Skip to content

Let’s Talk About Giants MVP Candidates

August 6, 2010

When baseball fans talk about who should be the Most Valuable Player of either league, there are a few different trains of thoughts that the fans use to support their argument. Some believe that an MVP should come from a winning team, as they believe – I’m probably saying this too loosely, but stay with me – “what good is a player if he can’t lead his team to a winning record?” Others may argue the opposite; players on inferior teams are more valuable because they are more important to their team; one might argue that Miguel Cabrera would be a bigger loss to the Detroit Tigers than say, Josh Hamilton to the Texas Rangers. Then, there are those that take the more statistical approach, just simply deeming the player with the best numbers the most valuable (and this falls within advanced statistical reasoning, as we believe in valuing things that the players have the most control over, their overall production). And the last criteria that voters and fans consider, is how “clutch” that player was; did an inferior player’s production in a season actually contribute to his team more than another’s simply because he produced “when it mattered?”

All are completely legitimate points, and I’m not going to argue that one is a better criteria than the former. However, being the statistical-minded, I do believe that the best method is simply looking at the player’s overall value (key word there), rather than intangibles and making assumptions about “what if his team didn’t have him?” That is the simplest method (and will produce the clearest result), as the other ways fall under the realm of opinion rather than objective analysis. However, I’m also a fan of considering how clutch a player was, because, even if not completely under a player’s control, it does in fact make a difference in baseball.

Perhaps I’m just lazy (“perhaps: should probably be edited out there, but I’m not doing it) and want to make things easier to analyze, but I’ve decided to just limit this to the 2010 Giants team, which of course almost all but eliminates the team factor for obvious reasons. So, I am going to value players according to this definition (which I just made up because I couldn’t find a suitable one):

The player who provides the most value to help his baseball team win games.

Simple enough right? Player. Value. Wins. That’s how nerds like it, straight-forward. But we need to decide what metrics we’re going to incorporate.

For position players:

  • StatCorner’s wOBA* turned into Batting Runs Above Average. Why? It’s park-adjusted (there are better park factors, but the differences won’t make a huge difference, especially since everyone here is on the same team), it doesn’t include baserunning (more on that soon), and they include ROE (reaching on error) while FanGraphs does not.
  • Baseball Prospectus’ EqBRR for baserunning
  • UZR and DSR from FanGraphs for defense
  • FanGraphs positional and replacement values
  • And here is where the fun comes in: valuing players on game state. I’ll be using both Win Probability Added and RE24Wins, both taken from FanGraphs as well.

What is WPA?

Sum of the differences in win expectancies for each play the player is credited with. Can be for a play, game, season, or career. This is denoted in wins and is of a similar scale to other wins-based statistics. It is highly dependent on the context in which a player played.  Note that it is relative to average, so a 0 WPA player is an average player.

And RE24?

Runs Added by 24 base-out situations. Sum of the differences in run expectancies for each play the player is credited with. Can be for a play, game, season, or career. This is denoted in runs and is of a similar scale to other runs-based statistics like linear weights. It is somewhat highly dependent on the context in which a player played. A player with a lot of runners on base ahead of him has more of a chance to create RE24 than a batter who always comes up with the bases empty. It is relative to average, so unlike runs created an average player will have zero RE24.

**If you want to read more about win probability, leverage index, clutch, etc. do a Google search for Tom Tango’s “The Book” and add any of those key words in, and you’ll definitely find some great reading material.**

The great thing about both is that they are expressed in wins and runs, and are “relative to average,” meaning they are on the same scale as linear weight runs which we’ll also be using. Now you may ask why I’m using both WPA and RE24, when they’re both very similar. I prefer RE24 as it doesn’t take into account game state, as those are even more out of a player’s control than the 24 base-out states (meaning some players just get better opportunities). RE24 also has to do with opportunity, but not to the extent that WPA does. What I’ll be doing is averaging RE24 and WPA, and then averaging that with BRAA. There is value in using one or the other, but for the most part there is no huge discrepancy in wins between RE24 and WPA.

