Archive - Sabermetrics RSS Feed

Post injury Chris Young: Regression or change in talent level?

December 12, 2007 in pitchf/x, Sabermetrics, San Diego Padres with Comments Off

edit: you should probably read this before reading this. 

In what is one of my least favorite posts on this blog, I said this:

Young is a case that deserves some real hardcore research. Perhaps, in the future, I’ll be able to look much deeper into why CY has defied the odds for some 260 innings or so now.

You could probably consider this at least my first attempt to look at Chris Young a little more in-depth, although it’s not exactly for the reasons I had earlier planned. On July 24th Young strained an oblique muscle and was placed on the disabled list. Courtesy of baseball reference, here are his pre injury numbers: 118.7 innings, 114 k’s, 39 bb’s, 4 hr’s, 1.82 era. And how about after coming back from the injury two weeks later on August 9th: 54 innings, 53 k’s, 33 bb’s, 6 hr’s, 5.96 era. His strikeouts stayed at a similar level, but everything else was way off. Now the traditional sabermetric thinking here would be simply that Young regressed toward the mean, at least to a degree. After all, nobody is a 1.82 era true talent pitcher and therefore nobody is going to maintain an era in that area. However, especially when an injury occurs, there could be another reason (or reasons) why Young’s numbers showed such a decline after returning from injury. He could have simply been a different pitcher with different stuff and a changed style. That is not as much regression as it is a change in talent level and that’s what I’m going to attempt to look for.

So I have finally loaded all of Young’s starts (that have PITCHf/x data) into a spreadsheet and now I can play with the PITCHf/x data a little bit. Note that this stuff has already been done by many brilliant researchers. Many have gone far more in depth than I could ever hope to, but I don’t believe anyone has concentrated specifically on CY and his injury. First we’ll look at Young’s performance overall, then we’ll take a look at the data pre and post injury. This first graph is perhaps the least interesting, but I’ll throw it up anyway. These are all of Young pitches with start speed on the y-axis and end speed on the x axis (remember to click all images for a better view … edit: definitely be sure to click now, as I had to rescale them to even fit on the blog):

cy2.JPG
Next up we have what has kind of become the standard for many of these graphs. Horizontal break on the x-axis, vertical break on the y-axis, and speed displayed by color. Check it out:

cy3.JPG
This stuff is admittedly sometimes tough to interpret and tough to display. There are a lot of 90-95 pitches but the are behind the 85-90 pitches. Anyway, it jibes pretty well with something like Josh Kalk’s CY player card, although he is using an algorithm to classify the pitches. Anyway, I think the majority of his fastballs have a negative horizontal break. The sliders range from ~75-high 80′s and have a horizontal break on both sides of 0. The curves are the slowest pitches and down toward the bottom right side of the graph. Finally, there are a few changeups overlapping the fastballs (they have similar break but of course a much different speed).

Now that we’ve looked at all of CY’s pitches, let’s break it down into pre injury and post injury and see if we can spot any noticeable differences. First the speed graph for pre injury (over 1,300 pitches):

cy4.JPG
And now for post injury:

cy5.JPG
There are around 835 pitches here. If you can’t see it in the graph, here are some interesting percentages:

Over 95 mph
Pre injury: 3.8% (51 pitches)
Post injury: .01% (1 pitch)

Over 90 mph
Pre injury: 54% (724 pitches)
Post injury: 19% (156 pitches)

How about average start speed
Pre injury: 88.59
Post injury: 86.43

Wow. Unless they were some major changes in the PITCHf/x system, this is pretty striking (I think). (I know there were changes throughout the season, but I am pretty sure that they wouldn’t have this large of an impact). Now, maybe Young went to more breaking and off speed stuff and that’s the cause for the lower velocity. I’m not sure. Either way, though, it appears he was definitely a different pitcher after the oblique injury. Now onto the other two graphs:

Pre injury:

cy6.JPG
Post:

cy8.JPG

Not really sure what to make of this. It appears that the many sliders that were 80-85 pre injury are now in the 75-80 bucket. The curves are pretty non existent, as well. But, with these graphs being relatively tough to interpret, I’ll leave it up to you to decide their relevance.

