The FightIQ model picked 10 of 12 winners on the UFC Abu Dhabi card last Saturday, an 83.3% hit rate for the night.

Both of our flagged picks won, and both won by the method we called.

Sam Patterson, flagged at 75%, won by KO/TKO in the second round. Our top method read was KO/TKO. Nurullo Aliev, flagged at 68%, won a unanimous decision after three rounds. Our method read was decision.

The main event was not a flagged pick. We had Magomed Ankalaev at 72%, a strong read that sat below our flagged bar. He won by KO/TKO in the fifth round, and KO/TKO was our top method read. There was no LOCK on this card.

Where the losses came from

Our two misses came off the bottom of the confidence range.

We picked Steve Erceg at about 51%, a coin flip and our least confident call of the card. Ramazan Temirov knocked him out in the first round.

We picked Dustin Jacoby at about 58%, also low confidence. Muhammad Said stopped him in the second.

So the ordering held on this card. The picks we flagged won and the picks we barely believed in lost.

That is the pattern we want. It is also one card and 12 fights, and one card proves nothing on its own. A model with good ordering will still throw up cards where the flagged picks lose and the coin flips land. The ordering is only worth trusting across a lot of cards.

The weak spot: method

We went 5 of 12 on method. We picked the right winner far more often than the right finish, and that is the clear failure of the night.

Submissions did most of the damage. Valter Walker won by first-round submission where we leaned KO/TKO. Axel Sola won by first-round submission where we leaned decision. Abdul Hussein won by submission in the third where we leaned decision.

For a reader who treats the method call as part of the pick, getting the winner and missing the finish is a real miss. We are logging it that way.

On the night, the winner column was strong and the method column was poor. Both go into the record as they happened.