Conditional Probability, Prop Bets, Curtis Blaydes' Absurd Stats, Andrei Arlovski and Durability
Welcome back, and thank you for checking out the MMA Beyond the Numbers newsletter!
Talk about a vindicating weekend for Kamaru Usman supporters. But no, he is not the Welterweight GOAT. And I don’t think that’s my Canadian bias talking, either. Give me another 3-5 title defenses, then we can talk.
Curtis Blaydes fight weeks are always a lot of fun for me. From a fan perspective, I really enjoy watching dominant fighters, even if they’re dominating in a grindy, slow-paced way. Something about completely controlling a fight and leaving your opponent incapable of doing anything except complaining or sighing is very impressive to me.
But he’s also one of the most statistically unique fighters in the UFC. Blaydes’ 59 takedowns landed rank 11th in UFC history, and every single fighter above him is a Welterweight or lighter (the only possible exception being Demian Maia, who had some UFC time at Middleweight of course).
We’ve never seen a UFC Heavyweight wrestle like Blaydes. The second-most takedowns in Heavyweight history are Cain Velasquez’ 34. It’s not going to be long before Blaydes has twice as many takedowns as anyone else in the history of the division.
I know a lot of people don’t find high-volume wrestling exciting, but those are some staggering numbers.
Which leads me to my first thought for this week …
Curtis Blaydes on DraftKings
To start this off, my focus on DraftKings UFC is almost entirely on large-field tournaments. There are some key differences between tournaments and cash games (more on that another time, perhaps), so keep that in mind here.
In tournaments, the score you get from a losing fighter really doesn't matter. You're gunning for first place, and the vast majority of the time a 25-point performance in a loss is just as disqualifying for that first-place finish as a 0-point performance.
With that being the case, it can be useful to look specifically at fighters' fantasy scores just in fights they win.
Under the new 2021 scoring system, Curtis Blaydes has averaged 115.77 fantasy points per fight in his victories. For context, that makes him one of only 16 fighters in UFC history with at least 5 victories and an average over 115.77. That is also 22.4% higher than the all-time UFC average fantasy score in a victory.
Derrick Lewis, on the other hand, averages just 80.18 fantasy points per fight across his 15 victories.
Andrei Arlovski …
… is a peak example of why I largely ignore any conversation about "chin." Yes, I absolutely think some fighters are more durable than others. And some are better at avoiding taking the full force of power punches than others. But so often "chin" is a results-based post-hoc conversation.
Don't forget the discourse in 2009 was that Andrei Arlovski's chin had "been cracked" by Fedor. Or at least, that became the narrative by 2011, after he suffered two more knockout losses. He then went 12 straight fights without getting finished.
Then he was knocked out 3 times in 4 fights. "The punishment is finally catching up with him, he's old and his chin is shot."
Since that stretch, he's been finished once in 11 fights.
This isn't me saying he has a great chin, or that he won't get finished by Aspinall. But rather that the bulk of the conversation around "chin"/durability carries very little water.
If anything, the best way to find betting value as it relates to "chin" is to just fade the public perception. That perception will be priced into the betting markets, and the perception usually isn't going to be very predictive.
More on descriptive vs. inferential stats another time.
New Advanced Stats Subscription Option Coming
Bringing the news to newsletter readers first — starting with UFC 259, I will have an advanced stats report available for every event. And newsletter readers will get a discount when I launch it.
I still haven't nailed down the exact specifics of what this will look like, but rest assured there will be *a lot* more to it than the free stat package I sent out for UFC 258. I'm not just putting something I already do behind a paywall. I will be releasing a lot of metrics I have never previously given access to beyond posting top-10 lists.
Rather than just raw data, this will include a ton of rate stats and metrics - ranging all the way from the simple (takedown accuracy, sig. strike accuracy, control rate) to my advanced stats (strike accuracy vs. expectation, control time dominator rating, my knockdown rate calculation).
You will have the option to purchase the package for individual events, or to have an ongoing subscription.
Don't worry DraftKings players, you will be covered here too.
Curtis Blaydes and Georges St-Pierre
Are the only two fighters in UFC history with takedown accuracy over 55.0% and at least 100 takedown attempts.
Blaydes has 59 takedowns landed through 12 UFC fights. GSP had landed just 30 through his first 12.
Conditional Probability Prop Bets
In last week's newsletter, I mentioned the Julian Marquez by submission prop bet at +900. Amidst the Miley Cyrus drama, that bet hit, so I decided it would be an interesting spot to talk about how I found the value. You can find some thoughts in this Twitter thread, but I'll get a bit more in-depth on the idea of conditional probabilities in betting here.
Marquez was a -165 favorite. Converting that line to an implied probability, that says he was 62.3% to win the fight. The flip side being the implication that he was 37.7% (100% minus 62.3%) to lose. (Note I'm ignoring vig and draws here because they only serve to distract from the concept).
As a general rule of thumb, over the long-term most fighters win about as often as their implied probabilities say. This is how sportsbooks make money. Obviously there are times the lines are wrong (this is how sports bettors make money), but it's an accurate enough trend to be a useful rule-of-thumb.
In this fight, I didn't really have any strong feelings about value on the moneyline, so I started from a place of operating as if it were accurate — that Marquez was really about 62.3% to win the fight.
Now, looking at his submission prop bet, the +900 odds implied that he had a 10% chance to win the fight by decision. Just trying to pick out "does he win by submission 10% of the time" when you don't have an overall win probability established is particularly tricky. But by using the moneyline as a starting point, we can get some additional context.
Taking those two numbers, the probability he wins by submission *conditional on his winning*, is then 16.05%.
For a visual aid, here is a finely-crafted doodle that shows off my high-level art skills.
Marquez can't win by submission unless he first wins the fight. So while it's implied that 10% of *all the possible results* will be "Marquez by Sub", we know that none of those 10% overlap with a Marquez loss.
So we can calculate the conditional probability by taking the 10% sub probability and dividing it by 62.3% win prob. 0.10 / 0.623 = 0.1605 16.0%.
We can do the same for KO/TKO and decision props, and the total will be 100% (again, with the sportsbook's vig in reality this will add up over 100% most of the time, if you'd like to get round numbers you can then calculate how much over 100% it goes and reduce the numbers accordingly).
So now even if we don't have a feel for how often Marquez wins this fight, we can analyze it from another angle — how often is a Marquez victory going to come by submission? By my analysis, that rate was more like 25-30%. That's not to say he wins by submission 1/4 of the time, but that *1/4 of the scenarios in which he wins are submissions*.
So again using that 62.3% implication from the moneyline as a jumping off point, we can calculate the overall implied probability that I handicap that prop bet at, to compare it to the 10%/+900 line.
(for those of you wanting to follow the math, the lower 25% of my range would be calculated: X / 0.623 = 0.25. That gives us X = 0.1558)
The lower range of my prediction (25%) would be an implied probability of 15.58%, and the upper end (30%) would be 18.69% (betting lines of +542 and +435).
So based on my analysis, I saw an edge then of somewhere from 5.58% to 8.69%. Using a conservative estimate and half-Kelly bet sizing (more on the Kelly Criterion and bankroll management another time), this made it a bet I was happy to place 3.1 units on, even as such a big long-shot.
Where I’ll Be this Week
I won’t be podcasting this week, and starting Wednesday I’m going to be more or less off the grid until Saturday.
Instead of doing MMA analysis, I will be very cold.
Thank you for reading! As always, if you have any questions, comments, feedback or topic ideas, I’m always interested to hear them. You can reach out by DM on Twitter/Instagram (@numbersMMA on both) or by replying to this email.