Archive - strategy RSS Feed

The Padres Are Not Optimizing Their Lineup

April 6, 2010 in Daniel Gettinger, Sabermetrics, San Diego Padres, strategy with 16 Comments

by Daniel Gettinger

At Fangraphs, Dave Allen took a look at which managers best optimized their lineup in their team’s first games. The San Diego Padres ranked third-worst in lineup optimization.

Sky Kalkman wrote a nice piece summarizing the Book’s findings on lineup optimization. Essentially a lineup should look like this:

Leadoff: High OBP guy

Two-Hole: One of the best batters on the team.  High OBP is nice.

Three: Not as important as traditionally thought

Cleanup: Best Hitter on the team with power

Five: Best hitter after 1, 2, and 4

Six-Nine: Decreasing order of ability after other slots are filled (with some adjustments for speed), but bat the pitcher eighth.

So, how did the Padres do at optimizing the lineup yesterday?

As Allen finds: not so well.

Using Baseball Musings lineup analysis tool, and imputing my own projections for the Padres batters, I found that the Padres will score 0.281 runs per game less using yesterday’s lineup compared to their optimal lineup.  Over the course of a season, that equates to approximately 4 wins.

Edit: I think the lineup tool may be overstating the impact of moving the pitcher from ninth to eighth.  The real impact should only be 2-3 runs over the course of a season.  But using the lineup tool, Allen found teams gain about 25 runs over the course of the season by making such a move.  I think the Padres optimal lineup may add something like 1-2 wins, rather than a full 4.  Still, a win is a win.

According to the tool, the optimal lineup looked like this:

1) Cabrera, 2) Blanks, 3) Eckstein, 4) Gonzalez, 5) Headley, 6) Hundley, 7) Venable, 8 Pitcher, 9) Gwynn.

However, I believe the seventh suggestion might actually be the best.  According to the tool, it would only cost the Padres 0.004 runs per game relative to the optimal lineup, but looks less funky:

1) Cabrera, 2) Blanks, 3) Gwynn, 4) Gonzalez, 5) Headley, 6) Venable, 7) Hundley, 8 Pitcher, 9) Eckstein.

The main problems with the Padres’ opening day lineup were:

1) Eckstein batted second.  The second spot in the batting order should go to one of the team’s best hitters, not one of its worst.

2) The pitcher should be batting eighth, not ninth.

3) Adrian Gonzalez should be batting either second or fourth, not third.

I would like to see Bud Black correct these lineup errors.  There are a number of reasonable choices for the lead-off spot, and how he orders 6, 7, and 9 is not entirely meaningful.  But Eckstein should not be batting second, the pitcher should be batting eighth, and Gonzalez should not be batting third.  These small corrections do not mean much in any particular game, but over the course of a season, they are likely to add up.

Bullpen usage

April 30, 2008 in clutchiness, San Diego Padres, strategy with Comments Off

One of the things I like to do is see how Bud Black is using his relievers. Fan Graphs makes that easy with leverage index. For this little analysis that I do, note that I use gmLI, which is the leverage index when the pitcher enters the game. I don’t think a pitcher should be able to create his own clutch situations or whatever. Or, more correctly, I don’t think we should worry about that when we’re evaluating how Bud Black uses these guys. If he brings in Kevin Cameron with a 3 run lead in the 8th, that’s a pretty low leverage spot. If Cameron proceeds to give up 5 hits in a row and ends up pitching in relatively crucial spot, well, Bud Black wasn’t planning on that. I suppose there could be an argument for using a different type of LI because he is leaving that guy in there, but for now I’m using this one. (edit: Now that I think about it, perhaps inLI would be better for next time, although I don’t think it would make much of a difference). Here’s a detailed description of leverage index from Tom Tango. So below we’ve got gmLI with innings and ERA in parenthesis:

Hoffman: 2.57 (8.7/7.27)
Thatcher: 1.84 (12.7/7.82)
Bell: …… 1.72 (15.7/2.87)
Meredith: 1.34 (12.7/4.26)
Rusch: ….. 1.04 (14.3/6.28)
Cameron: .95 (8.3/4.32)
Ledezma: .65 (15.7/1.15)
Gonzalez: .46 (3.3/10.80)

Hopefully you are quick to note that this season’s ERA is not all that useful, but I put it there anyway. What you would like to do in a perfect world is have your best pitchers at the top and your worst at the bottom. Here’s how I’d line it up — again, in a “perfect world” where you don’t owe anything to Trevor Hoffman.

