THE BOOK--Playing The Percentages In Baseball

Tom,

I’m glad you did that study because I heard Joe Morgan talking on ESPN the other night about Jose Reyes and how he’s more valuable at the leadoff spot than other guys with high OBP’s because he can, and I quote, “score from first on a single, go to third on any single, etc. and that other high OBP guys like Youkilies may get on base but they can’t score runs, and the the most important thing a leadoff guy can do is score runs.” I was taken aback by Morgan’s comments (as always), and your study proves that guys like Jose Reyes who don’t “clog the bases” hardly help their team in that aspect.

-Mike

I’m not sure we’re necessarily talking about the same thing.

To your point, I proved elsewhere how a great runner can add about .04 runs per time he’s on base.  It’s fairly straightforward to figure out.  So, a guy with about 200 times getting to 1B or 2B will add 8 runs with great running.  Of course, the guy who benefits from having a speedy runner is *not* the guy who hits HR!  You want a speedy runner in front of a singles hitter (preferably LH).  There’s real synergy there.

I recommend this article by Tom Tippett:
http://www.diamond-mind.com/articles/ichiro.htm

What value, baserunning?

If the average runner goes 1b to 3b on a single 33% of the time, a fast runner will do it 50% and a slow runner will do it 17%.  In essence, there is a .33 base gap between the best runners and the worst.  This is pretty much true on doubles with man on 1b and singles with man on 2b. 

A runner will be on 1B about 200 times, and about 20-25% of the time, that will be followed by a single or double (and have his speed on display).  That’s about 45 times that his speed will have an impact.  A runner will be on 2b about 140 times, and about 15-20% of the time, it’s followed by a single, or around 25 times his speed is on display.  This comes out to a total of 70 times where we satisfy the conditions of speed possibly being used.  Since there’s a .33 base gap between the best and worst, we are talking about a gap of 20-25 bases.  Each base is worth around .25 runs, giving us a final total of a 6 run gap between the best and worst baserunners, or +/3 runs from average.  Let’s bump this up to even +/-6 runs for getting bases on outs.

If the average runner scores around .30 times for each time he gets a walk, single, or double, then a fast baserunner will add 6 runs per 200 times on base, which is +.03 runs per time on base, giving us a scoring rate of .33 runs per time on base.

There is also the very occasional play where a slow runner on first allows the defense to record an out on a groundball that otherwise would’ve been an infield hit for the batter. Not sure how this could be picked up from game data though.

Plus a slow runner won’t score/advance on a wild pitch/passed ball as often. This could be quantified by, perhaps, wild pitch/passed balls rates with different runners on base. It would be tricky though…

As for getting bases on outs, I was wondering if you had obtained the extra 3 runs from a calulation, or just are guessing?

My “baserunning” lwts confirms that the true value of the best baserunner is around 3-4 runs above an average baserunner and the worst is around 3-4 runs less (per season).

This is important to know, as most people/teams vastly overrate the value of speed on the bases.  If you told a guy like Joe Morgan (or any player, manager, etc.) that the best baserunners in the league are worth less than half a win a year, they would look at you like you were out of your mind, right?

Tango and myself have also shown that a basestealer (at least one who attempts a lot of steals) adds little if anything, and might actually take away, from the batter at the plate (by disrupting his concentration and causing him to take some good pitches).

Hi Tom,

I agree with your general point (that the baseclogging effect is overestimated), but I think you underestimate it a bit.

Crisp will make an out about 65% of the time and he will be followed by another batter (unless Crisp’s was the third out) who might hit a single, etc. Without going through the probabilities, I’m guessing this will double your estimate. Of course, it is still laughably small.

-John

aweb: I recommend reading the Tippett link above.

John: I don’t understand your Crisp point.  Are you talking about baseclogging, or something else?

I think John is trying to say that you are only taking into account 1 single batter in the lineup being clogged by Ortiz.  He is saying in your example if Crisp gets an out with Ortiz on first the odds are another hitter will come up with the potential to also be clogged by Ortiz after a single.

I also think you underestimate because once Ortiz in on 2nd and Crisp is on first you only account for 1 single batter who has the chance to single in that instance.
If the first hitter strikes out for instance with Ortiz on 2nd and Crisp on first with less than 2 outs then another batter will come up and you’ll again have the potential for clogging bases. 

