[DSG]

Hey guys,

Since Chuck has disappeared from the face of the earth, I'm going to open this thread myself. For those who don't know, I am a writer for The Hardball Times and we're going to be shipping the fourth edition of our Annual book in a day or two. You can read more about the 2008 Annual here and buy it here. We're also putting out a Season Preview this year, which will include projections, player comments, team essays and more. I'm really excited about the book, which can be ordered here. If you order both the Annual and Preview together, you can get 10% off by entering the discount code HTC08.

Also, because it's a really good cause, I would like to remind you that today is the last day to donate to the SoSH Jimmy Fund-raiser, which you can do here. I myself will donate $1 (sorry, still a poor college student) for every Hardball Times Annual that a SoSH member buys and $2 if you buy both the Annual and Preview. Just send me an order confirmation from ACTA at daviddsg@gmail.com. A lot of people read this board, so lets see if you can empty my bank account.

Now on to the chat...

Fris edit here:

My apologies for the confusuion, David. I was waiting for you to start the answer thread, while you were waiting for me to open it.

[DSG]

Quote

Is there any hope that Schilling's slight decrease in fastball speed post-injury won't regress in '08? And what about that sligtly zippier changeup post-injury?

As I wrote in my article on Schilling, I think that an off-season of rest and recuperation should get him back to where he was before the injury. Also, remember that the decrease in fastball speed and increase in changeup speed could be small sample size symptoms: I just don’t know.

[DSG]

Quote

My question: What does he see in the future for Nick Hagadone

Way too early to tell. Hagadone was certainly good in Lowell, but all we have is 24 innings of low A-ball performance—not nearly enough to even make a guess about a prospect. Scouting-wise, Carlos Gomez, who now works for the Diamondbacks as a major league scout, really liked Hagadone when he was drafted. That can’t be a bad thing.

[DSG]

Quote

The article on Schilling focuses primarily on the declining differential between fastball and changeup, and refers to the changeup as his strikeout pitch. Are change and splitter combined for the purposes of this study? If not, does this conclusion apply equally to his splitter?

Schilling’s increased reliance on the curve and/or slider early in the count is striking. He seems to have made a transition from a 2-pitch (FB, splitter) to a pitcher with 3-5 pitches (FB, splitter, change, curve±slider). His ability to throw the curve for strikes early in the count gets him to a strikeout situation, even if his ability to put the hitter away has waned. Are there precedents for pitchers who have made a transition in repertoire late in career, and do they have similar dropoffs in K rates?

I did not look at Schilling’s splitter in my study (or I think I did, but found nothing interesting). It was not lumped in as a changeup. The thing is, Schilling didn’t throw all that many splitters this year; they comprised just 11% of his pitches, while Schilling threw a changeup 27% of the time.

According to Pitch f/x data (and remember, it only captured 756 of Schilling 2,270 pitches on the season), Schilling threw just four splitters in two-strike counts: Two went by for balls, one was hit for a single and one, a triple. He threw 67 changeups in that situation, eight of which ended the at-bat right there with a strikeout.

One possibility I should acknowledge is that the classification algorithm I’m using (actually, Josh Kalk, who provided me with all this data) is mis-specifying some changeups as splitters, which now that I look at the data looks like a possibility. I’ll have to look into it some more.

I have no idea what the answer is to your last question.

[DSG]

Quote

Dave were you surprised to see Jimmy Rollins win the NL MVP. In your article he wasn't listed as one of your top seven. Was that an oversight on your part? Was there To much weight being given to Defensive metrics, or do you think the voters got that one wrong?

I wasn’t too surprised that he won, but I definitely think the voters whiffed badly on this one. According to my numbers, Rollins was +34 runs as a hitter (this includes a positional and park adjustment) and -6 as a fielder, which combined was good for 32nd in the major leagues. He wasn’t especially good in the clutch, so you can’t make that argument either. Essentially, the voters gave him way too much credit for (a) Accumulating a lot of playing time (while ignoring all the outs Rollins made); (b) Playing in a hitter’s park; and © Being on a playoff team.

My personal choice was David Wright, but if the voters wanted a Philly, they should have gone with Chase Utley.

[DSG]

Quote

Recently, Theo Epstein at least paid lip service to the idea of a 6-man starting rotation. Presumably, the notion arose from an interesting mix of veteran and young Red Sox pitchers who might benefit from additional rest and limited innings. In the case of the Red Sox, there are certainly valid conditions to justify exploring the possibility. One of the theories you've floated as a cause for the relative reduction in IP's by starters over the last 40 years is salary inflation (free-agent era). Given the expected magnitude of Johan Santana's next contract and those of other elite starters in the future, as well as increased awareness of the injury nexus among young pitchers; do you see the 6-man rotation as an inevitable development in baseball's future?

