Laying the Ground Work for SoS Stats and Building a House
by szucconi on Apr 21, 2009 6:02 PM EDT
Email
PrintJCS from fightmatrix.com and Leland Roling formerly with MMA-Analyst.com now doing great work on BE in the fan posts have been posting on Strength of schedule, the fight matrix post and Leland's post. They really got my brain going about this useful stat. SoS is used in the calculation of BCS rankings and talked about in college basketball rankings and seeding for the torney, but it can be applied to other sports. I think it will be useful to judge a persons body of work at a glance in MMA. It is by no means a be all end all determining factor in a fight, but an indicator of how a person has faired over a career. A fighters record falls short in doing this for obvious reasons. Take a look at the record of Randy Couture and Jason Reinhardt to make this point clear.
The approaches taken by JCS and Leland are similar to those taken by other sports. Leland is close to the BCS calculation and JCS is a more standard approach, but MMA poses some unique issues. Namely, careers span ten or fifteen years and we don't have a nice tight season in which to analyze. JCS uses the sliding window of three years around a given fight to gain this "point in time" perspective. Another issue is the schedules lack uniformity. A fighter can be more or less active. A football team may play an easy schedule, but they all play about the same number of games. So there is a margin for error when fighters are less active. JCS uses a min of 100 opponent fights to gauge a schedule.
The BCS approach that Leland takes is good. It is very close to the real BCS calculation. He has done some filtering to address the issues I have stated that make MMA <> to college football, but it is a great start. There are, however, a few rules that the BCS uses that Leland didn't apply. Well, one really that stuck out at me. How D-II schools factor into the records of D-I schools. A win is dropped, but a loss counts. How can this be translated to MMA? What is the D-I and D-II of MMA? What I propose is that wins count if the loser has fought in a major organization and losses always count. What is a major organization is open for debate. And should a time window be applied to that question, I am not sure. It has everything to do with how an opponent's record is calculated.
The things I like about the method JCS uses is the sliding window put in place for a uniform schedule, but I don't think is solves anything. It makes the stat more timely, but it also trims the sample drastically. You can't know how big a win was for a fighter for 18 months. Logically that is true, but that's a long time. I think the proper way to do it is to look at all fights in the past three years. This will take some tweaking and I am sure JCS and Leland will make some changes and play with the numbers a bit, but a lot can happen in three years.
As in keeping with the BCS calculations, head to head match ups should be taken out when calculating an opponent's record. To see a sample of the BCS calculation. I will be running the numbers as soon as I can gather the data and get geeky with some sql. But I want to discuss some of the issues and mitigate the unique issue and complexities that MMA poses. Any ideas? Questions?
The FanPosts are solely the subjective opinions of Bloody Elbow readers and do not necessarily reflect the views of Bloody Elbow editors or staff.
8 comments
|
0 recs |
Do you like this story?
Comments
This is what I mentioned in my post,. Head to heads are supposed to be taken out, BUT… that is the argument that has been given by fans opposed to the BCS formula. A lot of people think those matchups included gives a better number.
I’m working on a different algorithm, albeit it’s going to take time, and I want to get the database in pristine condition to give more accurate results.
Follow my analysis of all things MMA on BloodyElbow.com
How do you feel about including some form of a D-I and D-II type distinction to eliminate some wins from records.
How would we figure this? Lumping promotions into distinct categories?
Follow my analysis of all things MMA on BloodyElbow.com
by Leland Roling on Apr 21, 2009 11:04 PM EDT up reply actions
Basicly I am toying with the idea of calling some promotions D-I and the rest D-II. Then calling a fighter D-I if they have ever fought in a D-I show or have Fought in a D-I show within the time frame we are looking at. To remove a lot of the rif rafe, and there is a lot of that in the 70K fights.
As a first pass, UFC, Pride, Dream, EXC, Strikeforce, Affliction. That covers most fighters that are the core of the sport. This also allows for movement without fighting. Like Huerta beat Wiman before either were in a big show. Huerta’s win over wiman counts once Wiman makes it to the show and Huerta’s SoS goes up or down when Wiman fights because huerta faced him.
I am not totall sure what you meant by..
eliminating head to head matchups.. but I did remove any looping scenarios that might have occurred in my analysis.
Creator of the FightMatrix rating system.
Look at the BCS sample calculation. The focus is 1999 FLORIDA STATE. the first opponet on the schedule was LOUISIANA TECH. There record was 8-3, but they played Flordia State and lost so you use 8-2. They also played a D-II school in SAM HOUSTON and won, but D-II wins don’t count. so you use 7-2 as there portion of the opponet record for Florida state. The raw opponets record for Florida State in 1999 was 69-55, but after you take out the record of Florida state’s opponets when they played Florida state and the D-II wins gained by all of Florida state’s opponets you come up with an opponet’s record of 65-44.
Right.
I handled the looping… but obviously not the D1/D2 situation
Creator of the FightMatrix rating system.
Its not so much looping as it is accounting for a record twice. Flordia State record 11-0, Flordia State’s opponets record when they were not playing Flordia 69-44, Flordia State’s opponets record when they were not playing Flordia and removing D-II wins 65-44.
And applying the D-I and D-II thing to MMA is only something I am thinking about. Like if a guy has fought in the UFC, Pride, WEC, EXC, ect. then they are “D-I”.
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Recommended FanPosts
Recent FanPosts
MMA Rankings
Certain photos copyright © 2012 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