So maybe we haven’t solved the post injury Chris Young debate (regression or change in talent level?). But perhaps we’ve at least laid out a bit of a methodology in assessing whether or not — and how — injuries impact pitchers.  Or maybe we’ve just wasted a Tuesday night. Hey, it’s better than school work either way ; )

Right on the sweet spot

December 11, 2007 in playoffs, Sabermetrics, San Diego Padres with 4 Comments

Dex linked to my last post yesterday and had some interesting thoughts on valuing players. What I tried to do yesterday was simply estimate what Bradley should make based on his established talent level and how free agents are paid these days. Dex brings a whole set of other things into the picture (like how much each win is actually worth for the team, how much revenue players bring in that isn’t counted in wins, etc). Anyway, it got me thinking about one of the favorite chapters in Baseball Between the Numbers (note: there are no “affiliate links” in any book links at this point). Nate Silver (fwiw, one of my favorite baseball writers) wrote a chapter called, “Is Alex Rodriguez Overpaid,” where he discusses a lot of these issues.

For now, though, I’ll ignore most of the stuff about player value, as another part is more interesting to me. Luckily, if you don’t have BBTN, a lot of Silver’s research is available in this article at BP. If you don’t have a BP subscription, then I guess you really are out of luck …. or not, because you can read the rest of this post ; )

Since it is behind the paywall, I’m just going to take one graph and a few of Nate’s points and put them here (I hope that isn’t disrespecting any copyright stuff). Here’s the relevant graph (click for a better view):

wins-2.gif

Silver uses some mathematical techniques that I don’t fully understand to get these playoff probabilities for a team that is predicted to win a certain amount of games. This version adjusts for the fact that predictions are inevitably tough and a predicted 88 win team could win 80 games or 95 due to luck, randomness, and many other factors.

But notice the sweet spot, as Silver calls it, that exists from about 82 to 87 wins. Teams in this area can add a few wins and do wonders for their playoff chances. Take a predicted 83 win team. They have a ~23% chance of making the playoffs. If the add a 5 win player, their probability jumps up to about 40%. That’s a 17% gain due to adding a 5 wins over what’s already on the club. Now look at a 65 win team. If they add 5 wins, they are adding 2-3% to their playoff chances. On both sides of the extremes (high and low win projections), it’s not very helpful to add a few marginal wins to your roster. You likely aren’t going to make the playoffs and even when adding a borderline star player, you still are going to have to get very lucky to creep in. However, teams in that mid-80s range have a ton to gain from adding a few wins to their projection.

So where do the Padres fit into this? Well, without running any actual projections, I think it’s fair to say that they are in that mid-80′s range right now, whether that be 83 or 87, I’m not so sure. But I’m going to venture a guess that it’s somewhere in that range with this current roster.  So what should they do? Silver’s advice:

Category II. Fringe Contender: 82-87 projected wins.
Optimal Strategy: Buy or Sell
Examples: Cubs, Mets.

Teams in this group have a natural tendency to stand pat. The thinking seems to be: we’re fielding a reasonable baseball club, and we think we can contend with a couple of good breaks. Look what happened to the White Sox last year. We certainly aren’t about to break the bank.

In fact, however, standing pat is the worst alternative for these clubs. Whether to buy or sell is conditioned on some of the same factors that we’ve described above, but either strategy is superior to holding. Buying is likely to produce a reasonably good return; although a team with 85-win talent will make the playoffs occasionally, a team with 90-win talent will make the playoffs more often than not. On the other hand, if buying isn’t feasible, then selling needs to be considered. Going from 85 wins to 80 doesn’t hurt as much as going from 85 to 90 helps, and there is nothing worse for a baseball team than to be caught in the 84-78 netherworld.

Silver makes it pretty clear that it’s probably silly to stand pat here. If you do, there’s a pretty good chance you’ll miss out on the playoff while still spending some pretty good dough. If you add talent, it is in a sense more valuable to you at this point, as it can have a tremendous impact on playoff probabilities (and thus, revenue). If you sell, you can go into a little bit of a rebuilding mode and not spend so much on a borderline playoff club (and, at the same time, you won’t completely kill playoff aspirations).