I would think both these examples may knock up the estimate to maybe 8 or 10 times a year instead of 1 time a year that Ortiz may clog for any runner and maybe 2 or 3 times clogging Crisp specifically.  These are only guesses as I didn’t take the time to do the math.  I agree with your basic point that baseclogging is really a very minimal effect in reality and overstated by many announcers.  I do think though that your math has left out some factors causing you to be too low on your estimate.

"Tango and myself have also shown that a basestealer ....adds little if anything, and might actually take away, from the batter at the plate”

The problem I see with that analysis, MGL, is that it didn’t seem to take account of the fact that fast runners on 1B frequently aren’t still on 1B when the batter’s PA is completed.  And if there is no longer a runner on 1B (because the SBA resulted in an out or advancement to 2B), then we expect the hitter’s wOBA to be lower, especially if it’s a LH hitter.  So to see if this “disruption” of batters is real, I think you need to look separately at hitters’ performance when a speedy runner remains on 1B for the full PA (compared to PAs with non-speedy runner on 1B), or make an adjustment for the fact that batters in the no-speedy-runner PAs enjoy the advantage of a runner on 1B at the completion of the PA a much higher % of the time.  (Or, if your analysis WAS limited only to PAs in which there was no SBA, then you also need to look at how the batter-w/-speedy-runner did in the SBA PAs.)

I don’t see how the estimate can go from 1 to 10. 

In any case, it’s a good point.  The average batter gets on first base 25% of the time, but he actually is planted on first base 33% of the time (because the batter gets an out, and a subsequent batter comes to bat).  So, bump up my estimate of 1 base by 33/25 (or 30%).

Guy, page 324 of The Book is about how the batter performs if the runner doesn’t attempt a steal. 

On page 326, I wrote

The focus so far has been on the situation where the runner is not going. Now, let’s turn our attention to the rest of the situations—the ones where a disruptive event has happened.

It sounds like what you are suggesting is to split Table 132, so that you have Table 132a, 132b, 132c, 132d based on the speed of the runner.

Pardon me but I’m lost by this passage:

“Of those 7 times, Ortiz will score 2-3 times and stay at 3B 4-5 times, meaning that the trailing runner will be stuck at 2B 2-3 times.”

Wouldn’t that mean “stuck 4-5 times” since Ortiz stays at 3B 4-5 times?  Or did you do mental math that I wasn’t aware of since I haven’t gotten your book yet and saying that half the time the runner could have made it to 3B, so you took half of 4-5 times and got 2-3 times?

Also, when you note what Crisp would do without a base-clogger, then applied it to the Ortiz situation, did you take into account that some of the 3-4 advance might have happened during the 2-3 times Ortiz scored?  My mind is doing the equivalent of a tongue twister trying to work the math out mentally, so I’m not sure what the conclusion would be if I am right (seems to reduce it to 0 bases), but I was hoping you could clarify that for me.

In any case, it all sounds good, whether 1 base or a few bases, doesn’t seem very consequential at all over a full season, which is the main point you were making.  Great work!

~3 Points to add:
The Lead-in is pointless. If Ortiz is at 3rd, and Crisp is at 2nd, a single will score both.
Ortiz at 2nd, Crisp at 1st, no stolen bases (Ortiz surely isn’t taking 3rd.) Crisp is 5-2 this year, 15-6 & 20-13 previously, so maybe this isn’t too bad of a thing (neg 0.6 sbruns overall).
Since we’re talking about runs, which is what this really alludes to, Crisp was .339 in 05, .377 in 04 on (r-hr)/(tob-hr). Ortiz was .253 in 04, .306 in 05. This, means little without context as well, but 63 in 04, 70 in 05 (runs) for Crisp, 72 in 05, 53 in 04 for Ortiz.

Overall, Ortiz isn’t the preffered _runner_, but Crisp still doesn’t make it on base as often.

PS. when does Ortiz ever hit right before Crisp anyway?

I said someone like Ortiz and someone like Crisp.  Feel free to substitute whatever players McCarver happens to be talking about at the time.

***

The 7:
1. long single, any runner will score from 2b
2. long single, any runner will score from 2b
3. many runners score from 2b, a few and Ortiz is stuck at 3b
4. some runners score from 2b, many and Ortiz is stuck at 3b
5. all runners stuck at 3b
6. all runners stuck at 3b
7. all runners stuck at 3b

So the above is what happens with the 7 singles with Ortiz on 2b.  This is what happens with Crisp on 1B.
1. all runners go to 3b
2. many runners go to 3b, slowpokes stuck on 2b
3. a few runners would have made it to 3B, but baseclogging keeps Crisp at 2b
4. all runners stuck on 2b, though maybe baseclogging happens here too
5. all runners stuck on 2b
6. all runners stuck on 2b
7. all runners stuck on 2b

So, Ortiz will score two times, situation #1 and #2.  Maybe #3.

