Palmer on DiPS

Pete Palmer was interviewed in April of this year, and was asked about Voros McCracken's Defence-Independent Pitching Statistics. Palmer is arguably the most important sabermetrician OF ALL TIME. Certainly his only rival is going to be Bill James, so reading Palmer's comments on DiPS theory, which James himself regarded as important, makes an interesting comparison and contrast.

James referred to McCracken's work in the New Bill James Historical Baseball Abstract (p 885):

3. This knowledge is significant, very useful. 4. I feel stupid for not having realised this thirty years ago.

Palmer, however, has a very different take on the matter.

I didn’t have a lot of faith in [DiPS]....[McCracken] said there wasn’t a great amount of correlation from season to season. But as I said, the variations due to chance and everything in sports, baseball in particular, is a lot higher than people think. Your average could drop 60 points from one year to the next, and it’s not really statistically significant because 500 at-bats isn’t that many at-bats to verify what your current batting average should be.

Whether this opinion is rooted in statistical analysis or not, it does conform somewhat with the analysis provided in "Solving DiPS", a compilation of an on-line discussion which you can find a copy of here. One key solution in "Solving DiPS" is that, given 700 Balls in Play, some 44 per cent of the outcomes are a consequence of random variance, the single largest factor. (Pitchers were assigned 28 per cent, fielding 17 per cent. Hold that thought for a moment.)

I have seen it suggested that Palmer does not understand DiPS, which has become a tool for projecting a pitcher's future. But from the perspective of evaluating a pitcher's season, Palmer's lack of "faith" makes more sense. BABIP's variance is irrelevant, because it is in the nature of the game. What is important is to convert extra-base hits into singles, and singles into outs.

When you don’t look at walks and strikeouts and home runs, you’re actually minimizing a difference between a good pitcher and a bad pitcher. And therefore, the gap in that category is going to be artificially low because some of the factors that would make it higher are not counted.

In other words, we shouldn't be surprised that pitchers appear to have limited or no control over the outcomes of balls in play. That has never been where the difference has been visible in the small sample size of a single season.

Finally, to return to those percentages from "Solving DiPS", what might be surprising from the traditional reception of DiPS is that pitchers have more control over the outcome than fielders. So, again, perhaps we should be a bit more sceptical, like Pete Palmer, of those making grandiose claims for DiPS. Insofar as anything has control over the outcome of the batted ball, it is the pitcher. Random variance in its nature is uncontrolled.

Should We Worry About Jordan Zimmermann?

Jordan Zimmermann has been my favourite National* since I started following the team closely in 2009. (Despite the pitiable record, still one of my favourite seasons, as Mike Rizzo coped with a roster he did not assemble.) I have found his starts this season, however, somewhat lacking in the standards previously established by him. I thought about looking into it, and I have assembled a few charts from BrooksBaseball.com to illustrate my concern.

First, here is his start on 18 May, against the Mets:

Then, a start from exactly one year ago, 18 May 2013, against the Padres:

And, finally, a start at about the same time of year against the Mets, 4 June 2013:

The first thing to note is that Zimmermann is throwing a lot more change-ups, and a lot fewer breaking balls. Theoretically, this is good, because breaking balls apparently lead to arm trouble. However, his fast ball does not appear to be moving as well as it did in his starts last year. In fact, even his change-up isn't moving quite as well as it did last season. His PitchF/X velocities as recorded at Fangraphs.com show his average fast ball travelling at a slower speed, while his breaking balls are a bit faster. Going back to the charts note that the fast ball spread is not as tight this year as it was last. The slower ones are a bit slower, and there aren't quite as many faster ones. Now, the charts also show his numbers so far for 2014 are comparable to the 2012 ones rather than the 2013 ones, and that overall he has gained velocity since 2009. It's all a bit of a muddle, and I have no conclusion to offer, only that question. Should we worry? Let's create a watching brief on this one.

EDIT: I stupidly forgot to mention that Dave Jageler's comment about Zimmermann being 'wild in the strike zone' during the radio brodcast on 18 May against the Mets was what sent me to look at BrooksBaseball.com In fact, it seems more true of 2013 games than the one on 18 May.

