Good stuff here, and the coins thing is a pretty snappy analogy.
Although, does that mean walks are sort of like... subway tokens?

Pretty cool, and very common sense. Obviously, batting average has been overvalued historically. However, I think BA gets unfairly derided amongst some sabr-inclined fans.

It's virtually impossible to be a worthwhile ballplayer without being to hit for contact. Rob Deer had a few years which basically prove this (1989 in particular stands out.) 30 homers and 100 walks mean very little if you can't keep your average over the Mendoza line. The same seems to be true for strikeouts. It becomes hard to hit for a decent average once a player is north of around 200 strikeouts in a season.

The cited F.C. Lane article proposes a crude system of linear weights - .30 for a single, .60 for a double, .90 for a triple, and 1.15 for a home run - and then proposes to divide the total of the linear weights for a batter by his AB to rank hitters.

Lane writes:

As it appears above, the batting average of Jake Daubert, reckoned on any sane basis, is not equal to that of Cactus Cravath by a very wide margin. In fact, the two are not in the same class. And yet, according to the present system, Daubert is the better batter of the two. It is grotesqueries such as this that bring the whole foundation of baseball statistics into disrepute.

since by his proposed system, Cravath's 1915 hit value per AB would have been .151, Daubert's would have been .106, even though Daubert had a higher BA and was considered to be a "better hitter" at the time. By OPS+, Cravath was at 171 in 1915, Daubert 125.

-- MWE

It's virtually impossible to be a worthwhile ballplayer without being to hit for contact. Rob Deer had a few years which basically prove this (1989 in particular stands out.) 30 homers and 100 walks mean very little if you can't keep your average over the Mendoza line. The same seems to be true for strikeouts. It becomes hard to hit for a decent average once a player is north of around 200 strikeouts in a season.

Isn't Rob Deer an example of being a worthwhile player without hitting for contact/average? Career .220 average, led the majors in strikeouts four times, a productive hitter even when hitting .210, major league starter for eight straight seasons. Deer took a lot of walks and had tremendous power, so most guys who hit that low aren't going to have similar success, but he was passable.

Not to say that I disagree with your general points - I just don't think that Deer is the best representative of a guy whose strikeouts were what kept him from being a viable major league player.

Isn't Rob Deer an example of being a worthwhile player without hitting for contact/average? Career .220 average, led the majors in strikeouts four times, a productive hitter even when hitting .210,

Not really. In general, he was a productive player when he was hitting .230+. Below that, the power/walks couldn't offset the putrid BA.

In fairness to Tango, he wrote about this article sometime back, either here or at The Book blog.

Looks like it might have been over at Baseball Fever, back in 2007...

Not trying to dump on Mike or anyone else that is picking up on this piece now (looks like there's also something at FanGraphs on it that was posted yesterday), but it looks like this is something that popped up earlier and was then re-forgotten.

'Rediscovering' Lane is always a worthwhile endeavor.
Used reprints of his book 'Batting' are available for a penny (+ s/h) at Amazon.

#4 Indeed. There was the cult of the Deer on RSB.

Worth noting though that DA saw Deer's glove as a positive. 3rd best in range in right fielders in the DA era (both in range and in terms of extra base prevention. IOW he was pretty good at something that as far as I know TotalZone takes no notice of)

Dale Stephenson gets him as +57 runs from 1987-1994.

Also, there's context: .210/.309/.428 (his numbers from 1989-90) looks awful, but that's a 106 OPS+. Nothing wonderful about that, but that's a worthwhile regular if he is in fact a good defensive player.

Thanks for the link Komminsk, I just ordered a copy.

It's virtually impossible to be a worthwhile ballplayer without being to hit for contact. Rob Deer had a few years which basically prove this (1989 in particular stands out.) 30 homers and 100 walks mean very little if you can't keep your average over the Mendoza line. The same seems to be true for strikeouts. It becomes hard to hit for a decent average once a player is north of around 200 strikeouts in a season.

Why is batting average unique in this respect?

Isn't it virtually impossible to be a worthwhile baseball player if you can't hit doubles? What's the best season with more than 500 PAs and less than 10 doubles?

Or, to be even more obvious, it's virtually impossible to be a worthwhile baseball player if you can't get on base.

Batting average is useful in some respects, but only really because it provides a big chunk of OBP and SLG. If you have access to those numbers, the only additional information you're getting from batting average is a qualitative story. That is: you can get a sense of the shape of the production.

I realize I'm not telling you things you don't know. I guess I'm just backlashing against the backlash against the batting average backlash.