Due to stuff like this, comments like, “a marginal win added for any team is worth the same,” are probably incorrect in a economic sense. A marginal win for the Royals may be worth very little, while a marginal win for an 88 win team is worth quite a bit (because of it’s increased chance of reaching the playoffs … and because the playoffs are the goal, both in an economic and baseball viewpoint).

Back to the book: In Silver’s chapter, he attempts to put an actual value on this in a chart called “Marginal economic value of 1 additional win.” For a 60 win team all the way up to a ~78 win team, the value is around $750k. He had previously calculated this number using a variety of variables like concessions, merchandise, gate receipts, and so on. On the other hand, a marginal win for an 88 win team is worth over $3.5 million. It peaks at about 90 wins and almost $4.5 million. Looking at the Padres as say a 84-87 win team, their marginal win value ranges from about $1.5m to almost $3m, depending on what their actual projection is. And that’s of course due to increased chance of a playoff appearance, which Silver estimates is worth around $30m.

Each extra win right now is worth more to the Padres than it is to many teams. Therefore, they could (and perhaps should) bid more for the services of someone because they “know” they are going to get extra returns over, say, what the Red Sox or Royals will get from the same player. He’s worth more to them because of their current situation.

From all of this, I’ll infer that the Padres are at a critical point this offseason. They could go one of two ways. They could spend the extra money and look to cash in on a playoff appearance. Or they could sell off some players and go into mini-rebuilding mode. With Peavy extending, Wolf coming in, and seemingly constant rumors about this guy or that guy, it looks like they aren’t going into rebuilding mode. At the same time, they haven’t added a ton of projected wins yet and the Peavy extension doesn’t do anything for the current team (he was already here), so they may not be going into the spend direction either.

Of course, there really is one more direction they could go that I haven’t mentioned yet. That’s standing pat. Hopefully, they’ll choose one of the former methods. For this team, standing pat is likely the least optimal thing they could do.

Evaluating Japanese prospects

December 6, 2007 in player evaluation, prospects, Sabermetrics, San Diego Padres, scouting with 2 Comments

No stats here … just some random thoughts. I may try to put together some type of quick-and-dirty MLE’s for Japanese players, though with my limited capabilities, I am not sure if it’d be worth it.

Anyway, over at Dugout Central, Mike Pagliarulo has a post up on how to evaluate Japanese prospects. FWIW, I’ve found that site to be pretty good, although I wasn’t sure about it early on. Anyway, Pags occasionally takes some unnecessary shots at stats/sabermetrics … like right here:

Some teams do incredible research while others use statistical projections (which never works), but there is little doubt of what is the most important factor when signing a Japanese player

I do not know what he’s talking about when he says that the statistical projections never work. Firstly, you can look at almost any player that has come over to mlb and they will most likely have put up good numbers in Japan. Now, that certainly isn’t a projection and it doesn’t mean they’ll do well here, but if players aren’t putting up good numbers over there, they likely (for the most part) won’t be signed and brought over to major league baseball. (note: as usual, I am essentially talking out of my ass here, as I really haven’t followed Japanese baseball closely). So whether you like it or not, right off the bat, stats are going to be a big part of any teams projections/evaluations.

Now beyond that is actual statistical projections, which are definitely different than just basic, unadjusted numbers. As far as I know, the most basic way to get a projection would be to calculate some sort of major league equivalencies (mle’s), based on how players perform when they switch over from Japan to mlb (or the other way around), and then go from there into your projection (with age adjustments, true talent, and so on). Anyway, my point is not just to disagree with Pags, but to bring some discussion about how the Pads should go about evaluating Japanese prospects, specifically Kosuke Fukudome (as they’ve been rumored to be very interested).

Fukudome is quite the hitter with a career .305/.397/.597 line with the Chunichi Dragons. Rally translated his numbers and then ran them through CHONE (his projection system) and got a .283/.373/.465 line. With help from Tango’s scale, he says that he’s worth around 4 years, 45m on the current free agent market.