______
*Jayson Werth is my favourite positional player, but I like pitchers best of all.

Riggleman's Pitch Counts

This study was half-finished a few days ago, but then I went to a conference. I hadn't even planned to do more work with the starting pitching, intending to look at another facet of the Nationals' spring training questions instead. But Mr Riggleman brought up suspicions some of us have had about whether he puts too much stress on his pitchers' arms.

To make a stab at seeing whether Mr Riggleman has adopted a new approach, I used Baseball Reference to assemble game-by-game data for all the starters used during Mr Riggleman's tenure. They included three starters used by Manny Acta, so looking at them together, we might get a sense of it all.

The study found the average pitch counts, the standard deviation of these and, based on the standard deviation, an anticipated maximum number of pitches per start, the number of outings and the number of times the anticipated maxium was exceeded. I also added in the average percent of Win Probability Added, to see whether more successful pitchers were kept on a longer pitch-count leash; and age, to see if older pitchers were allowed to go deeper.

Under Manny Acta:

Pitcher    Age   AvgPC   PCStDev     Max    Out    Max+    WPA%
Lannan       24    95      11.7      107    17      2        4
Stammen      25    88       9.6       98    10      1       -8
Detwiler     23    88      10.7       99    10      1      -12

Under Mr Riggleman:

Pitcher    Age   AvgPC   PCStDev     Max    Out    Max+     WPA%
Lannan       24    94      19.4      113    16      2       -1
Stammen      25    77      20.5       98     9      0      -10
Detwiler     23    75      25.2      100     5      0       12
Mock         26    93      12.4      105    15      3      -11
Martin       26    86      19.5      106    15      0       -3
Balester     23    73      15.7       89     7      1      -13
Livan        34    98      10.5      109     8      1       -8

This isn't quite what I expected to see. The big difference between Riggleman and Acta is not pitch counts (although there is a difference), but whether or not to leave a pitcher in. Acta appears to have quite a strict 'pitch count' mentality, setting a target for a pitcher and pulling him quite quickly once the target is reached, thus being a relatively 'slow hook'. Riggleman, by contrast, was more mercurial in the sense that he might pull a pitcher very quickly (Martin, 25 July, 28 pitches!) or leave them in for a very long time (Lannan, 25 September, 122 pitches!). This isn't in the table, but based on Win Probability, pitchers under both these men were more likely to run up maximum-busting totals if they were not pitching well.

I'd say Mr Riggleman, relative to Manny Acta, is a quick hook. I'm also not convinced that he's learned as much as he thinks.

Pitcher Options

You probably already know that Livan Hernandez has signed with the Nationals. Here's his Marcel projection, in line with those at this entry I posted yesterday. I'll throw in Miguel Batista as a bonus.

            ERA      FIP
Livan      5.29     4.79
Batista    5.28     5.18

On the basis of their projections, Batista looks the weakest candidate. Livan's FIP (Fielding Independent Pitching, which just counts home runs, walks and strikeouts) puts him ahead of Balester, JD Martin and Olsen(!), but his ERA is the worst of everyone's.

I actually came to write a post about rotation candidates with options still available, and the Livan news ensures that I get a longer post than I had planned.

The simple fact is most of our rotation candidates still have options. We can be confident that Lannan and Marquis are set. Olsen seems to be 'a lock' according to Mr Riggleman. The people who fill out the last two spots will presumably be those who do well in spring training, with one caveat — if they are out of options, they have to clear waivers to make it down to the minors.

This may affect two pitchers. One, I'm sure of: Matt Chico is out of options. I also think he's potentially good enough to draw a waiver claim. The other — JD Martin — might have option problems. The option rules are a bit complicated, given the kind of data one can access easily. Martin appears to have been a six-year free agent, so if he's been on a 40-man roster any length of time he can easily have used up his options.

As a consequence, I'm designating Chico as the likely #4 starter. He could get a bullpen role, but there's actually a logjam there. That's a story for another time.

6 March 2010 EDIT: Brian @ Nationals Farm Authority has a full options run-down here. Seems JD Martin still has all his options left.

