Paul Gift returns to crunch the fight numbers and estimate win probabilities for tomorrow’s UFC 175 card and Penn vs. Edgar 3 on Sunday.
It's UFC Fight Week!!! Ronda Rousey's busy getting mobbed, Luke Barnatt's cruising around Cosmo in a suit jacket and shorts and Justin Scoggins is finally on a main card! Thanks to the UFC for listening to this tweet and then one-upping it.
The UFC pool party is going on right at this moment and I'm tucked away in my hotel room typing. Man, I love you guys. As always, see the notes at the bottom for precise definitions of the statistics employed and read the introductory article for an explanation/reminder of how this works.
Chris Weidman (68.9%) over Lyoto Machida (31.1%)
We start with the fight we're all waiting to see - Weidman's first test post-Silva. Since entering the UFC, Weidman's been on a tear with his only somewhat close fight coming against Damian Maia back at UFC on FOX 2. The model had Weidman over Silva last December when I first set the whole thing up. Let's see why it's picking him now.
We all know Machida's Achilles heel of sometimes being too inactive. At distance, he throws 17.6 power strikes per five minutes (P5M), 35 percent below the middleweight average of 27.2 (keep in mind a good chunk of Machida's data is at light heavyweight). Due to his counterstriking style, Machida's allergic to throwing head jabs (4.3 P5M) while Weidman has a very jab-centric distance striking game (21.8 P5M). Machida's more accurate than Weidman in every key distance striking area except power to the legs, while Weidman throws substantially more volume than Machida to the head where the majority of strikes take place.
Our Bloody Elbow striking specialist, Connor Ruebusch, discusses attacking Weidman's body and legs in a great piece detailing the holes in Weidman's game, and he statistically appears to have some vulnerabilities there. At distance, opponents land 63 percent of their power shots to Weidman's body and 87 percent to his legs, both better than average, and these shots are attempted at a higher than average rate.
Perhaps surprisingly, Weidman and Machida both have the same knockdown percentage. Each knocks opponents down with 9 percent of their landed power head strikes while standing, three times the middleweight average. Weidman's knockdown rate is a little better than Machida's, but his real advantage is in busting up your face. Weidman busts up a face 27 percent of every five minutes he's not on his back compared to 8 percent for Machida and an 11 percent average. He also has three times the damage percentage of Machida and has never been damaged or knocked down himself.
Machida's only been taken down by one attempt out of 23 at distance (4%) but Weidman's takedown offense at distance is also incredible. He attempts 2.2 P5M and lands 67 percent, 2 ½ times the middleweight average.
In 5:37 of clinch time in his UFC career, Chris Weidman has never spent a single second being controlled against the cage.
Neither of these guys spends much time in the clinch (about 30 seconds P5M each) but for different reasons. Weidman often has control and he's taking your ass down while Machida's sometimes taking you down and sometimes breaking away. Here's a fun stat for you. In 5:37 of clinch time in his UFC career, Chris Weidman has never spent a single second being controlled against the cage. Machida, on the other hand, splits his clinch time almost evenly between controlling, being controlled, and being off the cage.
Machida's takedown defense is better than average in the clinch, but he's much more vulnerable here than at distance. Opponents complete 33 percent and 25 percent of their respective upper body and lower body takedown attempts on Machida. Meanwhile, Weidman completes 100 percent and 57 percent of his upper and lower body takedown attempts, and his volume of each is 170 percent and 59 percent more than average.
Here's another interesting stat. Chris Weidman has never been taken down at distance or in the clinch, and only one opponent has ever even tried to take him down - Damian Maia made seven takedown attempts and landed zero.
On the ground, Machida has control 82 percent of the time but he spends less than a minute there P5M, about half the middleweight average. This probably signals that he gets a lot of his takedowns towards the end of the round. Weidman has control 98 percent of the time and has only been controlled for three seconds out of 23:32 career ground fighting time. Machida stands up 4.8 times P5M being controlled (over twice as much as average) and Weidman is about average at having opponents stand up on him.
Both go to town with power shots to the head when in control and both make submission attempts at about an average rate and mix up their subs well between chokes and locks. Weidman's never had to defend a legit submission attempt and Machida's defended four of five, with his lone sub coming at the hands of Jonny Bones.
Our main event of the weekend is a battle here between Weidman's power, activity, physical dominance and control against Machida's technical accuracy, sometimes power, inactivity and craftiness. The model likes Weidman as do I.
Ronda Rousey (76.8%) over Alexis Davis (23.2%)
FYI, I don't have Invicta data, so this write up is based on Rousey and Davis' production in Strikeforce and the UFC.
Ronda Rousey is an insane stat fiend. Let me say it again. Ronda Rousey is an insane stat fiend! We're constantly reminded that she's only had one fight make it out of the first round, but her production P5M is just crazy. Dare I pull a Spaceballs and call it ludicrous?!? Statistically, her closest fight was against Sarah Kaufmann and that's only because it's hard to separate yourself statistically in 54 seconds (I don't have complete data for her Budd and D'Alelio bouts).
Rousey has a damage rate over five times the women's bantamweight average and infinitely more than Davis who has never busted up a face. Her bout closeness measure is a 20 while Davis' is a ginormous 81 (0 = in blowouts, 100 = in very close fights, women's average = 46). Rousey's dominated while Davis has tended to edge by or have tough battles before getting a finish.
Rousey spends 27 seconds at distance P5M compared to a 1:54 average and Davis' 2:10. She attempts 72.9 power strikes P5M to Davis' 35.9 and lands at a remarkably good clip of 31 percent to Davis' 36 percent. She strikes to the head a lot, the body twice as much as Davis, and Rousey has never attempted a single power leg strike in her documented career.