No way, starting pitchers are way too expensive. The authors of The Book actually found that pitchers were almost a run better per game as relievers than they were as starters and that pitchers got worse each time through the order, which led them to recommend that teams use their fourth and fifth starters as swingmen, going only once through the other team’s batting order before being relieved. I think that’s a much more likely direction.

Relievers are simply better pitchers than most starters because of all the advantages they hold, so there is no reason not to employ a big (and comparatively cheap) bullpen, while starting only superior pitchers. Imagine that you can make a player’s ERA drop by almost a run just like that; why wouldn’t you do it?! It would come at some expense of bench flexibility, but I think I could live with that.

[DSG]

Quote

On the whole "Pitchers pitch less innings now" tack, there seemed to be change in the early '80. This seems to correspond to the emergence of Wade Boggs and the whole 'foul off pitches/take pitches' movement. Have batters started taking more pitches in the last few decades? (I know pitch counts weren't statistics for a long time, but I wondered if there was any insight you could bring.) Of course, the 80's also had the 'steriods/HR' boom, so there could be opposite effects. Thoughts?

There’s so much game theory involved in what a ball is and what a strike is that without actual pitch charting data (like what Pitch f/x provides us with), it’s impossible to tell what’s happened, in my opinion.

[DSG]

Quote

Thoughts on who should/shouldn't get in? Rickey is up next year, so this probably a few of the borderline guys best shot.

I would vote Blyleven, Gossage, McGwire, and Raines, though I really only strongly care about Blyleven and Raines.

[DSG]

Quote

David,

It is illuminating that the methods that you use in "My 2007 MVP's" value Curtis Granderson, who, in my opinion, had one of the most incredible seasons I've ever seen last year. How does the baseball cognoscenti educate fans and popularize "new" stats that show the value of that particular type of player whose value is not expressed by the traditional statistics?

OPS. was a no-brainer and seems to have "crossed-over", so much so that I think most fans know what an acceptable range is (ie .600 OPS on a baseball card looks like a .211 avg 8 HR of yore while a .950 to 1.000 an above is your traditional MVP Candidate).
How do Runs above Avg., WARP, or DIPS make that leap? And how would you explain those stats to a fan new to these types of stats? (Or to me, because honestly I don't understand them that well)

Will you expand on the difference between the seasons of Jimmy Rollins and Curtis Granderson(which you mentioned briefly in My 2007 MVPs)? Which player was more valuable in 2007? Why they are different in value? In which ways did they differ?

Thanks!

I don’t know that they’ll ever make the leap, though DIPS is tracked on ESPN’s statistic pages. The first problem, I think, is actually coming to some sort of consensus on what numbers to use, and seeing them regularly published. There are so many numbers out there right now, and so many of them are fairly worthless, that it is almost impossible to separate the wheat from the chaff for a normal fan. Clay Davenport’s WARP, for example, is almost worthless, yet here you are asking about it.

Runs above average (otherwise known as linear weights) are very simple to explain, though Chuck should jump in here with his own lecture on the subject. Every situation in a baseball game can be described by how many outs there are and what bases are occupied. You can add a lot of other variables, but in terms of figuring out how many runs we expect a team to score in a given inning, that’s all you need to know. There are three different amounts of outs you can have (0, 1, and 2) and eight different base situations (no one on, man on first, bases loaded, etc.). That makes for 24 different base/out combinations. We can compute how many runs a team scores from a given base/out situation to the end of an inning fairly easily using play-by-play data, giving us a “run expectancy” (how many runs we expect a team to score) for each combination. So let’s say that at the beginning of an inning, we expect the average team to score .5 runs (that translates to 4.5 runs per game). If the batter hits a single, we now have a man on first and no outs, and looking it up, maybe we find that we now expect the team to score .9 runs. That means that the single increased our run expectancy by .9 -.5 = .4 runs. If the next batter hits a home run, the run expectancy goes back down to .5 (since there are still no outs and the bases have been cleared), but the team has scored 2 runs, so the change in run expectancy is .5 - .9 + 2 = 1.6. That home run was worth 1.6 runs. So you can do that calculation for every single, home run, whatever during a given season (or seasons) and find the average value of each event. It turns out that the average values are about:

1B = 0.47
2B = 0.78
3B = 1.05
HR = 1.40
BB = .33
IBB = .20
HBP = .35
SB = 0.18
CS = -0.45
Out = -0.27

These are known as linear weights. They tell you how many runs on average each type of event adds or subtracts from a team’s run expectancy. If you apply them to a player’s stat line, you can find how many runs above or below the average player he created. Average is, of course, 0.

