Late Game Management
We won! Woohah. However:
This isn’t anything we haven’t pointed out before, but since I said good things about Mike Hargrove’s line-up before the game last night, I have to point out how terribly he managed the end game.
Seattle – Bottom of 8th Score
Ron Villone pitching for New York
R Cano at second base.
E Perez walked.
W Bloomquist ran for E Perez.
Y Betancourt sacrificed to pitcher, W Bloomquist to second.
K Johjima hit for R Rivera.
K Johjima grounded out to third.
I Suzuki intentionally walked.
T Bohn hit for C Snelling.
T Bohn struck out swinging.
The inning starts out well, with Eduardo Perez getting the leadoff base on balls, which the announcers will tell you always comes around to score. They’re wrong, as you’ll see later, but it comes around to score even less often when the manager bungles the rest of the inning.
Using Tangotiger’s Win Expectancy chart, we see that the M’s are in a good situation. Tie game, bottom of the 8th, runner at first and nobody out, they have a WE of .673. In other words, the M’s would be expected to win this game two out of three times, considering they have 6 outs to go vs the Yankees 3 outs, and they have the potential winning run on base with no men out. It’s a good spot to be in.
Bloomquist pinch runs for Perez, which is totally defendable, especially if you’re going to have him steal second. If Bloomquist steals second, the WE raises to .738, a six point increase. If he gets thrown out trying to steal, the WE drops to .559, an eight point decrease. For the steal to make mathmatical sense in this situation, Willie would only need to be successful about 60% of the time. He’s 75% on the season and 85% for his career. Posada’s throwing out 36% of base stealers this year, and 30% for his career.
No matter how you break it out, the math works. Bloomquist is probably around 80% likely to swipe the bag there, which makes the reward well worth the risk.
Instead, the Mariners decide to have Willie stand on first and ask Yuniesky Betancourt to lay down a sacrifice, getting Bloomquist to second but using an out in the process. The WE goes from .673 to .656, a 1.5 point decrease. Yes, the runner is now in scoring position, but the Mariners are actually less likely to win the game now than they were before the bunt, because we’re running out of outs and are still tied.
The sac bunt to get the runner from first to second, especially when the runner is a base stealing threat, is a bad managerial decision. And Hargrove does it all the freaking time. If you want Betancourt to lay one down, have Bloomquist try to steal second first. Then, if he’s successful, bunt Willie to third. That’s okay, if you have to bunt. Getting the runner from second to third with less than two out is a big deal. But stop with the bunting a fast runner from first to second. It hurts the team’s chances of winning.
Then, of course, he compounds the problem with the absurd but totally predictable decision to pinch-hit T.J. Bohn for Chris Snelling. We like the fact that Bohn is here instead of Jones as the designated late game defensive replacement/pinch runner, but let’s not fool ourselves – he’s not much of a hitter. He’s vulnerable to breaking balls, chases pitches out of the strike zone, doesn’t make great contact, and has only gap power. Of course, he’s right-handed, while both Ron Villone and Chris Snelling are left-handed, so it was the by-the-book move to make.
It was just the wrong move to make. Chris Snelling can actually hit. He’s never shown a platoon split in the minors. He doesn’t have a swing that is exploitable by LHP’s. The M’s, needing a base hit to take the lead heading into the 9th inning, replaced a good hitter with a poor hitter, all in the name of getting a platoon advantage. A good LHB is a better bet against a good LHP than a bad RHB is. Mike Hargrove consistently fails to grasp this concept, and Bohn predictably struck out. Asking a kid to make his major league debut in that circumstance isn’t fair anyways, and having him pinch hit for a superior hitter is just foolish.
George Sherrill pitching for Seattle
T Bohn in right field.
K Johjima catching.
J Damon flied out to right.
J Mateo relieved G Sherrill.
D Jeter singled to right.
B Abreu struck out swinging.
A Guiel hit for C Wilson.
A Guiel walked, D Jeter to second.
A Rodriguez struck out swinging.
I’ve railed on this decision so many times, it’s not funny, but it has to be pointed out, because Mike Hargrove simply fails to grasp this concept, and it’s such an easy one.
When you’re the home team, and you go into the 9th inning tied, there can never be a save situation. Ever. It can’t happen. If you win, it’s going to be a walkoff win. You will never have a lead to protect. It’s fundamentally impossible.
Facing a situation where the Yankees had the middle of their order coming up, including two hall-of-fame RHBs, Mike Hargrove went with Julio Mateo instead of J.J. Putz. Why? Because it wasn’t a save situation, and J.J. is our closer.
This basic lack of understanding drives me insane, and Hargrove has done this repeatedly. Melvin did it too. It’s absurd, and the fact that a guy whose job is to make strategic decisions to help his team win can’t understand the concept is mind boggling to me.
Yes, it worked, and we won – huzzah. But we won despite the late game management of Mike Hargrove, who actually made it less likely the team would prevail. He made three lousy decisions in the span of 10 minutes, and the Mariners managed to overcome all three. But if you’re looking for reasons why the M’s suck in one run games this year, look no further than their manager. It’s nights like this that serve as glaring examples why Mike Hargrove just doesn’t belong inside the M’s dugout.
The Leverage Index (LI) of the top of the 9th of a tie game is 2.4.
To put that number in perspective, Mariano Rivera’s career LI is around 2.0. Essentially, once the LI crosses the 1.5 threshhold, you should think about (though not necessarily do) bringing in your best reliever. Once it crosses 2.0, you’d do well to bring him in. At 2.4? There’s no one else that even qualifies, other than your your best reliever.
Add in the fact, as Dave is saying, that the other team’s best hitters are also at bat, and there’s really no reason not to bring that guy in.
I also agree that you should never bunt over a speedster to 2b, if that guy is at least a 75% basestealer. WFB is 43 for 51 in his career, 84%, which qualifies him as one of the highest SB% of all time! I can’t really tell why he’s on the team, but that’s gotta be one good reason. And, you take away that weapon? Intentionally?
On the plus side, he had Ichiro batting leadoff.
My WAG is in two parts: 1) Hargrove was so paralyzed by the fear of asking WFB to steal off of a LHP that he couldn’t do it. 2) And, as you point out, Grover doesn’t realize that the probability of scoring one run with a runner on 2nd, one out is not higher than runner on first, no outs. He has probably never looked at conditional run scoring probability charts in his life. And for that reason alone, he should be fired (if he can’t take ten minutes to review numbers that can help him win more often).
