THE BOOK--Playing The Percentages In Baseball

I should note that the Batting Blocks requires that the batter not get a hit, walk or reach base on error.

A walk with 1B open and runner on 2B is not a block (and neither is it an assist).

Brilliant!  It is very Bill Jamesian.

Two questions....How did you get the +29 for Mauer? And what’s the unit of that number?

This is also kind of like the little things stat that devil_fingers wrote about on driveline mechanics.

I like the general idea of giving players credit for the positive things they do, but I don’t like the restriction that the base runner must score to have an assist counted for the batter.  He moves the guy over, he should not be punished because the next two batters strike out.

Also, runners at the corners nobody out, batter hits into a GIDP, and the runner on 3rd scores, that seems like both an assist and a block by your definitions above.

"He moves the guy over, he should not be punished because the next two batters strike out. “

What about a batter that hits a leadoff triple, and is left stranded there after 3 outs?  He gets no run scored.  Is he being “punished” by not getting a “R” because the next three batters strike out?

This is simply an accounting of runs that scored.  We have a tally in the “R” column and we have a tally in the “RBI” column.  I’m just adding a tally to the batters in-between those two guys who did something positive: “assist on the run scored”.  This is no different than an assist in hockey or basketball, where assists are awarded ONLY if a goal/basket is scored.

Now, nothing is stopping us from doing:
- batter got on base, no one drove him in
- batter moved runner over, next batters did not drive that runner over

etc, etc, etc

It is important to get out of the mindset of thinking of terms of “crediting” and “punishing” when creating a counting stat, and simply think of it in terms of giving a tally to whoever was involved in whatever it is that you want to count.

I’ll let the ANALYSTS worry about what this means.

***

Mauer: 69 - .167 * (69 + 170) = +29

Also, runners at the corners nobody out, batter hits into a GIDP, and the runner on 3rd scores, that seems like both an assist and a block by your definitions above.

Right.  A batter can get a single and be out stretching into a double.  He gets a tally in the “1B” column.  No one seems to mind when that happens.

So, I’m simply accounting for the fact that the batter was involved in a play that moves the runner forward a base, and I’m also accounting for the fact that he removed a runner from the bases.

At the end of the day, the idea is to account for each PA and each runner on some aggregate basis.  I’m suggesting each PA be aggregated as follows:
- RBI
- runner scored, no RBI (Assist)
- runner moved over, eventually scored (Assist)
- runner moved over, did not eventually score (Useless Teammates)
- runner stayed there, batter also safe (Neutral)
- runner stayed there, batter out (Block)
- runner out (Block)

That’s a nice and simple accounting, right?

That’s pretty cool. The one I’ve always wanted to see is extra bases taken (and times thrown out) on hits—taking two bases on a single, three on a double—sort of like a complement to stolen bases and caught stealing.

Add runs scored to that and you have a good version of R+RBI. There would be no need to take out the HR, right?

I agree with the point made in #4, though I don’t see it as a matter of “punishing” or “rewarding”. Correct me if I’m wrong, but shouldn’t the point of coming up with a stat like batting assists be to measure the ability of a batter to move base runners over? If it’s a simple accounting of runs scored, then it’s as context sensitive as RBIs. Wouldn’t it be better to take RBIs out of consideration altogether? As it stands now, there is no symmetry between assists and blocks. To get an assist the base runner must score, but the same isn’t true for a block. Of course, there’s then the question of whether any ability, or lack of ability, is actually being measured.

Good post.  Let me respond to each sentence.

Correct me if I’m wrong, but shouldn’t the point of coming up with a stat like batting assists be to measure the ability of a batter to move base runners over?

You are wrong.  The point of coming up with this particular stat (Assists) is to account for the middle man for the run that scored.  That is its ONLY point.

You CAN come up with a stat that measures the number of times that a runner was moved over, but that this particular batter didn’t drive home.  That’s a worthwhile stat.  But, that is NOT the stat I’m proposing. 