All Pitching, All the Time

That's the spirit of Nationals' coverage so far this season. We've had the excitement of Stephen Strasburg's bullpen session combined with the news that Mr Riggleman thinks he'll be starting the season in the minor leagues. We had Mr Riggleman's word that Drew Storen was headed in the same direction. Finally, we have all sorts of rumours.

I did a bespoke projection for Chien-Ming Wang just the other day, but this time, more as food for thought, I'm going to present the straight Marcel projected ERA for Nationals' rotation options as listed by CBS Sportsline. ERA is obviously earned-run average. FIP is one of those relatively newfangled sabermetric stats that gives a value to home runs, walks and strikeouts that can be turned into an ERA value. It has a better predictive ability than ERA. It's worth mentally adjusting it for defence, in that a good defence will reduce the FIP, and a bad one will increase it.

           ERA        FIP
Lannan     3.98      4.58
Marquis    4.23      4.35
Olsen      4.66      4.85 (shoulder injury)
JD Martin  4.41      4.81
Mock       4.76      4.21
Martis     4.37      4.76
Detwiler   4.26      3.95 (hip surgery)
Balester   4.83      4.90 (ribcage woes)
Stammen    4.42      4.45 (elbow chips)
Chico      4.65      4.72 (elbow surgery)
Wang       4.81      4.19 (shoulder surgery)

I've left off a few names, like Jordan Zimmermann, who isn't going to pitch this season, and some of the veterans lurking around the camp.

To say the least, we've got a lot of options. Also, there are some things that are hard to credit based on what we have seen so far, such as Balester posting an ERA under 5. Basically, going forward, one would expect on this basis to open the season with a rotation of:

Lannan, Marquis, Olsen, JD Martin and Mock.

Next man off the cab rank would be Martis, or maybe Chico. What will happen when Wang is fit is anyone's guess.

Of course, player options are going to play a part in the final outcome. And spring training will render these statistics more or less believable. After all, they are projections, not predictions.

DIERA??

Moving away from the Nationals postmortems for a bit, it's Hall of Merit time in the Sabermetric City. I was a latecomer to the Hall of Merit, but it's given me hours of fun. Just like last night!

To fill you in - I'd filed a preliminary ballot based on some people who'd been getting lots of votes, when I noticed from Rick Reuschel's Davenport Translationspage at Baseball Prospectus that he had an awful lot of WARP3. Anyway, I prefer to home-brew my own stats than use other people's. (Unless they are from the Big Bad Baseball Annual!) So, how to get a quick-and-dirty One Big Number for a pitcher?

I failed on the 'quick' part, but I made a stab at creating a Defense Independent Earned Run Average, based on Clay Davenport's article on his Fielding Translations in the 2002 Baseball Prospectus. Here's how it works.

Stats you need:
A pitcher's seasonal Games, Innings Pitched, Hits, Home Runs, Bases on Balls, Hit Batsmen, Strike Outs and Earned Runs.

1) Create a new category, called FldgHits. Subtract Home Runs from Hits.
2) Multiply the IP by 3. Subtract strikeouts. Take 30% of that number and add the strike outs back in. Divide by 3. This is 'PitchIP'.
3) Use a modified version of Equivalent Average and calculate the PitchingEqA and the FieldingEqA as follows
Pitching (2*(30% of FldgHits))+(3*HR)+(1.5*(BB+BP))/(IP*3)+(FldgHits)+BB+HBP
Fielding (2*(70% of FldgHits))/(FldgHits)+(IP*3)
4) Divide the PitchingEqA by the FieldingEqA.
5) Subtract 1 from the ratio in step (4). This gives you PitResp.
6) (PitchIP*3)+(30% of FldgHits)+HR+BB+HBP = PitchPA
7) (70% of ((IP-SO)*3))+(70% of FldgHits)=FldgPA
8) [(1+PitResp)*PitchPA]/[(1+PitResp)*(PitchPA+FldgPA)] = PERmultiplier
9) [(Earned runs)*(PERmultiplier)/PitchIP]/9 = DIERA

Ah, but you're not done there. The big number has to be a 'winning percentage', and to get that you need to perform all the same calculations for his league. But we'll come back to that after you've had some time to digest this.