And, to clarify, he could be fired for that reason alone, but there are also many others.
Problem is, most managers think that’s the logical thing to do in a tie game late with a man on first and no one out. The next manager we have will probably do the same thing.
At least there are managers who have figured out that there isn’t anything sacred about the save situation.
TJ Bohn season splits vs. Lefties in Tacoma in 2006:
.274/.384/.464 in 84 AB’s
Doyle season splits vs. Lefties in Tacoma in 2006:
.262/.329/.431 in 65 AB’s
As soon as i saw Doyle heading back to the dugout i knew there would be some good ol’ ass chewing today on the board regardless of the outcome of the game.
Doyle’s numbers vs. Righties this year are terrible .200/.325/.309. So why take a guy out who doubled in his last AB, traditionally hits Lefties just as good (if not better) than Righties, and has been getting on base like a yankee? Oh yeah the Lefty/Righty match up….that worked out well. I would almost appreciate Grover getting the axe after a win like last night to demonstrate that we don’t blame you entirely for the 11 game losing streak, we blame you for stupid decisions that the team has to constantly over come just to get a win.
By the way how great was Beltre last night? Did anyone hear what YES had to say about that hard slide to break up the double play?
Small Sample Size Splits are worthless.
Small Sample Size Splits are worthless.
Small Sample Size Splits are worthless.
I’m going to keep saying this until people start ignoring them.
#5 No. Did you? If so, can you tell us?
I think the criticism of the sacrefice bunt is incomplete without looking at the possible outcomes of Betancourt swinging away. I recon that he’s liable to make an out almost 7 of 10 times (his OBP is closer to .300 than to .400). There also is the chance of a doubleplay, although that is probably small since both Bloomquist and Betancourt are fast.
My point is: When criticising the bunt, we also need to look at the likelihood of all possible outcomes and the resulting win expectancies for each outcome.
Well, I imagine someone already may already have done the math on all these scenarios and concluded that the bunt is still a bad idea.
Slightly OT: Isn’t this a topic for “super reader” Paul Covert? I haven’t seen him post in a long time.
Problem is, most managers think that’s the logical thing to do in a tie game late with a man on first and no one out. The next manager we have will probably do the same thing.
And that’s STILL stupid–doncha at least wanna TRY to steal??? (And I’m thinking that there ARE managers who try the basestealing over the sacrifice…..)
I think the criticism of the sacrefice bunt is incomplete without looking at the possible outcomes of Betancourt swinging away.
Um, isn’t that the POINT of the Win Expectancy charts?
Correct me if I’m being stupid here, but it seems to me that the argument for stealing isn’t as simple as the skill of the base stealer and the base-out odds. I recall that a long time ago — and perhaps this has been contradicted since then — Bill James demonstrated that the success of an at bat declined dramatically during all steal attempts — successful or unsuccessful. If this is true, then having a man on second with no outs via a successful steal is not as likely to score runs as having a man on second with no outs via a double.
Is there modern research on the effect of a steal attempt on an at bat, and the consequent effect on run expectancy?
To be fair, I do remember Bloomquist once saying something to the effect of: “I know how long it takes me to get to second, I know how long it takes the catcher to get the throw to second, and the 1B coach tells me how long it is taking the pitcher to get the ball home. So if A is not less than B + C then I don’t go.” (Paraphrased, of course). Maybe Villone was too fast to home?
Sorry Dave, the intent of the splits (small sample size and all) was to show there is very little difference between how Doyle has been performing this year vs. lefties and how Bohn has. Everyone knew it was a terrible decision to pitch hit there based solely on the worn out notion that you have to bat your right handed bench guy vs. their LOOGY. Just trying to back it up with some numbers.
No i didn’t hear what YES said about the slide. I loved the slide, but the replay showed he was very close to a blowning out the defenders knee. I was just curious as to whether the YES guys could acknowledge it was a good hard slide, playing the game like it’s suppose to. Or if they thought it was dirty and dangerous.
(And, of course, the argument against sacrificing is made even stronger by the fact that we’re assuming a successful sacrifice. That is, it’s a bad play even if the sacrifice works, but the sacrifice doesn’t always work, either, making it an even worse play.)
Great, great analysis. Another game won despite the manager (there have been a few — unfortunately, many more lost BECAUSE of the manager).
I am beginning to think I would be happy if the sole qualification of the next manager of my favorite baseball team would be “never gives up an out for second base, ever (I know there are a handful of examples when it is an okay strategy, mostly in the N.L., but come on).”
But having a man on second as a result of a steal, does not effect the succes rate of the AB if–which you should in this situation–then sacrafice the runner to third, having put him there with less than two outs. In fact the result of the AB is the same in that, it was used for a sacrafice attempt.
Well, the NYTimes writer (“Squandered Opportunities Cost Yankees”) didn’t say anything negative about the slide or imply it was questionable in any way.
Mike Hargrove went with Julio Mateo instead of J.J. Putz.
…or leaving in George Sherrill, who once he gets past Jeter would give you lefty-lefty with Abreu. I was cringing watching Mateo pitch the 9th.
Even the woman sitting next to me, who seemed to be a casual fan, wondered why they took out Snelling to bring in a guy making his major league debut.
The bunt/steal thing I actually get. But how can an experienced manager be so brain-dead as to create batter-pitcher matchups that favor the opposition?
The scary thing is that I had thoughts along the lines of what you posted above, Dave, at about the same points in the bottom of the 8th and top of the 9th. I.e., WTF is Hargrove having Yuni BUNT WFB over to 2B for, Willie can STEAL that base!!!?, and WTF is MATEO doing pitching in this spot? Surely SOMEBODY else is available??!! (I think with more expletives than I talk aloud
)
The point being, I’m a nearly-62 year old lady. We old ladies don’t ordinarily know bad baseball management better than the guy who’s actually supposed to be managing the baseball team, do they?
Ack.
