The headliner was the one our model wanted, and it landed. Mike Malott stopped Gilbert Burns by KO/TKO in the UFC Fight Night main event on April 18, and the model had Malott flagged as a high-confidence pick going in. That call carried the night. Almost nothing else did.

The night in brief

Malott beat Burns by KO/TKO in the main event. The verified result gives the finish without a round, so the round stays out of this recap. What matters for the autopsy is the direction: the model sided with Malott, the less-favored fighter on paper, and Malott delivered the stoppage.

There are no other finishes in the verified record for this card, so this recap does not list undercard results. We do not invent fight narratives, and where a result is not in the facts, it is left out.

The model autopsy

The card grade was 7 of 12, 58%. That is a below-average night by our own standard, and we are not going to dress it up.

The flagged picks, the ones the model singled out at LOCK or high confidence, went 1 of 2. One of those two was the main event. The model picked Malott at high confidence, and Malott won. The other flagged pick missed. With only two flagged calls on the card, a single miss drops the flagged hit rate to 50%, and that is what happened here.

The headliner deserves a closer look, because it is the most revealing line in the grade. The model assigned Malott a 22% win probability and still flagged him as a high-confidence pick to win. Those two numbers sit oddly next to each other. A 22% probability says the model expected Burns to win roughly three times out of four. The high-confidence flag says the model trusted the underlying read enough to put its name on the outcome anyway. On this card, the read was right and the probability was the conservative number. One correct longshot does not validate the gap between the probability and the flag. It is the kind of result we log and revisit rather than celebrate.

So the honest scorecard: model picked Malott, Malott won, headliner correct. Flagged picks 1 of 2. Overall 7 of 12, 58%. The main event saved the flagged column from going 0 for the night. Strip the headliner out and the rest of the card grades worse than the 58% top-line already suggests.

We do not have a method-accuracy figure for this card, so we are not citing one. The win-loss call is what we can stand behind, and on the full slate it was a hair better than a coin flip.

One honest takeaway

A 58% night is below where this model lives over a larger sample. The canonical live 2026 record is 67.9% across all picks and 81.4% on the LOCK and high-confidence tier over 64 graded picks. This card sits under both. The flagged tier here went 1 of 2, which is well off that 81.4% line, but two picks is far too small a sample to read anything into. One bad flagged pick swings the rate 50 points when there are only two on the board.

That is the point of grading every card in public. A strong main-event call and a flat overall night both go in the same ledger. The 81.4% flagged number is earned across dozens of picks, not any single Saturday, and a 58% card is exactly the kind of result that average has to absorb. The model got the hardest call on the card right and still finished under its own bar everywhere else. Both things are true, and we report both.

No betting advice here. Just the grade, the miss in the flagged column, and the one call that worked.