Shameless Self Promotion. :) (SABR Matt)
I don't know how many of the readers here at DOV would be up for a little light technical reading on a new sabermetric tool I am finally 100% comfortable advertising, but on the off chance that a few of you have time and energy to spare to read the white paper on a new context-analysis tool called the Fiato/Souders Matrix, I've made it available at the following URL:
http://sabermetriclab.com/fsia.pdf
Quick verbal summary of what the F/S Matrix produces in terms of results:
- It tells us how each team would perform if their schedule strength was league average
- It tells us how an average team would play if they had the specific schedule of any real team
- It gives us more reliable and more consistent (year to year) park factors for runs and for HR
- It makes it much easier to compare teams across eras using logically "more correct" methods to correct for varying contexts
- It makes an important strength of schedule adjustment to the statistical analysis of starting pitchers
The central working formula within the matrix is quite simple:
In any game, each team's runs scored are equal to:
- League average run scoring rate +
- Park factor +
- Strength of opposing team defense +
- Strength of opposing starting pitcher (relative to his team's defense) +
- Stength of theteam offense
Feel free to read the full explanation if you are curious as to how the matrix works in more detail.












May 20th, 2008 at 9:11 pm Quote
I’m sure the details would be beyond me, but I’d love to see the results of some work using it. Thanks for sharing this Matt.
May 20th, 2008 at 10:03 pm Quote
I tried to write the paper for a non-saber-oriented readership…I started with some of the basics (including simple demonstrations)…though I’d certainly be curious whether I made it accessible at least in terms of the math theory.
I’m introducing it here in part because I intend to share results from it in future articles…it’s been through a long series of studies in the last several months and I am interested enough in some of the findings to be convinced it has value.
May 20th, 2008 at 10:20 pm Quote
Hey Matt-
Sounds exciting, and I’d love to help you out! Starting a read on it now, might take a few days to get through the full 18 pages but I’ll chip in with feedback as I have it. As a SABR-noob and average mathematician, hopefully I can give you a little of what you’re looking for.
May 21st, 2008 at 7:38 am Quote
Hey Brendan…thanks for taking the time to read through it.
Where I’m short on readability, I’d like to know for future improvements.
May 21st, 2008 at 7:59 am Quote
I think if you include park factor and make a 4*4 matrix I would think you should have made a 5*5 matrix including the Ump.
May 21st, 2008 at 8:40 am Quote
Hey DrNaka…I would agree if we had umpire data further back than 1957. That data appears to only exist in the play by play era…which is a shame, because the HP ump does appear to have a large impact on scoring.
May 21st, 2008 at 9:18 am Quote
So you can do it for the play by play era and now validate how good or how bad the 4 by 4 matrix is compared to 5 by 5 matrix.
May 21st, 2008 at 9:38 am Quote
that would require two different programs…not saying I’m against the idea…but I need to think about whether the programming time is worth it (not quite as easy to do as it is to say.
)
Thing is…I’m not the programmer…I came up with the concept…I don’t have the C++ programming skill to make that change without some consultation with someone who IS a programmer.
May 21st, 2008 at 10:40 am Quote
Hey DrNaka…you’ll be happy to know that I’m speaking with the program architect we paid to build the current release of the F/S Matrix about adding an umpire field for the PBP era…the BP dataset suggests that umpires have a HUGE impact on scoring so I’ve decided it could be a major benefit to the accuracy of the data returned.
May 21st, 2008 at 11:41 am Quote
Wow…the retrosheet.org crowd has been very industrious! The umpire data goes ack to the beginning of baseball history now…from 1871 to present we have home plate umpire data. That’s…stunning.
OK…this is DEFINITELY something I want to pursue now.
May 21st, 2008 at 6:44 pm Quote
Very quick
May 21st, 2008 at 7:40 pm Quote
Will probably be a week or so before I can get my coder to write the new code and get the results cleaned up…but yeah…I respond quickly when someone makes a good point and I have the data to use.