Bavasi has been doing a lot of good things for this organization in the last 10 days. But the biggest thing he didn’t do was firing Hargrove on Monday. I didn’t have a problem with Willie PH and Bentancourt bunting. 99% of the big league managers would probably make the same moves. But when I saw Mateo entered the game for Sherril, I almost wanted to grap the nearest sharp object and kill Hargrove and the TV on the spot. You have Jeter, Abreau, PH and AROD coming up in the tie game on the top of the 9th inning. If that’s not a high leverage situation, I don’t know what a HL situation is. Hargrove’s use of Mateo / JJ Putz this season alone should get him the pink slip. I don’t know how Bavasi can tolerate that. The M’s were lucky last night that Proctor and Rivera weren’t available. Otherwise, you wouldn’t see Villone out there in the 9th to face Beltre and Sexson. The fact that PH Snelling and Perez would leave Bohn in the 2 spot and Bloomquist in the 6th spot and if we were to go to extra innings, it would create 2 huge holes in the lineup. They clearly needed to hold the Yankees in the top of the 9th, and not bringing in Putz was beyond stupid.
I think the criticism of the sacrefice bunt is incomplete without looking at the possible outcomes of Betancourt swinging away.
I’d like to suggest that you read up a little more on win expectancy, how its calculated, and what it includes.
This article is a good start.
WE captures what you’re complaining that it doesn’t capture.
Thanks for the clear analysis, Dave. I was at the game and had the same reaction to Hargrove’s moves. I was particularly disgusted by the Bohn-head pinch hitting move, but the Mateo appearance in that situation seemed equally indefensible. I’m glad the numbers support my gut feelings. I just wish the Mariners were being run by somebody who gets it. I can’t wait for USS M’s analysis of managerial candidates (oh please fire Hargrove, please, please, please…)
Not if the batter had to take a strike or two to allow the runner to steal second: that would certainly affect the likelihood of a successful sacrifice.
Agreed on all points (of course). These were not subtle mistakes, they were GLARING. Another thought occured to me during the game — if you intend to bunt, why even pinch run Willie at that point at all? Why not bunt Perez over and then decide on a pinch runner if it is successful? You might need Willie’s speed/defense later in the game, as well as Perez’ bat.
I don’t know, the whole thing was just dumb. Glad we won anyway.
I can’t think of any reason a manager would pinch hit for your #2 hitter for a Rookie with 0 MLB AB in late inning tie game.
The sacrafice is frustrating, especially with a good base stealer on 1st. This is the only thing WFB should ever be used for. It is the situation he was born to excel at. Getting into scoring position without wasting an out. But I can over look that to an extent.
Not using Putz in a clear position in which it calls for his services, other than a save, is frustrating. Fortunately it didn’t cost us the game…this time. It is clear in this situation that there will never be a save opportunity and that it still calls for you best reliever. But JJ was coming off a blown save, maybe somewhere in the Hargrove universe he thought Putz needed some rest. Even though he has appeared in only 3 games since August 9th. Fine.
BUT! The pinch hitting incident of 2006 can never be let down as excusable. If you didn’t think Snelling could handle the duties, why hit him in the 2 hole? Who in they’re right mind puts a rookie with no major league experience in that situation? Poor Bohn, I wouldn’t wish that on my worst enemy.
And why did we give Hargrove a job? There was a reason he was out of the game and he should of stayed there. Yes, he hd some success with Cleveland, but did anyone bother to look at the 4 years he had in Baltimore, where he had a .425 winning percentage? I questioned a lot of BoMel’s moves while he was here, but I’d take him now over the sack of potatoes that currently sits in our dugout!
“Then, if he’s successful, bunt Willie to third. That’s okay, if you have to bunt. Getting the runner from second to third with less than two out is a big deal.”
Why, exactly? The underlying basis of your analysis is win expectency, but the win expectancy variance between runner on second, no outs and runner on third, one out (both in bottom of 8th) is negligible — half a “point” (to use your term). So why is it a “big deal” if it does not in fact help you win lots more games?
Because it wasn’t a save situation, and J.J. is our closer.
The most frustrating thing about that is that Grover’s allegiance to save situations is becoming less common. More and more, teams are using their closers in the eighth, or in tie games, simply because they’re really good pitchers and the situation calls for them. Grover’s trapped in a relatively brief but insane period of tactical decision making that stretched perhaps 20 years. And it’s over.
I disagree with the premise that having the Blooming Willie try to steal second is a good percentage move. With Farnsworth on the mound, you’re right. With Villone, not so much. He’s one of the hardest pitchers in the league to steal against. This year, runners are one for four. In his last 500-ish innings, runners have stolen 8 bases with him on the mound.
Speaking of WE, it drives me nuts that the articles I’ve seen that look at bunting always assume the expected outcome. There are different potential things that happen when you call a bunt. Like when Torre tried it, sometimes you get a FC. Sometimes, an error. Sometimes the batter doesn’t get it down, and starts an AB with an 0-2 count. I think any useful discussion about WE related to bunting would have to take all that into account, and when it did, bunting would look like an even worse % play than it does now.
I was listening on the radio last night. How did Soriano look?
If you put in Bloomquist, you have to send him. That’s what he is being paid to do. Even against a tough lefthander (with a career 38/38+31 = 55% success rate for stolen bases). Even when everyone at the stadium knows he’s going. It’s the one thing that Bloomquist is truly good at. You have to send him.
Dave, when does sample size become meaningful? 200 plate appearances? 400, 600?
It sounds an awful lot like you’re saying you have to send him even if it’s a bad idea to send him.
Base stealing is not the only reason to have a faster guy on base in that situation.
Dave, you are a literate guy, and always open to constructive criticism. So therefore, can you please remember to never again write “on the season?” Thanks.
Spencer, you are right. When we looked at it, there are several potential results to look at (reached on error, DP, etc). As well, you need to know the proficiency of the batter in swinging away, the speed of the batter and runner, and his bunting ability. There are dozens of things to consider. All these going into the WE tables. (i.e., the standard WE tables tell you if everythign was average… you need to construct a custom WE table for the particular context in question).
I see that Villone has a good record at preventing the steal, so maybe that played a large part in the decision-making process:
http://www.retrosheet.org/boxesetc/Pvillr001.htm
Over the last five years, runners are 9 for 21 on stolen bases. That percentage is way too low to yield a good return. You can fault Hargrove for a lot, but I agree with him here.
Dave, when does sample size become meaningful? 200 plate appearances? 400, 600?
Depends on what you’re trying to measure. For instance, historical research tells us that the reverse platoon split essentially doesn’t exist – we can’t find examples of hitters who have sustained better performance against same handed pitchers. So, before we decide that Snelling hits lefties better than righties, we’d probably want a few thousand at-bats over many years, becuase that would make him historically unique.