Anyway, back to Dugout Central and where I started: The Pags Rules on evaluation of Japanese prospects:

  • Never take the word of his agent on the player’s commitment level, or for anything, for that matter
  • Never use one scout to evaluate the player; double check and triple check your work
  • Identify the player’s commitment level
  • Grade the skill level of the player (tools)
  • Identify mechanics of the pitcher/hitter and project them to the team’s philosophy
  • Project the playing field change and effect (pitching mound, distance to the fence, etc.)
  • Project emotional fit with the coaching staff
  • Exactly identify the player’s role

He certainly makes some good points there … aggregation of scouting data, tools, mechanics, park effects, etc. Anyway, I don’t know where the stance against stats comes in. It is just another component in the tool box, especially when used on prospects (or guys who haven’t played in mlb) and not the end-all-be-all. But I can’t imagine any organization, especially the Padres, not utilizing a statistical projection for Fukudome or any other Japanese prospect.

In the end, I think evaluating these guys is much like evaluating well seasoned AAA prospects. There is a certain degree of uncertainly, of course, but numbers (when properly adjusted) can — and should – be used for projections. They should be combined with scouting data and weighed properly when all is said and done. I’m not exactly sure how you’d go about it, but that should be the goal.

*More on japanese translations/MLE’s from seamheads.com … actual MLE’s coming from that site soon.

Randy Wolf

December 1, 2007 in contracts, player evaluation, Sabermetrics, San Diego Padres with 1 Comment

Well, it look at if the Padres have locked up Randy Wolf to a one year deal. 

Wolf is a lefty starter who is 31 years old. He pitched with the Phillies from 1999-06, and then jumped over to the Dodgers last season. Here’s a quick look at Wolf’s career numbers:

untitled22.JPG

That’s k’s, bb’s, and hr’s per batter faced, courtesy of baseball reference. Interesting to see his numbers improve last year after his worst statistical year in 2006 (albeit, in only 80 innings). Part of it is probably due to regression and part of it to getting out of Philly and into the more spacious Dodger stadium. Wolf has been billed as a big fly ball guy, though that’s not really the case. His fly ball rate is 40.5% since 02 and the average is around 37%, or at least it was in 2006 according to the Hardball Times Annual. I’m sure there’s some fluctuation year by year, but my guess is that it’s pretty stable. So, Wolf is a fly ball guy, but he isn’t an extreme one by any means.

Wolf, however, did struggle in Philly with home runs per fly ball. Here are his numbers since 02 in that stat:

2002: 9.6%
03: 12.4%
04: 10.6%
05: 13.6%
06: 16.7%

Now, last year he put up an 8.7 hr/fb rate with the Dodgers. Surely, it’s a small sample, but there’s probably a good chance that that the change in ball parks helped Wolf. He could be a guy who gives up a lot of relatively long fly balls — which, I’m guessing, in general, isn’t a good thing. Moving to a park like Petco, though, could do wonders for a guy like that.

So far, I think all of this is pretty good. I don’t have much of a problem calling Wolf, per inning, about an average starter. Probably slightly below average, once you account for aging (especially, for a guy who is constantly dealing with arm troubles).

But … it’s the innings thing that, of course, worries you. Wolf was a durable pitcher from 2000-2003. Then in 2004 he threw a 136.7 innings. Since then, he’s thrown a total of 239.4 (80 a year). During that time, he had reconstructive elbow surgery and missed more than a year. And last year he had diagnostic shoulder surgery and was shut down in July. He is expected to return by spring training for the 2008 season.

Anyway, despite me looking at the numbers here, Wolf’s another case where you want to use a good deal of major league scouting. For one, he’s coming of arm troubles so his talent level may have changed. And secondly, we don’t have much good data to work with. I mean, you could go back to his days when he was throwing 200 innings a year, but he was very likely a different pitcher back then.

Foxsports scouting report:

One scout calls Wolf a poor man’s Tom Glavine. Wolf has an extensive repertoire, throwing a fastball, curve, changeup, slider and cutter. He sets up his fastball with a big, looping curveball that is sometimes thrown in the mid-70s. He also features an above-average changeup. That makes his 90-MPH fastball appear to get on hitters more quickly than it actually does. Wolf has a deceptive motion and knows how to work hitters: up and down, in and out, soft and hard. Command is his key, as he walked just 2.4 batters per nine innings last year. Wolf almost always remains poised while on the mound, but he is a fierce competitor.

