THE BOOK--Playing The Percentages In Baseball

incredible!

I wonder who did that graphic, Popular Mechanics or Sportvision.  I’m guessing PM.

It will be interesting to see…
1) When FIELDf/x actually gets implemented league-wide.

2) What gets done with the data.  It’s going to take a lot of expert human input to turn the data into understandable information, and I’m not sure the principals comprehend the extent to which that is true.  It’s like PITCHf/x in that regard, only WAAAAAY moreso.

I can’t wait for this to come

The batter isn’t in the box when he swings

The batter isn’t in the box when he swings

I noticed that, too.  That’s one of the many reasons I’m pretty sure this a PM artist’s conception rather than based on any actual data from SV.

That’s beautiful.

I thought that Sportsvision had a FieldFX up and running since it was revealed at their first “summit” in May 2008. What am I missing?

Greg Rybarczyk made a presentation at the first summit in 2008 that first proposed (as far as I know) the concept of FIELDf/x system.

Sportvision worked with some ideas over the 2008-2009 offseason, one of which was the source for the data in the demo that was presented to MLBAM and later partially published in the NY Times (http://www.nytimes.com/2009/07/10/sports/baseball/10cameras.html).

At the July 2009 summit, Sportvision revealed that they had an experimental system running at AT&T Park in San Francisco.

Then it’s a long way from experimental to production quality and implementation in every major league park.

Mike, how much data are we talking about? Have you had a look at the experimental fieldf/x raw data?

Jeremy/9, I have not seen any of the data from the system they have at AT&T.

I did the analysis for the video and graphics that are attached to the NY Times article I linked in #8.  That was from a precursor to the system at AT&T, so it’s not exactly the same, but I imagine that the type and complexity of the data will be similar in the ultimate FIELDf/x. 

I can’t go into the specifics of that system beyond what was published in the Times due to NDA. However, between looking at that video and a little imagination, you can get a good idea of what would need to be in the data set. 

A PITCHf/x pitch can be described with some game state information (batter id, pitcher id, inning, outs, ball-strike count) and the nine pitch trajectory parameters (initial position, initial velocity, and acceleration, in three dimensions).  What would you need to track the batted ball and each thrown ball throughout their flights, the real-time position of every fielder, and the real-time position of every baserunner?

Well, if field F/X are this close, that must mean we’re pretty remarkably close on hit F/X, right?

A question, though, are those concentric circles around the left fielder’s original position generated using time or distance?

Peter - Sportvision has been providing Hit f/x for 2009 to the teams. Actually, the developments in Field f/x seem to be delaying Hit f/x, as most of the available resources at Sportvision are being used on Field f/x rather than processing the backlog of 2007 and 2008 video for Hit f/x.  The question of if and when the earlier Hit f/x information will be made available to the public depends upon the financial model that Sportvision is negotiating with all its paying customers.  For some silly reason they have this need to make a profit from their work.

A question, though, are those concentric circles around the left fielder’s original position generated using time or distance?

If that’s just an artist’s conception, as I believe it is, it’s whatever the artist felt like drawing.  If you want something that’s data-based, look at the Times article I linked in #8, particularly the video that accompanies it, and that should answer your question about the system.

The way it would have to work, I think is time.  Think of things as 4 dimensions: the 3-D world you know, plus time.  So, at each second or split-second, where is everyone.  And at the next second, where is everyone.

Tango/14,

Agreed.  Unless it’s generated from time, it’s nothing more than a pretty visualization.  If it’s time-generated, there’s plenty of useful information portrayed--which is pretty much the goal.

Peter/15, I really do believe your questions will be answered if you view the video with the Times article.  Some people don’t notice the video the first time they read the article.  It’s about half way down on the left side, like so:

That was made as a demonstration of the system’s capabilities, and it will tell you a lot more than the Popular Mechanics article or dissecting the artist’s mistakes in the PM graphic.

You are right, I never saw the video.  Very cool.

Yeah, I didn’t notice that, either, thanks Mike.

Thanks for the video reminder.  I notice it has Miguel Cairo running 20 mph.  Does that seem fast for Miguel Cairo?

Again, I think this is going to emphasize “hang time” as the key component to defense in the outfield.

Chris, it’s also got the ball crossing high above the infield arc after more than two full seconds, at a speed of 45 mph, and moving with a more or less constant horizontal speed from home plate to the LF wall, which isn’t how the ball really moves after leaving the bat.  A ball like this one slows from ~100 mph off the bat to ~50-55 mph at landing - if you’ve ever watched a blimp-shot of a home run, or even a golf drive, both of which, visually, seem to rocket off the bat/club, and then gradually die at landing, you’ll recognize the difference. 

This video was a demonstration of how the various elements (pitch, hit, field) could come together, and how the synthesis of these various elements can provide a really immersive presentation of the play.  Never mind the exact ball trace, or the numbers…

Chris/19, Miguel Cairo was, once upon a time, among the fastest (i.e., upper quintile) runners in baseball, at least according to speed scores.

Believe me, I was surprised to see him running that fast, too, and I double and triple-checked those numbers before we included them in the video.