For something more common, the necessary sample isn’t going to be as large. For instance, we can say that Bloomquist is a good baserunner, even though our sample is only 51 stolen base attempts, because they’re the product of a skill that is clearly defined. It doesn’t take nearly as big of a sample to suggest that a guy can steal bases as it does to suggest something contrary to the expected norm, such as a reverse platoon split.
If it’s too tough to steal 2nd with Villone on the mound, why didn’t Bloomquist steal 3rd after being bunted to 2nd? It’s easier to steal 3rd than 2nd off a LHP, and at the end of the sequence, you’d still have the go-ahead run at 3rd with less than 2 outs.
#20 – I am starting to believe that firing Hargrove is something Bavasi has to get the two headed monster known as Linstrong to sign off on.
RE 30: Just ’cause the situation isn’t optimal doesn’t mean having Bloomquist attempt a steal is a bad idea. A bad idea would be leaving in Perez and sending him. Even against Villone, Bloomquist is likely to pass the 70% success rate which is what you need to make it worth while.
The emphasis on grammar, spelling, and punctuation at this site caught my attention the first time I read it. I understand and applaud the need for clarity, but it intrigues me nonetheless. Professional linguists make a distinction between prescriptive grammar, (for example your third-grade teacher insisting that ain’t, a perfectly good English word, is bad grammar), and descriptive grammar, an analysis of the underlying rules native speakers use naturally when they form sentences. Variation in such rules is normal, and few things vary more than the use of prepositions. Thus the difference between ‘on the season’ and ‘for the season’ would be viewed as a variation between different speech communities, and it could be argued that baseball fans form such a community, and that said community has been using the expression ‘on the season’ for at least as long as I can remember (1970 season?).
For instance, historical research tells us that the reverse platoon split essentially doesn’t exist – we can’t find examples of hitters who have sustained better performance against same handed pitchers.
Uh, don’t 5+ years of Ichiro in MLB serve as the counter-example to that? He’s been consistent in hitting LHP better year after year.
For crying out loud, there is nothing wrong with the common and easily understood idiomatic phrase “on the season”. This is conversation, not a formal paper. You might as well object to “for his career” and “break it out” and “stop with the bunting”.
If it’s too tough to steal 2nd with Villone on the mound, why didn’t Bloomquist steal 3rd after being bunted to 2nd? It’s easier to steal 3rd than 2nd off a LHP, and at the end of the sequence, you’d still have the go-ahead run at 3rd with less than 2 outs.
Any information to back this up?
Tangotiger, do you have numbers on what steal attempts do to the at bats during which they occur? When we speak of the win expectancies in the base-out siutations, is it assuming that all situations with a man on second and no one out are equal, or does it take into account that after a successful steal the man at the plate doesn’t have an 0-0 count?
#41: I don’t have stats to back it up, but I’ve heard it as conventional baseball wisdom for years. (I know, I know.)
Consider that the runner taking a secondary lead off 2nd is behind the pitcher’s back, whereas the runner on 1st is looking at the LH pitcher eye to eye.
That’s not quantitative, bit it’s information.
Uh, don’t 5+ years of Ichiro in MLB serve as the counter-example to that? He’s been consistent in hitting LHP better year after year.
Yea, he’s pretty much established himself as the exception to this rule. The research on the issue I remember was done about 2 years ago, so at that point, I don’t think he was included because of the still-too-small sample. But yes, Ichiro qualifies as the rare exception who has a reverse platoon split.
#44: I look forward (in a very small way) to Ichiro’s retirement, when perhaps he will tell us his secret. Right now, his answer to why he hits LHP better than RHP is, “I don’t know.” Well, maybe not, but if he did know, I wouldn’t expect him to tell the league how to get him out just yet.
#37 – Let me put it another way. Last year, Villone picked off three runners, or one more who actually tried to steal second base. That’s Terry Mulholland territory (terrytory for short). It’s a poor percentage play. You could argue, as a poker player might, that you need to do this anyway to keep the opponent guessing. But clearly, there is a strong argument for not having the runner go.
scraps: yes. Not to be a pimp, but I have it all laid out in The Book. The steal attempt drops the batter’s OBP by 8 points. Having a guy with a threat to steal drops the batter’s performance as well. Having a LHH at bat with a runner on 1B increases his performance.
As for your WE question: it all depends on your parameters. The minimum parameters is inning, score, base, out. That’s your state. You can have a WE table that also adds the count. That’s your state. So, to answer your question, if the count is important, then you need a WE table that includes the count as part of your state.
#37 – Let me put it another way. Last year, Villone picked off three runners, or one more who actually tried to steal second base. That’s Terry Mulholland territory (terrytory for short). It’s a poor percentage play. You could argue, as a poker player might, that you need to do this anyway to keep the opponent guessing. But clearly, there is a strong argument for not having the runner go.
Villone’s not the only variable in play here. Last year, Villone didn’t have Jorge Posada catching him. And not every runner is Willie Bloomquist. You can (correctly) argue that I overstated the odds of Bloomquist’s success by omitting Villone from the equation, but you can’t just say “never steal off Ron Villone”.
Again, the necessary success rate in that situation for the SB attempt to be a good idea is about 60%. I’d say that, looking at all the variables (Bloomquist, Posada, Villone, anticipation of a pitchout in that situation), it’s still a good play, though probably not the home run I made it sound like earlier.
#46 – Are you stating that Villone “picked off” that many runners, or that they were thrown out stealing?
#37 I agree with you here. As pointed out at the top of the thread, Willie has done a great job over his career at picking optimal times to steal a base, mainly when A (WFB’s time to 2nd) is less than B+C (pircher’s time to plate plus catcher’s time to second), but sometimes you have to take a chance. I think the reward, here, outweighs the risk.
As far as WE .738 with 0 outs and man on second versus .743 1 out man on third, call me ‘ol fashioned but I think there is a greater chance for this run to score–especially if there is speed at third–than there is with the runner at second. You still need a base hit to drive in the man on second, which is not the case with a man on third and less than 2 outs. In my lifetime I have seen more runs scored from third and 1 out that second and nobody–especially from the M’s.
Terrytory. Hah! Such drollery.
Let’s cut to the quick here. Hargrove is a purblind* moron who has no business managing a peanut stand, much less a major league baseball team.
In fact, I’d guess there were peanut vendors right there at Safeco last night who would have made better decisions than these.
(*According to Dictionary.com
2. slow or deficient in understanding, imagination, or vision.