Bell, Meredith, Hoffman, Thatcher, Cameron, Ledezma, Rusch, Gonzalez

Obviously, you could quibble here or there and I think the difference between Meredith, Hoffman, and Thatcher is quite small. If you can establish a hierarchy like this – preferably based on projections that incorporate minor and major league numbers, scouting data, etc — well, that’s great. But it really comes down to a situation by situation basis, and that’s where you can use superior knowledge of match ups to maximize the talent that you have. Thatcher, for instance, would obviously face a lot of lefties. Our projections might say that Meredith is better than Thatcher overall, but against, say, Todd Helton it would be Thatcher that would be the optimal choice.

I hear some people say that if Hoffman can’t close, what can he do. I don’t really get that. In my utopian bullpen, he wouldn’t be the “closer” because he isn’t the best overall pitcher. That doesn’t mean he can’t be a valuable guy in the 7th and 8th (or 6th, or 9th, or extra innings). And you know his best pitch is the changeup. Who does he do best against? Fly ball, power hitting righties? Pitch him against them.

Anyway, at this point of the season, this stuff is probably somewhat fruitless. But I wanted to take a look at how Buddy’s been using the relievers. He seems to have them is pretty solid order, considering that I expect Hoffman to continue to close no matter what. I would expect (er, hope) that Bell rises to number #2 and that the gap between Bell and Hoffman gets much smaller. I presume those things will happen.

By the way, who do you think has the highest gmLI so far? Yeah, Hoffman (well, that’s if you discard Mike O’Connor’s 1 inning).

Snyder’s blast

April 28, 2008 in Jake Peavy, San Diego Padres, strategy with Comments Off

In Sullivan’s article, Peavy’s talks about his pitch selection to Chris Snyder in the second inning:

“It was just a bad pitch selection,” Peavy said later. “I shook off Josh there thinking way ahead of myself, thinking of (Snyder’s) next couple of at-bats. I threw the wrong pitch. It wasn’t high enough and it wasn’t in a good enough spot. Unfortunately, that one pitch cost us the ballgame.”

Later, Peavy lamented his misguided planning — “If I throw a slider, I have no doubt in my mind I can get him out,” he said — and cited Bard’s admonition about overthinking things at the bottom of the batting order.

“You don’t have to save pitches,” Peavy said. “Maybe to a 3-4-5 hole hitter. That’s not taking anything away from Brian Snyder — Chris Snyder, sorry.”

Sullivan concludes:

When you have trouble remembering a hitter’s name, it’s probably prudent to stick with what you know. When in doubt, throw him what he can’t hit.

I know that game theory is like an entire branch of mathematics, so it’s another area I will tread lightly in. And with my last inning on the mound coming as an 11 year old in little league, you might also say I probably don’t know much about pitch selection. You would be right. But I did stay at a Holiday … I mean, I have read The Book as well as some lengthy posts by MGL et al. on game theory and it’s application to baseball, so I must be an expert, right? (answer: No, once again; I’m just trying to be sarcastic … it probably doesn’t help that I’m pointing this out, but I’m not good with words … or sarcasm ; )

Okay, on to the point. Let’s say that we can look at Jake Peavy’s pitches and evaluate each type. So we go through and get this order: Slider, fastball, cutter, change, curve ball. His curve and change up are terrible, relative to his other pitches, in my hypothetical example here. You might say, then, why the hell do you throw them, Jake? My guess as to why is simply because they make his other pitches look better. Right now, he may throw the curve and change 10% of the time. Clearly, Jake would rather throw other pitches and he usually does. However, if he completely eliminates them from his arsenal, then the batters have two less pitches to worry about each time at the dish.

