Stats. Giggity!

[comish4lif]

Greatest catches measurable? 

http://hardballtalk.nbcsports.com/2014/03/01/video-mlbam-unveils-revolutionary-new-player-tracking-system/

[comish4lif]

Average Player which is actually definable

Those stats use a "replacement level player" because when a good player gets injured, an average player isn't available to replace him, you need to find a AAA/replacement level player.

[blue911]

Those stats use a "replacement level player" because when a good player gets injured, an average player isn't available to replace him, you need to find a AAA/replacement level player.

You want to build the most accurate method of rating players, the whole "replacement player" argument is nonsense.

[nobleisthyname]

You want to build the most accurate method of rating players, the whole "replacement player" argument is nonsense.

I hate to go through this again but it's really not. Read through the article I posted. There are plenty more too.

Also this MLBAM product... 

[PC]

For Washington Nationals, there are eyes, deep flies and statistics

http://www.washingtonpost.com/sports/nationals/for-washington-nationals-there-are-eyes-deep-flies-and-statistics/2014/03/04/93bfd108-a3e5-11e3-84d4-e59b1709222c_story.html

[JCA-CrystalCity]

For relievers, I like to spot them by looking for a K/9 > 7 and then sorting by K-BB.  K:BB is nice, but I think relievers need to strike out guys.  I spotted Steve Delabar when he was with Seattle this way, and Jared Burton a couple of years ago.

[PowerBoater69]

Baseball Prospectus Day at Nationals Park - June 21, 2014

Baseball Prospectus and the Washington Nationals invite you to join us for a great day of baseball on Saturday, June 21 at Nationals Park. Thanks to the fine folks in the Nationals front office, we are proud to be able to offer our guests the following:

--Admittance to a private "meet and greet" with special guests and Baseball Prospectus personalities (including Jason Parks, Ben Lindbergh, Harry Pavlidis, Bret Sayre, Joe Hamrahi, Mike Gianella, Mauricio Rubio, Ben Carsley, and Craig Goldstein) in the Roosevelt Room 90 minutes prior to game time
--Question and answer session with industry guests and members of the Washington Nationals Baseball Operations department
--Ballpark buffet including hamburger bar, hot dogs, salads, chips and soft drinks
--Reserved seat in the LF corner section of Nationals Park to watch the Nationals take on the Atlanta Braves with Baseball Prospectus fans and staff
--A $15 coupon to be used toward a new Baseball Prospectus Premium subscription or your next Premium renewal
--Game time is 7:15 PM.
--Price: $67.99

http://www.baseballprospectus.com/article.php?articleid=23046

[blue911]

Baseball Prospectus has a working agreement with MLBAM. If you have a BP subscription, you can get a 20% discount on MLB.TV.

[HalfSmokes]

http://www.businessinsider.in/Mystery-MLB-Team-Recently-Spent-At-Least-500000-On-A-Supercomputer-To-Use-During-Games/articleshow/32028947.cms

some team bought a cray- from the article

Specifically, the supercomputer will allow the team to process information during a game quickly enough that they will be able to use the information to influence strategy during the same game.

my guess is that isn't the reason, but who knows

[dracnal]

The bit I find interesting is that the team that bought it reportedly is one where five years ago, you never would have dreamed they would have need of a computer like that.  So how many teams sucked horribly in 2009 and are considered serious contenders this year?  If that statement isn't just smoke, I think it's a pretty small pool of teams that a: can shell out 500k, b: have eschewed sabermetrics in the past, and c: went from worst to first in 5 years.

Can a STH go offrez on their tour and look for the locked room with the shiny shiny IT wonders locked deep inside?

[varoadking]

Think they had enough room in Davey's old office for it?

[welch]

http://www.businessinsider.in/Mystery-MLB-Team-Recently-Spent-At-Least-500000-On-A-Supercomputer-To-Use-During-Games/articleshow/32028947.cms

some team bought a cray- from the articlemy guess is that isn't the reason, but who knows

Hmm...processing power is easy. Human input is slow. Unless the mystery team has (1) determined what real-time inputs really matter, and (2) has installed sensors that detect those inputs, then this Cray thing sounds like a story meant for release on April 1.

There are about a half-dozen other questions this story should provoke: some team developed a real-time stat system without another team knowing? Who wrote the software, in secret? Is it portable? If not, then what does the team do on the road? If the manager comes to depend on The System, aren't they something like a modern "developer" forced to write a system in assembly language?

OK. I'll now return to my regularly scheduled job as a semi-retired computer programmer.

[blue911]

Werth and his damned algorithms

[dracnal]

Hmm...processing power is easy. Human input is slow. Unless the mystery team has (1) determined what real-time inputs really matter, and (2) has installed sensors that detect those inputs, then this Cray thing sounds like a story meant for release on April 1.

There are about a half-dozen other questions this story should provoke: some team developed a real-time stat system without another team knowing? Who wrote the software, in secret? Is it portable? If not, then what does the team do on the road? If the manager comes to depend on The System, aren't they something like a modern "developer" forced to write a system in assembly language?

OK. I'll now return to my regularly scheduled job as a semi-retired computer programmer.

With the cameras they've got in parks these days, it shouldn't be too hard to get a real time analysis of things like pitch break, etc.  You can also very quickly note things like how often the catcher calls for it low and in when the count is 1-2 and have the machine start noting those patterns.  Yes, you can do it with pen and paper but it's easier with a computer.  Assuming you have something like a quick form that tracks things like count, infield in, guarding the baselines, shift, whatever, and toss in how fast the pitcher is to the plate from the windup, the stretch, etc., you could probably put together a pretty impressive probability spread.  A computer that fast means they're getting automated input with a LOT of data.  Again, that means likely they're scanning video feed and pulling data out of it. 

