UFC Shanghai parlay: our model says one of these loses
Two model-priced UFC Shanghai parlays with every leg explained, plus why our own math says the six-leg favorites ticket is a losing bet.
Our model just priced two parlays for UFC Shanghai, and by its own arithmetic one of them is a losing bet. I am going to lead with that, because it is the part every other picks column leaves out.
The six-leg favorites ticket pays 10.16x. The model puts its combined probability at 8.4 percent. Multiply those together and you get 0.85, which means that for every dollar you put on it, the model expects 85 cents back. The two-leg underdog ticket pays 14.63x against a combined probability of 23.0 percent, which comes out at 3.36. Same card, same model, same night, and a four-fold difference in what the two slips are actually worth.
Both are now on the event page. I would rather publish the number that embarrasses the first one than quietly round it off.
The favorites ticket, and what it costs you
Six market favorites, all of them sides the model also likes. Nothing here is a contrarian pick and that is the point: this is the ticket a reasonable person builds after reading the card.
| Leg | Price | Model | Books |
|---|---|---|---|
| Kai Asakura over Aoriqileng | -443 | 74.0% | 78.5% |
| André Lima over Namsrai Batbayar | -269 | 72.3% | 70.2% |
| Sean Woodson over Jack Jenkins | -147 | 58.2% | 57.4% |
| Xiao Long over Francesco Nuzzi | -176 | 64.8% | 61.4% |
| Lawrence Lui over Hector Santiago | -283 | 72.4% | 71.3% |
| Ding Meng over Cameron Nelson | -144 | 57.2% | 56.8% |
Four of the six carry a positive edge, meaning the model rates that fighter higher than the devigged market does. Xiao Long is the widest at 3.5 points, bought almost entirely on the other corner's inactivity: Nuzzi has not completed a fight in 505 days, his last outing was a no contest at Cage Warriors 204, and his three-fight rating trend is the steepest fall on the card. Lima and Lui are the model agreeing with the books to within two points, and it likes both because their opponents are debutants it can see nothing on.
Ding Meng is the leg I would talk you out of if you asked. He is 1-4 in the UFC with no quality wins and the lowest average opponent rating on the card, and the model's 57 percent for him rests mostly on him having 26 career fights to Cameron Nelson's eight. Two of the model's other leading factors, age and career win rate, both point at Nelson. What keeps it on the slip is the money: this opened with Nelson a slight favorite at -118 and now sits at -144 the other way.
Why stacking favorites is a tax
Photo: ホタテマン0010 / CC BY 4.0
Here is the mechanism, and it is not subtle once you see it. Every price a sportsbook posts carries a margin, a few percent shaved off the true odds in the house's favor. On a single bet you pay that margin once. On a six-leg parlay you pay it six times, compounded, because the six prices are multiplied together and so are the six little haircuts.
Asakura at -443 is the clearest illustration. The model likes him at 74 percent and the books have him at 78.5 percent, so on that leg alone the price is already 4.5 points worse than our number. He is still the right side of the fight. Aoriqileng arrives on 91 days' rest off a second-round knockout loss to Cody Haddon, sits 4-5 in the UFC, and owns a late-round dominance figure of -14.8, the worst on the card, so the model reads it as a fight Asakura mainly has to survive early. The pick is fine. The price is not, and the ticket contains five more like it.
That compounding is why the favorites slip lands at 0.85 and the two-leg underdog slip lands at 3.36. It is not that the model dislikes those six fighters. It likes all six. It just does not like them 10.16 times as much as the books are charging, and stacking six short prices is the most reliable way to hand a bookmaker money while feeling clever about it.
The underdog ticket is two legs for a reason
| Leg | Price | Model | Books |
|---|---|---|---|
| Denise Gomes over Yan Xiaonan | +129 | 62.8% | 42.1% |
| Nilson Rojas over Bilal Hasan | +539 | 36.6% | 15.1% |
Gomes is the model's largest disagreement with the market on the entire card, and the heaviest single factor behind it is the age gap: 37.2 against 26.7. She arrives on a four-fight UFC win streak with takedowns at the 91st percentile of her division and ground striking at the 97th. Yan has not fought in 504 days, comes in off a decision loss to Virna Jandiroba, defends 64.9 percent of takedowns and carries 195 career minutes across ten wars to Gomes' 103. The books are pricing a former title challenger's name. The model is pricing a 16-month layoff against a younger fighter whose best route is the one Yan defends worst.