Crisp advances on situation #1, #2, and maybe is stopped at #3 and #4.

about 15% of singles are infield singles or bunt singles anyway with little chance for “clogging” to occur. This would strengthen your point [your advancement rates look like advancement rates on outfield singles], but another dimension to clogging not discussed is that “Ortiz” may prevent “Crisp” from stretching his own hit to a double or triple. Obviously that really isn’t going to happen very much either, but it might be enough to double or triple your one base estimate [pun not initially intended].

I have not followed the base clogging discussion, but there is no way that if a baserunner, even a slow one, only advances one base (or two on a double) that the batters could have stretched his hit.  If a batter can stretch a single into double, even a slow baserunner is going to make it to third.  Same thing for a double, possible triple.

Guy, sure we looked at when no attempt is made and when the runner was not going on a batted ball.

My analysis (separate from that in The Book) suggested that the possible disruption to the pitcher and the greater number of fastballs and occasional pitchouts was cancelled out by disruption to the batter and him taking some pitches that he would not ordinarily take such that the batter had exactly the same OPS or wOBA (when the runner was not going) whether there was a basetealer on first or not.

Now what we did not look at and AFAIK, no one has looked at (the data is spotty, even in PBP databases such as retrosheet), is the value of the runner going when the ball is put into play (whether it is a straight steal or a hit and run).  Whereas there will be a few more line drive and fly ball DP, there will be fewer GDP and a few more hits I suppose.  That should enhance the value of a straight steal that is not picked up by the SB/CS numbers.  As far as the hit and run, of course the value of the runner being in motion is offset by the batters HAVING to swing, even at bad pitches, so I would guess that it is a wash at best (so why do managers use the hit and run?).  But there still should be some value on a straight steal which should in fact enhance the value of a basestealer beyond his SB/CS numbers and despite the fact that he does NOT enhance the value of the batter (contrary to what you hear from commentators).

"no way” is too absolute - you’ve seen it happen on popups falling in unexpectedly I’m sure. If we were limited to ground ball hits I’d agree with you, but not balls in the air. A fast runner needs what, 7.5 seconds to reach second, and 10.5-11 to reach 3rd? To be clogged only requires that the lead runner decide he can’t make safely make his extra base in that same time, which can occur with the combination of some tentativeness about whether the ball will be caught and some slowness. A slow runner can’t afford to depart as far from his initial base if he thinks the ball might be caught and he will have to return.
Meanwhile the fast trail runner is running all out and catching up.

Joe, yes “no way” is an exaggeration, but come on, we are both Cornell grads!  How often do you think that a batter could have stretched his hit but that the runner could not advance the extra base?  One time a year?  Let’s not get silly in this discussion.

with a few minutes on mlb.com ‘s searchable video,
I found a Carl Crawford double on 5/12/06 vs Toronto in the first inning in which you can briefly see him rounding second in the background before Julio Lugo has reached 3rd! And Lugo isn’t slow. I’m not saying that Lugo kept Crawford from getting a triple on this play [he didn’t], but it’s enough to illustrate my point; Crawford was 80 feet behind Lugo, and Lugo would “clog” 3rd base for Crawford if Crawford could have made his 85 more feet to 3rd but Lugo judged he couldn’t make his 95 more feet to home and didn’t try to advance.

cross posted -didn’t see yours before I hit submit on mine - my original guess was once or twice a year- small ? yes. silly? Maybe, but interesting on the scale Tango was discussing, which was a total of one base a year without considering these scenarios.

Joe, my guess is that these types of plays don’t have THAT much to do with base cloggers (slow runners) or not.  They are almost always on plays where the runner has to hold up because he is not sure that the ball is going to be caught or not.  But your point and example is well-taken.  I actually did not think of plays where the runner has to hold up until he knows that the ball is going to drop.  The slower runners probably hold up the batter more than the faster runners on plays like these.

Sure, in times like these, you’re going to double my estimate of 1 base!  But, it certainly has less to do with the baseclogging that McCarver is talking about (lack of speed), and more with just the situation (not wanting to be doubled-up). 