So that’s runs above average. Now what about DIPS? All DIPS tells us is that there is a lot of luck involved on what happens when a ball is put into play against a pitcher. Because of that, while balls in play have a huge affect on ERA, they often make it harder to judge a pitcher’s true ability, not easier. To combat that issue, we can look at only pitcher-dependent statistics (SO, BB, HR, HBP) to compute a pitcher’s expected ERA. In the short run, DIPS is a much better indicator of a pitcher’s ability than is his actual ERA.

Granderson is more valuable than Rollins because he is just a much, much better fielder. Rollins is about average, maybe a little below; Granderson is one of the best defensive center fielders in the game (if not the best).

[DSG]

Quote

David,

We recently had a good discussion going over in the Minor League Forum LINK on the marketability and value of statistical analysts to major league baseball teams, with Tango and Eric Van more or less at the center of the conversation from the analyst point of view.

I wonder what if you think the general consensus opinion that baseball can't or won't compete with the corporate world for the best analytical talent is valid, or subject to change. Since you're a young guy, I was also wondering if you plan on, or would take a lower-paying position with an MLB team, or go where the money and prospects are better and be doomed to continuing baseball analysis as a passionate hobby like most of the rest of us.

I doubt baseball ever will, for the simple reason that there are a lot of people willing to work in baseball for cheap. Is that smart? I don’t think so; a good analyst can add millions in value to a baseball team just by making some very small improvement. If teams are paying $5 million a win on the free agent market, there’s almost no salary that’s too high for a Mitchel Lichtman or Tom Tango.

As for me, working in baseball has always been a dream of mine, but I’m not going to do it for a third of what I could get on Wall Street. If someone wants to offer me a decent salary (less than corporate America, but not absurdly less), I’ll take it.

[DSG]

Quote

David,

has a definitive answer been reached on the exact manner in which pitchers exert their (highly limited) control over their BABIP against? i.e. do they do so by preventing line drives, or by lowering the probability with which any or all of LD/GB/FB turn into hits?

I know that this was discussed in a previous edition of the Annual, but my copy is currently on loan, and IIRC, the article didn't reach a definite conclusion.

Thanks!

A large part of it is a pitcher’s ability at forcing different kinds of batted balls. A small part is fielding ability. Pitchers also control to a large degree the direction in which a ball is hit off them (even after controlling for handedness), and that’s probably a part of it. And some of it is likely just stuff. The Pitch f/x data is going to allow us to probe this question in some really interesting ways.

[DSG]

Quote

Thanks for agreeing to do this, David.

Which metric for assessing offensive performance do you think is most accurate and why? Also, could you give your opinion on LWTS, WPA, and WPA/LI?

I use linear weights, which are by definition what a hitter contributed without regard to context. WPA is a cool statistic and useful for MVP discussions, but I need David Appleman to incorporate park factors into his calculations at Fangraphs.com before I make more use of it. Tom Tango really likes to use WPA/LI, but I don’t really get that—it seems to me like some weird bastard child solution between WPA and pure linear weights.

[DSG]

Quote

What do you think are the implications for MLB of total revenues rising to $6.08 billion while player salaries drop to just under 42% of that number, down from 56% as recently as 2001?

I think the reason this is happening is that the revenues are mostly coming from national sources (MLBAM in particular), which are split up evenly among the teams regardless of how much a team wins or loses. Since you only want to pay a player the marginal revenue which he creates for your team, there’s no reason to be giving any of that away in salaries. However, I don’t think the players are going to accept that explanation so easily, and I do expect some labor tensions eventually.

[DSG]

Quote

David, it seems that every year we have more and more defensive metrics come out to measure defensive proficiency on an individual player basis, yet less and less of a consensus on whether such a goal is even attainable. How close are we to measuring defense in a way that is accurate and makes sense?

Fairly close. All the play-by-play metrics mostly agree at this point, especially when you’re talking about projections rather than retrospective looks at single seasons (where there is more variation, especially in the magnitude of individual numbers). With that said, there are some individual players on which the metrics disagree completely (Ichiro, Sizemore). Part of that is probably data quality and park factors. With more data coming in each year, both those issues should be easier to solve as time goes on.

The other issue is just that there are limitations to what we can do with the numbers we have now. When someone finally starts tracking hang time and trajectory for every ball in play (as well as positioning), we’ll be able to completely solve the issue of inaccurate defensive ratings (well, almost completely).