Rojas needs its caveat stated before its case. He is a UFC debutant and Hasan has 45 seconds of tracked UFC time, so neither corner has a real file and the model's 36.6 percent is closer to an admission of ignorance than to a read. Every driver the model can actually name favors Hasan: his rating, seven inches of reach, two years of age. I am not telling you Rojas is the better fighter. The claim is narrower than that. Hasan fought on August 11 and fights again on August 29, an 18-day turnaround after a first-round knockout, Rojas is 8-0 with every win by stoppage, and 85 percent is a lot of certainty about a man the sport has watched for less than a minute. The line has moved against us since it opened, from -500 to -757, so the market has grown more sure, not less.
There is no third leg because there was nothing worth adding. The next-best combinations either dropped the combined probability below 19 percent or leaned on a side the model does not back, and buying multiplier with a leg you do not believe is the same tax described above, paid voluntarily.
The main event is not on either ticket
Photo: Новости ННТ / CC BY 3.0
Umar Nurmagomedov against Song Yadong is the best fight on the card and it appears on neither slip, which surprised me until I looked at the prices. The model has Umar at 74.2 percent. The market has him at 82.2 percent, which is -593. That is an eight-point negative edge and it buys you a multiplier of 1.169, so a leg that costs more than it is worth and barely moves the payout.
The other side is more interesting than the number suggests. Song at +441 carries a positive edge of eight points and the model still only gets him to 25.8 percent, which is not enough to anchor a two-leg ticket next to Gomes without dragging the combined probability down to 16 percent. The common opponents cut against him too. Both men have beaten Deiveson Figueiredo, Song by submission in round two this May and Umar on the cards in January, but the pair also share Cory Sandhagen, whom Umar beat over five rounds and who stopped Song on a doctor's stoppage in 2022.
So the headliner sits out. That is usually the honest answer for a main event the market has already priced correctly, and it is worth saying out loud that "our model has no edge here" is a finding, not a gap in the work.
What the model refused to be sure about
Every fight has to clear four trust rules before it is allowed onto a ticket, and the count this week is unusual. All 13 bouts survived pricing, which is rare, but seven of them came out carrying a warning. Shanghai is a prospect card: five fighters on it have never fought in the UFC at all, and three more have a single appearance.
Five of the seven warnings are the thin-data flag, which fires when a corner has fewer than three UFC fights and means the model is comparing a real file against an empty one. Two are the model-market gap flag, which fires past 15 points of disagreement and which we treat as evidence of our own ignorance rather than the bookmaker's error. Both underdog legs carry a gap flag. The other rules are that a price needs at least three sportsbooks behind it before it counts as a market, and that no probability is allowed outside the 5 to 90 percent band. Nothing on this card hit that clip.
Our style-percentile layer covered both corners in only 2 of the 13 bouts, and the single style rule we trust enough to cite in a pick fired zero times. On a card this young there is not enough tracked film to run it. I would rather write that sentence than manufacture an angle out of nothing, which is what the alternative would have been.
One caveat that actually changes the math
Both combined probabilities above multiply the legs as though each fight were unrelated to the others. They are not. These bouts share a card, a cage, one night's officiating and one set of judges, and the errors in our own model are correlated across them as well: if we are overrating how much a long layoff matters, we are overrating it in more than one fight tonight. Read 8.4 percent and 23.0 percent as ceilings rather than forecasts.
Every number here is our model's opinion against a sportsbook's price, pulled roughly two hours before I wrote this. It is not betting advice, odds move, and this is 18 and over only.
Doors at the Oriental Sports Center in Pudong. If the favorites ticket cashes anyway, I will hear about it, and the 0.85 will still have been the right thing to print.