Complimenting your traditional scouting report, Josh Kalk has given us a detailed look at each pitchers repertoire. According to the Pitchf/x data and Kalk’s classification algorithm, here’s what Wolf featured last year (pitch, speed, percent used):

Fastball: 89.5 mph, 59.6%
Curve: 67.2 mph, 19.3%
Slider: 80.9 mph, 11%
Change: 79.1 mph, 10%

Against lefties, he throws the same amount of fastballs, but uses the slider far more often. Keep in mind, of course, the sample is pretty small. Anyway, check out that player card yourself — awesome stuff. Here are the breaks:

randy_wolf.gif

So, what’s a guy like this worth? ZIPS projects Wolf at 93 innings and a 5.03 ERA. I’d say that’s pretty pessimistic, but we’ll stick with it. Dan has the league average for a starter at 4.69, so we’ll use my back-of-the-hand calculation of replacement level, which is simply average plus 1. That means, Wolf is projected to be about 7 runs above replacement next year.

Tango’s salary scale says that a 7 RAR player is worth about $3-3.5 million for a year. The Padres are paying him $4 million guaranteed, so it appears to be a pretty fair deal. Incentives could make it worth much more for Wolf, but that will also mean that his performance will most likely be much more valuable. It’s like a built in safety net and I think it’s a wise thing to do with injury risk type pitchers like Wolf.

So, on one hand, I think the Wolf deal is the prefect low risk, high reward deal. If Wolf breaks down in April, you’re down 4m and that’s it. If he throws 180 innings with a 3.80 era, you give him 9m and call him one of the best back of the rotation starters in the league.

However, on the other hand, it’s only low risk in that there’s not a ton of money or commitment involved. If the Padres think they can rely on a full season of Wolf, they are really probably kidding themselves. If you look at him as your 4th starter, then I think it almost becomes a high risk move, because you don’t really want to see Justin Germano, Jack Cassel, Josh Geer, etc. rack up inning after inning in a tight pennant race.

I think the Padres have some other moves in store and will perhaps bring in one or even two more starters. That, in my opinion, makes this deal that much better. We’ll see what they do as the winter meeting start up next week.

Estimating Peavy’s innings

November 16, 2007 in Jake Peavy, player evaluation, projections, Sabermetrics, San Diego Padres with 1 Comment

In Tim Sullivan’s article on Peavy, he quotes Kevin Towers saying this:

“The risk on our part is injury. I don’t think this guy (Peavy) is going to regress in the way of his performance. The key thing for us is: Is he going to be healthy for the next three or four years?”

So, I thought I’d take a look using Peavy’s comparable players. I did that using PECOTA and ended up settling on using 16 of the 20 guys available. I didn’t count four guys because they either are only slightly older than Peavy or they became relievers. Then I simply looked at how the other 16 guys aged from their age 27 through 33 seasons, in terms of innings. Some retired, some got injured, and some others may have changed roles (I’m not sure), but I left all of them in there. This is very unscientific and should certainly be taken for what it’s worth (not much). Anyhow, here’s a graph with all 16 pitchers “innings path” on it:

peavy-ininngs.JPG

Pretty confusing, so here’s the average for each age amongst this group of pitchers:

peavy-inings.JPG

Now, that’s of course with a lot of zero’s mixed in from injury or retirement. Here are the most optimistic career paths:

potimistic.JPG

And the least:

pess.JPG

Like with all pitchers, there’s a significant chance at injury over the length of any deal. And for what it’s worth, no pitcher increases his workload over this time period (out of these 16). This is strict decline time, even for the best pitchers in the game. 

Some of the comps include Pedro, Tom Seaver, Bret Saberhagen, Alex Fernandez, and Kevin Appier. I don’t want to list many more because it’s behind the BP paywall. I should also note that these pitcher are Peavy’s comps before 07, so they could change a little this year.

I’m not sure if this tells us anything. They gave at least 150 innings though their age 31 years, but after that it gets really dicey. It’s just hard to predict how long a guy is going to remain effective and stay healthy, especially when you’re getting 6-7 years into the future. I think it may be optimal if they could lock him up until 2012 (his age 31 year) and then let him walk after that. Of course, that all depends on dollars and many other factors.

… just a few numbers on a friday afternoon (not to be taken too seriously).

Page 24 of 28« First...1020«2223242526»...Last »