Tangotiger Blog

Monday, December 02, 2013

WAR versus ERA+ and IP

It's interesting how people will not look at WAR, but rely on ERA+ and IP, and not make the connection that they are one and the same.

Take as a for-instance, Mike Mussina.  His ERA+ is 123 and he has nearly 3600 IP.  How do you quickly convert that into WAR?

(5.5 - (450/ERA+)?) * (IP/9) /9

The "5.5" is replacement level runs allowed per 9IP.  The "450" is 100 times the league average runs allowed per 9 IP.  ERA+ is as you'd find on BR.com.  The first "/9" is to convert IP to games.  The second "/9" is to convert runs to wins.  For Mussina, the above comes out to 82.

Maybe you don't like this particular construction.  That's ok!  Come up with one of your own.  And apply it consistently to all pitchers.  Don't be weak and say you do not look at WAR, and then say you do look at ERA+ and IP, and then... not tell us how you combine it.

WAR is telling you how it combines it.  When you say you don't like WAR,  you are saying you don't like the way ERA+ and IP are combined.  But at the very least, then present the way you DO like to combine them.

For you Fangraphs fans, you'd do this instead:

(5.5 - (0.0450*ERA-)?) * (IP/9) /9

***

Maybe you don't like the "5.5" as the baseline comparison level.  So, change it to 5.0 or 6.0 or whatever you want.  That's really the only thing you can tweak really.

Thursday, November 21, 2013

Indis v Win Shares

A few years back, I introduced a player's Individualized Won-Loss records (or The Indis).  It never got traction, at least, not yet.  I remember Bill James had a great line: all his ideas are orphans, that once he lets them out, it's up to the rest of us to decide what to do with them.  It's a good perspective.

Anyway, so he's talking about the new HOF ballot, and he posts some Win Shares / Loss Shares numbers for Concepcion and Garvey (others are forthcoming tomorrow). 

As you may know, there are two main fundamental differences between WAR and Win Shares.  One is the split between nonpitchers and pitchers (Bill has the split at around 64/36, or maybe 62/38, while I have it at 57/43), and the other is that the spread in fielding numbers is wider in WAR than it is in Win Shares.

The net result in all this is that great fielding players, and great starting pitchers, (WAR loves both) will show a greater difference between the two systems.

Anyway, to compare, if I divide all the Win Shares numbers by 3, we get:

Concepcion

88W - 81L (Win Shares)

89W - 65L (The Indis)

Garvey

97W - 66L (Win Shares)

83W - 69L (The Indis)

The Indis gives a slight edge to Concepcion, while Win Shares gives a stronger edge to Garvey.  From a high-level view, we've got broad agreement.  It's hard to construct a system that won't give you something close to that.  This is why it is very odd for those who have never tried to create a system to decry the results.  All I can say is: try to do it yourself.  And, you will almost surely find that whatever I've done, or Bill has done, or others, that you'll come to a broad agreement.

The arguments will just end up being on the periphery.

Thursday, November 07, 2013

szWL

Bill James a few years ago introduced a nice toy, that simply converts a pitcher's K/BB numbers into a W/L record.  It's easy enough to explain with an example: Cliff Lee had 222 strikeouts, which is 1.2% of all strikeouts in the NL.  The NL had 1211 wins, so, we give Cliff Lee 1.2% of those wins, or around 15 wins.  He also had only 32 walks, which is 0.45% of all walks.  Giving him that rate of the NL's 1219 losses gives Lee 6 losses.  Lee's szWL record is therefore 15-6.

On the flip side we have perpetural replacement-level pitcher Jason Marquis, whose 72K and 68BB represent a 5-12 record.

There are ways to improve it slightly (subtract IBB, include HBP), and with more complexity (see my link).  But otherwise, the beauty of it is in its simplicity of conveying the information in an easy-to-understand representation. 

And you can try it for hitters as well.  Bill James creates many such toys, and this is one that deserves a bit more mention.

Monday, November 04, 2013

wOBA24

Great job by Matt in presenting the wOBA24 values for the walk, and applying it specifically for Joey Votto's walks, to show that his walks are just as valuable as anyone else's.  And that his overall production using the run expectancy chart is just as well as we'd expect without using them.

