Graph of the Day: Expected Run Values (beta)

by Harry Pavlidis on Jul 2, 2009 12:01 PM EDT Comment 16 comments

If it wasn't for bad luck, I wouldn't have no luck at all. But some of these guys have good luck. Using batted ball types (ground/fly/line/pop), I ignore the actual outcomes (single/double/triple/homer/out) and apply park adjusted expected outcomes. That gives me RVAA and rv100 expected, which I compare to actual to come up with "luck".

This is four graphs in one, click to enlarge. Numbers are in "runs" - positive numbers mean better than expected outcomes (or good luck).

0 recs | Comment 16 comments | Share

Story-email Email Printer Print

Comments

A cool next step with this

would be to use Hit FX to try to identify what is bad luck (i.e. who hit ‘em where they ain’t) and what is bad defense (i.e. balls hit where they are normally fielded, but the defense didn’t do their job).

Also, I think Haren’s “good luck” has been more than balanced out by the bad luck of his offensive support and bullpen “help.”

"It seems like we're not hitting because we're not getting hits." - Dusty being Dusty

by Slyde on Jul 2, 2009 12:14 PM EDT reply actions 0 recs

I'm working on that.

Peter Jensen’s approach is good, I’m taking a slightly different cut at it.

by Harry Pavlidis on Jul 2, 2009 2:39 PM EDT up reply actions 0 recs

Aggregation

Group by batted ball type, calculate the likelihood of each event.

by Harry Pavlidis on Jul 2, 2009 2:40 PM EDT up reply actions 0 recs

So they're based on run values of singles/doubles/etc?

Which year and league are those from?

by Graham on Jul 2, 2009 6:15 PM EDT up reply actions 0 recs

2007-2009

Consume this thread:
http://www.insidethebook.com/ee/index.php/site/comments/linear_weights_by_batted_ball_speed/#comments

by Harry Pavlidis on Jul 2, 2009 10:15 PM EDT up reply actions 0 recs

Could you explain the park adjusted expected outcomes?

or direct me somewhere.

by JBrew on Jul 2, 2009 1:28 PM EDT reply actions 0 recs

I'm using a "standard" approach

From the KJOK mailing list, this is the basics:

2007 St. Louis Cardinals/Busch Stadium Triples Factor calculation:
Cardinals 3Bs at Home = 7
Cardinals 3B’s Allowed at Home = 19
Cardinals 3Bs on Road = 6
Cardinals 3Bs Allowed on Road = 19
Cardinals ABs at Home = 2,704
Cardinals ABs against at Home = 2,866
Cardinal ABs on Road = 2,825
Cardinal AB’s against on Road = 2,729

Numerator = (7+19)/(2,704+2,866) = .004668
Denominator = (6 + 19)/(2,825+2,729) = .004401

Unadjusted Triples Factor = .004668/.004401 = 1.037
OPC = 16/(16-1+1.037) = 0.9977 (16 teams in National League)
Adjusted Triples Factor = 1.037 x -.9977 = 1.03

by Harry Pavlidis on Jul 2, 2009 2:41 PM EDT up reply actions 0 recs

Just to check a calc (with imaginary numbers)

100 GB
Expected Outcomes: 20-1B, 10-2B, 1-3B
Park Adjusted: 20-1B, 11-2B, 1-3B
Linear Weights (runs): 9.4-1B, 9.35-2B, 1.02-3B
Total = 19.77

Actual Outcomes: 16-1B, 14-2B, 0-3B
LW: 7.52-1B, 11.90-2B, 0-3B
Total = 19.42

“Luck” = 19.77-19.42 = 0.35

Then the same thing for the other batted balls. Is that close?

by JBrew on Jul 2, 2009 4:02 PM EDT up reply actions 0 recs

I think you've got it

it’s park adjusted expected outcomes * value of outcome

so, if a GB in Park x results in the following mutually exclusive events at these rates:
1B .070
2B .020
3B .005
HR .000
Out .905

So expected value = Val1b*0.70 + Val2b*.020 + Val3b*.005 + ValHR*.000 + ValOut*.905 (an out has a negative value, btw).

The Val1b etc are based on the count

by Harry Pavlidis on Jul 2, 2009 5:30 PM EDT up reply actions 0 recs

Close enough

Since I assume you are using your run values by count chart from your blog, what is the added benefit by accounting for the count? I know this was discussed some in the above thread at The Book blog, but I don’t fully grasp the need to use the count in determining the run value of a batted ball.

For instance, if I get ahead 0-2 and serve up a HR the value is 1.50, but if I serve it up on 3-0 the value drops to 1.18. I can see using it when determine individual pitch value but not when determining batted ball outcome value.

by JBrew on Jul 3, 2009 9:59 AM EDT up reply actions 0 recs

Batted ball outcomes are count, out and baserunner neutral.

The value of a hit is the same no matter the count, it’s just that the count has some run expectancy built into it. The count is not used to determine the value of a batted ball. It’s the value of the pitch I’m after, not the batted ball. That’s an input to the pitch value, not the other way around.

by Harry Pavlidis on Jul 3, 2009 11:20 AM EDT up reply actions 0 recs

But, you are right

The first chart shouldn’t go by count, it should use 0-0 (through) no matter what. The 2nd chart, however, mixes balls not put in play in the denominator. Should probably have keep that split out for purposes of this one…

by Harry Pavlidis on Jul 3, 2009 12:10 PM EDT up reply actions 0 recs