My stat is exactly like the “2nd assist” in hockey. And one of the conditions to get an assist is that there be a goal.

If it’s a simple accounting of runs scored, then it’s as context sensitive as RBIs.

Yes, definitely and absolutely. 

Wouldn’t it be better to take RBIs out of consideration altogether?

Why would that be better?  The definition of the stat is very simple: determine the hitter that moved the runner over who eventually scored, that this hitter did not get an RBI for.  That is its definition.  The definition cannot be wrong, since that’s the point of defining terms!

You can argue that the value of such a stat is very limited, and I will not disagree at all with that assessment.

The only thing this stat does is fill in one of the holes in the accounting for the actual run that scored.  That’s it.

As it stands now, there is no symmetry between assists and blocks. To get an assist the base runner must score, but the same isn’t true for a block.

I don’t know that we need symmetry.  Blocks are blocks and assists are assists. 

I attempted to try to link the two, but that is really irrelevant for the definition of the two terms.  Indeed, I probably would have been better off not linking the two.

Of course, there’s then the question of whether any ability, or lack of ability, is actually being measured.

That of course has nothing to do with the definition of the stat, any more than R and RBI may or may not measure ability.  Or that the number of SH a pitcher allows measures any kind of ability.

The job here is to simply account for one particular facet of play.

Howdy, Tango.  I’m not sure if I ever posted here before, but it’s been a long time if I had.  John Holway, who’s more noted for Negro League research (and baseball astrology) came up with the assist idea 25 years ago.

The hyperlink in my name leads to the Baseball Digest article where he mentions it.

Ah, those Baseball Digests.  Good times.  I used to subscribe for several years.  I definitely had that issue.

Yes, he has the same idea, except he goes overboard to my definition. The basic idea is to give an aggregate of the data in some logical fashion.

***

I was thinking whether a HR (or triple) should cancel out the assist, like Holway suggested.  But, I decided against it.  After all, in hockey, you get the assist regardless of how needed the pass was.  (I know in basketball it’s not like that.)

The reason I did that is say you have the following:
- single
- sac bunt
- strikeout
- triple

The sac bunt wasn’t needed, since the guy on 1B would have scored regardless.  But, that sac bunt did change the course of events.  And, I didn’t want to get into the ridiculousness of the ER/unearned run that you get with the silliness of reconstructing the inning as if something didn’t happen.

So, I settled on the idea that it is what it is, and not try to analyze past what I’m trying to record.

There’s data recording and there’s data analysis, and the data recorder’s job isn’t to try to analyze any more than what he is recording.

Cool.  I like “assist” as a name, but “block” sounds more like a hitter is clogging up the basepaths, or somehow obstructing a teammate’s performance.  How about “hitting strands,” or something like that?

Tangotiger,

I think I understand what you’re doing, and I just have two further comments.

Regarding a symmetry between the stats, I think that if you’re going to create a ratio you need to have a symmetry between the stats if you expect them to be comparable. That is admittedly a preference on my part and not a fact, but it makes sense to me. To create a ratio as things stand now is like creating a ratio of stolen bases divided by RBIs. Does it really mean anything? (That example is an obvious exaggeration of what you’re doing, but I use it to make my point, not with any intention to insult.)

When it comes to baseball statistics, I like to ask myself the following questions: If I were a GM, would this stat help me to evaluate a player, would it give me any new knowledge? With these stats as defined, I can’t say that they would. Yes, Justin Morneau had more assists than anyone else, but he’s a heckuva hitter so it’s no surprise. As for Francoeur, well, I live in Atlanta and I’m a Braves fan, so the fact that he led in blocks comes as no surprise.

I like this, and I’m not the smartest guy in the world when it comes to the saber stuff.

One question I have (and hopefully it isn’t too off base) is instead of using a Linear Weight type result, couldn’t it be expressed as a simple percentage of assists/blocks divided by opportunity.