I bought mine when someone else (I forget who) posted it here some months back. :)

What's the best season with more than 500 PAs and less than 10 doubles?

I'd go with Roy Thomas, 1904.

Batting average is useful in some respects, but only really because it provides a big chunk of OBP and SLG. If you have access to those numbers, the only additional information you're getting from batting average is a qualitative story.

Sure, but why stop there.

OPB and SLG are only useful in that they provide a big chunk of WAR (or OPS+ or EQA).* If you have access to those numbers, the only additional information you're getting from OBP or SLG is a qualitative story.

You ask why BA is unique in that respect. It's not. But no one says "doubles are worthless" either. BA is historically overvalued throughout the baseball world and frequently undervalued at places like this. In all cases, it's never been BA's fault.

* For illustrative purposes only. I'm actually more of a component guy than an all-enclusive number feller.

OPB and SLG are only useful in that they provide a big chunk of WAR (or OPS+ or EQA).* If you have access to those numbers, the only additional information you're getting from OBP or SLG is a qualitative story.

There's one major difference: OBP and SLG are directly calculated from events in play, while WAR/OPS+/EQA are adjusted stats. That's a major leap, and the adjustments are far from universally agreed upon.

You ask why BA is unique in that respect. It's not. But no one says "doubles are worthless" either. BA is historically overvalued throughout the baseball world and frequently undervalued at places like this. In all cases, it's never been BA's fault.

I don't think it's any more valuable than knowing how many doubles a player hits in a season. Knowing the number of doubles might have greater predictive value than knowing the BA, actually. (And I'd say that knowing just doubles generally isn't a very good measure of player quality, much like BA.)

'Rediscovering' Lane is always a worthwhile endeavor.
Used reprints of his book 'Batting' are available for a penny (+ s/h) at Amazon.

That "Hitting" reprint was a SABR premium to its members 10 years ago. Great stuff.

Lane died at the age of 98 in early 1984. Bill James had been writing his Abstracts for seven years at that point. Thorn & Palmer were about to publish "The Hidden Game of Baseball" which elaborated on Lane's 70-year-old theories. Heck, I joined SABR in 1984.

I don't think it's any more valuable than knowing how many doubles a player hits in a season. Knowing the number of doubles might have greater predictive value than knowing the BA, actually. (And I'd say that knowing just doubles generally isn't a very good measure of player quality, much like BA.)

Every argument against Chadwick Ratio comes down to two things: It doesn't do a good job of measuring value and/or it doesn't measure what the commenter thinks it should measure. Neither of those is actually an argument against BA.

It remains what it's always been: an important component of offensive baseball )and pitching/defense, for that matter) and a statistic that adequately meausures what it intends to measure (the rate a batter is successful at earning a base hit compared to the number of opportunities he's given).

"Hitting"="Batting"

my bad

Just imagine if this had been written by someone with a manly name like Damon or Ring, instead of Ferdinand. The Chadwick Ratio would have been banished before any of us were born!

Used reprints of his book 'Batting' are available for a penny (+ s/h) at Amazon.

Looks like somebody saw your endorsement and jacked up the price. $1.68. With water damage. Might as well spring for the five bucks for a new copy.

Or the $3.65 Prime Eligible one, for free shipping. If you're not a Prime member, I don't want to talk to you ;)

Batting average is useful in some respects, but only really because it provides a big chunk of OBP and SLG. If you have access to those numbers, the only additional information you're getting from batting average is a qualitative story.

There are 760 batter-seasons from 2007 to 2011 which qualified for league batting title. The correlations among the batting rate stats are:

      BA  OBP  SLG  OPS  
BA 1.00
OBP 0.70 1.00
SLG 0.53 0.65 1.00
OPS 0.64 0.83 0.96 1.00


#21 Correlations aren't the way to approach this. Run a multiple regression with BA, OBP and SLG against runs scored per game.

In any long study BA will come back as insignificant. This is complicated by the fact that the relative value of OBP and SLG varies with the offensive context.

Further, in very high offensive contexts it looks like BA does in fact matter somewhat. Basically it's almost always true that if you have OBP and SLG and weight them properly (OBP*1.7 + SLG is a good general rule. The higher the offensive context, the greater the relative importance of OBP. The lowest I've found is the NL, 1968 when OBP*1.5 +SLG seems to work best -- Coors in the late 90s, OBP*2.25 + SLG works best but doesn't work as well as usual. BA absolutely matters) then BA is of only narrative value. (Yes, there are issues with young players who have "old player" skills. They're generally overstated though.)