Given the foray into drones, I could see a video stream being analyzed for positioning of the players, depth, etc and some sort of database being built up around that.  As for away games, I don't see why you couldn't take the same video feed and run it over a decent broadband connection back to home base.  Given RDP and modern bandwidth, this shouldn't be even remotely a problem.

Granted, video analysis like this isn't my field of expertise (networking, tunnels between sites, etc. is) but I'm pretty sure that anything that doesn't involve video wouldn't need a Cray.  That kind of power means realtime analysis of a huge chunk of data.

[dracnal]

Werth and his damned algorithms

Ironically, algorithms would be exactly how you'd analyze a data stream like that looking for patterns.  I wonder if that comment was something of a Freudian slip by Werth.  ie the team has been told about the plan to use a algorithms and a hell of a computer to get real time adjustments to their approach at the plate and he's expressing his scorn for the value of that sort of analysis.

Of course, that's probably the Kraken telling me the insight is more clever than it is.

[blue911]

Ironically, algorithms would be exactly how you'd analyze a data stream like that looking for patterns.  I wonder if that comment was something of a Freudian slip by Werth.  ie the team has been told about the plan to use a algorithms and a hell of a computer to get real time adjustments to their approach at the plate and he's expressing his scorn for the value of that sort of analysis.

Of course, that's probably the Kraken telling me the insight is more clever than it is.

I'd imagine it would be helpful for Pitchf/x, Hitf/x and Fieldf/x data. All of which is use cameras to capture data. I know that baseball prospectus writes a script to analyze pitch framing using pitchf/x data.

[welch]

I'd imagine it would be helpful for Pitchf/x, Hitf/x and Fieldf/x data. All of which is use cameras to capture data. I know that baseball prospectus writes a script to analyze pitch framing using pitchf/x data.

The data would not be heavy enough to use a Cray. I'm familiar with various bits of hardware used in the financial industry, including real-time trading. I can't remember meeting a Cray. The various types of input for "f/x" data would be nothing fancier than typical analog input converted to digital and read by something. An airliner on automatic pilot probably receives and analyzes more inputs than we have in a ballgame. I saw some classy calculations done on Mickey Mantle's 540 - 550 foot homer out of Griffith Stadium, calculations that compared the trajectory of all home runs hit in the last ten years or so -- and that's about as far back as the data goes. The calculations also allowed for the weather: temperature, humidity, and, especially, wind direction that day, with an adjustment for weather as reported at National Airport against weather at the Stadium. I think someone even tried to guess the wind-speed at roughly ground-level and a change when the ball got above the grandstand.

It's still hard to know what a manager would do with the data. Move the LF in a step and to the right because of a hitter's tendency and the pitch that had been called? What happens when the curve doesn't break as much as the pitcher intended?

[JCA-CrystalCity]

10 lessons Mitchel Lichtman has learned about defenisve statistics

Developer of the most common version of UZR.   Nice read on what is behind a defensive metric and issues in developing a metric.

[welch]

Interesting especially for what Lichtman says he could not accurately measure: runner on 1B, hit&run, right-handed batter, for example. Another: infielders might adjust (should, anyway) depending on fast-ball in or outside, breaking ball, etc. Might not position themselves in a way that would tipoff the hitter, but out to be prepared to jump. 

[Squab]

10 lessons Mitchel Lichtman has learned about defenisve statistics

Developer of the most common version of UZR.   Nice read on what is behind a defensive metric and issues in developing a metric.

I really enjoyed these as well:
10 Lessons I Learned From Creating a Baseball Simulator
10 Lessons I Learned From Creating a Projection System
I've been wanting to try something similar myself (the simulator).

[blue911]

10 lessons Mitchel Lichtman has learned about defenisve statistics

Developer of the most common version of UZR.   Nice read on what is behind a defensive metric and issues in developing a metric.

He fumbles when it comes to errors. He has already built in the "easy out" made or not into his formula, then wants to assess an additional penalty. At the end of the day, good fielders have to be to show that their contributions can be quantified. Grading the degree of difficulty without ball speed is terribly simplistic.

[JCA-CrystalCity]

This generation of defensive metrics might be like comparing civil war surgery to just leaving the wounded on the field.  It's an advance, but 10 years from now with hit f/x and field f/x, the UZR will be thought of like trying to decipher how honus wagner fielded off of box scores.  You will still have some sort of notion of penalizing a guy for balls not converted to outs that should have been, some measure of difficulty of fielding a ball and converting it to an out, and some sort of notion of which fielder should have gotten to a ball.  It's just it will be based on precise positioning data, take account of the speed and angle of the contact, and perhaps incorporate more data on baserunner and out state.

[imref]

http://fivethirtyeight.com/datalab/the-spot-in-mlb-lineups-where-managers-are-still-ignoring-sabermetrics/

[JCA-CrystalCity]

http://fivethirtyeight.com/datalab/the-spot-in-mlb-lineups-where-managers-are-still-ignoring-sabermetrics/

fortunately for this team, Willliams may have stumbled into one of the better choices for #2 in Rendon.  If Span were dropped in the order when Harper gets back, then we might have close to an optimum order.

[HalfSmokes]

Here's a question about UZR that I was wondering- how will it deal with more frequent shifts? Right now, at least according to the fangraphs primer, it appears that plays where there is a shift are ignored. That's fine if they are rare, but since they are becoming more frequent and are usually used against an opponent's best hitters, doesn't that both lessen the quantity of data and skew it?