Reuschel's best season on this measure is 1985. ERA+ agrees but says his second best is 1977. I think I might take a look at 1981 with the Yankees.

Catching Up #3 The Zimmermann Injury

When the Washington Nationals' Jordan Zimmermann was diagnosed as needing Tommy John surgery, I possibly took it more philosophically than other bloggers.

I dealt with Zimmermann's PECOTA comparables before. Two of the three didn't have particularly long careers, and the third, John Maine, is by no means a stud. His ERA+ has been in decline since 2006, and his K/9 since 2007. I suspect he'll join Erik Hanson and Bobby Jones on their career paths, not getting much past 30.

First-Half Nationals' Review: the Pitching

Pitching is 75 per cent of the game. Or something like that. Maybe 52 per cent, but there's some defence in there, too. Actually, where the Washington Nationals and the fabled 'Plan' are concerned, pitching is probably 75 per cent, if not more. As Definitely-Not-Interim-or-Acting President Stan Kasten said the other day, 'It's hard to amass a core of young future starters. That took us time. I've always believed that that was the most important thing, that was Job One.'

And, frankly, it's still a work in progress. There are signs of a rosy dawn in the future, but not necessarily where you'd think. But right now it's still night.

Rather than trot out some traditional statistics, or even some old-fashioned sabermetric ones, I'm going straight to the 21st-century granular approach here. I'm going to talk about LD% (Line-Drive percentage) and HR/FB ratio (home runs as a percentage of fly balls). I'll throw in K/9 and BB/9, just for retro effect.

Basically, LD% has an impact on the batting average by a pitcher's opponents. If the LD% is high, the BABIP (batting average on balls in play) will probably be high, too. Young pitchers who don't make quick progress with getting a high LD% lower are likely to have short big-league careers. HR/FB is pretty constant for most pitchers, so if there's some deviation from the league average there, for a young pitcher it's safer to assume that it will drift towards that league average. Taking the pitchers currently on the 40-man, this is the statistical snapshot.

Starter      K/9   BB/9    LD%   HR/FB
Olsen        6.0    3.6   22.5   13.7
Martis       3.6    4.1   15.4   10.2
Detwiler     5.7    3.8   24.5    5.6
Zimmermann   8.8    2.9   25.3   13.9
Stammen      4.0    1.8   19.1    8.8
Lannan       4.0    3.1   18.7   11.8

League Ave   6.9    3.2   19.2   10.3

Martis, Stammen and Lannan are doing a Mark Fidrych impersonation. If they don't get that K/9 up, their careers will be short. Otherwise, Martis needs to get his walk rate down, and the other two are doing a fine job. You can fit guys like this into the 4 and 5 slots in the rotation, and not complain.

Olsen's a league-average innings eater type, with numbers like that, great for spot 3 or 4 in the rotation.

Detwiler and Zimmermann are supposed to be the future studs. Unfortunately, they look pretty uneven here. Zimmermann's got the Ks, but that LD% worries me. It's remained stubbornly high all season. Detwiler's not quite got the stuff, and also has a high LD%, but something about that line fills me with more confidence that you're looking at a future Scott Olsen type, who might turn into a #2 for a season or two. That's actually pretty good, because one never knows. Sometimes people raise their game even higher. But, being realistic, a young pitcher's best seasons for strikeouts are at the start of their career. Then they go down.

Well, this review is partly about the coming 'Riggleman Era'. My advice? Keep Stammen in the rotation. At one time their was a rumour that he was headed for the bullpen, but he's doing too good a job starting. I think that plan has been set aside with Detwiler's return to the minors, but you never can tell.

As for the bullpen, I'll just go straight to the chart

Starter      K/9   BB/9    LD%   HR/FB
Beimel       5.6    3.8   16.7    6.2
Bergmann     6.2    3.9   23.5   15.3
Burnett      7.1    1.4   11.9   17.0
Clippard     9.0    3.9   10.0    7.8
Kensing      3.6    7.2   29.2   14.6
MacDougal    3.7    7.4   14.3    8.5
Mock         4.2    5.5   10.9    0.0
Tavárez      8.4    6.8   17.2    3.4
Villone      4.9    6.4   20.3    6.2

League Ave   7.4    4.0   18.7    8.8

Well, what a mess. Walks are killing this bullpen. Look at MacDougal, he's got the numbers the wrong way round. 7.4 would be acceptable as a K/9, but as a BB/9 it's a sentence to transportation. Here's where Mr Riggleman has his work cut out for him.