More to the point, it doesn't change the standard error to re-run the study using BA, ISO (SLG-BA) and IWR (OBP-BA). If BA was carrying any kind of important value, it should make the standard error smaller.

Every argument against Chadwick Ratio comes down to two things: It doesn't do a good job of measuring value and/or it doesn't measure what the commenter thinks it should measure. Neither of those is actually an argument against BA.

It remains what it's always been: an important component of offensive baseball )and pitching/defense, for that matter) and a statistic that adequately meausures what it intends to measure (the rate a batter is successful at earning a base hit compared to the number of opportunities he's given).

This is a silly semantic game. Of course BA adequately measures what it is defined as; that's a tautology. Measuring the number of hits per at-bat that a player gets on Thursday night games "adequately measures" the rate at which a player gets hits in Thursday night games. That doesn't make this statistic useful as much more than a curiosity.

What people mean when they criticize BA is that it isn't a good measure of value, and despite that, people are using it in support of an evaluation of the value of players. You may be able to bang a nail into a wall with the back of a screwdriver, but it doesn't mean that you've used a particular good tool for the job, and sometimes, you'll do a much crappier job if you use that instead of a hammer. I like seeing BA as commentary on the shape of a player's performance, but for actual contribution to winning, it's not particularly useful.

This is a silly semantic game. Of course BA adequately measures what it is defined as; that's a tautology.

Not really. Some things do a very good job measuring what they intend to, but are limited in scope. Hits, for example. Other stats try to do a much more complex job, say something like pitcher wins (as a traditional statistic) or WAR (for the advanced metric set), but are far less accurate in the process. BA happens to be very good at measuring what it was designed to measure, more so than some other statistics.

What people mean when they criticize BA is that it isn't a good measure of value, and despite that, people are using it in support of an evaluation of the value of players. You may be able to bang a nail into a wall with the back of a screwdriver, but it doesn't mean that you've used a particular good tool for the job, and sometimes, you'll do a much crappier job if you use that instead of a hammer.

And criticizing BA because it does a poor job of measuring value makes about as much sense as criticizing the screwdriver for being ill-suited for banging in a nail. The criticism should be directed at the nitwits who are using the tools incorrectly.

Batting average doesn't do anything useful well.

Does anyone really care about a player's ratio of hits to plate appearances minus walks, sac flies, sac bunts, and times hit by pitches?

Isn't it virtually impossible to be a worthwhile baseball player if you can't hit doubles? What's the best season with more than 500 PAs and less than 10 doubles?

John McGraw once had a season where he had eight doubles in 652 PAs and 8.5 WAR. But I guess that was kind of a long time ago.

Batting average doesn't do anything useful well.

If you only use baseball statistics for the purpose of assessing value, then you are correct.

Looks like it might have been over at Baseball Fever, back in 2007...

I think Cyril wrote about it first.

Speaking of Chadwick ratio, here's Chadwick's logic:

"Long hits are showy, but they do not pay in the long run. Sharp grounders insuring first-base certain, and sometime second-base easily, are worth all the hits made for home-runs which players strive for."

Henry Chadwick, 1868

I have to believe that a big part of this is because he came to baseball via cricket, and the risk/reward equation is very different in cricket. He was a very sharp guy, but formative influences are hard to shake.

"Long hits are showy, but they do not pay in the long run. Sharp grounders insuring first-base certain, and sometime second-base easily, are worth all the hits made for home-runs which players strive for."

Henry Chadwick, 1868

Since walks were virtually nonexistent at the time, he's basically arguing in favor of OBP. The first sabermetrician.

Besides, if we didn't have the Chadwick ratio, how would we get to the all-important BABIP?

Batting average doesn't do anything useful well.

Does anyone really care about a player's ratio of hits to plate appearances minus walks, sac flies, sac bunts, and times hit by pitches?

I would say a significant portion of the casual fans care about this...

I think it's silly to dismiss batting average, as it's arguably one of the most important and hard skills to maintain in the game.. Fans think the ability to hit the ball(which is ####### hard) at the major league level is something to be celebrated.

Only around here do people dismiss it out of hand. Yes it's not about value, but it's definitely a skill.

Sorry, but I've been railing against this for years ...

#21 Correlations aren't the way to approach this. Run a multiple regression with BA, OBP and SLG against runs scored per game.

In any long study BA will come back as insignificant.

BECAUSE THAT'S A STUPID REGRESSION TO RUN.

Please, people, if you're going to run regressions, please learn how to interpret them. A coefficient is the impact of one variable controlling for the others.

