BE Analytics: How would decisions be different if MMA fights were scored as a whole?

Stephen R. Sylvanie-USA TODAY Sp

Johny Hendricks gets a chance at redemption on Saturday after his dominant ‘loss’ to GSP. In remembrance of that event, Paul Gift examines how MMA decisions would be different if bouts were scored as a whole.

Tomorrow night Johny Hendricks gets his second shot in a row at the UFC welterweight title that many - myself included - thought he rightfully snatched from Georges St-Pierre's clutches four months ago at UFC 167. All three judges gave Hendricks rounds 2 and 4 and GSP rounds 3 and 5. The deciding round was the first. Glenn Trowbridge gave the first round to Hendricks while Sal D'Amato and Tony Weeks gave it to GSP for the split decision victory. Over at MMA Decisions, you can see that all the listed media members (15 in total) plus FightMetric scored the fight 48-47 for Hendricks.

This leads to a few questions. Who won the first round? Who won the entire fight? Should MMA fights be scored round-by-round with the 10-Point Must System or should they be scored as a whole? Answers to "should" questions are matters of opinion and we've got plenty of outstanding writers at Bloody Elbow and SB Nation who will gleefully share theirs. Today we're looking at what the fight data and my model of MMA judges say about the GSP-Hendricks fight - the first round and the fight as a whole. Then we'll go back in time and see what percentage of decisions likely would've had a different winner if the fights were scored in their entirety and check out some notable fights that may have changed the course of MMA history.

To be clear, I'm not advocating for any particular position on scoring. I personally thought Hendricks beat GSP in the fight as a whole and in three of five rounds, but that's irrelevant. What we're doing is using MMA analytics to help answer "The Mother of All Judging Questions," how would decisions change if we got rid of the 10-Point Must System and scored fights in their entirety?

Before we dive into the stats here's BE's Kid Nate, Dallas Winston and Connor Reubusch previewing UFC 171's Hendricks vs. Lawler bout:

Let's take a look at the key performance stats in the first round of GSP-Hendricks. In addition to the striking statistics, both fighters had one takedown with a few seconds of ground control. GSP also had a guillotine attempt early in the round.

Analytics_-_one_big_round_-_gsp-hendricks_round_1_medium

If you've seen some of my previous articles, you'll know why jabs to the body and legs are excluded and know that the most valuable strike for scoring points is a power shot to the head. In the first round, Hendricks crushed GSP with power shots to the head. GSP's advantage in power shots to the body was mostly nullified by Hendricks' advantage in power shots to the legs. Head jabs score points but they're worth very little.

If you're saying to yourself, "Yeah, that's what judges count, but I don't care what they think," you may also be interested to know that when it comes to getting knockouts, power shots to the head and legs are respectively 510 percent and 27 percent more valuable than power shots to the body (if we could separate true liver shots from other body shots the statistics would probably look different, but we can't right now).

All told, every major media outlet I've seen gave Hendricks the first round. My judging model gave Hendricks a 67 percent chance to win the first round, and it increases to 78 percent if we don't put any weight on misses.

What about the fight as a whole? GSP still had the only submission attempt and he also had three takedowns to Hendricks' two. Hendricks had one and a half more minutes of clinch control and one more minute of ground control than GSP. All other key performance statistics are shown below.

Analytics_-_one_big_round_-_gsp-hendricks_whole_fight_medium

That's a lot of green on Hendricks' side - and incidentally, green is something he missed out on with the loss. In its entirety, Hendricks had a 94 percent chance of winning under the "Who won that fight?" criteria. He won the striking battle, the control battle and the "Damn, look at the other guy's face!" battle, but lost the 10-Point Must war.

One of the cool things about having a model that explains judging decisions is you can use it to learn about scenarios we don't see in the real world. We'll never know with 100 percent certainty how fights would have been scored under a different system, but we can do a lot more than just guess.

I set up the model to treat the entire fight as one big round. It doesn't give any advantage to what happened earlier or later. While not exactly the same, we can use it to get an idea of how decisions would likely be different if athletic commissions were to use something like Pride-style scoring or Andy Foster's "Who won the fight?" system. I did three versions that each look at a different question.

Version 1: When is the predicted winner for the whole fight different from the actual winner?

Version 2: When is the predicted winner for the whole fight different from the predicted winner with round-by-round scoring?

Version 3: When is the predicted winner for the whole fight different from both the actual winner and the predicted winner with round-by-round scoring?

Version 1 is the least conservative because the model could just disagree with the judges in round-by-round scoring and for the whole fight. Version 3 is the most conservative since it first requires the actual and model round-by-round scoring to agree and then the model winner for the whole fight to be different.

I ran the model for all Zuffa-promoted events since Sept. 2001 and find that 14.4, 9.0, and 5.4 percent of fights that go to a decision would have a different winner in versions 1, 2, and 3, respectively. How does this happen? Because all 10-9 (or 10-8) rounds aren't made the same. If you win two 10-9 squeakers and I clearly win the other round (but not enough for a 10-8), you win by the 10-Point Must system but I won that fight.

So who else likely "won that fight" but went home with a loss or draw? Here's a sample of some notable fights predicted to have a different winner if they were scored as a whole.

Analytics_-_one_big_round_-_past_fights_that_change_winners_medium

Let's see how history might have changed. B.J. Penn doesn't lose his lightweight title to Frankie Edgar. The Korean Zombie gets a win over Leonard Garcia and we don't get to see the Twister because there's no rematch. Lyoto Machida helps send Rampage's career packing earlier. Gray Maynard becomes lightweight champion (yes, I see the logical inconsistency with the Penn-Edgar result above but work with me here). Ian McCall challenges Joseph Benavidez for the inaugural flyweight championship. Frankie Edgar gets his belt back from Bendo in the rematch. And, perhaps most importantly for UFC fans this weekend, Johny Hendricks gets set to make his first welterweight title defense tomorrow night.

We already knew the scoring system in MMA matters. What we couldn't be sure about is how much it matters. The analytics tell us that anywhere from 5.4 - 14.4 percent of MMA decisions would be different if fights were scored as a whole instead of round-by-round. I'd put the truth towards the lower end of the range for methodological reasons and because in real life judges may find it difficult to psychologically separate themselves from round-by-round thinking - they're even reminded of it in Andy Foster's system. The analytics give us a baseline to better understand how many fights end with one person being the more dominant fighter but going home with a loss because of the scoring system. Now it's up to the "should" pontificators to take over.

Are there any a fight questions you'd like analyzed? Send them to paul.gift@pepperdine.edu or @MMAanalytics on Twitter. Fight data provided by FightMetric.

X
Log In Sign Up

forgot?
Log In Sign Up

Forgot password?

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

Forgot password?

Try another email?

Almost done,

By becoming a registered user, you are also agreeing to our Terms and confirming that you have read our Privacy Policy.

Join Bloody Elbow

You must be a member of Bloody Elbow to participate.

We have our own Community Guidelines at Bloody Elbow. You should read them.

Join Bloody Elbow

You must be a member of Bloody Elbow to participate.

We have our own Community Guidelines at Bloody Elbow. You should read them.

Spinner.vc97ec6e

Authenticating

Great!

Choose an available username to complete sign up.

In order to provide our users with a better overall experience, we ask for more information from Facebook when using it to login so that we can learn more about our audience and provide you with the best possible experience. We do not store specific user data and the sharing of it is not required to login with Facebook.

tracking_pixel_5349_tracker