The flip side is Rousey's opponents engage in this madness and attempt 115.6 power strikes P5M and land 33 percent. Yet Rousey's never been knocked down and has a much higher than average knockdown rate while Davis' knockdown rate is zero. Davis has never been taken down at distance, but Rousey's never attempted one. She does her damage in the clinch.
Davis is incredibly active in the clinch. She throws 47.3 power strikes P5M, lands at a slightly below average rate of 65 percent and mixes things up nicely between the head, body and legs (24.3, 14.4 and 8.5, respectively). Rousey's also active and she's even more evenly distributed in power strikes at 15.2, 12.1 and 12.1 to the head, body and legs while you know she's looking for and setting up a takedown.
Rousey attempts 6.1 and 8.1 upper and lower body takedown attempts P5M (1,120% and 200% more than average), landing 67 and 88 percent, respectively. How's Davis' clinch takedown defense? About average for lower body and non-existent for upper body, but with a small sample (her opponents are 1-for-1).
Rousey has been vulnerable to lower body takedown attempts and Davis has been about average at it but, if you get Rousey on the ground, good luck keeping her there. She gets 2.4 standups P5M of being controlled on the ground (compared to 0.7 for Davis and 1.3 average) and 1.2 sweeps (1.4 for Davis and 0.5 average). However, Davis has been very effective at keeping opponents on the ground (0.3 standups and 0.3 sweeps P5M of Davis' control).
Rousey's submission rate is ridiculous. She attempts 3.0 submissions per five modified ground minutes (see definition below) to Davis' 1.0 and a 1.5 average, completing 57 percent to Davis' zero and a 21 percent average. Davis has never completed a submission (in Strikeforce or the UFC) and she's also never been sub'd. However, no one's ever tried to submit her. Something tells me that's changing tomorrow night.
While being an excellent grappler, Davis' statistical fight game is striking at distance and in the clinch. She rarely goes for takedowns other than lower bodies in the clinch and she's about average defending them in the clinch. When on the ground, the grappler comes out as she usually has control and often with half guard or better. She's not a power striker (even though one of her four wins is by TKO) and hasn't been an effective submission artist - except for her two Invicta bouts. Her statistical fight game has essentially been to have battles wherever the fight may go: distance, clinch or on the ground.
We'll see how well it works out tomorrow night.
Matt Mitrione (60.0%) over Stefan Struve (40.0%)
I wanted to provide a little more detail for this fight but I'm running out of time. In addition to having to deal with a heart condition and a broken jaw, Struve's been out of action since March of 2013. There seems to be a negative statistical effect on fighters who are out for a year or more (although it might not be causal and could be spurious to something else). Mitrione's game is striking and Struve likes getting to the ground where he has a 4.7 submission attempt rate compared to a 1.3 average. He completes at about an average rate of 21 percent, but Mitrione has been submitted in one of the two attempts against him.
Urijah Faber (54.1%) over Alex Caceres (45.9%)
At +700 as of this writing, a bet on Caceres would normally be a no brainer long-term edge bet. But the model doesn't do very well in the scenarios where a predicted loser seems to have a good edge. As mentioned in the introductory piece, the model has had around a 59 percent return on units bet in about 5 1/3 years since 2009. But it has lost around 12.5 units in the 32 situations in which a fighter seems to have a good edge but is the predicted loser. Excluding those situations, the model's return has been 71.6 percent (this does not mean it will definitely hold in the future). No bet here.
Kenny Robertson (69.2%) over Ildemar Alcantara (30.8%)
At +130 as of this writing, there probably will be a Robertson bet here, but I think I'm going to hold off confirming these until close to the last minute on Twitter.
Frankie Edgar (76.6%) over BJ Penn (23.4%)
My judging model gives Penn four out of five rounds in their first encounter at UFC 112 for a nice highway robbery at the hands of the judges. Thus began B.J.'s slide. It gives Edgar all five rounds in their rematch at UFC 118 for a clear and convincing win. One draw, two thorough defeats and a retirement later and we've got Penn vs. Edgar 3 in what I'm calling the rubber match.
Here's a stat you may not have known. B.J. Penn has won only one decision in his UFC career - against Matt Serra all the way back at UFC 39 - and lost six. He's going against a man with 11 decisions in 14 documented bouts.
I want to share some meaningful stats, but Penn's decline has been so hard that it's a little sad and I have no idea how relevant the information is with his time off and competing at featherweight for the first time. We shouldn't need estimated win probabilities for this one - I'd be amazed if B.J. wins.
The great thing about being an op-ed and special piece writer is my work's done for the rest of Fight Week. Off to enjoy Vegas and all the action!
Notes: Strike attempts are for an entire five minute round in each position (P5M) and are categorized as jab or power. A jab is just a non-power strike. Strikes are documented based on where they land or are targeted (head, body, legs), not the type that is thrown (punch, elbow, kick, knee). Visible damage rate is per five minutes the fighter is not on his back. It's hard to bust up someone's face while lying on your back. Damage percent is per power head strike landed. Knockdown rate is per five minutes at distance or in the clinch off the cage. Knockdown percent is per power head strike landed while standing. It's really hard to knock someone down if they're already on the ground. Clinch control is having the opponent pressed against the cage. Ground control is having top position. Submission attempts are per five minutes of ground control minus time spent in the opponent's guard plus time spent with the opponent in guard.
Are there any fight questions you'd like analyzed? Send them to firstname.lastname@example.org or @MMAanalytics on Twitter. Fight data provided by FightMetric.