That’s more appropriate than I knew when I wrote it.)
A note about grammar and conversational writing:
Most of the folks who comment here are easily understood, whether or not they observe the formal grammatical niceties. Some have more viivd voices than others — every one of the authors on this site has a vivid voice — and often a vivid writing voice is actually less grammatical by the formal rules.
Every once in a while, someone passes through here who cannot make himself clear. We’ve had one of those in the last few days. The writing is incoherent, a collection of bleats carrying fuzzy ideas that we try to guess at (and usually the content of the thoughts is no better than the expression). When one of those folks starts commenting, he’ll often get told to fix his grammar (and spelling and punctuation) if he wants to get taken seriously.
But the grammar isn’t really the issue: it’s the thoughts. The fellow who has been commenting hasn’t got coherent thoughts to express. Writing grammatically is well beyond him. Writing, grammatically or otherwise, is the shape of thoughts. Whether you choose to express those thoughts formally or idiomatically is a matter of style. When we tell an incoherent writer that he lacks grammar, what we are really trying to tell him is that he needs to sort out his thoughts before he writes. But if you have no difficulty understanding what someone is saying, there’s no need to harp on conversational choice in phrasing. Formal grammar is one road to clear writing: but the goal is clear writing (and clear thinking), however one gets there.
As for the reverse platoon split, we also looked at this. Of all players from 2000-2004, Ichiro is #1 in terms of having the smallest measured split. (Next is Mark Grace, Larry Walker, Alex Cora, Johnny Damon.) And Ichiro is way ahead of them.
Until this point, we would establish Ichiro as having a positive platoon split of 4 OBP points. (For a typical LHH, it’s around 25 points or so). That is, even though he had a measured REVERSE split of 33 points, we have to infer, because of sampling, that it was a +4. That still leaves him as the guy with the smallest platoon split.
However, if we were to add in his 2005 and 2006 seasons, he likely would come in with a reverse platoon split, and be the only MLB to be like that.
It’s also possible that if we knew more about Ichiro, that his true platoon split would be even more in reverse. As it stands, until that happens, we have to treat Ichiro as a LHH, drawn from that LHH sample, and regress his performance heavily towards that mean.
If you want to include an additional parameter, one that is fairly unique, that may override the fact that he is a LHH.
Ichiro is a problem case for half of all research studies because of his apparent uniqueness.
Could one make an argument that based off mechanics alone players like Ichiro and Doyle (short stride and very quick hands) would be affected less by the LOOGY than traditional LH pull hitters like Thome or Ortiz?
Not to say they are better hitting Lefties, just not as bad as other left handed hitters vs. left handed pitchers.
Tangotiger: Thanks! Okay, I need this book.
#5,
I watched the YES broadcast here in NY (I return to Seattle in 3 weeks) and it was one of the first mariner games I’ve seen this year (been out of the country) and Bobby Mercer and Ken Singleton both said it was a ‘good clean slide.’ They even talked about Ron Fairly and how the Mariners seemingly don’t have the will to win. gee, ya think? Mercer actually reminds me of Fairley but he was the PBP guy and I’d take him over Rizzs any day in that capacity. Singleton was better than Valle and Hendu. Although I cannot STAND the other Yankee broacasters. John Sterling and Michael Kay are just annoying.
Sure, the 3rd/1out vs. 2nd/0out isn’t a huge advantage. If you look at it in percentages, it’s 0.677% increased odds of winning. One might, essentially, consider that equivalent to one game per season (fully played out under such circumstances). 1/162 would be 0.617%. I’ll still take that over 0.000%. Make a lot of sound, smart plays over the course of a season to marginally improve your odds when you can, and you’ll separate yourself from where you would have been, over time.
#53: Would you be able to validate Ichiro’s reverse split by including his numbers in Japan?
Could one make an argument that based off mechanics alone players like Ichiro and Doyle (short stride and very quick hands) would be affected less by the LOOGY than traditional LH pull hitters like Thome or Ortiz?
Sure, that’s a scouting argument. Shin-Soo Choo is always going to struggle to hit lefties because he has a long swing. Snelling’s short stroke and quick wrists should minimize his problems against LHP’s, at least in comparison to other LHB’s. But his uppercut swing will make him somewhat vulnerable to pitches up and on the inner half. Karstens went after him there several times last night with success.
58: The only place I know to get the data is here:
http://www.japanesebaseball.com/players/player.jsp?PlayerID=28
And there’s no split data. If you have it, send it over, and I’ll ask Andy will run it through his programs.
I think in comment 32 tango makes the point I was trying to make about the bunt situation. When critizising the decision to have Betancourt bunt you need to look at all possible outcomes both for bunting and hitting for Betancourt facing a LHP of Villone’s calibur, establish their respective likelihood and take it from there. In the words of the hardballtimes article linked by Dave:
“Of course, the application of actual strategy (should he bunt or not?) depends on a lot of other factors, such as the skills of the batter, the pitcher and the baserunner, the following batters in the order, the game conditions and probably a number of other things. But Win Probability sets the baseline for evaluating each event on the field.”
If one wants to defend Hargrove, one would argue that when batting, Yubet is about twice as likely to make an out which would hurt the win expectancy to a stronger degree than when he sacrifices. And he’s not highly likely to hit an extra base hit that would score Bloomquist.
I don’t have the data or time to do a calculation for all possible outcomes and their likelihood (Tango, I suspect it might be in your book) but imagine that when you consider sample sizes, its basically a toss-up whether you have Yubet bat or bunt.
BTW: In updated win expectancy tables, this game will count in favor of the home team winning for the scenarios before and after the bunt.
The bunt thing doesn’t bother me too much. Like someone pointed out, there are quite a few managers that would do that and seems like a reasonable thing to do when you have an average hitter (Yuni) bunting followed by two good hitters (Kenji & Ichiro) trying to get a single to score the run. Playing it safe, so to speak, also eliminates the chance of you (meaning Hargrove) looking like an idiot to casual fans and the media when Willie get’s picked off or thrown out and you’ve “run yourself out of the innning.”
Of course, it doesn’t mean Hargrove can’t still look like an idiot for pinch hitting for Snelling and bringing in Mateo.
#60: Thanks, I’m surprised that the split data for Japan isn’t easily available. (Don’t they play some kind of fantasy baseball over there?)
The front office loves the concept of “Productive Outs”. Firing Mark Hargrove will not change this philosophy. Chances are good that the new manager will play these situations the same way.