Instead of sitting there in a 2-2 count saying, “well, I bet the slider is coming low and away, or the high fastball, or the cutter, or a change up down and away (or a curve ball even!),” they can simply eliminate those last two pitches from their thought process. This makes, at least in theory, his other pitches less effective.

For an extreme example, imagine Jake eliminating everything but the slider. Hitters are now waiting for the slider every single time. When it starts out in a certain point, they’ll know it’s going to end up outside (because there’s no way it’s a fastball or anything else). When it starts out at another point, they’re going to crush it. Peavy, an elite pitcher, would probably lose his job in the bigs.

Anyway, back to yesterday’s game. Peavy, in his comments above, seems to think that the best way to get Snyder out was with the slider. But if he throws the slider to Snyder every time in that situation, it will begin to lose it’s effectiveness (even though, on average, it gives him the best chance of getting him out). It’s only effective because the other pitches could still be coming to the plate, and Snyder knows that.

I guess my point is that I disagree with Jake and Sullivan. The mistake there was (perhaps) not pitch selection, rather it was execution. Had the fastball been a little higher or a little more inside or a little faster, perhaps Snyder k’s … or sees another pitch and then k’s or pops up, etc. It may sound somewhat paradoxical, but you can’t give ‘em your best pitch every time. 

Inefficiently effective

April 16, 2008 in San Diego Padres, strategy with 7 Comments

Padres beat the Rockies 6-0 — Randy Wolf allows 1 hit, 4 walks, and no runs in 7 innings of work

Often times I hear people say that they wished a pitcher was working more efficiently on a given night. This happens when, let’s say, a guy is through 3 innings with 50 pitches, but has allowed 0 runs and k’ed 5 — or something like that. You want that type of performance, but you want it for 9 innings. I usually wonder if it’s correct to want the guy to be more “efficient” — and by efficient, I mean limiting the number of pitches thrown to each batter. I suppose there could be multiple definitions of efficient pitching (like pitches per inning or out), but I’m going with this one because I think it’s what people mean when they say that they want a pitcher to be more efficient. Of course, in a perfect world, a pitcher would throw 27 pitches and record 27 outs (or if he was on the Giants, he would continue to get an out for each of his pitches until his team scored a run).

Anyway, here’s my ‘theory’ (that is certainly, I’m sure, far from original): some pitchers, especially ones like Randy Wolf, benefit from throwing more pitches to each hitter. That is, they are better going deep into counts, just missing on a corner, allowing a guy to foul a pitch off, etc. Sure, you’d rather they get outs in more economic fashion, but as soon as they try to, they get hit hard.

Wolf’s career average pitches/PA is 3.84, according to Baseball Reference (league average = 3.78). Greg Maddux, for instance, the model of pitching efficiency, is at 3.30. Just for a little fun, I looked at Wolf’s ’07 game log: in his 5 lowest P/PA starts, Wolf’s RA was 7.55. In his 5 highest P/PA outings, it was .89 (yes, there is a decimal point there). This is surely not worthy evidence — 10 starts in 2007 by Randy Wolf – and I’m definitely not trying to make any conclusion based any it … just pointing it out. Anyway, if you look at it in terms of P/out or inning, you get a different story. It could also just be a case of selective sampling — or something –where, because higher pitches per plate appearance generally mean more strikeouts, we’re getting all of his good performances. But, I guess, that is kind of my point. For Wolf to be successful, perhaps, he can’t just go right after guys; he has to nibble and work the corners. If Wolf wants to succeed, he, unlike a Greg Maddux, can’t just go right after guys.

With that, I thought we’d take a quick peak at the PITCHf/x data. In the 3rd and 4th inning, according to MLB’s pitch classification system, Wolf had a stretch where he threw 17 consecutive fastballs. All of the pitches during that stretch ranged between 88.3 and 90.8 mph. This has no significance; I just thought it was interesting. I hear people talk about mixing pitches a lot … you know, to confuse the hitter and upset his timing. I wonder if sometimes it’s correct to throw the same pitch that many times in a row, by game theory or whathaveyou. You are certainly confusing the hitter, as he’s probably saying, “well, he’s throw that pitch 5 times in a row to me, something off-speed is coming here.” At the same time, if you keep throwing it, they’ll eventually figure out that’s all you have … during that stretch, by the way, Wolf struck out Tulo, Helton, and Holliday (although Holliday’s k, a 10 pitch ab, eventually came on a changeup).

