no LAND MASS is slower than Bengie Molina
LOL. Love that Joe.

Excellent post, Joe. I share your smarter-than-most-about-the-stats-but-in-no-way-want-to-do-the-math-myself limbo.

This is interesting, and confirms my impressions:

Score from second on a single:
Crawford: 14 out of 23
Papi: 18 out of 31

Score from first on a double:
Crawford: 4 out of 8
Papi: 1 out of 9

Go 1st to third on a single:
Crawford: 5 out of 20
Papi: 8 out of 36

Crawford did take 26 bases on wild pitches, passed balls, balls, sac flies. But Papi took 27 bases.

Big Papi, while not fast, can get up a pretty reasonable head of steam. He always runs hard, and I think he's effectively aggressive on the basepaths. Especially given that a bunch of those doubles are almost certainly wallballs, that's a very good baserunning record.

I like Papi's chances in a race against Paul Konerko.

Especially given that a bunch of those doubles are almost certainly wallballs, that's a very good baserunning record.

Wouldn't that make it easier to score since you don't have to wait as long to tell that the ball isn't going to be caught?

Wouldn't that make it easier to score since you don't have to wait as long to tell that the ball isn't going to be caught?

Yes and no. Generally, yes, for long flies to the outfield you can get a good read on it - and if you don't (and it's caught) the LF has a long throw back to first to double you up.

On a number of wallball doubles at Fenway - particularly the line drives - it's a close play at second. The LF, when he gets the ball, is only 220 feet or so from 2B. It's a pretty quick relay to get the ball back in. If the batter has trouble making 2 bases, a runner on first - even with a good read on the ball and a full-steam run - doesn't have much time to get past third base before the ball is back in the infield.

Most line drives to left aren't doubles at Fenway---they are only singles. You have to get it to left-center to get a double, pretty much....people trying to take two on liners are part of why Manny's assist totals are pretty healthy.

A double to left is usually a pretty high pop fly that scrapes the wall, or a shot that hits hard and high on the wall and bounces back a good bit.

a player with a 113 OPS+, for instance, performed 13 percent better than league average

The preposterous decision to subtract 100 rather than divide by two bites another rational person on the ass.

I admit it. I heart Joe Posnanski.

I didn't realize Eddie Murray was so good in 1974.

Also, what are Springsteen's non-mainstream albums?

I would guess he's talking about bootleg albums

The preposterous decision to subtract 100 rather than divide by two bites another rational person on the ass.

I'll admit it: I don't get this. What's the correct way to interpret OPS+, if not Posnanski's way?

I'll admit it: I don't get this. What's the correct way to interpret OPS+, if not Posnanski's way?

OPS+ on BB-Ref is computed as 100 * (OBP/lgOBP* + SLG/lgSLG* - 1).

OPS+ on BB-Ref is computed as 100 * (OBP/lgOBP* + SLG/lgSLG* - 1).

I see. So really, there is no correct way to interpret it. That's disappointing.

I would have thought that OPS+ was calculated as 100 * ( ( OBP + SLG ) / ( lgOBP* + lgSLG* ) ). Why is it not done this way? To increase the weight given to OBP?

To increase the weight given to OBP?

It elminates the scaling issues so that there is an equal relationship between OBP and SLG.

I think he cherry picked his answer here. Ortiz is known to be an effective baserunner due to baseball intelligence which he's high on, and can be surprisingly fast. In fact, I'm sure Bill James +/- number on him isn't all that bad. So obviously it won't look like Crawford is much faster than him on paper. Pick a guy like Frank Thomas, who last year had a +/- of -25 or so, and then compare him to Crawford, and you'll see a huge difference.

It elminates the scaling issues so that there is an equal relationship between OBP and SLG.

you could also do it as (obp/leagueOBP)*(slg/leagueSLG)
which I've seen some claim is more accurate since OBP and SLG should be multiplied together rather than added... (If I really cared I'd check to see which correlates better with actual runs scoring)

or you could do it as: ((OBP * 1.8) + SLG)/ ((leagueOBP * 1.8) + leagueSLG)
(essentially normalized GPA)

personally I think the stathead community should agree on some run estimator
divide by outs, adjust for park
and use pythag to come up with an OWP (offensive winning %) that we can all use as a handy refernece point rather than OPS, OPS+, GPA, EQA etc etc etc...

It elminates the scaling issues so that there is an equal relationship between OBP and SLG.

Right, effectively increasing the weighting of OBP, since it usually has a smaller magnitude than SLG.

What was the rationale behind the choice to subtract 100 rather than dividing by two? I can't see any way in which subtracting 100 is the superior choice.

personally I think the stathead community should agree on some run estimator

Have you ever tried to get the community to agree on ANYTHING??

-- MWE

Have you ever tried to get the community to agree on ANYTHING??
Plaschke sucks?

Plaschke doesn't suck, he blows.

he swallows.

I'm a fan of OWP myself. I'm w/ JPWF13.

What was the rationale behind the choice to subtract 100 rather than dividing by two? I can't see any way in which subtracting 100 is the superior choice.

It's how Total Baseball did it (PRo+) and I assume BBREF simply followed suit from there.

subtracting 100 rather than dividing 2 is the superior choice if the result correlates more closely with actual run scoring.

ok real Q & D
I did "OPS+" for each team (not park adjusted- I want to compare with actual runs scored, raw OPS+ is just as good- otherwise I'd have to park adjust each team's runs scored) and I calculated OPS+ by dividing by 2 rather than subtracting 1.