#64: If only we could get them to refine their definition:
Runner on 3rd, sac fly. Scores a run. That’s the best kind of productive!
Runner on 1st, less than 2 out, strikeout. Stayed out of a GIDP. Kept the inning alive. That’s productive, too!
re. #29, 34
“Dave, when does sample size become meaningful? 200 plate appearances? 400, 600?”
I disagree with Dave’s take on this. I would say that it doesn’t matter what you look at, you need an adequate sample size to determine ability statistically. If you want to say WFB is a good base stealer based on 51 attempts, well, I think you can say he’s no worse than a decent base stealer statistically (see below). Which you probably could have arrived at with less steal attempts from scouting!
Rule of thumb: take the square root of the number you’re interested in to get an idea of the likely random variation in it. (This is a good approximation for anything you count, like hits, home runs, steals.) That is to say, if you think Player A is likely to average 200 hits in a season, take the squareroot of 200 (approx. 14). You should not be surprised if Player A gets anywhere between 200-14 = 186 and 200+14 = 214 hits.
So applying this to Bloomquist’s steals: he is 43 of 51, or 84%. Excellent. If we take the squareroot of 43, it’s about 6.5, which suggests that statistically we cannot be too sure that Bloomquist is better than a 36.5 of 51 stealer – 72%, so still pretty good.
So the answer to the question “what is a big enough sample size” depends on the what you’re looking at. For separating good base stealers from bad base stealers, 50 attempts appears to be plenty. However, for separating excellent base stealers (I’ll call that 80% or better) from decent base stealers (70%), it’s not enough.
If we look at the example with hits above and assume that Player A had 650 at-bats a season, we see that his batting average could be expected to range from 0.286 to 0.329 just based on statistical fluctuation (i.e., not even considering injuries, changing abilities with age, etc.). There’s a pretty big difference between 0.286 and 0.329 you might think – but really it is a difference of only 4% of the at-bats. It takes more ABs to determine the batting average than steal attempts to determine the stealer’s ability because we’re interested in a much more exact answer. In the stealer’s ability a difference in success of 10% of attempts (70% to 80% success, for example) is big, but not huge. In batting average a 10% difference is huge (0.250 to 0.350, for example).
(For those readers who understand a bit of statistics/probability jargon: for the Poisson distribution, the standard deviation is the squareroot of the mean, so +/- the squareroot would be the error bars you’d typically use for a Poisson distribution. While the distributions of things like hits and steals are not Poisson distributions, they are similar in some regards, similar enough to make the squareroot a decent yardstick for normal random variation.)
#66 is right-on. However, if we know more about WFB, like, he’s actually a fast and smart runner, then we have to use that information. In a vaccum, WFB’s 43 for 51 and say some other guy’s 43/51 has to be treated exactly the same way. But, if one guy is fast/smart, and the other guy is not, then we know that we need to regress WFB to a different mean than the other guy.
Ichiro hitting .350 over a 2-month period, and WFB doing the same is different, because we “know” Ichiro.
So, even if we can’t accept his 84.3% rate as his “true” rate, it’s likely to be above 75% (just guessing).
rlharr–
I think this is a common misconception amongst people coming from a strong stats background. Yes, it is possible, we can calculate the variance in the sample mean for a stat assuming that we know its underlying distribution (binomial is natural for lots of baseball stats), but that’s only the first step. After the statistical variance, we need to look at variance caused by systematic effects–park effects, strength of opponent effects, platoon effects, etc. The smaller a sample size gets, the easier it is to have an extreme set of systematic effects.
Say over the course of a week, if the offense is going decent, Ichiro could get about 35 at-bats. Assuming he’s maybe a “true” .350 hitter the standard deviation in his sample mean for the week would be something like .081. But we know from experience that his average fluctuates a lot more than that statistical variance would expect. That’s because over the course of the week, Ichiro might wind up having to face some of the top run prevention teams in the league (say, the Tigers and the A’s). But that’s only the tip of the iceberg. Maybe those teams are usually tough, but then on top of that it’s possible that he was at a platoon disadvantage for most of the week. And maybe one of those two teams each had a key defensive player come off the DL. So now we’ve got a lot of reasons to expect him to hit a lot less than .350, but that aren’t going to be captured in the statistical variance.
By assigning only a statistical variance to a small sample, and not accounting for each and every systematic effect in play, you are overstating the confidence you should have in the conclusions you’re basing on that small sample size. That’s a complicated business, and until you do all of that work, you’re better off being extremely skeptical of conclusions based on a hundred or even a couple hundred at-bats.
Wow, interesting arguments! The Managers dilemma. Let’s suppose the Manager is aware of the WE but doesn’t like the Villone/Posada pairing to steal on. If he steals and fails, he’s got to start over with no one on. I don’t know the numbers but it is believed the bunt is a safe play with the right guy. So the question might actually be, was Betencourt, despite getting it down the right guy? Oh, I have never liked the sac bunt to get a decent speed runner to 3rd. He is in scoring position, hit the SOB in! At any rate, today is why I read the site, great discussions.
Great discussion! So many people have interesting and informative things to say in this thread. The USS Mariner is the best baseball blog.
I thought I could go back to MLB.tv and see the replay of this sac bunt. Apperantly this year it is a pay feature to watch archived games. I don’t believe that was the case last year.
I would be curious to see it, though, as I didn’t catch the game live. How was the bunt? Did Alex charge from third? Does Jeter move over to cover third? And if WFB takes off with the pitch does he have the kind of speed to go 1st to 3rd on a sacrafice?
#69: He is in scoring position, hit the SOB in!
If hitting were that easy, MLB would be full of .400 hitters. The point of getting the runner to third with less than two outs is that you can score a run without “hitting the SOB in.”
#71: if WFB takes off with the pitch does he have the kind of speed to go 1st to 3rd on a sacrafice?
Michael Johnson doesn’t have the speed to go from 1st to 3rd on a sacrifice, running with the pitch or not.
I disagree with Dave’s take on this. I would say that it doesn’t matter what you look at, you need an adequate sample size to determine ability statistically. If you want to say WFB is a good base stealer based on 51 attempts, well, I think you can say he’s no worse than a decent base stealer statistically (see below). Which you probably could have arrived at with less steal attempts from scouting!
Certainly, this is a case where scouting information is going to be available and accurate long before we have a big sample of SB/CS attempts, so scouting data for baserunning will probably trump statistical analysis in most cases.
