Obviously, I haven’t been terribly active on this blog recently, just doing what might be considered the bare minimum as far as covering MMA goes. SILVA has been pretty bad rating fighters in recent events, and when SILVA is doing poorly, I tend to get irritated and want to do something about it. Over time, SILVA has been getting better in small increments, and so I want to take the next step to make it just a little bit better than it is right now.
After looking at a couple ways to possibly improve the system, I think I’ve settled on the changes I want to make. SILVA is proprietary, so there’s only so much I can share, but I’ll explain the changes I’m making and why I’m making them.
SILVA scores are essentially an estimate of how good a fighter is, based on his record, using only the wins and losses of the fighter and his opponents as data. SILVA first rates individual fights using a proprietary formula called Victory Score. It then takes a fighter’s ten most recent Victory Scores, and determines his FQ and FC: Fight Quality and Fight Consistency. A high FQ is achieved by defeating at least two high-level opponents. A high FC is achieved by defeating quality opponents on a consistent basis. FQ and FC are the statistics that determine a fighter’s SILVA score.
One problem I’ve found is that Victory Score tends to overrate wins against opponents who are not very experienced. For example, Mike Russow earned high Victory Scores with his wins over Jon Madsen and Todd Duffee. Neither Madsen nor Duffee is rated highly by SILVA, but wins against them were rated highly by Victory Score, because Madsen was previously 7-0 and Duffee was previously 6-0. As a result, Russow has been given a very high FQ, and because Russow’s FC is extremely high based on being 10-0 in his last 10 fights, his overall SILVA score is an extremely high 80.66, good for 5th best in the heavyweight division.
I plan on regressing wins earned against relatively inexperienced opponents like Madsen and Duffee. Doing this should help Victory Score to rate such wins appropriately. If I made only this change, Russow’s SILVA score would fall to 68.80, and he would be rated as the 10th best heavyweight in MMA. He would still be virtually tied with his opponent at UFC 147, Fabricio Werdum (69.03), but at least that’s a rating that seems more accurate.
(By the way, this change would also once again result in Junior dos Santos being the top heavyweight with a SILVA score of 97.21. Daniel Cormier would be a very close second with a SILVA score of 96.97. Regardless of who wins the championship rematch between dos Santos and Cain Velasquez, I really want to see dos Santos vs. Cormier at some point.)
The second change I want to make concerns Victory Scores on the other end of the spectrum: wins against opponents who have fought in professional MMA for a long time. Right now, the Victory Score formula still treats wins over legends like Antonio Rodrigo Nogueira and Mirko “Cro Cop” Filipovic as if those fighters are at the top of their game. My plan is, instead of having Victory Score account for their whole careers, to only look at their last 20 fights. This way, when a fighter like Frank Mir manages to beat Nogueira, Victory Score sees Mir as beating a fighter who is 15-5, instead of a fighter who is 33-7-1.
Here’s the problem. These changes are enough to compel me to basically build new spreadsheets from scratch, because the data in the spreadsheets I have now would need to be vastly changed anyway. Building new spreadsheets will take a lot of time, so I’ve decided to stop updating the current version of SILVA scores, and instead dedicate my time to developing the new SILVA scores. The resulting system will be considered SILVA 2.0.
While I won’t be able to unveil complete lists of fighters in SILVA 2.0 for a while, what I can do is at least show SILVA scores for upcoming fights. This means the first two events I plan on using SILVA 2.0 for are next week’s UFC shows: UFC on FX 4 and UFC 147.
Even with these changes, SILVA will be far from a finished product, but like I said, it’s all about making it better in small increments. Here’s hoping the result is a system that is truly worth taking a look at when breaking down high-level MMA fights.