A lot of not saber people can deal with/understand a percentage formula a lot easier than a Linear Weights type one.

Just curious.

Chuck said:

Regarding a symmetry between the stats, I think that if you’re going to create a ratio you need to have a symmetry between the stats if you expect them to be comparable.

I’ll just repeat what I said:

I attempted to try to link the two, but that is really irrelevant for the definition of the two terms.  Indeed, I probably would have been better off not linking the two.

Chuck:

If I were a GM, would this stat help me to evaluate a player, would it give me any new knowledge? With these stats as defined, I can’t say that they would.

Again, why would the target audience be 30 people?  Why evaluate the interestingness (that a word?  it is now!) of a stat based on the non-target audience?

The stat exists strictly to close the loop of R and RBI, to acknowledge the guy who contributed to the actual run.  That is it.  Nothing more than that.  Just as you can’t criticize OBP for giving the HR the same weight as a walk in the numerator, you can’t criticize a stat for living to its own definition.

...not with any intention to insult

No insult taken.

...is instead of using a Linear Weight type result, couldn’t it be expressed as a simple percentage of assists/blocks divided by opportunity

The truth of the matter is that I coded this last night, posted the results shortly thereafter, and haven’t given it much thought.  I was just trying to come up with some ways to express leaders and trailers.  The “opportunity” factor becomes more interesting when you account for the fact that an RDI (runners driven in) becomes a non-play, when really you’d probably want to do RDI+A+Blocks as the denominator, and maybe RDI+A in the numerator.

At the moment, I’m just going to let it hang out there as counting numbers, R, RBI, Assists, Blocks (or whatever other word might be better), and anything else I come up with, I’ll post the 2008 and 1993-08 numbers in full, and I’ll leave it to you guys to do something with it.

I enjoy visiting your site, but it breaks everytime Tango uses a blockquote tag, and the site then becomes unreadable.  Using IE6, if anyone wants to corroborate the brokeness.

I am presuming you work in an office where you are restricted from upgrading to IE7?

If someone can tell me how I can give the option to disable my style sheet, that might help.

Also, try Tools/InternetOptions/Accessibility, and see if any options there helps.  I expected the Advanced tab to have the option to ignore the stylesheet, but I don’t see it.

Otherwise.... IE7 and FF3 is all I can support.

Ok, I’ll code this tonight.  So, if anyone has any opinion, now is that time to say it.  I will record, for every batter, the number of runners on base, and for each one, they will be placed in exactly one of the buckets below:
1- runner scored, RBI to current batter
2- runner scored, no RBI to current batter (Assist)
3- runner moved over, eventually scored from other batter (Assist)
4- runner moved over, did not eventually score (Unrewarded)
5- runner stayed there, batter also safe (Neutral)
6- runner stayed there, batter out (Block)
7- runner out (Block)

Do you guys agree to these buckets?

Do you also agree that these 7 buckets can be collapsed into these 5:
- RBIs
- Assists
- Unrewarded
- Neutral
- Blocks

Still not loving “blocks,” but maybe that’s just me.  How about “LOB,” if you don’t like “strand.”

It includes runners doubled-up, so he wasn’t “left” or “stranded”. 

How about “Punishments”?

Turnovers?  (I’ve got basketball on the brain.)

Oooohhh… in hockey we say “Giveaways”. How about that?

"Unproductive outs”? 

“Blown RBIs”?

Gee, don’t know if I like, “Punishment”, even if you are trying to give me credit!  ;>)

Another complication; in your list of 7 situations the last is “7 - runner out (Block)”.  Suppose you have a runner on first and the batter gets a hit with the runner being thrown out.  But the batter ends up at 2nd or 3rd, so bases have been gained.  If the runner gained the bases and the batter was out, as in a sac bunt, he would get an assist or an unrewarded, seems sort of harsh to give the batter a block, or whatever we end up calling it, in this situation.