Control BA and SLG and a change in OBP is (essentially) a walk. That's something meaningful.

Control BA and OBP and a change in SLG is (essentially) an extra base. That's something meaningful.

Control OBP and SLG while changing BA ... and that's completely meaningless. It's a stupid thing to estimate. It's something like trading a HR and 2 BB for a triple, a double and a single. Why would anybody want to know this? Why would anybody think this is a meaningful coefficient to estimate? The very notion of "changing" BA while keeping OBP and SLG constant is absurd -- how would you run such an experiment if you could?

What is the ACTUAL effect of BA in the proposed regression?

runs_hat = b0 + b1*BA + b2*OBP + b3*SLG

= b0 + b1*BA + b2(BA + ISO_OBP) + b3*(BA + ISO)

= b0 + (b1 + b2 + b3)*BA + b2*ISO_OBP + b3*ISO

dy/dBA = b1 + b2 + b3 ... which you'll notice is a lot bigger effect than walks or extra bases have.

Now you've isolated the value of getting a hit from getting a walk and getting an extra base. BA is way more important. Not to mention that, with a few exceptions, BA>ISO and, with no exceptions I'm aware of, BA>>ISO_OBP.

Go ahead, try it Ron, regress runs on BA, ISO_OBP and ISO and report back. Yes, it's all cleaner if you divide everything by PA rather than BA and ISO being based on AB. Feel free to do it that way.

You find essentially the same thing in xRuns or linear weights.

xR = .51*1B + .8*2B + 1.14*3B + 1.46*HR + .33*(UIBB)

which is approximately equal to

.33*bases + .15*hit

Linear weights is a bit different (triples in particular being worth less)

basic RC = (times on base)*(total bases)/PA

= (H + other)*(H + XB_)/PA
= (H^2 + H*XB + H*other + other*XB) / PA

'Rediscovering' Lane is always a worthwhile endeavor.
Used reprints of his book 'Batting' are available for a penny (+ s/h) at Amazon.

For anyone who doesn't know Batting, it's a compilation of interviews about the art of hitting with all the top batsmen in the game, arranged by topic. The subtitle is "The Secrets of Major League Batting and Useful Hints for Hitters of All Ranks and Ages Collected Over a Period of Fifteen Years** From Nearly Three Hundred Famous Players". There's never been anything quite like it before or since.

**from 1910 through 1925

Since walks were virtually nonexistent at the time, he's basically arguing in favor of OBP. The first sabermetrician.

And "showy' or not, XBH were pretty damned rare events in Chadwick's day too. League ISOs in the NA (1871-1875) were .097, .063, .067, .060, .056.

I have to believe that a big part of this is because he came to baseball via cricket, and the risk/reward equation is very different in cricket. He was a very sharp guy, but formative influences are hard to shake.

OTOH, he switched from runs to hits because he recognized that run scoring is largely a team function in baseball, but an individual function in cricket.

#28 McCoy

I checked and I mentioned something about Lane and this article on the SABR list back in 2003 and I have had links to it on my page for awhile

#32 If you actually troubled to read #22 you'd see I have. The standard error in the estimate of team runs is essentially unchanged whether you use BA, OBP and SLG or BA, ISO and IWR.

Yes, in the latter BA comes back as significant, but it leaves unchanged the central point. With certain very specific exceptions (all having to do with very high offensive contexts), BA adds no meaningful information as to the value of a player's offensive performance if you have OBP and SLG.

I know you simply don't accept this, but it genuinely does not matter how the total bases and times on base are achieved. .260/.350/.450 is of indistinguishable value from .300/.350/.450

That said, it gets a tad weirder if you dig deeper. BA with RISP is of minor significance, and the best way to estimate BA with RISP going forward is to use the estimate of BA itself. If there's any kind of ability to hit with RISP (and there does seem to be something going on with RISP -- the correlation between BA and BA with RISP is lower than I'd expect if BA with RISP is nothing more than an artifact of overall BA -- something like 85% in long careers) then it's beyond my ability to model.

So BA itself doesn't generally matter, but BA with RISP does and the guys who do well in RISP (in absolute terms that is) are almost always the guys who do well in BA.

But the impact is (over time at any rate) small, and most people don't want to include situational stats in their offensive metrics.

And criticizing BA because it does a poor job of measuring value makes about as much sense as criticizing the screwdriver for being ill-suited for banging in a nail. The criticism should be directed at the nitwits who are using the tools incorrectly.