May 21st, 2008 at 11:08 pm Quote
Knock yourself out on the promotage Matty :- )
Sorry we can’t join you guys more this week… 3 jobs and they are all mass crazy… will be back soon though… play nice :- )
May 22nd, 2008 at 8:40 am Quote
DrNaka…you might be interested in this study conducted by Tom Tango on umpires’ strike percentages:
http://www.insidethebook.com/ee/index.php/site/comments/do_umpires_have_their_own_strike_zone/
Demonstrates pretty clearly that the umpires do indeed screw with the normal run scoring rate. Since the swing in run value between a single ball and a single strike is about 0.15 runs on average, and the range of balls called in the 150 pitches per game per side thrown is as much as 4-6 from the lowest number (74 or so) to the highest (80)…umpires could easily be adding or subtracting half a run of ERA.
May 22nd, 2008 at 11:21 am Quote
Now that I think about it…a comprehensive sabermetric study of the impacts of umpires on the game has NEVER been done…so I’m rather excited about including that variable in the matrix. Work on that begins next Wednesday according to my coder…new results to be had by the beginning of the following week.
May 24th, 2008 at 1:15 pm Quote
Hiya again mateys…
Doin’ 80-hour weeks but, assuming anybody cares :- ) will be back in days, not weeks …
Brother Baker’s intuition that the early-May ballclub was on the verge of giving up, seems to have been borne out here…. ’s ok by me, ’cause for me, a club in rebuild is almost as interesting as one in contention… if ‘08 is going to be a dry run for ‘09, the focus shifts to Wlad, Clement, Bedard, Morrow and the other amigos with whom the M’s will win their next pennant…
Back to normal D-O-V shtick levels afore long… keep da faith :- )
May 24th, 2008 at 2:22 pm Quote
Re: Bavasi’s comments today—
We’ve all seen this before. A decent situation turns bad, then worse, and the upper level people who created the situation throw the worker bees under the bus while completely ignoring their own culpability. “If only we had the right footsoldiers…the management team is fine.” What a crock!
Bavasi has got to go. If there was a last straw in this regard, he just handed it out.
May 24th, 2008 at 2:30 pm Quote
Bavasi’s lengthy interview before today’s game indicates that he’s not there yet though, Doc. His emphasis is on finding that “one guy” to bring in that will galvanize the team. The prime characteristics that he’s looking for in that guy are character, intangibles and clubhouse leadership. He would also like them to be better than the guy he’s replacing.
Given Sexson’s struggles and Bavasi’s trade history, can a Clement for Millar trade be far off?
May 24th, 2008 at 8:27 pm Quote
Bavasi has not ONCE traded a top prospect for a bad player. People love to knock his trading history, but no, he’s not going to trade Clement for a rent-a-player. All of the prospect-for-veteran deals have been third string prospects like AsCab, Choo, Snelling…not guys you build teams around.
I would not, however, be surprised if he traded, say, Chen for Millar.
May 26th, 2008 at 12:15 am Quote
Little progress update:
There’ve been 189007 games played in the majors from 1871 to 2006. In 189002 of them, the home plate umpire was known.
The other five are just getting labeled as UmpID of 0, which is fine.
I have my coder working on making the changes to the F/S Matrix program starting today (I sent him the changes to the SQL Queries and the new umpire data he’d need and he’s free to work). He estimated 8 hours of work tweaking the code and running the standard debug tests, so I should have results before the end of the week.
May 26th, 2008 at 10:10 pm Quote
Nice!
Perhaps we can get some information if some Umps got pocket money from some teams or not.
May 26th, 2008 at 10:36 pm Quote
HEH.
I can in fact get team/umpire splits…IOW…I can compare how each team reacts to each umpire with the umpire’s normal affect on the game and see if some teams (or starting pitchers for that matter) have a pattern of doing better with certain umpires (or ALL umpires) than they should.
May 27th, 2008 at 10:55 am Quote
My C++ guy is done making changes…I should have the deliverable (the revised program) on my system within the next day…expect reports soon…
BTW, early indications are (from looking at the test database) that umpires’ impacts on run scoring are as much as 75% as strong as the parks’ impacts on scoring. Also indications are that today’s umps have a larger spread of performance than the umps did 20 years ago.