3-D Baseball

2011-09-29T23:18:55.293-07:00

This comment has been removed by a blog administrator.

Hi, I came across your site and wasn’t able to get...

2011-07-10T18:06:03.013-07:00

Hi, I came across your site and wasn’t able to get an email address to contact you about some broken links on your site. Please email me back and I would be happy to point them out to you.

Thanks!

Frank
frank641w@gmail.com

I made an Excel macro to make calculating Bayesian...

2011-05-20T13:14:28.926-07:00

I made an Excel macro to make calculating Bayesian W% estimates for a group of teams easier. You'll need an Excel spreadsheet with each team's W and L totals, and then paste the linked code into Excel's VBA environment. Read the comments embedded in the code to see what values you might have to alter.

https://docs.google.com/document/d/1xdtx4H_upEzAqkspRwWJl_OAmuqIq9to7sgpHn0Uld0/edit?hl=en_US

Indeed, excellent work! I believe the general for...

2011-04-22T16:02:53.066-07:00

Indeed, excellent work!

I believe the general form for calculating the posterior distribution from 2 standard distributions (when we're not necessarily using a binomial distribution for the variance) is:

Posterior = (A/stdevA^2 + B/stdevB^2)/(1/stdevA^2 + 1/stdevB^2)

Where A is the mean if distribution A and B is the mean of distribution B.

And the posterior standard deviation is then

PostStdev = 1/sqrt(1/stdevA^2 + 1/stdevB^2)

I derived all of this from the calculus when I was working up my various Bayesian NBA rating systems.

Nicely done!

2011-04-22T14:47:03.777-07:00

Nicely done!

Thanks for doing analysis on our projections (Roto...

2011-02-13T10:26:20.893-07:00

Thanks for doing analysis on our projections (RotoChamp). We didn't provide FanGraphs with IBB or wOBA data, so the wOBA projections on their site might not be accurate. I'm not sure how they are calculating wOBA for us.

I would throw out any comparison for guys like Dan Johnson who have low reliability scores in Marcel.

This is our first year doing projections and we may or may not be regressing for age enough.

Magglio is an interesting case. His xBA (by our calcs) was .335 last year versus a .303 actual BA. Since we ignore BA in our projections and use xBA, we will be more optimistic for him than other systems.

I haven't looked at everybody, but I know Bautista and Dan Johnson have both been 'unlucky', and that is why they will be seen more optimistically in RotoChamp than other systems.

Marcel will probably beat our system if you look at cumulative stats over the entire pool, but our system should be stronger in providing relative value for fantasy purposes. This is because our projections are optimistic across the board in terms of projected ABs and IPs.

Thanks,
Mike - RotoChamp

Bo Hart is an interesting case, so I've used y...

2010-11-22T22:51:35.912-07:00

Bo Hart is an interesting case, so I've used your blog post as a jumping off point for one of my own.
http://bayesball.blogspot.com/2010/11/bo-knows-probability.html

For what it's worth, rWAR has Longoria as #1 i...

2010-11-18T21:09:42.892-07:00

For what it's worth, rWAR has Longoria as #1 in the AL.

2010-11-18T21:06:15.842-07:00

This comment has been removed by the author.

"Let's just say, we really don't want...

2010-09-25T14:45:10.060-07:00

"Let's just say, we really don't want to have to use the actual distributions, so we'd better hope that there is a simpler distribution that is a good approximation.

Lo and behold, there is."

The reason for this (as I'm sure you know) is the Central Limit Theorem. I don't know what we would do without it...

Really interesting. One thought, however: The actu...

2010-09-24T13:45:26.411-07:00

Really interesting. One thought, however: The actual distribution of errors on one given play may be narrower than assumed because of the way scorers work. Scorers technically don't call out the gain of the play. That is deduced from where the ball is spotted. Any one single spot itself may have a +/-.5 yd error, but all gains must reconcile to the total gain of a drive.

Take one example: a team that starts precisely on its own 20 (following a touchback) drives 80 yds for a TD. All the individual spotting errors must reconcile to exactly zero. Other drives, including ones that don't start or end at a precise yd line must reconcile with a total absolute error of less than 0.5.

