Gamrot vs Salkilld parlay: the legs our model refused to add
Our model built two Gamrot vs Salkilld tickets, at 9.03x and 15.00x. The math offered a bigger number. Here is every leg we left off, and the rule that removed it.
The search that builds our parlays handed back a five leg ticket paying 10.14x this afternoon. We published one paying 9.03x instead.
That is the whole job, and almost nobody who publishes picks will show you the part where the number goes down. So here is tonight's, in full: the two model tickets on UFC Fight Night: Gamrot vs Salkilld, the rules that built them, and every leg we took off on the way.
What the tickets are
Every UFC card, the model prices all twelve or thirteen bouts, and two tickets go up on the event page before the walkouts. One is built only from market favorites, one only from market underdogs, and they never share a fight. They are timestamped so they can be graded afterwards, and they are, win or lose.
The rules that govern them are written in code, not in the mood of whoever is holding the keyboard:
- Both corners need at least three UFC fights on record.
- At least three sportsbooks must be pricing the bout.
- Probabilities are clipped at 5 and 90 percent, because the model has never demonstrated it can tell 93 from 97.
- Any bout where model and market sit more than 15 points apart is treated as our ignorance, not the book's error.
- Nothing gets added to a ticket that the model does not back, ever, and least of all to make the headline multiplier bigger.
That last one is the rule that cost us 10.14x tonight. Separately, I wrote up where the model and the board disagree fight by fight this morning, which is the diagnostic view. This is the decision view.
Ticket one: favorites, 9.03x
| Leg | Price | Model | Market |
|---|---|---|---|
| Quillan Salkilld over Mateusz Gamrot | -137 | 61.0% | 55.6% |
| Alexia Thainara over Amanda Lemos | -256 | 70.6% | 69.3% |
| Diyar Nurgozhay over Bruno Lopes | -162 | 63.9% | 59.6% |
| Louie Sutherland over Jose Montanha | -202 | 72.8% | 64.5% |
| Miles Johns over Gianni Vazquez | -181 | 62.5% | 62.1% |
Model probability 12.5 percent. Market probability 9.2 percent. Fair price 7.98x against 9.03x offered.
Two of those need their warning flags read out rather than buried. Sutherland faces a UFC debutant, so the 72.8 percent is built off a 6-1 regional record and nothing else; we took it because Sutherland has five UFC appearances and beat Tai Tuivasa on the cards in May, and because the known quantity is the known quantity. Johns is on a three fight losing streak, which is exactly why a man with 142 UFC minutes is only -181 against an opponent with 15, and none of those three losses came by knockout.
The one I want to defend properly is the leg we removed. The probability-optimal ticket included Diego Ferreira at -178, which is what got it to 10.14x. The model prices Ferreira at 58.6 percent while the market prices him at 61.7, so it is a losing proposition before anything else is said about it. He is also 41.6 years old, coming off a second round knockout loss, and 329 days out of the cage. Swapping Alexia Thainara in for him costs price and drops the ticket below the 10x we aim for.
There were exactly two ways back over 10x. Put Ferreira back, or staple on Yadier del Valle at -818, which the model does like at 79.2 percent but which would have dragged the ticket's expected return from 1.13 down to 1.00. A leg that adds a rounding error to your chances and a digit to your headline is padding, and padding is how this genre earns its reputation. We took 9.03x.
Ticket two: underdogs, 15.00x
| Leg | Price | Model | Market |
|---|---|---|---|
| Billy Ray Goff over Ty Miller | +325 | 32.0% | 22.7% |
| Guilherme Pat over Steven Asplund | +253 | 38.2% | 27.3% |
Model probability 12.2 percent. Market probability 6.2 percent. Fair price 8.19x against 15.00x offered.
Miller is -407 on the strength of five minutes of UFC footage, one first round knockout of Adam Fugitt in January. That is the entire sample, and it is why this fight carries a thin-data flag. Goff has 22 UFC minutes, the better striking defense of the two at 59 percent, and four knockout wins of his own. His 434 day layoff is real and is most of why he is +325 rather than +200.
The heavyweight leg is thinner still, flagged on both corners, because Asplund and Pat have two UFC fights each. Pat lands at 62 percent accuracy and absorbs 2.03 significant strikes a minute. Asplund throws far more, 9.70 a minute, and absorbs 6.63 while defending at 50 percent. Neither man has been taken down in the UFC, so this stays standing, which is where volume against efficiency actually gets settled at heavyweight.
Here too the arithmetic wanted something else. Its preferred underdog pair was Ravena Oliveira with GiGi Canuto at 10.57x. Canuto is the debut-versus-debut fight where the model and the board are 18 points apart, which our own rule defines as a hole in what we know rather than a hole in the price. Oliveira is 0-3 in the UFC, coming off a knockout loss 140 days ago, and carries the lowest rating of anyone on this card; the model likes her at 38.3 percent only because it likes Juliana Miller less. Neither is a bet. We replaced both, and the replacement pair happens to price higher anyway.
What five refusals actually cost
Five of tonight's twelve fights ended up on neither ticket: Ferreira against Billy Quarantillo, Darren Elkins against del Valle, Manoel Sousa against Richie Miranda, Miller against Oliveira, and Canuto against Carol Foro. Six of the twelve tripped at least one trust rule. On a card with four UFC debutants, that is the machinery working rather than failing.
It is worth being blunt about what the refusals buy, because it is not certainty. The favorites ticket is 12.5 percent to hit. The underdog ticket is 12.2 percent. Both are far more likely to lose than win, and saying so out loud is not modesty, it is the number. What discipline buys is that when they lose, they lose for reasons we wrote down in advance, and the record stays worth reading.
One caveat matters more than any of the above, and parlay content almost never states it plainly: these legs are not independent. Every fight happens on the same card, in the same cage, with the same officials, on one night. Multiplying five probabilities together assumes five unrelated coin flips, and same-card outcomes correlate for reasons nobody has fully modeled. Read 12.5 percent as a ceiling, not an estimate.
This is a model's opinion, published for the record and graded afterwards, not betting advice. 18 and over.
Both tickets are live on the event page with the per-fight breakdowns behind them, and if you would rather put your own name on the card than ours, the pick'em is open until the first bell. The model does not get to revise once the cage door shuts. That is the point of publishing early.