Tuesday, October 29, 2013

He had me, until he said “batting average”

Good work, good thinking, good effort... and then to scale to batting average?  Forget the common fan.  Forget trying to appeal to the broad group.  You are leading people somewhere.  Lead them on your terms. 

The whole idea to scale to batting average is absurd.  By scaling to OBP, you maintain identical denomiators.  And that means the weighted sum of your numerator will be identical to the unweighted sum.  (Though, OBP would be defined with the IBB removed, and ROE added in, etc.)?

Please, stop citing batting average.

Anyway, interesting that he didn't cite RE24 on Fangraphs, since his metric looks like it's halfway between RE24 and wOBA, in terms of "granularity".

Tuesday, October 22, 2013

MLB Rule Changes

Thirty years ago, Bill James proposed a solution to the scourge of the IBB. 

It should be remembered that the reason for a strike zone is to make sure the pitcher gives the batter something to hit, and the batter swings at something that is hittable.  Taking a strike is a penalty to the batter, and getting a called ball is a penalty to the pitcher.  It's a pretty simple concept.  That is, until managers determined that in certain situations, it is more advantageous to the defense to walk the batter than to give him a hittable pitch.  And the reason that the managers aren't bothered by giving out the occassional IBB is that the ultimate penalty for 4 called balls is one base, and unless the runner is on first, without moving any of the runners over.

Anyway, so Bill proposed a simple solution: if a batter gets four balls (or is hit by a pitch), he has the additional option to reset the ball count.  If he once again gets four balls (or is hit by a pitch), it's now a "double-walk".  Bill originally proposed it as the batter goes to second base and everyone moves two bases.  I don't think we have to go that far.  I asked Bill, and he agreed that just moving the runners over enough to allow the batter to get to second base is sufficient.

We can even restrict it further, and limit it only to situations where there is a runner on base, with 1B open, and only if a batter never got a strike.

In any case, instead of us just sitting here and theorizing what would happen, we simply do what the NHL and other sports do, and what all of corporate america does: test it.  MLB has an extensive minor league network.  Why not change the rules in one of the leagues?  After all, for what purpose would a minor league manager issue an IBB?  Would anyone in MLB like to see an opportunity to pitch to a quality batter be removed?  Would anyone in MLB like to see an opportunity for a quality batter to be lost?  No, of course not.  There's no reason to have an IBB to begin with in the lower levels. 

So, do away with it, and let's see what happens.  And as you study what happens, you adapt as you need to, to respond to the unintended consequences.  You test, you refine, you study, you adapt, all the things necessary to ensure you get it right.  But at the very least, you have to do it.

All we need now is a speech from Bluto, and we can get this thing done.?

Thursday, September 26, 2013

Blast From the Past: Proof of the Modified OPS

I ran across this old thread.?

Thursday, September 19, 2013

Off + Def = WAA

Fangraphs has introduced the Off and Def as rolledup values.  Previously, you'd  have to add the hitting and baserunning portions to get Off, as well as the fielding (aka UZR) and positional value to get defense.  Now, it's nicely rolled up in the leaderboards.

A good comment in the comment section: we don't need a qualifier for playing time, as this may knock out guys who were great (or terrible) in limited time. WAA is a combination of a rate stat and playing time stat, essentially: PA * (rate performance - league average rate).

***

And note that WAA + Repl Level = WAR.  Therefore, this presentation of Off and Def definitely gets my thumbs up (to whomever it is that that's important anyway).

?

wOBA24: wOBA equation by base-out states

A few years back, I had a pretty technical article on wOBA, LI, WPA, and WPA/LI (i.e., situational wins).  Buried inside that article was the 24 wOBA equations, one for each of the base-out states.  You can find it here, as Table 8:

http://tangotiger.net/wpali.html?