Funny you should say that.  In trying to account for all situations, I have a gap of 0.3%, and that is one of them (runner out, but base situation better than before… like a sac bunt, but batter ends up at 2b).

I’m going to have to come up with a better algorithm tomorrow.

Love it, Tango.  I believe your “assists” are what BJOL calls “ghost runs.”

Studes, thanks.  I had forgotten about James’ stat.

Ghost RBI: Batter advances runner who later scores, and who would not or might not have scored at the same time had the batter not advanced him.

It looks like he doesn’t give an assist if the guy subsequently scores on a HR or triple, (or double if runner was pushed from 2B to 3B).

You can make a case for either way as discussed in post 12.

Some alternatives for ‘block’ that quickly came to mind:
hindrance
holdup
delay
curb
check

Hmmm… I’m partial to Batting Giveaways, as in “he gave away an out, without moving the runner over”.

I was also thinking of Batting Stoppers ("stopped him from advancing"), Batting Freezers, or Batting Blockers.

Ok, I think I’ve figured out the algorithm, if someone wants to play along.  Here are the buckets:
- RBIs
- Assists
- Unrewarded
- Neutral
- Blocks

Step 1: Any runner that scores in this at bat, or scored in rest of the inning (and moves up a base in this at bat) will get counted as an RBI or Assist.

Step 2: Now, remove all those runners from what is about to follow.  That is, all the remaining baserunners count as Unrewarded, Neutral, or Blocks.

Step 3a:  If no outs on play (batters or runners), then any runner that moves over counts as Unrewarded; any runner that doesn’t move is counted as neutral

Step 3b: If 1 out on play, then:
i) if one runner on base
- if the final occupied base is greater than the original occupied base, then Unrewarded, otherwise Block
ii) if two runners on base
- if the final occupied base value is +2 from the original, then 2 Unrewarded, if +1 from original then 1 Unrewarded, otherwise 2 Blocks
iii) if bases loaded, then 3 blocks

Step 3c: If 2 outs on play, then:
i) if one runner on base, then 1 Block
ii) if 1B/2B
- if final dest is runner on 3B, then 1 Unrewarded, 1 Block, otherwise, 2 Blocks
iii) if 1b/3b, 2b/3b: 2 Blocks
iiii) if three runners on base: 3 Blocks

Step 3d: If 3 outs on play then:
i) if two runners on base: 2 Blocks
ii) if three runners on base: 3 Blocks

***

The one little hick, which I think I will live with, is if there is an out on the play, the batter gets on base, and the base situation is more attractive than before (i.e., the kds/25 situation) and the batter ends up scoring later in the inning.  This SHOULD be an assist.  In this algorithm, it counts as an Unrewarded.

This has got to be so rare, that I’m not going to justify coding for it.

***

Ok, we good with this?

I root for “blocks” to be called “unproductive outs”.  I already call them that in my head sometimes when I’m watching a game--usually when a batter does it for a team I’m invested in.

Heh, maybe “unrewarded” should be termed “noble effort”, as in “It was a noble effort, but...”

Have the results of this been posted online yet?

I love counting stats.  What someone does with them is their own business, but it’s always good to have them.  I’m waiting for the first manager to say something like “I’m moving him up in the lineup.  He doesn’t have a lot of RBI, but he’s had a lot of assists and he should be up more with guys on base.”

Dan: no I haven’t posted them yet, because I haven’t finished coding it.

The algorithm is getting excessively complex in trying to figure out the unrewarded/neutral/giveaways.  (Assists are straight forward.)

Some examples are singles whereby the runner is thrown out at home, while the runner from 1B reaches 3B, who never scores.  That should count as two unrewarded, but it’s getting to be a tough coding job.

Tom, why not call advancing a runner a Gretszky?  My name links to a BTF discussion of a True Hoop article that mentions Gretszkys in basketball.