However, getting back to the point about variable sample sizes – in my opinion, one of the main reasons we need a large sample is to extract the contribution of players other than the person we’re evaluating from the totals. For instance, when Willie Bloomquist gets a single through the hole, some percentage was his skill, some percentage was the pitcher’s lack of skill, some percentage was the fielder’s lack of skill, and some percentage was chance (ball takes a weird hop, fielder blinks at contact and gets poor first jump, whatever). As we get a larger sample with a collection of other variables, we’ll be able to isolate Bloomquist’s hit skill more easily. Since his skill is only a fraction of the result, we need a big sample to get good data.
A stolen base, however, is significantly effected more by the baserunner than a single, in my opinion. In a successful steal attempt, the credit breakout might be something like 70% runner, 20% catcher, and 10% pitcher. Or whatever – you can argue with the split, and maybe its 50/40/10, or some other combination, but the credit for the steal is certainly higher for the runner than it is for a batter getting a single.
Thus, since the result is more representative of the runner’s skill than the hitter’s skill, the sample doesn’t have to be as large to filter out the impact of the other variables.
Some results just show skills earlier than other results. Thus, I don’t think we can use a blanket “correct sample size” for each statistic. They’re all different.
LB-No real argument as from 3rd with less than two outs you can sac fly the guy, passed ball, go on contact, etc. etc. My dislike of it stems from you already have a guy in postion, I always hated to give up outs when managing. (I also pitched to a bunter under the theory that if he was going to give me an out, I’d take it)I like the WE theory, wish knew about it when involved. It adds information but is one part of the decision making.
LB, are you telling me you haven’t watched enough baseball to see this play executed? That’s why I asked if Jeter was covering third. I have seen it plenty of times when the base is left unmanned and the runner never stops at secound and takes the extra base before anyone can get back to cover it.
Actually, rl, re-reading it, I’m not sure we disagree. I certainly wasn’t trying to claim that Bloomquist’s 85% rate was his true value. But I’m a lot more comfortable claiming that he’s a good baserunner based on 50 steal attempts than I would be making any claims about his hitting ability based on 50 at-bats, for the reasons I stated above.
As we get a larger sample with a collection of other variables, we’ll be able to isolate Bloomquist’s hit skill more easily. Since his skill is only a fraction of the result, we need a big sample to get good data.
You said it better than I did, but this is basically what I was trying to get across using more technical language.
A stolen base, however, is significantly effected more by the baserunner than a single, in my opinion. In a successful steal attempt, the credit breakout might be something like 70% runner, 20% catcher, and 10% pitcher. Or whatever – you can argue with the split, and maybe its 50/40/10, or some other combination, but the credit for the steal is certainly higher for the runner than it is for a batter getting a single.
One thing I would add, though, is that if we are sufficiently clever and have sufficiently good data, we might be able to figure out the breakdown pretty accurately and then adjust for it. A small but adjusted sample could be just as useful as a large but unadjusted sample.
#76: I misunderstood you. Clearly the defense has to have someone covering the base. I have already zapped the game from my TiVo, so I can’t tell you who was covering.
I believe that a typical bunt defense with a runner on 1st is for the second baseman to cover 1st, the shortstop to cover 2nd, and the catcher to be near enough to 3rd to prevent the runner from doing what you describe.
Speaking of decisions (good or bad):
Andrew Baldwin last night in Inland Empire:
7 IP, 1 H, 0 R, 0 ER, 1 BB, 4 SO, 9 Groundball outs vs 8 Flyball outs picking up his first win for us.
It also looks like Bavasi hasn’t given up on Jorge Campillo, as he makes his first start for San Antonio tonight.
#75: The trouble with applying WE to lower levels of the game is that the numbers in the tables are derived from a league filled with elite players. As bad as, say, Charles Gipson is/was, he’s likely a better hitter than the guys you were managing. Of course, the pitchers in your league were probably a bit weaker, too.
Ichiro also has a crazy swing and a slightly less crazy approach at the plate. I wouldn’t put alot past him.
Speaking of sample size, have we had enough to determine that Rivera is a terrible catcher or what? It seems he can’t hang on to the ball very well back there. We know he can’t hit, but I thought he was supposed to be a good defensive catcher. Of course, that’s how people usually describe a catcher who can’t hit. It sure doesn’t seem like he’s good defensively to me. I’m not going into the whole CERA thing which has already been discussed, I’m simply trying to access his ability to field the position.
Dave — Loved this post and all of this type of “in-game”; thank you.
LB-Well of course! Information is always useful if you understand it. OTOH, baseball is such a cool game in that there is some relativity. Skill sets at lower levels are not as diverse as you would think. With the odd all star thrown in, pitching, hitting and defense stay in some balance. I am not capable of managing a pro team but I can see similarities and situations that use the same solutions. I guess that is why I love it. A 12 year old on 60 ft bases get’s thrown out at first by two steps. A major leaguer gets thrown out by two steps at 90 ft!! Willie B gets thrown out either place…
“I would be curious to see it, though, as I didn’t catch the game live. How was the bunt? Did Alex charge from third? ”
I, too have deleted it from Tivo, but if I recall correctly, Alex was a non-factor, the bunt was up the first base line.
“Does Jeter move over to cover third?”
I’m guessing Jeter was covering second and Alex third as soon as the bunt went to the first base side of the pitcher.
“And if WFB takes off with the pitch does he have the kind of speed to go 1st to 3rd on a sacrafice?”
No, not in my opinion. And certainly not in a way you could possibly plan on.
Hargrove seems to be noticeably worse than even your average MLB manager when it comes to things like this. And we all know that there are plenty more bad ones out there.
I wonder if he really is that much worse or just seems so because Dave does such a good job of breaking it down.
Let’s assume that WFP takes off on the pitch (a “bunt-and-run” play) and the bunt is placed so that both the third baseman and the catcher have to play the bunt. Then Willie has a chance to take third if he looks in as he is going to second, sees that the bunted ball is placed so that both the 3rd baseman and the catcher are committed to the ball (and that the pitcher is not in position to cover third), and, observing all of thst, rounds second without slowing down and proceeds to third base.
Alternatively, the situation is read by the rhird base coach, who signals Willie as Willie is en route to second that he should continue to third instead of sliding into second.
But I’m a lot more comfortable claiming that he’s a good baserunner based on 50 steal attempts than I would be making any claims about his hitting ability based on 50 at-bats, for the reasons I stated above.