Greg/20, the player tracking data that’s included in the video is all from real data.  The ball tracking data includes some estimates/models/guesses on my part as we didn’t have full-trajectory tracking data at that point in the project.  The more or less constant horizontal speed is me ignoring/kludging the drag effect in the interest of knocking out this analysis in the space of a few days.  The overall speed of the ball should be roughly accurate, however.  The ball will in fact be going its slowest speed at the peak of its flight, having exchanged much of its kinetic energy for potential gravitational energy and will speed up again toward landing as it changes that potential gravitational energy back into kinetic energy.

So you can blame the inaccuracies in the ball trace on the crudeness of my quick model, although the hang time should be accurate and the ball trace shouldn’t be horribly off.

For that matter, Chris or Greg, you can time Cairo around the bases yourself and check my work.  This video and data is from the 6th inning, September 20, 2008, Seattle vs. Oakland.

20 MPH would get you from home to first in 3.1 seconds.  Faster than Ichiro.  But home to first times start with the runner stationary, then accelerating to top speed (not sure how many feet it takes to get there).

So 20 MPH might be right.  That’s only 2/3 of Usain Bolt speed, right?

4th inning I think.

The first 50m (about 165 feet or the distance from a standing lead at 2B to home plate) of Usain Bolt’s Olympic run was 5.50 seconds, or 20mph.

His top speed was 0.82 seconds per 10m, or 27mph. 

(In order to convert, take metres/seconds times 2.237.)

So, if Cairo’s top speed is 20mph, that’s 1.12 seconds per 10m.

In other words, Usain Bolt, out of the gates, and running for 50m on a perfect track is the same speed as Miguel Cairo, with a running start, running for 50m on a dirt track, timing with the base, and with equipment.  Both are 20mph.

The key is the running start, since Bolt with the running start has a 27mph speed.

I think we discussed this that I had Raines at 3.1 seconds per 90 feet (on the run), which is 27.432m or 20mph.

Hmmm, something doesn’t seem to be right.  Cairo can’t be that fast can he?

Somebody get a stopwatch please and test this.

Mike, no sweat, I recognize that the hit trace is an estimate, and I totally understand the imperative to prepare that trace quickly.  The demo did its job very well, no need at that stage to agonize over the exact ball path & time, that will come later… and in any event, the ball and trace did both arrive at the fence at the same time, so yes, the overall average horizontal speed is fine.  Once they are actually tracking the ball the versimilitude on the ball path will come.

On the slowest speed point, turns out that is actually a bit past the vertical peak of the ball, as the bleed-off of horizontal speed of the ball near its peak, due to the drag force, outstrips for a while the gain in vertical speed as the ball passes peak and begins to move vertically back down towards the ground. 

Once the ball gets well past its peak and further into its descent, the drag force rotates and starts pointing more at an angle up into the air, rather than straight horizontally as it does at peak, at the same time that the vertical speed is still growing larger, and the ball begins to speed up overall. 

Eventually (not in a ballpark, but if you were to hit a ball off the rim of the Grand Canyon), the drag force will entirely eat away the horizontal component of velocity, and the vertical component will stabilize at the ball’s terminal velocity.

Sorry for the length of this, but I love discussing this stuff!

Tango, yes, 4th inning, thanks for correcting that.

Cairo ran about 180 feet in 7.4 seconds, for an average running speed 16.6 mph.  (180 feet, because he ran a circular path, but he started with his lead foot already off second base, and he cut the inside corner of the third base bag.) He took about 2 seconds to accelerate to 17 mph.  The remainder of his run, his speed was between 17-20 mph.

Tango/24, the MLB video doesn’t show Cairo tag up and leave second base.  However, if you time from when Cunningham missed the catch until Cairo crossed the plate, both of which are in the video, it’s about 7-8 seconds.  I can assure you that I checked that before I released that data. 

If you think my 180-ft estimate of the distance is valid (I see now that in the video I say 181.5 ft), then the running speeds should also be valid.

I can’t click the video from work, I’ll have to check later.

Wouldn’t a runner’s speed suffer a bit while he’s rounding third and changing direction?  Even taking the wide turn I think you lose something there compared to running a straight line.

Usain Bolt has to be the most appropriately named athlete ever.  I hope he remains the point of reference for top human speed for a while.  But he’s still not as fast as the domestic BrianKaat.

Wouldn’t a runner’s speed suffer a bit while he’s rounding third and changing direction?  Even taking the wide turn I think you lose something there compared to running a straight line.

You might think so, but it doesn’t.  Or didn’t for Cairo.  (I might not want to generalize that too much to other situations.) Cairo was accelerating the majority of the way from second to third.  He began to turn well before he got to the bag.  When he crossed the bag he was running at top speed and wasn’t really changing direction at all (i.e., he was running parallel to the second base to home plate line).  Then after he “rounded” the bag he began to turn again to make it toward home.

To reiterate something from #24, if I’ve learned anything from Tom Tango and this blog, it’s the value of a stopwatch.  If I screwed up on something that basic, you can take my sabermetrics card away. 

Btw, Cunningham’s speed in the outfield going after the fly ball peaked around 13-14 mph, but I don’t think he ever got up to top speed.  After reaching 13-14 mph, he quickly slowed down and drifted back toward the wall at about half that speed.

The range rings were estimated based on his reaction time and acceleration in response to this fly ball and an estimate (based on speed scores and I forget what else) that he could achieve and maintain a top sustainable speed of 16.5 mph for the several seconds required to chase down a fly ball. 

We also made the simplifying assumption that his range in all directions would be equal.  I doubt this would be completely true in reality.