Matt has independently created the same thing, but seeing some of his results, he may have misfired somwhere.  In any case, I think wOBA24 is a great thing to track.  First, it's very similar to RE24.  But, the one thing it does differently is that rather than rely on the actual ending state as RE24 does (that metric will give credit to the batter, for anything the runner does), wOBA24 will give each batting event a standard value based on the base-out state.

wOBA24 will for example give a HUGE penalty for K with a man on 3B and less than 2 outs.  A standard wOBA (or standard Linear Weights) will simply assume that the K occurred in "typical" situations, even though: (a) we know it didn't and (b) we know both the pitcher and batter are actively approaching that plate appearance to increase/reduce the chance of a K (depending on the perspective of the pitcher/batter).

Whether wOBA24 in year T is more predictive of wOBA24 in year T+1 than wOBA in year T is, that's for you guys to figure out.  Someone sent me research that suggests there may be a slight advantage to wOBA24.

Friday, September 13, 2013

Your personal all-encompassing system

What the WAR framework does is create, well, a framework to evaluate a player.  It also denominates the evaluation in a common and easy-to-understand currency of runs.  It's particularly focused on actual performance, the context of that performance, and translating that performance into runs.  You can even include things like drive, heart, and other intangibles (that don't directly lead to the outcomes that are already considered).

If you don't like the results?, you might want to adjust the framework, or the context... but make sure to do this for ALL the players.  You can't pick and choose results, then alter results on a per-player basis.

Anyway, so WAR is solid, and it puts it out there.  The point is that YOU can do the same thing, create your own system to put it out there in whatever form you want.  Bleacher Report did that, except they use "points" and they judge the players on 5 categories with whatever subjective elements they want, and they weight each of these five categories in whatever way they want.

And I challenge everyone out there to come up with something better than WAR.  Go ahead and do it.  I don't care if it's a completely subjective system like Bleacher Report.  I want to see you put it out there before you criticize what is out there.  Because most of the criticisms of WAR are just gut-level reactions to results that don't pass their own sniff test.  But these very same people, if they constructed their own all-encompassing system would not even be able to pass their own sniff test.

Finally, even if you want to throw your hands up that NO ONE can create an all-encompassing system, everyone in fact has an all-encompassing list in their head, be it in terms of salary values for players or trade value for players.  You've already decided who you would, and would not, trade Andrelton Simmons for.  You in fact do have an all-encompassing system.  You just aren't admitting to it, nor can you confirm its own internal consistency.

Until someone wants to do better, WAR is the best framework available.  And the Fangraphs (fWAR) and Baseball-Reference (rWAR... not bWAR... it's rWAR) implementations of WAR are the best available to reference.

Tuesday, September 10, 2013

WAR misconclusion

In this thread, there was this comment:

Then WAR isn't really measuring value over replacement level. It's an indispensible prerequisite to the concept of WAR that the (anticipated) production of the replacement level player in all three phases be identifiable and identified in a way that makes sense.

That is a false conclusion.  His second sentence is correct, and actually, very well articulated.  The key point is this:

identifiable and identified in a way that makes sense

And this is done by comparing the performance against the league AVERAGE player.  This is the key part of the WAR framework as I've shown it many times.  Everything is compared to the average, every little component.  Heck, you can even break out the HR and BB and whatever, and compare it against the league average.  Everything.

Then, at the end, you compare the total of all these components relative to average, and apply an overall average-to-replacement level adjustment.  It's at the PLAYER level, not the component level.

And therefore the conclusion is that WAR *is* measuring against the replacement level... it's the replacement level player.

?

Friday, September 06, 2013

Reader Mail of the Day: how do the wOBA values change by run environment?

That was basically the question.  Here is my response:

You can see the annual wOBA weights here:

http://www.fangraphs.com/guts.aspx

Sort by R/PA (runs per PA) to get a proxy for run environment.  You will see that the numbers do change, but in a very very gradual way.

As R/PA approaches infinity (or a very very large number), all the events coalesce toward "1".  That's because if everyone is getting on base, it doesn't really matter how you get on base, because you will end up scoring.

So, the run value of the HR actually will equal the run value of the walk, in an extremely high run scoring league.  And it's in a very low run scoring league that the *relative* value of the HR will be much much higher than that of a walk.