**Also of note, WPA doesn’t take into account park factors, or opportunities as I mentioned above; however, with all of these players being Giants the variability there is reduced. It’s also why I’m placing value in linear weights as well as game situation. I do realize the danger in mixing park-adjusted stats with park-neutral stats, but fortunately AT&T isn’t too extreme to skew anything too much, especially win probabilities**

For pitchers:

  • WAR from StatCorner based on tRA* (I like FIP, but definitely prefer tRA and the inclusion of xIP)
  • For relievers, we’re always dealing with small sample sizes and of course the volatility of relievers is extremely high year-to-year. So to account for that, we’re going to use WPA. Again, we’re dealing with extreme sample size issues, but that’s the life of a reliever. One day I’m going to do a bit more research into leverage and look at a much better way at evaluating the bullpen, but for now, this is what I’m going with. Also, the great thing about WPA, is that it’s usually “right,” in the sense that said player did indeed help or hurt his team win. The only problem is it tells us nothing about the root of the player’s performance and whether or not he was lucky.

And after all of that, here are the results:

Position Players

Player WAR
1. Aubrey Huff 4.8
2. Andres Torres 4.5
3. Buster Posey 2.2
4. Juan Uribe 1.8
5. Pat Burrell 1.5

Pitchers

Player WAR
1. Matt Cain 3.5
2. Tim Lincecum 3.0
3. Brian Wilson 2.8
4. Barry Zito 2.5
5. Jonathan Sanchez 1.9
  • The big surprise is of course Brian Wilson being the 3rd most valuable pitcher and 5th most valuable player on the team overall. This is a combination of Wilson just being very, very good (ERA, FIP and tRA all hovering around 2.00) and Bochy utilizing Wilson quite well by pitching him in high leverage situations as much as possible (not only 9th inning save situations).
  • On the other end, Sergio Romo comes in at replacement level. This comes from a couple game-changing homeruns he allowed in the beginning of the season which then led Bochy into a knee-jerk reaction of using Romo in mop-up duties only for a good month or two. He’s worked himself back into the set-up role where he belongs (3.44 FIP, 2.47 tRA). Also, his RE24Wins lists him 1 above average, so he’s definitely pitched well, he was just hurt by small sample sizes early in the year.
  • The other surprise might be Pat the Bat who’s definitely been a pleasant surprise in returning to the NL. Bochy’s also managed him well (whoa, this is getting ridiculous now) in starting him mostly against lefties and replacing him with Nate Schierholtz after a couple at-bats to avoid his awful glove from being exploited. Small sample sizes aside, his glove hasn’t cost the Giants too many runs and he came into yesterday’s game with a .387 wOBA*. His value comes from when he’s producing. His “clutch” performances this year represent a near half-win addition to his WAR, and both his WPA and RE24Wins are near 1. Kudos to Jonathan Broxton for that.
  • Everything else shouldn’t be of any surprise, with Huff and Torres leading the way with Huff’s tremendous offensive output this season and Torres’ great glove in the outfield. Although neither will win the MVP this season, they should hopefully at least receive a few votes as they (along with Posey) have been performing far above any expectations and are the reason the Giants are holding a playoff position in August.

**I’ll update these at the end of the season with other NL players to create my MVP rankings for the 2010 season.**

Advertisements
10 Comments leave one →
  1. Kool permalink
    August 7, 2010 6:54 AM

    I myself do not like FIP that much. I think that a pitcher has more control over the hits that he gives up. I think I read somewhere on fangraphs that every pitcher (or at least the majority of them) in the HoF have a BABIP under .300. Coincidence? Besides, it doesn’t like Matty. Yet, it makes walks sound bad. I’d rather give up a walk than a hit if I’m a pitcher. More often that not a game is decided by a hit. Where can I find this tRA? I’d like to see what it says about a few pitchers.

    • TheGig permalink*
      August 7, 2010 8:46 AM

      I like FIP, but it definitely has its limitations. And while it is true that a pitcher has more control over his hits, the factors that FIP uses are typically more predictive.

      And yes, FIP and xFIP have consistently underrated Cain. However, FIP is only about .4 runs per 9 off from his career ERA, and he’s had good defenses and a good park for flyball pitchers.

      Cain is one of the anomalies though in that he’s always been a guy that has been tough to hit, and that’s why tRA is much more favorable.