I converted OPS+ into runs (simple formula: league average runs/100 times team OPS+, as run estimators go its; surprisingly effective btw)
mean error regular OPS+ = 2.12% (worst error was on Texas- 6.1%)

Calculating OPS+ by dividing by 2- mean error is 4.04% (worst error was on Yankees 12.2%)

so subtracting 100 works MUCH better than dividing by 2.

BTW GPA's mean error by the same method is 4.13%

Have you ever tried to get the community to agree on ANYTHING??

Roger Clemens and Barry Bonds are innocent.

Have you ever tried to get the community to agree on ANYTHING??

Criticism of dusty Baker is racially motivated.

Hee Seop Choi would be the second coming of Lou Gehrig if given the chance.

Suri Cruise was not conceived using L Ron Hubbard's frozen sperm

I converted OPS+ into runs (simple formula: league average runs/100 times team OPS+, as run estimators go its; surprisingly effective btw)
mean error regular OPS+ = 2.12% (worst error was on Texas- 6.1%)

Calculating OPS+ by dividing by 2- mean error is 4.04% (worst error was on Yankees 12.2%)

so subtracting 100 works MUCH better than dividing by 2.

Don't you need different run estimators? The "divide by two" method results in a much smaller range of values. For example, Barry Bonds' OPS+ in 2002 is 268 by the "subtract 100" method but only 185 using the "divide by two" method.

Furthermore, and this could be a lack of statistics education talking, but wouldn't the linear correlation between both methods be 1.0? I mean, if you are looking for a linear formula that relates the two methods, it's pretty easy to see that they are perfectly correlated: A = 0.5 B + 50 or A = 2.0 B - 100 (depending on which way you're going). And if the two methods are perfectly linearly correlated to each other, doesn't that mean that they will linearly correlate equally well with another measurement?

I'd run the numbers myself, but I don't have the time at the moment, maybe I will this evening.

Also, what are Springsteen's non-mainstream albums?

Nebraska.

Don't you need different run estimators?

yes, but Q & D = quick and dirty

hell if I wanted a run estimator I'd use Baseruns or ERP or the latest version of RC before coverting OPS into runs...

someone asked why normalize OPS (which is essentially what OPS+ is) by adding normalized SLG to normalized OBP and subtracting 100 rather than adding SLG and OBP and dividing by 2

the Q & D answer is that OPS+ of 150 really does mean that a player created approximately 50% more runs than average - while an OPS(/2) figure of 150 means a player created approximately 75-80% more runs.

The "divide by two" method results in a much smaller range of values.

Yes it does, and that is it's flaw.

If you normalize a figure to 100, most any person who looks at a score of 115 will say, "that means 15% better than average", well talking about a combined normalized figure like OPS+ it's not strictly accurate to say that 115 is 15% better than average, but as a rough approximation it is, as the old saying went, close enough for government work.

Any way I tolerate OPS+ because I grew up hearing people endlessly recite AVG-HR-RBI as the statistical gold standard, and I quote OPS+ figures because
1: Everyone here knows what it is
2: A surprising # of non-stathead fans know what it is
3: It's the best sorting stat available on BBREF PI

the Q & D answer is that OPS+ of 150 really does mean that a player created approximately 50% more runs than average - while an OPS(/2) figure of 150 means a player created approximately 75-80% more runs.

Yeah, but that's because OPS isn't denominated in runs. If I saw an OPS+ of 150, I would have assumed that this means that someone's OPS is 50% higher than an average OPS, not that they produce 50% more runs. How OPS converts into runs is another question entirely. If OPS+ is trying to solve a problem other than adjusting OPS to remove league or park bias (etc), then I submit that it should not be called OPS+, since it is measuring something else. It's all semantics though; I still like OPS+, since I am used to its current scale. I'm not going to campaign for changing its name, I just think that the original name choice happened to obfuscate what it actually represents.

I just think that the original name choice happened to obfuscate what it actually represents.

Which original name choice?
In my old [really old] copy of Total Baseball it was called:

/A

Now WTF was that supposed to mean???? [in case you are wondering the A stood for "adjusted" as in the number was park adjusted)

Later editions of Total Baseball called it Pro+
Total Baseball called OPS "Pro"- short for production... I don;t know what "+" stood for in the Total Baseball editors' minds, some have taken it to refer to the fact that the number was normalized against league averages, some that it was park adjusted, I'm sure Total Baseball's glossary straightens that out.

BBREF calls it OPS+ which seems to have caught on (and despite what Jesus Melendez seems to think the BBREF and Total Baseball stats are the same thing- the only difference is how each accounts for park adjustments)...

Personally I think OPS+ is just the latest (and so far most successful) attempt to replace Ye Olde Chadwick Batting Average with something that actually works... 25 year ago I thought Total Average would catch on... 20 year ago I thought the idea of OWP was going to succeed. (Actually I still think OWP was and is the "right" idea, but James' execution of his own idea was botched and since then he's seemingly abandoned it).

I agree with John Lynch. His post 30 states the issue clearly. It seems silly to call it OPS+ when that's not what it is. As far as I can tell, the "Foo+" notation generally indicates Foo/lgavgFoo -- I'm happy to be corrected if there are exceptions besides OPS+. Just because it correlates better with runs created doesn't justify calling it "OPS+" -- it should be called something like "RC+".

Does it accurately map the degrees of separation between players? If it does, I'm still happy as a lark with it. If it doesn't, I'll stop using it. I don't care what it measures, as long as the measures accurately reflect the distinctions inherit in reality. All that matters to me is to get a reasonably accurate picture of league/park adjusted production. Everything else is details and false precision, IMHO.