This is why I'm ok with going with the Standard wOBA equation.  Having a complex version like Fangraphs uses may be good for 1% of the people, but really, the standard one gets you pretty far and makes it far easier for people to rally around.  Remember when we introduced wOBA in The Book, we really didn't position it as something for the mainstream to use.  And even when David wanted to put wOBA on Fangraphs, I didn't think twice about trying to push a standard version.  But seeing how wOBA has become popular, it only makes sense that a standard wOBA equation be the thing that gets pushed.

Think about FIP.  The coefficients of -2/3/13 only really applies for a specific run environment.  It won't apply for Mariano Rivera.  At all!  But, there's no way FIP becomes popular unless I push the standard version out there.?

Thursday, September 05, 2013

RE24/LI24

Colin looks at RE24 and boLI (which I now think is better named as LI24, for consistency's sake).  He focuses on Robinson Cano in 2009.  You can check Cano's data on BR.com:

http://www.baseball-reference.com/players/c/canoro01-bat.shtml#batting_win_probability::none

His standard linear weights for 2009 was +18 runs (that's BtRuns column).  His RE24 was -8 runs.

His Leverage Index by base-out state (boLI) is 0.98, which looks like he's faced average base-out opportunities.  But, then we go to RE24/boLI, and we see +7 runs.

So, we see based on his RE24 that he was terrible with men on base, and with men on base, your swings are double that than with bases empty.  By "deleveraging" those results, we are ensuring that each plate appearances counts as "1", and Cano ends up at +7.  So, RE24/LI24 shows his "situational" performance by base-out state, but ensuring that each PA counts equally.

If you can imagine what 24 different wOBA equations looks like (one for each base-out state), that's what RE24/LI24 is. ?

Thursday, August 29, 2013

RE24 is the same thing as RA9 (for starting pitchers)

One of the side-effects of information explosion is that those who consume it far outstrips the ability of those to explain it.  Such is the case with RE24.

Let's think about what RE24 does, if you look at it from the inning perspective: you start the inning with a run expectancy of around .48 or .51 runs (or whatever the league average happens to be).  If the inning is a shutout, the offense has earned -.48 or -.51 runs, while the defense has earned +.48 or +.51 runs.  And this is true whether the inning ends with the bases loaded or it was a 1-2-3 inning.  The intermediate parts may change within the inning, but the sum of all those parts will be IDENTICAL for all shutout innings.

(Well, almost identical, depending on how park factors are handled.  The league runs scored per inning may be .48 runs, but in some park it could be .53 and in another park it could be .44.  So, not all shutout innings earn the same across the league, but all shutout innings do earn the same for that park.)

Anyway, so all that RE24 is measuring, if you look at an individual inning, is the number of runs generated, relative to average.

If you take a pitcher's RA9, and compare it to the (park-adjusted) league average, you have... number of runs relative to league average.

These two metric will give you the identical "runs relative to average" only if a starting pitcher is never pulled mid-inning.  Otherwise, it will be SLIGHTLY off.  But in any case, RE24 is the better measure than RA9, since RA9 clumsily handles the assignment of runs for runners bequeathed to other pitchers.  RE24 is perfect in this regard (among many other regards).

It's why we learn very little with RE24 compared to RA9 for starting pitchers: the number of mid-inning pulls pales in comparison to the total number of completed innings.

But for relievers?  This is where RE24 shines.  Shines!  It perfectly handles the mid-inning entrances and mid-inning exits.  We don't need to rely on RA9 for relievers.  That's where RE24 should really be used.

If it helps, we can recast RE24 into an RA9 scale (i.e., similar to ERA) as follows.  Say the league average is .48 runs per inning.  Say you have a pitcher that has an RE24 of +40 runs and has pitched 200 innings.  That means the league average is .48 x 200 = 96 runs, and our pitcher here is 40 runs better than that, or 56 runs allowed.  So, his (RE24-based) RA9 is simply 56/200*9 = 2.52.

Easy right?