I thought I made it clear that this is what people are doing, generally speaking. If you're looking for value, using BA is somewhat like using the back of a screwdriver to bang in a nail (although the analogy breaks down a bit because the screwdriver is more effective at driving nails than BA is at measuring value). If you're simply trying to describe the shape of a player's performance, then there shouldn't be any problem with using BA.

So BA itself doesn't generally matter, but BA with RISP does and the guys who do well in RISP (in absolute terms that is) are almost always the guys who do well in BA.

Are you sure that this isn't just noise? I was under the impression that nearly everyone does better with RISP. At the very least, we are selecting a situation where the pitcher has already shown signs of weakness.

You'd really have to control for power and discipline, I would think.

Are you sure that this isn't just noise? I was under the impression that nearly everyone does better with RISP. At the very least, we are selecting a situation where the pitcher has already shown signs of weakness.

It's not just that. A flyball to the outfield that would result in just an out and an at bat can be a sac fly*, and a ground ball that would have produced a 4-3 can slip through a drawn-in infield for a base hit. Generally speaking, BA should be higher in RISP situations.

* My biggest gripe with BA. Sac flies should be counted as at bats, dammit. It's not as bad as ROE not counting as reaching base in OBP, but it's a true flaw in the measurement.

Are you sure that this isn't just noise? I was under the impression that nearly everyone does better with RISP

Yeah I'm sure it's not noise. It's not a large effect, but it does show as statistically significant.

And while the league hits better with RISP, that's partially an issue of selection. Non-hitters are somewhat less likely to be allowed to hit with runners in scoring position. There's also the not insignificant issue that there's no reliable way to hit a sac fly without having runners in scoring position, so a certain percentage of outs don't count with RISP.

From an old study I did:

I have 187 players with 1,000+ PA with RISP. They hit .275/.346/.432/.340 overall and .278/.369/.430/.345 with RISP. (The last number being OBP with IBB removed. Didn't think to remove SF and I don't feel like re-doing this. Won't change the essential conclusion.)

Despite remarkably similar overall numbers, there's only an 86% correlation between career BA and career BA with RISP. And the standard deviation of BA with RISP is around 20% higher than expected. (An indication to my mind that it's not random. But you couldn't separate luck from ability with any confidence)

As you can see, most of the observed difference in OBP (the primary change) comes from IBBs. And there are some obvious semi-intentional walks included as well.

#30 As I'm sure I've mentioned before, Chadwick did in fact experiment with OBP (including reached on errors).

But he was doing everything by hand and just found it too much work. Every time he tracked something new he had to drop something else. So it was a one year experiment.

But yeah Chadwick was obviously the first sabrmetrician.

EDIT: As Don Malcolm pointed out, he did pitch counts in the 1860s.

And while the league hits better with RISP, that's partially an issue of selection. Non-hitters are somewhat less likely to be allowed to hit with runners in scoring position. There's also the not insignificant issue that there's no reliable way to hit a sac fly without having runners in scoring position, so a certain percentage of outs don't count with RISP.

Right, so you are more likely to replace a lousy hitter with RISP, the pitcher is more likely to be struggling (or replaced with a less capable pitcher), and the fielders are likely to be positioned differently.

But you couldn't separate luck from ability with any confidence

I'd call that noise. Maybe we're just using the term differently?

#41 We may be talking past each other. At the individual level you really can't identify any ability to hit with RISP. Though I've found a few players who meet a definition of clutch. 95(+)% confidence that their increased BA with RISP is an ability. Tony Fernandez and Paul Molitor from the study I did. Interestingly none of the players in the study met that same definition in reverse. IE none of the players hit significantly (using the word loosely -- as David Grabiner very gently pointed out statistical significance has a meaning and it's not 95% confidence) worse with RISP.

At the team level, it's not just individual ability. There is the issue of who is getting those at bats with runners in scoring position too. If everybody hits exactly the same with RISP and overall and a higher percentage of PAs go to the guys with high BA, the team BA with RISP will go up without any clutch performances.

And regardless of whether it's luck or ability, it matters to team run scoring. Just not a heck of a lot. The single greatest source of error in team run scoring estimates comes from luck/clutch/timing -- whatever you want to call it, but you can only chop about 2 runs off the standard error (to be clear, that's not insignificant -- right now I think the best metrics are in the range of a 15 run standard error) by including situational stats.

Though I've found a few players who meet a definition of clutch. 95(+)% confidence that their increased BA with RISP is an ability.

Do those players represent more than 5% of the sample? You studied what, nearly 200 players, and only 2 of them deviate enough to appear to have this ability? At that point, it sounds like it might just be random variation.