their NBA coverage (especially now that they have TrueHoop). And I do really like their MLB coverage, especially the blogs (Insider-only, unfortunately) - Buster Olney and Rob Neyer both have outstanding blogs. I read anything I can written by Neyer and Keith Law. For some reason, it's become fashionable for bloggers to express their hatred for "mainstream" sites like ESPN.com - I don't feel that way at all. (I'm not a fan of their NFL coverage at all, with the exception of Tuesday Morning Quarterback, but that's neither here nor there...) So what I'm going to say is not out of hatred for ESPN.com - it's about how ridiculous this particluar article was.

But Monday afternoon, I came home and found this on the ESPN.com homepage:



Abacus-Cadabra

ESPN.com found a scientific method to sort out baseball's best players.

Put your slide rules down. We've got something more cutting-edge. Want baseball's top 100 players, with a mystical number rating their worth? Done. How about All-Star projections? Presto.
Player Ratings

Now, I was intruiged. A scientific method to sort out baseball's best players? You mean, like using statistics? We already have a number of ways to do that, and those could be applied to All-star projections as well. VORP (Value Over Replacement Player) is a very good rating of a player's value. EqA is probably the best measure of a player's hitting ability. Bill James spent five years or so working on his Win Shares statistic, which not only rates players but also expresses their worth in a meaningful way (wins contributed to their team), not just a "mystical number". But you're saying these stats are now archaic? Your stat is "more cutting-edge"? Cool! I'd love to see what new and creative stats you came up with!

Unfortunately, the explanation was less impressive. Here's the link to the story
explaining the rankings, and it's not nearly as "cutting-edge" as it seems. Instead, the formula basically takes a bunch of counting stats and mixes them to produce one super-stat, a smart idea except it's been done hundreds of times (and usually much more effectively) by other people before. Here's how they figured out how batters rank:



Batting bases accumulated: 20%
Runs produced: 15%
OBP: 15%
BA: 10%
HRs: 10%
RBIs: 5%
Runs: 5%
Hits: 5%
Net Steals: 5%
Team winning percentage, defensive position: 5% each

Let's take these one-by-one, to make things more organized:

Batting Bases Accumulated: 20%
This is very reasonable - the formula given is TB + BB + HBP. It's kinda like OPS, except actually a bit better, since it doesn't count hits twice. It's not perfect, but it's a good start.

Runs Produced: 15%
Oh, boy. They define "Runs Produced" as Runs + RBI - HR. So, you're saying that 15% of a player's worth is based on how good his teammates are? Raul Ibanez finished seventh in RBI in all of baseball last year. Raul Ibanez. Jimmy Rollins OBP'd only .334 last year, but still managed to finish tied for third in runs scored because he had Utley and Howard hitting behind him. There are dozens of statistics that are a better measure of how good a player is than runs or RBI (though apparently nobody's told the Most RBI MVP voters that yet...)



OBP 15%; BA: 10%
Not really a whole lot to pick at here, although both should be worth much more than they are. Hits are worth more than walks, so BA should be a category, along with OBP. One thing to notice, though, is the curious overlap of stats - you've already really accounted for both OBP and BA in the "Batting Bases Accumulated" category...it seems they could have condensed things some and made it less confusing (or maybe more confusing)...

HRs: 10%
Um...what about doubles? What about triples? You really think you can determine 10% of a player's worth directly from how many homers he hits? Okay, then...

RBIs 5%, Runs 5%
Again? Didn't I go through this already? First of all, "RBI" stands for "Runs batted in", so there's really no need to add an extra "s" there. But that's just nitpicking. Still, so far 25% of a player's value is apparently determined by how good his teammates are. And more strange overlapping of stats.


Hits: 5%
Okay...how are "Hits" and "Batting average" different? Couldn't they have just weighted BA by at-bats, eliminating the need for another "Hits" category? This is really starting to confuse me...

Net Steals: 5%
I'm assuming they mean (SB - CS). But, according to numerous experiments, a player is actually hurting his team if he converts less than 75% of his steal attempts. So, shouldn't you have used 75% as your baseline? Plus, steals aren't often worth that much anyways - certainly not 5% of a player's value, with the possible exception of ridiculous Juan Pierre-types. (And, actually, his success rate was under 75% last year, so the value of those steals was negligible. Not that his .330 OBP added much value either...)

Team Winning Percentage, Defensive Position 5% each
If you're counting at home, that's 30% of a player's total value that is based on how good his teammates are. Really, winning percentage? You're saying that Oakland's hitters are more valuable than Tampa Bay's because Oakland's pitching staff has given up 150 fewer runs? How does that make sense? To be fair, the actual impact of this ends up being not much - the difference between 36-26 Detroit and 23-40 Texas ends up being only about .4 points (out of 100). Which prompts the question, why have the category anyways? But the more egregious one is the second column. Only 5% of a position player's worth comes from what position he plays? A player's net steals is worth the same as his position? This is completely absurd. The league-average 1B last year hit .285/.365/.495 (BA/OBP/SLG). Pudge Rodriguez' career numbers are very similar - .304/.342/.483 - but, because he's a catcher, he's a future 1st-ballot Hall-of-Famer. Are you telling me Pudge, for his career, was only 5% better than the league average first baseman? Secondly, they count all outfielders as "OF", and don't count where they play in the outfield. Center fielders are much more valuable than corner outfielders, because they must cover much more range. In fact, according to the
defensive spectrum, center fielders are more valuable than third basemen as well. So this system unfairly hurts center fielders, and that is obvious in the rankings. Finally, there is no accounting for quality of defense, either. So, according to ESPN's "cutting edge" stat, it doesn't matter at all how good a defensive player you are! Derek Jeter must be thrilled...Update: According to ESPN.com, they now differentiate between the types of OFs, though the difference still isn't that much, and CF is still behind 3B. But it's a start...