It would be an interesting study to see the OF singles and doubles rate of Carl Crawford and other speedsters, based on the speed of the runner on 1B.  My guess is that you won’t find anything, but, it’d be interesting to do, nonetheless.

Tango and MGL:

My bad—when I read the disruptive runner section I missed the pg. 326 quote Tango cites—clearly it’s limited to times when runner stays put.  (Bad idea to comment when you don’t have book in front of you.)

Still, I have a few questions/reservations about the analysis.  On the disruptive runners (staying put), shouldn’t you compare hitter performance to what these hitters do w/non-disruptive runner on 1B, rather than their overall wOBA?  I say that because while your disruptive runner sample is large in terms of PAs, it’s not in terms of hitters (I’d guess 40-50 players account for majority of PAs).  It could be that these #2 hitters are disproportionately young, fast, and/or RH, so we’d expect little benefit from runner on 1B.  Also, the pitchers are not random:  disruptive runners will stay put more when facing LH pitcher and/or great defensive catcher.  So you should also adjust for these hitters’ performance against this particular array of pitchers.

There could also be a selection bias in that a disruptive runner is more likely to still be on 1B at completion of a short PA (batter puts ball in play on 1st or second pitch) than a long PA.  Since that precludes a BB, wOBA could be lower.  So the best comparison I think is between these hitters’ wOBA for ALL PAs that begin with disruptive runner on 1B (whether or not SBA ensues) vs. all their PAs that begin with a non-disruptive hitter on 1B.  That should tell us if a hitter is really worse off with Rickey on 1B.

* * *

On disruptive events, you report a -.008 impact of SBAs on hitters’ wOBA (and a 22-point swing compared to what happens if runner stays put).  Here I think the impact of H&R plays may be quite substantial.  On a 2-strike H&R attempt, if the hitter Ks it will be captured in your study, but if batter puts ball in play it won’t be. So that will artificially depress the hitter’s wOBA on recorded SBAs.  Also, when the result is a SB, there will be more IBBs for the following hitter, which I don’t think you include in wOBA.  If you could exclude 2-strike SBAs, and account for IBBs, I’d guess the 8-point loss is much smaller.

Finally, the discussion of win probability.  In last graf of the chapter you say that “break-even point would need to rise a few points” (over your previous calculations) to account for disruption of the hitter.  However, at least two-thirds of the 22-point wOBA swing is already captured by your WE calculations: when you calculate initial WE with runner on first, that already incorporates advantage to hitter of runner on 1B; when you calculate new WE (runner on 2b, or no runner/extra out) it already factors in the loss of that advantage to the next hitter.  So the WE calculations already capture most, perhaps all, of the SBA impact on the hitter.

Guy, good points.

In terms of your second paragraph, the sample could be biased for the type of hitters and pitchers you cite, but I would be very surprised if that’s the case.  I could look into it, though.

I hadn’t thought about your third paragraph (the count), but I don’t see how a “short count” is necessarily a good or bad thing in terms of wOBA for the hitter.  I would bet that the short count is neutral to that effect.  In any case, I doubt the premise is true.  These runners only attempt to steal 25% of the time.  You are suggesting that perhaps they would attempt 35 or 40% of the time if the count was longer.  Your point is very intriguing.  My gut tells me we won’t find anything, but, certainly worth looking into as well.

***

As for the WE, that is also intriguing.  I’ve gotta run, but this is one of those things that is tough to separate out.  You may indeed be correct.

In the early posts on this thread, Tango estimated that the best/worst baserunners add +/- 6 runs per season compared to average runners, and MGL said he’s come up with a figure of +/- 3 or 4.

May I ask which one is ‘right’?

It would be interesting to do this math with teams that featured multiple slow-footed high-OBP guys.  Like the 2001-2003 Mariners with Olerud and Edgar.

Actually, I said +/-3 runs on hits, and maybe double that if you count getting bases on outs.  It all really depends on what you include.  Tippett looked into even more things.  He concluded the following:

By looking at Ichiro’s 2001 season in detail, we found the following:

3 more times reaching on error than the average hitter
6 more errors on steal attempts than the average
2 more errors forced on pickoff throws
6 extra bases taken on singles and doubles
7 extra bases on grounders while at third base
1 extra stolen base for a teammate while at third base
8 trailing runners advanced on double steals
1 extra base on a grounder while at second base
-2 fewer bases taken on fly balls while at second base
-2 fewer bases taken on WP/PB while at second base

That’s around 7 or 8 runs in all.  So, it all depends on what you are counting.