Mariano Rivera for his career has 1268.2 IP, an RE24 of +356 runs, and we can surmise a park-and-league RA9 environment of around 4.90.  So, 356 runs better than average in 1268.2 IP is .28 runs better than average per IP or 2.50 runs better than average per 9IP.  Since the average is 4.90, that sets his RE24-based RA9 at 2.40.  His actual RA9 is 2.39.

We don't really learn much with someone like Mariano using RE24, since he doesn't have many incomplete innings.  But, I'm sure you guys can find some relievers this year that had many mid-inning entrances and mid-inning exits, and I'd be very interested to see how their actual RA9 compares to their RE24-based RA9.

Anyone up for the challenge?

Wednesday, August 28, 2013

If the park adjustment is what puts Trout ahead, how much faith can we have that it’s real?

?Poz does a bang up job of educating the masses.  The park adjustment is some 13 or 14 runs, which is a gigantic adjustment, that swings the pendulum from Cabrera having the better season to Trout having the better season.

But, I don't like these one-size-fits-all park adjustments.  While I agree that something is better than nothing, I still think we can do more and better.  In no way do I believe that a typical LHH is representative of how Bonds was affected by 3Com.  Those LHH hits two-thirds as many HR at 3Com than away from 3Com, while Bonds hits as many at 3Com as away.  And, we can understand the reason: dude is a monster.  3Com just doesn't affect him the same way.

And if you go through other extreme hitters, we can understand why these standard park adjustments may not illuminate as much as we'd like.  And if you have to hang your hat on Trout being ahead of Cabrera on the idea that we understand how the parks affected them, that's a tough thing to have faith in.

On the other hand, the other idea is that we don't care how it affected THEM, but rather how everyone else is affected, as that describes the run environment they played in. So, it's not so much how much Trout/Cabrera stats were inflated or deflated as the case may be, but rather how ALL players were affected, setting a new run environment, and therefore, a park specific runs-to-win conversion.

Anyway, lots of consideration, and I can understand Poz's friend at the end.  For most people, this is too much to think about, and it's not fun.

Me? I LOVE this stuff, and I can never think enough about this.

Uncertainty in offensive metrics

Colin takes a stab at explaining the uncertainty in offensive metrics.

The "uncertainty" that Colin is reporting is the difference between the standard linear weights run values (the 1.4 for HR, the .3 for walk, etc), and the run values based on the 24 base-out matrix (the example he gave for the bases loaded strikeout, etc).

Fangraphs tracks both values (see wRAA and RE24, though I'm not sure if there are park adjustments to contend with), as does Baseball Reference.

So, what Colin is saying is that there's a 90% chance that Trout's RE24 in 2012 is going to be higher than Cabrera's RE24, if we rely on their standard linear weights values as being 12.5 runs apart.?

Sunday, August 25, 2013

oWAR + dWAR does not equal WAR

?The correct way to handle WAR within the WAR framework is that each component is compared to the average, and then you have a replacement level component (tied to playing time) at the PLAYER level.  This is the right way to do it.  Not "a" way, but the right way (if you subscribe to the WAR framework).  Fangraphs presents it as such at the bottom of each of their player pages, as does Rally on his site.

If you want to start to merge components into "offense" and "defense", you still need a THIRD component (the replacement level or playing time).  So, it's offense relative to average, plus defense relative to average, plus the replacement level (average relative to replacement level).  Andrelton Simmons, using rWAR, breaks down as follows for his career:

-10 runs offense relative to average

+65 runs defense relative to average (which includes his positional adjustment)

+22 runs average over replacement

===

+77 runs above replacement

In the BR.com presentation, it references "oWAR", which you might think is "offense above replacement level", when the reality is that THERE IS NO SUCH THING as replacement level offense.  It's simply something that we must get away from talking, referencing, or otherwise whatever.  It's anathema to the WAR framework.  What "oWAR" should be is the "offensive component of the WAR framework".

But like I said, at its barest, the WAR framework has THREE components.

***

Now, if we must and insist that we only have two components (offense and defense), you then have to decide how to distribute the third component, the playing time one.  There is zero reason to simply distribute it all into the offense component. 

***

All of the above is all about the WAR framework, and it's not up for discussion.  What is an opinion is what is about to follow.  If we insist on distributing the playing time portion onto the offense and defense, how to do it.