I'd do something like this:

Closer: Burnett
Setup: MacDougal, Clippard, Tavárez, Beimel
Mop-up, Long Relief: Bergmann, Villone.

Villone I'd keep on a short lease. The Nationals have Dave Williams, a former big league left-hander, at Syracuse. If Villone doesn't pick up his game, I'd DFA him and try Williams.

Beimel has tolerable numbers, but his performance doesn't measure up. I believe he's added a few goats to his existing collection.

Tavárez may be done. If he could get his walks down, though, that's a closer there. MacDougal may well be in the same boat, to be honest.

The rest are a bit mix and match at the moment. I have a feeling they are going to miss Kip Wells. He somehow could dial it up for an inning, but he was screwed if he had to go two. Then he got hurt, and didn't come back the same pitcher. They won't miss Kensing, though. Didn't he cost them a real, live player?

St Claire vs McCatty - Some Data

So, on 2 June 2009 Randy St Claire was fired and replaced by Steve McCatty. There's been some favourable comment about the new regime — allegedly pitching has improved. The trouble with making direct comparisons of 'St Claire 2009' with 'McCatty in the Big Leagues' is that McCatty is working with significantly different personnel over a shorter period of time. So, before it's too late, I decided to take a look while the names they both had to deal with were much the same.

From 2nd June to today is about the same length of time as from 20th May to 1st June. (In fact, McCatty has had an extra game in that time.) By 20th May, Daniel Cabrera had been removed from the rotation, and Scott Olsen was on the disabled list. The youngsters Stammen and Detwiler entered the rotation and have remained since. The bullpen hasn't change dramatically, with Kip Wells going on the disabled list. So 20th May to 13th June involves much the same personnel.

First, let's use the old Big Bad Baseball Annual Quality Matrix (QMAX) to analyse the starters. (I like QMAX, because it actually uses individual starts, as opposed to lumping everything together, and offers a more nuanced picture than the Quality Start.)

Coach       Success (Elite 
            Square   Square)   Hit Hard      --- 
St Claire      3       0           5          2
McCatty        6       2           1          2

The Elite Square is a subset of the Success Square, so don't add those together. In this case, it's a solid win for McCatty. This makes sense, after a fashion, because St Claire built his reputation working with veteran pitchers, and suddenly there weren't any in the Nationals' rotation. McCatty's more 'instinctual' approach might work better than St Claire's 'technical' methods with younger pitchers still learning the broad principles of pitching.

For relief, I took two elements. The Win Expectancy (WE) situation when the relief pitcher entered the game, and the net plus/minus after comparing the ideal result (no change to the leader's advantage in runs if the change occurs at the start of an inning; or, if the change is made mid-inning, all batters faced make outs) with the actual result.

Coach      Average WE             WE       WE+
           at entry              effect
St Claire      .254             -1.378     -1.79     
McCatty        .330             -3.230     -2.49

The WE+ is an attempt to translate the win expectancy effect to a common standard. The problem with the raw numbers is that the better performance of the starters under McCatty mean there is more win expectancy to lose than there was under St Claire. (Think of it as a reliever having an 8 per cent better chance of entering a close game under McCatty; there's just more win expectancy to lose to begin with.)

Basically, the bullpen under McCatty is having an even more difficult time than it was under St Claire. Some of that may be down to the fielding. If you take away the 12th June game, lost in part by Nick Johnson misplaying a foul pop-up, the WE+ number comes down to -1.88. Maybe the difference isn't so big.

On balance, from this evidence, the improvement in the rotation probably justifies the switch from St Claire to McCatty. I'd go so far as to say that a pitching coach really must stand or fall on how effective the rotation is.

Nats' xFIP Horror Show

For me, the best thing about the Hardball Times' web site is the statistics. (Well, next to some excellent articles from time-to-time, a subject I shall return to, I hope, over the weekend.)

