Fight Lab did the digging. Now it is time to make the calls.
Below are all 13 fights from UFC Shanghai, with the TOE Winner Model alongside Fred’s final prediction and expected method of victory. The model is one part of the process rather than the final answer, and this week there are three fights where the full matchup read takes us in a different direction.
For the deeper analysis of Umar Nurmagomedov vs Song Yadong and four of the most interesting fights on the card, read: UFC Shanghai Fight Lab: Nurmagomedov vs Song.
UFC Shanghai Picks
| Fight | Model Pick (Win %) | Fred’s Pick | Fred’s Method |
|---|---|---|---|
| Umar Nurmagomedov vs Song Yadong | Umar Nurmagomedov (81.3%) | Umar Nurmagomedov | Decision |
| Yan Xiaonan vs Denise Gomes | Denise Gomes (72.4%) | Yan Xiaonan | Decision |
| Aoriqileng vs Kai Asakura | Kai Asakura (78.7%) | Kai Asakura | KO/TKO |
| Alex Perez vs Sumudaerji | Alex Perez (67.2%) | Sumudaerji | Decision (TKO/KO very close) |
| Liu Ce vs Levi Rodrigues Jr. | Liu Ce (71.7%) | Liu Ce | KO/TKO |
| Bilal Hasan vs Nilson Rojas | Bilal Hasan (53.0%) | Bilal Hasan | KO/TKO |
| Namsrai Batbayar vs Andre Lima | Andre Lima (73.6%) | Andre Lima | Decision |
| Rei Tsuruya vs Kevin Borjas | Rei Tsuruya (83.5%) | Rei Tsuruya | Submission |
| Jack Jenkins vs Sean Woodson | Sean Woodson (76.9%) | Sean Woodson | Decision |
| Xiao Long vs Francesco Nuzzi | Xiao Long (52.9%) | Xiao Long | Decision |
| Lawrence Lui vs Hector Santiago | Lawrence Lui (57.7%) | Lawrence Lui | Decision |
| Xiong Jingnan vs Julia Polastri | Julia Polastri (92.2%) | Julia Polastri | Decision |
| Ding Meng vs Cam Nelson | Cam Nelson (53.3%) | Ding Meng | KO/TKO |
Where Fred Goes Against the Model
Three winner predictions split the model and the final card.
Yan Xiaonan over Denise Gomes
This is comfortably the biggest disagreement of the week. The model makes Denise Gomes a strong favourite, and it is not difficult to see what it dislikes about the Yan side of the matchup. Yan is 37, has been inactive for 16 months and is facing an opponent 11 years younger who is improving quickly, carrying more physicality and increasingly willing to mix wrestling into her game.
The final read still comes back to what happens if Gomes cannot consistently turn that pressure into sustained control. Yan remains the cleaner striker, manages range better and has spent years competing against the elite end of the division. Gomes can absolutely make the age gap relevant by forcing a physical, high-tempo fight, but if too much of this stays at kickboxing range, Yan should still be the fighter landing the cleaner work.
The model’s warning is taken seriously. The final call remains YAN XIAONAN BY DECISION.
Sumudaerji over Alex Perez
The model prefers Alex Perez at 67.2%, and there is a very reasonable case behind it. Perez has the more complete MMA game, has fought stronger opposition over a long period and owns an obvious wrestling route against an opponent who has historically been far more comfortable at range.
There is also something here that no set of career averages can fully capture: they have already started this fight once.
Their May meeting ended in a no contest after an accidental low blow, but not before Sumudaerji dropped Perez cleanly with the left hand and showed that his length and straight striking translate directly into this matchup. Perez should wrestle much more aggressively in the rematch, and if he does this becomes a far more difficult fight for Sumudaerji. Even so, the first meeting gave enough matchup-specific evidence to keep the final pick on the other side of the model.
SUMUDAERJI BY DECISION, with a finish very much in play if he finds Perez cleanly again.
Ding Meng over Cam Nelson
This is technically an override, but 53.3% is hardly the model planting a flag for Cam Nelson. It is essentially describing a fight with very little separation between the two.
Nelson brings youth, finishing ability and upside, but there is still a huge gap in professional experience. Ding Meng has been through 45 fights, has 28 career knockouts and has already had to navigate a full UFC contest. Nelson has considerably more to prove at this level.
The winner call remains close. Where the picture becomes clearer is the method. If Ding wins, the evidence points much more strongly towards KO/TKO than a decision. The Method Model makes the stoppage his most likely route, and the market does the same, pricing Ding by KO/TKO shorter than Ding by decision.
That also fits the wider career profile. Ding has spent years proving he can hurt and finish opponents, while Nelson is still an unknown quantity against this level of power and experience.
The model sees a coin flip on the winner. We do too. But if Ding gets his hand raised, KO/TKO looks like the likeliest route.
DING MENG BY TKO/KO
Method Calls Worth Watching
The Winner Model and Fred agree on most of the card, but there are a couple of fights where the route to victory is much more interesting than the winner itself.
Kai Asakura is one. The Method Model is more conservative about the finish and leans towards a decision if Asakura wins, while the Fight Lab read comes back to KO/TKO. Aoriqileng is tough and experienced, but he also absorbs a lot of clean offence and is willing to exchange in exactly the areas where Asakura’s speed and explosive countering become most dangerous. Historical method tendencies are useful, but this particular matchup looks more volatile than the broader numbers suggest.
Rei Tsuruya is the other. The Method Model prefers a decision, while the final pick is SUBMISSION. Tsuruya averages more than five takedowns per 15 minutes and builds his entire offence around repeated contact, scrambles and positional transitions. Borjas has respectable takedown defence, but he has not faced many opponents who are likely to keep asking the same grappling question this often. If Tsuruya gets enough prolonged sequences, the submission opportunities should eventually arrive.
The Rest of the Card
There are a few other predictions where the headline number deserves context.
Julia Polastri carries the highest Winner Model probability on the card at 92.2%, but that is not a number I would treat as literal certainty. Xiong Jingnan has an unusual career profile, a huge amount of high-level experience outside the UFC and only a small amount of UFC-specific data for the model to work with. Polastri remains the pick, particularly given the age gap and her greater familiarity with the current UFC environment, but the raw percentage looks more confident than the fight feels.
Bilal Hasan is almost the opposite case. Bet365 has treated him like one of the strongest favourites on the card, while the model sits at just 53.0%. That caution makes sense. Hasan looks extremely talented and has been finishing people quickly, but both he and Nilson Rojas bring limited high-level data into the UFC. The tape points towards Hasan, particularly because of his speed and finishing ability, but there is nowhere near enough evidence to treat the fight as a formality.
Xiao Long against Francesco Nuzzi also sits very close to 50% in the model. Xiao has the UFC experience and Nuzzi arrives on short notice, which is enough to keep the final pick on Xiao, but it is another fight where the uncertainty should be taken seriously rather than hidden behind a confident-looking prediction.
Final Word
The headline call remains UMAR NURMAGOMEDOV BY DECISION. Song Yadong has enough power, defensive wrestling and five-round experience to make the main event difficult, but Umar has more ways to control the fight and more ways to change the question when one phase is not working.
Across the rest of the card, the most interesting predictions are probably the ones where the numbers do not make the decision for us. Yan Xiaonan and Sumudaerji are the two clearest examples of the final matchup read overriding a strong model preference, while Asakura and Tsuruya show why predicting the winner and predicting how the fight gets there are two different problems.
Thirteen fights. Thirteen calls.
Now we find out which reads survive the cage door closing.