      Anyways, you can find tERA on FanGraphs.com to compare it with FIP, xFIP and ERA, but the best place for it is statcorner.com. The links are in my post. For the most part, the statistics aren’t too far apart, but tRA seems to explain a lot of which the other metrics cannot.

  2. triplesalley permalink*
    August 7, 2010 9:39 PM

    The more I think about it, the more I’ve come to the conclusion that WPA should not be figured in WAR calculations. This is due to the fact that WAR is meant to be entirely context-neutral; WPA is context-dependent. RE24 works fine for this, as it compares each player to the league average performance, but WPA doesn’t do that. It just looks at the literal value a player provided at the plate based on the inning and game state, but it doesn’t compare this to how the league does in this situation. WPA/LI *might* do this, but I haven’t looked in to it that much.

    Just something to think about. 🙂

    • TheGig permalink*
      August 7, 2010 10:20 PM

      This is what I was worried about.

      But at the same time, the definition of WPA does say that it is relative to average, so that’s what I was unsure of.

      I mean, it still all adds up to 0, so the scale has to be the same right?

      Also, the difference between the two metrics didn’t make too much of a difference anyways. Perhaps there’s some adjusting that should be made, even if the difference is minimal.

      • triplesalley permalink*
        August 7, 2010 10:44 PM

        WPA isn’t relative to average- RE24 is. WPA doesn’t compare the player’s performance to others in that situation; it simply looks at the amount the player adds or detracts from his team’s probability of winning a game. If a player is “0 WPA,” that means he hasn’t helped or cost his team anything when all is said and done. It doesn’t have anything to do with the league averages, which is what you’re looking for in WAR.

      • TheGig permalink*
        August 7, 2010 10:48 PM

        Well, I guess my definition sucks?

        Check B-R’s definition, it does say relative to average.

        Perhaps B-R changes something to make it relative?

      • triplesalley permalink*
        August 7, 2010 11:46 PM

        Their definition is awfully confusing- I honestly have no idea what they’re doing.

        On the team level, it is relative to average. This is because the sum of wins added/cost add up to the team’s wins above/below .500. Whether or not that works for individual hitters, I do not know.

      • TheGig permalink*
        August 8, 2010 12:00 AM

        Well the good news it didn’t make more than a difference of 0.1 win for any player. Perhaps I’ll make relativity changes when I do it again at the end of the year.

      • triplesalley permalink*
        August 8, 2010 1:06 AM

        If you’d like, we could design something simpler to account for “clutch” performance rather than having to worry about using the proper win probability metric. Perhaps look at the player’s performance in C&L situations as measured by wOBA compared to the player’s overall performance, above the league’s “clutch” performance.

        That probably sounds a bit confusing, so let me illustrate with an example:

        Let’s say Aubrey Huff has an overall wOBA of .419 and a wOBA of .395 in C&L situations. The league average is .330 and .310, respectively. That means that while the league hits .020 points worse in those situations, it’s still better than Huff, who is at .024. We could then write our “clutch runs” as…

        (Player C&L wOBA – Exp. C&L wOBA) / Scale * C&L PA

        In this case, Huff would be expected to hit .399 in C&L situations. So that would make him a -0.3 clutch runs hitter in 2010. Derek Jeter, on the other hand, has a .340 wOBA and a “clutch” wOBA of .355. That would make his “clutch runs” at +1.7 on the year.

        It doesn’t have to be Close and Late, either. I just chose that because it’s a standard definition of “clutch.” This way, we could avoid any potential issues with WP stuff and keep it to what we know will be context-neutral.

        Or, we could just use the simple idea of clutch performance being “above/below” the player’s standard level of performance, rather than compared to the league’s clutch ability. I suppose there’s merit to both, but the latter idea is the more standard definition of clutch score.

      • TheGig permalink*
        August 8, 2010 1:13 AM

        I actually like that, it’s pretty similar to what BtB use to do with their PR, “clutch runs” I think.

        My biggest problem with WPA and RE24 is that they’re both opportunity based and doing this would leave the linear weights alone and then add in whatever clutch situations we see fit.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: