Tangotiger Blog

MGL is pointing out on Twitter, correctly, that it’s wOBA that matters, not some combination of BA and SLG.

And he is correct.  Indeed, the metric was developed with wOBA.  When it came time to setting boundaries, I could have selected anything.  I could have said “at least .500 wOBA”, or “at least .800 wOBA”, or “at least .950 wOBA”.  Or anything really!

So, how did we end up where we did?  Well, one of the requirements is that we wanted something simple.  That instantly meant that we had to keep it in one zone.  I couldn’t use a .800 wOBA as my minimum, even though .800 wOBA is a tremendous outcome.

Therefore, we kept increasing the boundary until one zone started taking shape.  Once it became clear where this metric was heading, it became in issue at the margins.  Should we start at 99mph or 98mph or 100mph? 

I looked at HR rates, since now this metric was starting to become a proxy for HR or a proxy for hard-hit Extrabase Hits.

As we were playing with the boundaries, we saw that the leaderboard for Barrels had the same count totals as the leaderboard for Extrabase Hits.

Once all that was materializing, we then tried to figure out how to explain this.  While wOBA is the actually guts of it, if we can explain it without saying wOBA, that helps everyone.

And as it turned out, the boundaries were coalescing around a .500 batting average and 1.500 slugging.

So, if someone says, “what’s a barrel”, I can say “wOBA of at least .950”.  But I can ALSO say “BA of at least .500, SLG of at least 1.500”.

***

As a side note, wOBA is proportionate to 60% batting average, and 40% SLG.

The actual equation is this:

where (launch_speed * 1.5 - launch_angle) >= 117
and (launch_speed + launch_angle) >= 124
and launch_angle <= 50
and launch_speed >= 98

As you can see, there is no mention of batting average or SLG, or even wOBA.  It’s simply drawing lines.

And that region represents that ABSOLUTE MAXIMUM WOBA for the league, if we require a certain total number of Barrels.

That is, this is the question: “We want to have X number of barrels.  Construct me boundary lines that maximizes wOBA.”

The above equation is the result.

OK, but I’m still not happy with BA and SLG criteria. All it does is needlessly promote BA which we should be killing unless there is a compelling reason to do so. As I said on Twitter, you are dumbing down a metric developed from the most technologically advanced data in sports. I find that unpalatable. Take a risk. Explain and promote wOBA! Someone has to do it. You don’t see Apple dumbing down their phones because the average consumer doesn’t really know how to use them.

One thing that I haven’t done is set a lower boundary.  That becomes a bit trickier if only because the sample size at 115-120 mph, at 0-10 degrees is so low.

As it stands, a 120mph will be considered a “barrel” at a range of 4 to 50 degrees launch, where 50 is hardcoded as its limit.  Maybe instead of 4 the limit should be 6 or 7 or something.  I don’t know yet.  For now, there’s no lower limit. Once someone hits a ball at 124mph at 0 degrees launch, I’ll rethink that right away. 

Otherwise, I have this one the backburner and will revisit it before the 2017 season.

But BA and SLG are NOT criteria!  The criteria is what I have in post #2.  And that’s based on asking the very question you want answer: the zone where wOBA is maximized.

So, your issue is in its presentation.  And I think that once things start getting developed, there will be a natural progression toward wOBA.  wOBA is inevitable because drawing a BA chart AND a SLG chart AND a SLG-BA (ISO) chart is dancing around the whole issue.  wOBA will be it.  Just will take time.

Also don’t forget these charts:

http://tangotiger.com/index.php/site/article/statcast-lab-quality-of-contact-description

They are all based on wOBA.  wOBA is the key, no question.

The other concern is more interesting, and that is: what denominator to use?  You can use “contacted PA”, but you can also use “swings”.

Even the numerator is in question: should we use foul balls?

After all, this metric is agnostic as to the spray angle, and so, a ball that is launched at +35 degrees that slices in one park at +44 degrees, but slices in another park with the RF pole a bit farther out at +46 degrees SHOULD be treated the same way.

Right now, they are not.

So, I think those kinds of issues should be the real talking points.

Off the top of my head, foul balls should be ignored. I understand that it may be part of the “talent” component (e.g. a hard hit ball just foul has some relevance) but we start with performance (not talent).

As far as the metric being based on wOBA, if I’m understanding correctly, you are basing it on a wOBA minimum and then claiming it is based on a minimum of .500/1.500?

Which is it? It can’t be both. There has to be at least one ball at the margins that would qualify under one criteria and not on the other.

If you’re going to tell me that it’s actually based on a wOBA minimum but you tell the public that it’s based on a .500/1.500 so that they can understand it better, that’s incredibly disingenuous is not out and out lying. But I’ll reserve that criticism until you clear this up.

Yes, I understand that it almost all cases it’s probably the same and that, like the quality start, it’s just a minimum so it doesn’t make much of a difference either way, but as I said, it’s either one or the other, not both.

OK now I get it. I hadn’t refreshed the site and I didn’t read your intermediate posts.

This, however, on the MLB web site:

“Created by Tom Tango, the Barreled Ball classification is assigned to batted-ball events whose comparable hit types (in terms of exit velocity and launch angle) have historically led to a minimum .500 batting average and 1.500 slugging percentage.”

clearly suggests that it is a fixed minimum criteria. It should have read “around” a .500/1.500.

Not a big deal, but I am a stickler for accuracy of words.

I think the way Mike worded it is perfect:

A “barrel” is defined as a well-struck ball where the combination of exit velocity and launch angle generally leads to a minimum .500 batting average and 1.500 slugging percentage

You have to remember that technically, that EVERY SINGLE COMBINATION of velocity and angle is unique.  For example, you’ll have one ball at 101.3487 mph at launch angle 27.8734.  No two are identical.

So, now we get into rounding, binning, and smoothing.  If we ask “what is the historical wOBA at 100mph, 28 degrees”, well, I can say that EXACTLY at 100.0000 and 28.0000 it is undefined.  We have to expand that and the more we expand it, the less weight we place at the ever increasing circle.

Therefore, 100/28 would be comprised of say 98/26 through 102/30 where the closer the batted ball is to 100/28, the more weight we give it.  (Think Marcel.)

After I do all that, I get wOBA values for each integer pair.  But even so, I then use those values as a GUIDE to specify my boundary lines. If my “line” breaks because I have values of .955, .945, .960, I’m not going to change my line.

Hence, Mike’s description above is perfectly accurate.

The word “historically” will be changed to the word “generally” in the glossary.  Thanks for pointing it out!

Yes I like generally!

This feels like the concluding scene to a 1950s family sitcom:

Husband - I really should have said we “generally” have meat loaf on Thursday, not “always”.

Stay at home wife - Yes I like generally!

Cue the laugh track and fade out.

Really love the way you’re discussing these metrics on the blog.

Tango and MGL,

Do you think its possible for hitters in the future to use this kind of data and manipulate their swings so that the launch angle of their hits come at the ideal range based on stats. Lets say 25-35 degrees is a .700 wOBA. Would it be possible that a hitter sees that and could change their swing to take advantage of that?

I don’t think there’s an easy way to improve offensive (or pitching) performance but I do think that players and coaches will be able to use Statcast in order to help them to optimize their individual approaches.

The ideal launch angle for batting average is 12 degrees.  That’s what you would call “line drive”.

The ideal launch angle for slugging is 28 degrees.  That’s what you would call “fly ball”.

The ideal launch angle for wOBA is close to either 12 degrees or 28 degrees, but NOT 20 degrees (i.e., fliners).

So, that’s the real challenge for a hitter, to decide if they want their midpoint launch angle to be 12 or 28.

Now, you can split the difference, and make the ideal launch angle at 20 degrees, on the idea that when you “mis-angle”, you might mis-angle toward 12 or mis-angle toward 28.  Indeed, I would think that should be the ideal launch angle for a hitter.

Of course, you have to do this on a case-by-case basis. The reason that 28 degrees launch angle is ideal is only true because we have plenty of 95+ mph hit at that angle.  If say Billy Hamilton can’t have the needed launch speed to take advantage of 28 degrees, then he definitely should not try to hit at 28 degrees.

Therefore, what’s going to happen is that for every hitter, you need to know their “maximum launch speed”, where “maximum” may need to be established based on his 20% or 30% hardest hit balls, something like that.

Once you have that, then you can establish each hitter’s ideal launch angle.

Something like that.  So, yeah, absolutely, Statcast will end up telling us which hitters are launching ideally, and which should consider changing their approach.

My guess is that 4 out of 5 hitters wouldn’t need more than a 1 or 2 degree angle adjustment.  The fun is in finding that 1 in 5 hitter that might need a drastic change.

17/Tango -

Those are the averages of all batters, who possess a variety of mean batted balls speeds. The optimum vertical angle for any particular batter will depend on his distribution of BBS. Hitting even an average number of balls into 20-28 won’t make sense if all those balls are 95 and below - they’ll just be fly outs (look at Josh Harrison).

The guys who have below average BBS need to hit the ball lower to get it in the white zone on your chart.

It’s been my belief that young sluggers who lose bat speed and thus BBS as they age increase their VLA to maintain their HR%, but that crashes their BABIP (see Andruw Jones et al)

Brian: you didn’t read all my post:

Of course, you have to do this on a case-by-case basis. The reason that 28 degrees launch angle is ideal is only true because we have plenty of 95+ mph hit at that angle.  If say Billy Hamilton can’t have the needed launch speed to take advantage of 28 degrees, then he definitely should not try to hit at 28 degrees.

19 - whoops

I do want to use the data to see what the optimal VLAs are for each mean BBS, given the distributions of each. I’ve been thinking along the lines of what you say in the 2nd half of your post #17 which somehow my eyes glossed over.

By the way we are including foul balls… if it ends the PA. 

So, I am getting pretty uncomfortable saying that we’ll exclude a foul ball… unless it is caught.  To me, that’s starting to not make any sense.

Furthermore, if the proper denominator is “swings”, again, it’ll seem pretty silly to exclude a swing where the ball was an uncaught foul.  The whole thing is supposed to be outcome-agnostic.

Therefore, I strongly anticipate that any swing that satisfies the exit speed + launch angle criteria gets counted, regardless of outcome.

I’ll include a spray angle of say -50 to +50 degrees or whatever the range is for HR.  Remember, the spray angle is off the bat, not from 0,0 (tip of home plate).

This is what aspiring saberists should be doing:

http://www.ussmariner.com/2016/09/24/yet-another-update-on-safeco-field-home-run-haven/

So, do you envision a point at which barrels allowed or something like the concept of barrels becomes a part of FIP (similar to how Fangraphs includes IFFs in FIP)?

Is an .800 BA on barrels defense independent enough?

I wouldn’t say it would be “part” of FIP, so much that it would be FIP 2.0.

The question to ask yourself is this though: if barrels has ZERO predictability, do we want to bother to count barrels for FIP?  Answer is: no, definitely not count it.

Now, if barrels has a SLIGHT amount of predictability, do we want to count it for FIP?  And herein lies the fight between regression and not.

Then, beyond Barrels, we have Texas Leaguers, Almost-Barrels, and so on.

And maybe eschew categories and go straight to “estimated wOBAcon” for each speed/angle combination.  But if that only has tiny predictability, do we want to bother with it?

This is the fight, between retrospective and predictive.

Don’t forget, I do have FutureFIP:
http://www.insidethebook.com/ee/index.php/site/comments/tangos_lab_futurefip/

So, we may end up with:
FIP Classic
FIP (using estimated wOBAcon)
FutureFIP (using regressed wOBAcon, BB, SO)

Update:

https://twitter.com/tangotiger/status/781566349881540608

Please suggest better names.

Going back to this:

By the way we are including foul balls… if it ends the PA.

So, I am getting pretty uncomfortable saying that we’ll exclude a foul ball… unless it is caught.  To me, that’s starting to not make any sense.

Furthermore, if the proper denominator is “swings”, again, it’ll seem pretty silly to exclude a swing where the ball was an uncaught foul.  The whole thing is supposed to be outcome-agnostic.

Therefore, I strongly anticipate that any swing that satisfies the exit speed + launch angle criteria gets counted, regardless of outcome.

Every Barrel and Near-Barrel, that landed fair, was hit within +/- 37 degrees of spray.  For the two balls that were barrelled at +37 and -37 degrees, there was another 150 of them that were foul balls.

If our focus is on performance, but also being fielder- and park-agnostic, then we have some decisions to make.

Remember that the threshold is to have a wOBA of .900+ or so, which means around +.50 runs per PA.

If you crush a ball that is JUST foul, that’s an extra strike (or a “let” in the case of a 2-strike foul).  So, let’s say that such swings are worth -.05 runs.  But if it’s JUST fair, that’s going to be worth as much as +1.4 runs (HR).

In the above case at +37 and -37 degrees, it’s a fair ball under 2% of the time.  That means the run value of such a swing (say 100mph, 28 degrees launch, -37 degrees spray), a ball that hooks just foul over the fence:
+1.4 x .02
-.05 x .98

That’s -0.025 runs.  That’s a NEGATIVE.  But, let’s just call it neutral.

If we included spray angle, then we’d have to conclude that such a ball would not be “barrelled”.

In order to be barrelled, it would have to have a chance at being +.50 runs or better.

***

Since we are not including spray angle, my inclination is as follows:
- include foul balls in the calculation for wOBA, making this the first time the calculation for wOBA is being reconsidered

- recreate the boundary lines with this information

- include an additional parameter that a barrel must be within 38 (or 40?) degrees of spray

Thoughts?

Instead of wOBAcon, I’d probably call it: SwingWOBA.

We don’t care about the fielder or the park.  We just care about the swing. That’s what we are measuring here… swings.

In Post #2, that is the definition of the barrel.

This is .... well, I don’t have a good name for it… it’s the “perfect hit” (so feel free to suggest a good name), meaning it’s a subset of a barrel.

when (speed * 1.5 - angle) >= 129
and (speed + angle * 2) >= 156
and speed >= 106
and angle between 4 and 48

Whereas the minimum threshold for a barrel is 98mph, the minimum for the perfect hit is 106mph.

Someone on twitter suggested “bomb”.

All Barrels:
Estimated wOBA 1.428. 

Broken down:
Perfect Barrel: 1.939.
Rest: 1.360.

The maximum wOBA value is set at 2.000 (HR).

On a similar note to the discussion about including foul balls—why not replace the barrels/PA metric on Savant’s Statcast leaderboard with one that looks at all barrels (including fouls) per swing? Barrels/PA makes little sense to me, since hitters are getting penalized for BB and HBP. That substitution would leave two metrics telling two different stories: one showing performance ((barrels on BIP)/BIP) and the other one possibly providing a better view into barrel consistency and talent ((all barrels including fouls)/swings). As Tango noted in #7, a foul-inclusive metric is more aligned with the idea that barrels should be held independent from spray angle, anyway.

Right. 

On the other hand, we could say we want the spray angle to be within +/-40 degrees (and hooks/slices will take it to 35 to 55 degrees).

So, we can include fouls, but only those that “had a chance” at being fair.

It’s really a tough call.  However, the data will eventually be there for the user to decide how he wants to aggregate it.

Fouls are a real pain for us right now, if only because it breaks our rule of one event per PA.  So, it’s something we have to handle.  I mean, can Giancarlo Stanton be credited with THREE barrels on ONE PA?  Do you only count a maximum of one per PA?  Do you count all three as swings, so it was one barrel on three swings?

Lots of questions to consider…

If you’re restricting barrels to a one-per-PA basis, then I’d say fouls need to be left out because the focus should be on the actual BIP events. But if that one-per-PA rule is loosened, allowing swings to be used as a denominator as well, it helps us (and less saber-oriented fans) better understand what may have been bad luck. It opens the door to fun comparisons too (ex: Stanton’s foul barrels compared with teammate Dee Gordon’s BIP barrels).

Either way, it’s good to hear that the raw foul data will be made available.

I don’t use the word “raw” because that means different things to different people.

when (speed * 1.5 - angle) >= 129
and (speed + angle * 2) >= 156
and speed >= 106
and angle between 4 and 48
then 7—‘Perfect Barrel’

when (speed * 1.5 - angle) >= 117
and (speed + angle) >= 124
and speed >= 98
and angle between 4 and 50
then 6—‘Barrel’

when (speed * 1.5 - angle) >= 111
and (speed + angle) >= 119
and speed >= 95
and angle between 0 and 52
then 5—‘Near-Barrel’

It may not be readily apparent, but perfect barrels are a subset of barrels, which are a subset of near-barrels.

After that, flares-and-burners covers up everything else that is “Good”.

The code for the six categories.  This was based on 2016 data.  2017 is a bit different, so if someone wants to tweak these, by all means:

SET SpeedAngle_Code 
 = case 
  when (speed * 1.5 - angle) >= 117
  and (speed + angle) >= 124
  and speed >= 98
  and angle between 4 and 50
   then 6 -- 'Barrel'

  when (speed * 1.5 - angle) >= 111
  and (speed + angle) >= 119
  and speed >= 95
  and angle between 0 and 52
   then 5 -- 'Solid-Contact'
   
  when speed <= 59
   then 1 -- 'Poorly-Weak'
 
  when (speed * 2 - angle) >= 87
  and angle <= 41
  and (speed * 2 + angle) <= 175
  and (speed + angle * 1.3) >= 89
  and speed between 59 and 72
   then 4 -- 'Flare-or-Burner'
   
  when (speed + angle * 1.3) <= 112
  and (speed + angle * 1.55) >= 92
  and speed between 72 and 86
   then 4 -- 'Flare-or-Burner'
   
  when angle <= 20
  and (speed + angle * 2.4) >= 98
  and speed between 86 and 95
   then 4 -- 'Flare-or-Burner'
   
  when (speed - angle) >= 76
  and (speed + angle * 2.4) >= 98
  and speed >= 95
  and angle <= 30
   then 4 -- 'Flare-or-Burner'

  when (speed + angle * 2) >= 116
   then 3 -- 'Poorly-Under'

  when (speed + angle * 2) <= 116
   then 2 -- 'Poorly-Topped'

   else 0 -- 'Unclassified'
 end