[DSG]

Quote

David, how do you see park factors as affecting the measurement of defense? Does Fenway's unique dimensions present undo challenges to looking at Sox outfielders? Are there any other parks that present interesting challenges for analyzing defense?

There’s definitely a big effect; I had to wrestle with that in doing park factors for the THT projections to be published in the Hardball Times Season Preview 2008 (by the way, the cover isn’t up on the ACTA website yet, but it is the most beautiful thing I have ever seen not named Scarlett). Baseball Info Solutions has started tracking which balls hit off the wall and throwing them out of their dataset, which is the right thing to do. Hopefully, others follow suit. It’s not impossible to measure defense in Fenway, but you do need to be mindful of the wall and other park factors.

Other parks that are weird for defense include Colorado (or it used to be; now, it might just be neutral for outfielders but really good for infielders), Minnesota (sucks for center and right fielders), Pittsburgh (also really tough on outfielders), and left field in Houston (same issue as Fenway).

This post has been edited by DSG: 30 November 2007 - 04:05 PM

[DSG]

Quote

David, thanks for doing this. Here are two Covelli questions for you. According to both Win Shares and Baseball Prospectus’s metrics, Crisp had a defensive season for the ages in 2007, thus making him a very valuable player in spite of his sub-mediocre offensive performance. Do you agree with this conclusion?

Also, it seems to me that, given Crisp’s great defensive skills and the Sox’ relative absence of glaring needs at other positions, it makes sense to keep him as a part-timer in 2008 unless he can be traded for the team’s only significant weakness: a good catcher who could supplement or replace Varitek. What are your thoughts on this?

Depends on your definition of valuable. According to my numbers, Crisp was -10 (position adjusted) runs as a hitter and +9 as a fielder. So overall, he was about average, a little below. An average player these days is worth around $9 million on the free agent market. Seriously. On the other hand, a team full of average players won’t get you to the playoffs (well, it might, but it’s not too likely to).

I think that keeping Crisp wouldn’t be the worst thing in the world given that Manny and Drew are not the most durable players in the world, and he could be useful as a defensive replacement in left. With that said, if Crisp is part of a trade that nets the Sox Johan Santana, I would part with him in a millisecond.

[DSG]

Quote

RE: your batting projections, on page 11 of the THT 2007 Season Preview Book, you (or studes or Chris Constancio) write:

“We assumed all minor leaguers are average fielders.”

Why?

Because we didn’t have reliable fielding statistics for minor leaguers. That’s all going to change in this year’s edition, where we will use zone-based fielding ratings for minor as well as major leaguers.

[DSG]

Quote

Which leads into another question I have, what do you feel is the most effective calculation for park effects when measuring offense?

This stuff can get extremely complicated if you want to do it totally right. For the vast majority of purposes, I think Patriot’s park factors work just fine.

[DSG]

Quote

Hey David:

A) A common theory, which has been discussed in the Voros thread, is that we've reached the point of diminishing returns as far as innovations in batting and pitching metrics go. That is, the bulk of the formulas to measure those facets of the game have already been concocted and refined to the point where only negligible improvements in accuracy can be gained with any further refinement. Do you agree? If not, where do you see the greatest need or opportunity for those innovations to occur? What can be measured in pitchers and hitters that isn't already being looked at?

B) There are a plethora of player projection systems out there, all claiming to be more accurate than the next. Typically their authors will cherry-pick a few players whose output closely matched the forecast and say, "See? We told you so." Do you feel there's some other system, or an adaptation of a current system, that will finally bear some fruit that's worth eating?

Thanks in advance!

I think that one thing we don’t know about either or pitchers is how to accurately measure when someone is about to drop-off or breakout. Every year, tons of players establish a new performance level and we ascribe that to luck or something that equally can’t be measured. I think that there’s still a lot to be done in the area of predicting future performance using similarity scores, traditional statistics, and pitch-by-pitch data that is only now becoming available.

On a similar note, one of the things we’re trying to do with the Hardball Times projections is not only incorporate as much play-by-play data as we can, but also to use that data to explore new areas. I doubt our projections will ever be significantly more accurate than anyone else’s, but hopefully they cover more ground and do a better job of it.

[DSG]

Quote

David,

Can you talk about what you think Pitch F/x data's biggest contribution to analysis will be?

I think Pitch f/x will let us understand the effect of stuff on pitching and better predict how a pitcher’s stuff will affect his future performance. Of course if they ever install Pitch f/x in minor league parks, that’ll create a whole revolution in player analysis.