Here’s Wolf’s ‘pitch mix’ for the night (frequency of pitch with velocity in parenthesis).

Fastball: 82 (89.4)
Curve: 16 (66.6)
Change: 9 (79.8)
Slider: 4 (79.5)

He really went primarily with the fastball on this night (74% of the time. According to Kalk, in ’07 he averaged 59% FB, 19% curves, 11% sliders, and 10% changeups).

***

Regarding the rest of the game, I thought the turning point came on Wolf’s sac bunt attempt, where he reached base in the 5th to load the bases. I was thinking to myself at that point, whether the decision to leave Wolf in here to bunt, rather than opt for a pinch hitter, was a good one. I figured if Wolf reached base enough it might be a decent one (but that going with a better hitter would probably be a wiser choice). As soon as I thought that, Wolf laid down a perfect bunt that was fielded by an off balance Jimenez. Wolf hustled down the line and beat the high throw. After that Brian Giles ripped a double off the wall in right, which was actually the biggest play in the game (according to wpa; +.178). Kouzmanoff and Edmonds both added doubles and the Padres were on their way to a nice 6-0 victory.

Bard’s bunt

April 7, 2008 in Sabermetrics, San Diego Padres, strategy with Comments Off

Dex has a terrific analysis of the sac bunt in the 7th inning yesterday (I recommend checking it out, along with the comments over there). The situation was a 2-2 game, 7th inning, 0 outs, Hairston on first, and Bard at the plate.

A run expectancy table gives us this:
man on first, 0 outs: .953
man on second, 1 outs: .725

As Dex notes, however, what we are more concerned in is the win expectancy:
man on first, 0 outs: 64.5%
man on second, 1 out: 62.4%

So the bunt play there reduced our chances of winning … but, as you might expect (and as everyone at GLB would say), that analysis is far too simple. For one thing, the win expectancy is based on average situations. This was certainly not an average situation. It was Scott Proctor on the mound, Bard at the plate, Hairston on first, Clark on deck, Dewitt at third, in Pecto, and on and on.

Furthermore, if you’ve read MGL’s great sac bunt chapter in The Book (or watched some baseball), you know the sac bunt attempt does not always result in a successful sac bunt. According to MGL’s research (check p. 249), a sacrifice bunt attempt is better than a successful sacrifice, due to the fact that the positive events that can happen on the attempt (hit, error, etc) outweigh the negatives (primarily making an out without getting the guy over).

So, one thing that’s important to consider is that the batter can reach base on a sacrifice attempt. Now, Bard’s speed is not very good and I’m honestly not sure about his bunting ability. If I’m reading this correctly, he’s a whopping 1 for 7 (14.3%) getting hits on bunts. It’s tough to tell if he’s a good bunter, but considering the fact that he’s a relatively slow catcher who has only bunted 7 times in his career, I’m going to assume he isn’t one.

Also, watching the bunt again, Bard squared around on the first pitch of the at bat (and did so pretty early). You can’t see the defensive alignment, but considering the game situation and the fact that he wasn’t afraid to square early, I would guess they were playing for the bunt (especially on the second pitch of the ab). With the defense playing the bunt, Bard had less of a chance of reaching base or getting the sacrifice down. If he decides not to bunt on the second pitch, he’d have a better chance to squeak a hit past Dewitt (again, assuming he was playing in for the bunt). Along these lines, there is also the game theory aspect to consider, which The Book also discusses in depth.

Anyway, take a look at Bard out of the box. He isn’t running hard at all. He kind of takes a little hop out of the box and jogs down the line. He’s out by a mile. It appears that this was a clear case where they traded an out for a base, and that’s all they wanted. I’d guess it was, ever so slightly, the wrong decision if you incorporated all the variables. But the main point here is that there is so much to consider, unless you actually go through the exercise, it’s very difficult to pin down what the correct move in many of these cases. That doesn’t mean that there can’t be great discussions about them, though, as Dex’s post and the ensuing comments display.

Page 1 of 212»