Guy, my study before the study in the book was done exactly the same way you describe. I compared all batters with a basestealer on first and a non-basestealer and took a weighted (by the lesser of the PA) average of the difference in all categories (s,d,t,hr,bb,ho,so).  While the individual component rates were different (suggesting things are going on), the overall OPS or wOBA was exactly the same.  I do agree that there is a potential selection bias when the basestealer runner is not going.  That would not be the case for the non-basestealer, as they are rarely in motion anyway.

May as well check the actual numbers for 2005. By my count, Ortiz was on second 85 times with first base occupied (by Ramirez 53, Millar 13, Varitek 7, Nixon 6, others 6). A single was hit 11 times, with Ortiz scoring three times, and the runner going to third once when Ortiz scored.

For comparison, the numbers for Johnny Damon are (with numbers just presented for Ortiz in brackets):

On second with first occupied: 107 (Ortiz 85)
Singles: 21 (11)
Scored: 14 (3)
Runner on 1st went to third: 2 (0)

Interesting that of the 14 singles which scored Damon from second, the runner on first only went to third (Renteria 1-for-7, Ortiz 0-4, Ramirez 1-2, Nixon 0-1) twice, and held up at second 12 times.

Small correction, the last line of that mini-table should be:

Runner on 1st went to third: 2 (1)

Tango:
On the WE issue, you can break the 22-point wOBA “swing” on SBA into two parts.  Part A is the hitter loses 14 points because there’s no longer a runner on 1b (he’s out, on 2b, or on 3b).  Part B (about which I’m more skeptical, but...) is the hitter loses another 8 points because the SB/CS disrupts him (presumably, mainly by forcing him to take a strike).  My theory is that WE captures Part A already, so the only additional impact, if any, is part B.

Also, when you do the WE analysis you report the starting WE on SBAs (.580) but not the total WE after all attempts.  What is the total net effect in terms of WE?

Another speed question is how much it contributes to hit results. How often do fast runners beat out grounders, stretch 1b into 2b, and 2b into 3b--as compared with an avg runner. IOW, what would guys like J Pierre or T Womack hit if they were just avg runners? (Probably .000/.000/.000, since they wouldn’t be in the majors )

Doing a back of envelope calculation, I’m getting something like 10 runs per season. And most of the value is on beating out grounders (and bunts).

dackle, great work!  It should be apparent, when you think about it, that baseclogging doesn’t make sense.  The average runner goes 1st to 3rd 33% of the time, meaning the range is 17 to 50%.  The average runners goes 2nd to home about 60% of the time, meaning the range is 45 to 75%.  There’s just so little possibility for overlap that it doesn’t make sense to think about it, much less make a strategic decision based on it.

***

Guy, the starting and ending WE for all steals (not just those of 2B), can be found in Table 11.  Was this what you were looking for?

Tango:  I was talking about the calculation on page 334, where you say the starting WE on all SB events is .580 and then calculate a revised break-even percentage.  I was wondering what the ending WE was. 

Table 11 may give me the answer.  If I look at SB, CS, and PKs, it looks like the net effect is -22 wins over 4 seasons (9720 games), pretty close to a wash.  But, at least some of the PKs are not true SBAs, and some of the CSs (and SBs) are busted H&Rs, as we’ve discussed.  And I’d think that at least some Balks result from threat of SB.  So it seems pretty likely that the net W/L effect of all true SBAs is in fact positive, not negative (though of course not nearly as important as the Joe Morgans of the world would suggest).

Some good data from Dan Fox on baseclogging:

http://danagonistes.blogspot.com/2006/10/quick-takes-and-clogging.html

Posted by: G
Email:
URL: http://mlbnation.net/forums/

Tim McCarver must be cut from the same cloth as Dusty Baker, b/c Dusty believes the same thing:

“I think walks are overrated unless you can run. If you get a walk and put the pitcher in a stretch, that helps, but the guy who walks and
can’t run, most of the time he’s clogging up the bases for somebody who can run.

“Who have been the champions the last seven, eight years? Have you ever heard the Yankees talk about on-base percentage and walks? Walks help. They do help. But you aren’t going to walk across the plate, you’re going to hit across the plate. That’s the school I come from.

“It’s like when I see kids in Little League and they make the small kids go up there and try to get a walk. That’s not any fun. Do you ever see the top 10 walking (rankings)? You see top 10 batting average. A lot of those top 10 do walk, but the name of the game is to hit.”
-- Dusty Baker.