I don’t have 50 on-hand, but his first 38 PAs after coming up in 2002 gave us this:
.455/.526/.576
Such a perfect example of how irrelevant it is going forward. I don’t recall exactly but I’d wager at the time a few announcers were frothing in delight.
The Gritty Splinter!
How about this thought on the bunt. Hargrove knows that having Betancourt at the plate in the late innings of a tie game with a runner on base is ripe for Betancourt to swing at any pitch that is remotely near the strike zone. Since by now every team in baseball knows that this situation amplifies Betancourt’s hacktasticness, Hargrove also figures the chances are real good that Betancourt will hack at a pitch out of the strike zone and hit a weak pop up or a groundball out (or DP).
Hargrove tries to counter this by having Betancourt bunt Willie to second.
Hey – it’s possible, isn’t it?
#88 – I have seen that happen before. I’ve also seen it happen where the runner starts on 2nd, the bunt is laid down, fielded and thrown to first, but the runner rounds third and keeps steaming home (little league style!).
They’re rare plays, but occasionally through either laziness or lack of concentration on the defense, they do happen.
A thread about Hargrove’s obvious – even to the fans – mismanagement (and he isn’t even lucky)? Duh!
Dear Mariners’ FO,
Please remove Mike Hargrove from his position as manager. If he is released in the dead of night and dressed in one of the beer vendors uniforms perhas he can be safely whisked away to an undisclosed “safe house” until Mariner fans’ anger abates.
(re. statistics stuff) Yes Dave, Mat, I think we all agree. Mat, I did put in parentheses somewhere that this kind of analysis did not include injuries, etc., and I would certainly include quality of competition – Yankees or Royals – within the factors not covered by my argument. It only covers the random variation. And yes, the binomial is the actual distribution, I didn’t want to get into limiting values.
I still stand by the squareroot as a rough first order tool, however. I do not claim that it gives you the actual standard deviation, but I think it a rare case when it does not give an appropriate amount of “uncertainty”. For instance, Mat, you argued that you need hundreds of at bats to judge something like batting average, but my earlier message showed that using the squareroot you would not even put much credence in a 650 at bat sample – after all, a 40 point difference in batting average is huge. The squareroot rule for 200 at bats for a 0.300 hitter would yield a range of 0.261 to 0.339. I expect if you looked at 200 at bat samples of 0.300 hitters you would find something like half in that range – less than you would if it were a true standard deviation, certainly, but at the same time representative enough to give a handle on the expected size of variation.
That’s all I’m suggesting. A rough, easily-computed, first-order tool that can be used by the average reader.
re. Josh, #89
“I don’t have 50 on-hand, but his first 38 PAs after coming up in 2002 gave us this:
.455/.526/.576
Such a perfect example of how irrelevant it is going forward.”
There’s a difference here. What are we counting? The 0.455 is hits, so we’re talking about something like 16 hits in 35 at bats (+ 3 walks/sac bunts/whatever for the other 3 plate appearances). With stolen bases, the success rate is much higher. In WFB’s case there were 43 successful steals – that is a considerably higher number that 16 hits! You would need to get over 100 at-bats to get similar precision for a batting average estimate – and even then, you would find your estimate less than satisfying. Knowing that WFB’s successful steal rate is likely to be somewhere between 70% and 90% is enough to know that he is a decent to excellent base stealer. Knowing that his batting average is likely to be somewhere between 0.245 and 0.355 (about the range you would guess if he was hitting 0.300 after 100 at bats) would tell you almost nothing.
Dave made a point above that we also have historical information to help us set limits. We know that 0.455 is extremely unlikely to be any batter’s true ability. (Of course, by that argument we would have said there was no way Bonds would ever have a OBP over 0.550 for a season, when, in fact, he did so twice, peaking at 0.609.)
I was at the game last night and sitting behind Mariners dugout. I kept checking the boxes where Chuck, Howard and Bavasi hang out with my binoculars. Up through the Baek innings Chuck sat looking half interested. O’Faherty comes in and he jumps up and disappears.
Flash forward to Bohn coming in and he (Chuck) is pacing back and forth making what looked like “sharp†comments to Bavasi who is along one wall in the box with body english that says he is not very comfortable. They were clearly not happy with Bohn coming to the plate. It appeared that they kept telling someone in the back of the box to go do something (hold the press release announcing Grovers release you suppose?).
I often will check the activity in those boxes esp when trades are in the works as you can get a indication something is going on via the phone activity. In the years at Safeco, I have never seen Chuck or anyone else “pace†like last night. It continued to the end of the game. Something was up but didn’t happen.
95 – Interesting observation. Considering the circumstances, I hope that your impressions are correct!
That’s a really interesting observation #95. I was wondering today if the FO could really be that oblivious to Hargrove’s mismanagement, fan perception, and/or extent of the damage he’s done, or just what the deal was. Giving them the benefit of the doubt, it occurred to me that maybe Chuck and Howard are obviously waiting til the end of the year to give themselves time to decide whether Bavasi goes as well.
#96 and 97: Well we know they will never fire themselves so the manager is always going to take the complete hit (deserving or not). With a majority owner who ceded control to a corporation, it would take a majority vote or effort from the minor owners to force a change in the Howard/Chuck/Bill axis of evil. I assume Howard and Chuck would offer a compromise and offer Bill up like Grover and that might keep the owners at bay – if they really care. I think the only way the FO changes its dynamic is not a complete scorched earth approach but a surgical strike to take out the person(s) who set the charter. I think that is either Chuck or Howard or both. I guess one can dream….
LATE GAME MANAGEMENT – [Perhaps this has already been mentioned. (I haven't read all of the previous comments.)]
Another thing that concerns me is providing for enough pitching should the game go into extra inings.
IMO, it’s wrong to have a reliever pitch only part of an inning in the late (or extra) innings, but managers have them do it, hoping that the game will end before they run out of pitchers. (You use up your relievers faster than you realize.)
BTW, does anyone know how many innings most managers allow for? 12? 15?
I’m not sure that I agree that Putz is the obvious choice for a top of the 9th tied game. Yes, there can never be a save situation, but the chances of going into extra innings are good, and for all you know there is going to be a bases loaded no out situation in the top of the 16th you’ll need him for. Granted, putting in your worst relief pitcher is never a smart way to go, and it surely was worth putting Putz in once there were runners on.