So, one idea I have is that since the spread in offense is about 1.7 times the spread in defense, then we'd split the playing time component accordingly.  So in Smmons' case, the +22 runs will get spread at +14 for offense and +8 for defense.

Therefore, oWAR would be +4 runs and his dWAR would be +73, and that gives him his WAR of +77.

Friday, August 23, 2013

Explain FRA to me

I have nothing good to say about Fair RA.  Momma told me I should stop there, but, who listens to their mother really?

***

While the idea behind it seems reasonable, the results are frankly ludicrous (see the whole thread, but especially the comments starting at comment 21 if you are short on time).? 

So in the end, I can't explain the methodology nor the results.  And yet FRA is the central component to WARP.  Outside of Colin, I doubt there's a single person at BPro that can explain FRA that can be understood by any one of us. (That's a challenge.  Crickets will be here until it's taken on.)  But it is being referenced by its authors, most visibly in their book (where they highlight the metric by showing that Doc Halladay wasn't as good a pitcher as the rest of us think). 

The thing is that FRA has been around for three years now (Happy belated Birthday!), and I haven't seen any effort at all in moving the understandinging of the metric forward.  It simply exists, and it is simply being referenced.  I think outside of the original article, no article has ever been devoted to it since.  The sellers can't sell the metric, and the buyers aren't buying the metric.  So, why does it even exist?

SIERA in contrast made a splash at BPro when it came out, and Matt had several articles that delved into SIERA and what made it tick.  And almost as quickly, it was gone, in a takedown by Colin himself.  (SIERA survives at Fangraphs.) 

The last time I took out the daggers was to eviscerate the PECOTA percentiles.  They were wrong from the start, and they were wrong for several years.  And yet, there they stood, existing without explanation.  Until Colin proved that they were wrong.

When will the FRA obituary be finally written?   Until then, someone, anyone, explain FRA to me, in a way that I can explain it to someone else.

Wednesday, August 21, 2013

Reworking WARP

WARP is the BPro version of WAR, which in my nomenclature would be called pWAR. Colin says he's going to rework it. It'll be interesting to see how different it'll be from the framework I developed, or from the implementations (rWAR, fWAR) of that framework.

Colin is right that the place where it's going to get interesting is the pitcher-fielder interactions.  Currently, WARP is based on FRA.  FRA might be rooted in logic, but there's an obvious problem, given the results we see of Felix, Doc, and Maddux.  That makes FRA something that I discard (see this thread from last year, and if you are pressed for time, start at comment 21).  Anyway, I presume we will get drastic new results for these three pitchers among others, and they will fall in-line somewhere around fWAR and rWAR.

Thursday, August 01, 2013

Standard wOBA

OBP's formula doesn't change, regardless as to the relative value of the HR to the ngle, given the circumstances.

SLG's formula doesn't change, regardless as to the relative value of the HR to the single, given the circumstances.

The core of FIP doesn't change, even though it should.  (It does have a fudge factor though.)

It seems to me that what helps these metrics is that they have a standard formula, and then analysts can tweak them to suit their purposes.

wOBA should be like that, have a standard formula.  This is what I use:

Version 1 - Basic

numerator:

0.7: UBB+HB

0.9: 1B+ROE

1.3: 2B+3B

2.0: HR

denominator: ?PA - IBB - SH

 

Version 2 - Speed

numerator:

0.7: UBB+HB

0.9: 1B+ROE

1.25: 2B

1.60: 3B

2.0: HR

0.25: SB

-0.5: CS

denominator: ?PA - IBB - SH

If you don't have ROE handy, then ignore it.  If you don't have walks split between UBB and IBB, then ignore the distinction.  If you don't have SH, ignore that too.  Same deal with steals.

So, that's what I'd like to see for non-Fangraphs sites.  It's nice and simple.

The key point by the way is that you start with "1" for any positive event (i.e., the numerator of OBP), but adjust it up and down relative to the event.  The overall average of these events will come out to "1", which is why it is called a "weighted" on base average.  Both metrics (unofficial OBP and wOBA) share the same denominator.