Of course, if one is a fan of the Washington Nationals, this can create some upsetting reading. When I first stumbled over baseball on 'teh Intertube', as they say at Baseball Primer, DIPS had just hit version 2.0, I think, or was about to. The theory of DIPS is that pitchers have very little influence on the outcomes of balls in play. Out of the DIPS debate emerged FIP, or Fielding Independent Pitching, a way of calculating an ERA that set to one side the balls in play. And then the chaps at Hardball Times developed xFIP, which adjusts the home runs allowed in relation to the ratio of fly balls allowed by the pitcher. (Their research seems to indicate that a pretty consistent percentage of fly balls go for home runs, once one adjusts for the effects of the ballpark, and random variations.)

The objective of xFIP is to find a number that will more accurately predict a pitcher's future ERA. It works a bit better for starters than for relievers, although you can use it with caution for relievers. So, how about those Nationals?

Pitcher                  xFIP
Zimmermann               3.36
Hanrahan                 3.90
Detwiler                 4.46
Colome                   4.66
Tavárez                  4.81
Lannan                   4.92
Olsen                    5.04
Wells                    5.05
Stammen                  5.06
Villone                  5.20
Beimel                   5.32
Martis                   5.61
Bergmann                 5.97

Remember, these numbers don't include fielders booting balls, or just not having the range to get to a grounder. This is strikeouts, walks and home runs. It's pretty horrific, especially a closer with a 3.90 xFIP — relievers, to be useful, generally need to be at least half a run better than an average starter, and closers more like a run or more.

Basically, going by xFIP, Zimmermann looks good going forward, Detwiler is about a league average pitcher, and Lannan and maybe Stammen can fill out the back of a competitive team's rotation. And, as of 5 June 2009, that's all Nationals' fans have in the whole bullpen that's really worth keeping.

Martis vs Cabrera

For the past few days I have posted the Quality Matrix, or QMAX, stats for the Nationals starter on the Game Chatters at Baseball Think Factory. For those of you who've forgotten, QMAX was a somewhat controversial method used by the old Big Bad Baseball Annual to evaluate pitchers. It could lead to some odd results, such as the possible equation of Jose Lima and Randy Johnson as pitchers of similar value c. 1998.

In exactly that vein, I observed that Shairon Martis, not exactly a Young Phenom™ but still a promising pitcher, and Daniel Cabrera were more alike than they might seem. To which an astute Think Factory reader responded to the effect that I should get my head out of my spreadsheet and actually watch some games, you know.

It's a fair point. Martis has much more self-possession on the mound — in other words, The Good Face™. Now, I actually put a lot of value in that kind of mature handling of difficult situations. But, my main point was that in terms of what the Nationals had got so far out of the two pitchers this season, there isn't all that much between them. Let's use the QMAX method to compare the two pitchers' overall effectiveness prior to 13 May's game.

Daniel Cabrera
Power Precipice starts (2) - April 19 & 30.
'Meh' starts (3) April 8 & 13, May 6.
Hit Hard starts (2) April 25, May 11.

Shairon Martis
Elite Square (1) May 2.
Success Square (1) April 16.
'Meh' (1) April 21.
Hit Hard (3) April 10 & 27, May 8.

Today's start added another to the Success Square, so with this he surpassed Daniel Cabrera in terms of contributing valuable starts to the 2009 Nationals. (He's also showing an 'iambic' pattern of good start/bad start.)

But there's another matter here. Daniel Cabrera is about to turn 28, Martis is just 22. When you see young pitchers putting out Power Precipice starts, you think 'wow, there's a good risk for a future'. Daniel Cabrera, however, is a little old to still be flirting at the Power Precipice. It's got that name for a reason—the pitcher might fall off, but might turn into Randy Johnson. Cabrera may be a serviceable fourth or fifth starter on a second-division team, but my Game Chatterer is quite right to express scepticism over any attempt to avoid a harsh judgment that Cabrera, on the basis of his 2009 starts, is anything other than a stopgap, a $2.6 million one-year gamble that at the moment looks like not paying out much more than a dime.