Here's the formula for starting pitchers:

ERA vs. league average weighted by IP: 40%
Wins weighted by win percentage: 20%
Innings Pitched: 10%
Defensive independent bases allowed per IP: 10%
Strikeouts: 10%
Opponents' BA: 10%

These are actually okay categories - ERA is probably one of the best stats to use here, and it's good to see they used the league average to account for the DH. IP is important, because pitchers who throw innings are helping their team get through games, even if they aren't very effective innings. The defensive independent bases are defined as (4*HR + BB + HBP), which is a very good measure of a pitcher's true effectiveness (i.e., not determined by luck or defense). Unfortunately, they then have to count opponents' batting average (very influenced by luck and defense) at the same weight, which doesn't make much sense. But the one category that really sticks out is wins. Wins are loosely related to how good a pitcher is, but are at least as related to how good the team is. Case in point:

Randy Johnson, 2006 - 17 wins, 5.00 ERA
C.C. Sabathia, 2006 - 12 wins, 3.22 ERA

Who had the better year?


Relief Pitchers:


Saves and wins with blown save penalty: 40%
ERA vs. league average weighted by IP: 20%
Opponents' BA: 20%
K/BB: 15%
Inherited runners stranded percentage: 5%
ERA is tough to use for relievers, as one bad inning can really damage your ERA because you don't throw very many innings, but there aren't that many better metrics to use. K/BB = good stat, Opponents' BA = bad stat. The save is quite possibly the dumbest stat ever; is your best reliever more valuable with one out and the bases loaded in the seventh in a tie game or with nobody on and a three-run lead to start the ninth? Wins are almost completely random for relievers; often they come because a reliever actually has a bad outing and blows a save, but ends up with a win when the offense saves them. And what is this "steep blown save penalty"? The formula given is (saves * 2) + wins - (blown saves * 3). That means that if a closer converts around 67% of his save chances and picks up the lucky win or two, he gets some points. (What are points, anyways? Is 0 a replacement player? Is 50 average? What the hell do these numbers even mean?) I can't find the exact number, but I bet that teams with a lead entering the ninth win much more than 67% of the time. You should not get points for saving 7 out of every 10 games.


Overall, these ratings aren't awful, but they're worse than most of the sabermetrics used today. This would be fine if it were just some guy's creation, but ESPN's trying to actually use it; I saw them being mentioned on Baseball Tonight as if they actually proved anything. ESPN's baseball writers (Olney, Neyer, etc.) have been noticably silent on this topic - they don't want to out-and-out disparage their employer's stat, but if they really thought it was good, don't you think they would have mentioned it by now?


Here are the main problems, as I see it:

1. No park factors - Jake Peavy is rated as the best pitcher right now, mostly because he's having an amazing year, but also partly because he gets to pitch half his games in spacious Petco Field. Meanwhile, stats in mile-high Coors Park count exactly the same as those in Petco or RFK. It's not that difficlut to add
park factors into ratings, so it really surprises (and disappoints) me that they didn't go that extra step. This early in the season, park factors usually aren't worth using because they fluctuate so much, but using last year's won't really disrupt anything.

2. No defense adjustment - A great defensive shortstop is worth more than a poor one. That's just fact. Unless you're working with this system. (To be fair, VORP doesn't have a defense adjustment, but it at least bills itself as such.)

3. Not enough adjustment for position - Chase Utley, Joe Mauer, and Jose Reyes can tell you that the position you play determines much more than 5% of your value. The difference between center field and the corner outfield spots isn't enough, and the difference overall isn't nearly enough.

4. Too much reliance on teammates - Really, team winning percentage? Runs and RBI? Wins for a pitcher, and saves for a reliever? These are the best stats you can come up with to evaluate an individual's performance?

5. No meaning to the nubmers - I still can't figure out what the numbers mean. 0 is the floor, and 100 is the higest possible...but where do normal people score? What's a normal MVP-caliber score? 70? 80? What does a score of 0 mean? A replacement player? A bad minor-leaguer? Is 50 average? Is the difference between 60 and 65 big or small? There are way too many questions to answer.

Here's a better system, in my opinion:

Batters MLB ranks in:

VORP: 100%

The ESPN.com "Rating" is an okay stat, and I'm sure it took a lot of effort, but it's not nearly as comprehensive or meaningful as stats such as VORP, EqA, and Win Shares. If this had been invented/popularized five years ago, it might have been useful, but today we have too many other stats that are better. There are too many flaws - such as the fact that it only compares relievers to relievers, position players to position players, etc., and the fact that it doesn't adjust for park factors - it just doesn't measure up to the more comprehensive stats.

Baseball Prospectus' Nate Silver has
an excellent take on this as well...