It's true that a RB would only have only a subset of those carries, so errors won't necessarily reconcile to zero for him. But a star RB will get the vast majority of carries--perhaps about 40% of his team's plays. Those errors aren't independent of one another. A big + error on a previous play will likely convert into a - error on a later play in the same drive.

One other way to state this is that if a team has 2 plays in a row, and one goes for 4.5 yards but is scored as 4, and the next goes for 5.5 yds, it can't be scored as 5. It must be scored as a 6 yd gain because the ball is very clearly 10 yds further down field, not 11.

In other words, the errors are not independent, which (I think) is an unstated assumption of the analysis. Also, they are capped at +/-0.5 on any one drive. (Actually, they would be capped not just on any one drive, but on any exchange of drives between scores (or the expiration of a half.)

Here are the Marcel projections for the Derby grou...

2010-07-16T03:35:59.576-07:00

Here are the Marcel projections for the Derby group and the second control group to compare to their second-half figures:

http://www.insidethebook.com/ee/index.php/site/comments/hr_derby_hangover_effect_with_control_group/#1

One correction, the umpires didn't strike in 1...

2010-04-13T00:32:50.008-07:00

One correction, the umpires didn't strike in 1995, they were locked out.

Ron, umpires are supposed to keep order in a game they are working, not in a game where the league has locked them out, cut them off from their permanent source of income (yes, paychecks are an important part of someone's career-imagine that!), and replaced them with whatever non-Union workers were willing to do the job. There's a line where a worker is no longer bound by the code of his employment within the workplace, and this would be pretty far past that line. Keeping the order of a game as an umpire does not extend to require them to docilely accept whatever treatment their employer wants to hand down or to stand by meekly when their employer tells them they will no longer be working because they can find people who don't care about working conditions or pay or silly things like Unions who will replace them.

Thanks in part to the visibility of the umpires who were locked out making sure MLB couldn't just sweep the issue under the rug, they soon got their careers back. I tend to not think of the difference between working in your career and being out of employment with no idea when your family will get it's next paycheck as a little more than silly and petty, and I think causing a little ruckus at a baseball game in order to try to get out of that limbo is pretty understandable.

I've umpired for 30 years. But somehow, the id...

2010-04-06T01:46:58.153-07:00

I've umpired for 30 years. But somehow, the idea of umpires, who are supposed to keep order in a game, causing a fuss that gets them kicked out of a game, seems silly and petty.

It also goes against the entire grain of what umpires do. Especially when it was for a paycheck. If that's why they're doing the job, we were better off without them.

They should go back to a 9-game W.S.

2010-03-23T12:36:42.867-07:00

They should go back to a 9-game W.S.

Thanks, Mike. I think working from black-and-whit...

2010-03-17T09:36:17.665-07:00

Thanks, Mike. I think working from black-and-white photography really suggested the chiaroscuro-like approach. The photography from that era really pushes the highlights and shadows, and I wanted to keep some of that kind of look in the painting even after adding in more mid-tones to better define the form.

I love it! I wish I had the spare cash right now ...

2010-03-16T09:23:55.567-07:00

I love it! I wish I had the spare cash right now to commission a piece.

I like the way you used light and dark in this piece. It reminds me a little of Rembrandt chiaroscuro.

Thanks. I posted one I did of Tony Gwynn a while ...

2010-03-15T19:35:16.194-07:00

Thanks. I posted one I did of Tony Gwynn a while back:

http://www.3-dbaseball.net/2009/05/art-of-hitting-tony-gwynn.html

And another one I've done for my dad in the past, of Josh Gibson:

http://i16.photobucket.com/albums/b37/adorhauer/PICT0149.jpg

The ones I've done already are all owned by someone, but I'm always open to doing commissions. If you're interested, you can email me at kitikami (a) gmail (dot) com

This is great, if you have others I'd love to ...

2010-03-15T16:40:29.302-07:00

This is great, if you have others I'd love to see them, and would definitely consider trading some of these dead green presidents to take some off your hands.

-Chris

The moment I saw that butterfly ballot I started l...

2010-01-14T07:11:59.229-07:00

The moment I saw that butterfly ballot I started laughing. :)
-j

Another one to add to the list: When pressed in a...

2010-01-12T22:17:38.504-07:00

Another one to add to the list:

When pressed in an MLBNetwork interview to put an estimate on the number of times he and McGwire shot up steroids in a bathroom stall together (he says only that this incident occurred "more times than [he] could count" in his book), he admits that it was maybe only one or two times. His defense was that his statement about "more times than [he] could count" was misinterpreted, and he meant something about playing for 7 different teams in his career or some such explanation that made no sense. Here is the quote from his book:

"...McGwire and I spent a lot of time together. Of course, we didn't talk much. What we did, more times than I can count, was go into a bathroom stall together and shoot up steroids." (pp7-8)

If that was misinterpreted, then Canseco is a sufficiently bad communicator that we should not be taking anything he says as meaningful anyway.

Right, it's mostly just a matter of how much m...

2009-12-08T07:02:49.911-07:00

Right, it's mostly just a matter of how much math you're willing to do and how precise you want to be. (2*OBP+SLG)/3 is good because it is simple to calculate, whereas 1.75*OBP is a pain to try to estimate in your head, so it loses a little functionality. (1.75*OBP+SLG)/3 is better if you have a spreadsheet or calculator or something, and it's still simple to remember the formula and type it in, which is probably the biggest advantage over something more precise like the equation in the article. If you are doing a whole list of players and are running the calculations in Excel, the complex formula wouldn't be much more cumbersome than the simpler formula, except there's no way someone will have the more precise formula memorized for immediate use. Also, as long as you are using 1.75 instead of 2, you'd probably want to use a divisor closer to 3.05 to scale the mean a bit better, which goes back to how precise do you want to be compared to how much you want to have to remember or rely on computational aids.

Honestly, I don't think the formula in this article is of much practical use except in a few rare cases (though if you do want to be more precise and don't mind a little extra work, I guess it's up to someone to decide if it's worth using). If there is something of use to take from this exercise, it's probably that OPS (or other incarnations thereof) are a messy operation that doesn't really work off of a clean set of logic like something like wOBA. Combining unlike terms such as OBP and SLG distorts value to some extent even if you weight them properly, and they happen to work well enough at valuing most players mostly by accident.

Not that you should really care about that if you are just trying to get a quick and decent estimate of a hitter's value. The shortcut works well enough, so you might as well take advantage of that.

The biggest problem with (2*OBP+SLG)/3 is that the...

2009-12-06T23:45:44.723-07:00

The biggest problem with (2*OBP+SLG)/3 is that the "2" is an oversimplification. It's been known for awhile that the best fit is a multiplier around 1.75. Just use (1.75*OBP+SLG)/3 and you will find the league average is very close to .330.

Of course your fitted curve will give a better fit than a linear equation, but it's not always worth the extra math.

Very big thanks.

2009-11-27T11:45:05.231-07:00

Very big thanks.

Team UZR is the number of runs the team saves over...

2009-11-25T15:52:22.065-07:00

Team UZR is the number of runs the team saves over average on all balls in play, so to figure out how many runs they save on a single ball in play, you would take team UZR and divide it by the number of balls in play the Rays allowed. BIP can be estimated by (BF-HR-SO-BB-HBP), which comes out to 4323 for Tampa Bay. They were +70 in the field this year according to UZR, so you would divide 70 by 4323, and get that each batted ball in front of that defense was worth .016 runs better than average. Ideally, you would want to use a regressed value for team defense instead of the full 70 runs, but it won't make a huge difference. If you want, you can knock it down to 50 or 60 runs and see what that does in your final outcome.

Since the Rays are an above average defense, you subtract .016 from -.04, which is the average value of a ball in play, because a BIP against the Rays is worth .016 runs less for the offense than an average BIP. You would get the following for your new values of each of the 4 events in FIP:

HR 1.4, BB .3, SO -.27, BIP -.056

Subtract -.056 from each of the values (or add .056, same thing), and you would get

HR 1.456, BB .356, SO -.214, BIP 0

Multiply each of those by 9, and you have your coefficients.