Moicano vs Nolan predictions graded: model vs odds vs fans

All week before UFC Fight Night: Moicano vs Nolan I printed three opinions about every fight on the card: the sportsbook consensus, the people making picks on our pick'em, and our AI model, which works from fight data and never sees a betting line. All 12 bouts that made it to the Meta Apex on Saturday have now been settled. Four other bookings fell through before fight night, and they are not counted here. So it is time to grade all three, using the numbers each of them had before the first bell.

The short version: our model picked 10 winners out of 12, including Tom Nolan in a main event the market had talked itself out of. The closing favorite won 9 times. The fans also went 10 for 12. And when you score the probabilities instead of just the winners, the order flips: the fans first, the market second, the model third.

The scoreboard

Every number below is the last one each side had before the prelims began at 5 p.m. ET. Model is our model's win probability from its final pre-fight run, at 15:01 UTC on fight day. Line is the 18-book consensus, shown as the favorite's price and turned into a probability with the bookmakers' margin removed; on our board it last moved in the early hours of Thursday, October 29 (UTC), and stood there until the first bell. Fans is the count of pick'em entries made before the first bell.

A note on the model column before anyone checks it against the full-card preview. That piece quoted an earlier run. The final run came in closer to even on several favorites: Lucia Szabová from 87 percent to 70.7, Francis Marshall from 63 to 57.3, Zach Reese from 72 to 66.3, Rodrigo Sezinando from 59 to 54.6, Piera Rodriguez from 60 to 54.8. No pick changed sides. I'm grading the final run because it is the last thing the model said.

Counting winners: the model missed Jean-Paul Lebosnoyani and Dakota Hope. The closing favorite missed Renato Moicano, Marshall and JeongYeong Lee. The fans missed Moicano and Marshall.

Counting winners treats a 50.8 percent call the same as an 84 percent one, so I also scored the probabilities. The Brier score is the average squared gap between the forecast and what happened; lower is better, and a forecaster who said 50 percent on every fight scores 0.250. The fans scored 0.143 (reading their vote share as a probability), the closing line 0.164, our model 0.178. Log loss, which punishes a confident miss harder, ranks the line ahead of the model too, 0.506 to 0.537, against 0.693 for the coin.

Two caveats on the fans. Only 86 people made a pick on this card before the first bell, and the bouts drew between 17 and 66 votes each. And a 21-0 vote is not really a 100 percent forecast. It does show how sure the room was, and on Saturday the room was sure about the right fights.

The main event: the model stood alone and won

Photo: Show Me The Money / CC BY 3.0

On Tuesday I wrote that the line flipped and our model didn't. Nolan opened as the -130 favorite on October 23, the books turned the fight around over the weekend, and by the close Renato Moicano was -122 and Nolan +101. The fans went further than the market, 41 to 25 for Moicano. Our model gave Nolan 60 percent before the move and 60 percent after it, mostly because Moicano is 37 and Nolan is 26.

For a round and a half, Moicano fought the fight the market was paying for. He took Nolan down in the first round, held him for 4:07 of it and tried a submission. He added another 3:02 of control in the second. Then Nolan knocked him down in that second round, outlanded him 25 to 4 in it, and landed 40 significant strikes to Moicano's 3 in the third before the stoppage at 1:28. The totals read 69 to 9. Nolan earned a Performance of the Night bonus, and he still has never fought past the third round, because he ended this one inside it.

That piece counted 14 UFC lightweight bouts since 2015 between a fighter aged 37 or older and an opponent under 28, and found the older man had won three. Make it 3 and 12. I picked Nolan by decision, so I'll take the winner and give back the method.

Where the model was wrong, and how wrong

The model's two misses were not the same kind of miss.

Lebosnoyani vs Hasanov was the widest gap on the card: 71.4 percent for Jean-Paul Lebosnoyani on the line, 49 percent in our model. That piece explained where the model's number came from: one fight, in which Farman Hasanov landed 17 of 27 takedowns against Eric Nolan. On Saturday he landed 2 of 6 and spent 36 seconds in control. Lebosnoyani outlanded him 104 to 53 before the doctor stopped it at 2:11 of the third round. The market read 15 minutes of tape as a single sample. The model read it as a style, and it was wrong.

The other miss was a model that was less wrong than the market. The model gave Marshall 57.3 percent; the line gave him 66.8; the fans 76.5. Hope won. On the probability score, that bout cost the model less than it cost the books.

Then there are the two fights the preview flagged as the model's boldest away from the main event. I wrote that if Nick Klein and Talita Alencar both won, the model would have read two fights better than 18 sportsbooks. Neither did. Joseph Kropschot submitted Klein at 1:25 of the second round, and Piera Rodriguez won 30-27 on all three cards. The model still had the right winner in both, at 52.9 and 54.8 percent, so they count as hits above. But the market was at 73.7 and 71.6, and it was right to be that sure.

Those three fights, Lebosnoyani, Kropschot and Rodriguez, are the whole gap between the model and the line. In the other nine bouts the model scored 0.161 and the market 0.193. In those three, the model sat near a coin flip while the market said 71 to 74 percent, and the market was right every time.

Where the market missed

The closing favorite lost three times, and the model had the right side in two of them.

The first was the main event. The second was Julian Erosa, a +140 underdog against JeongYeong Lee. Our model gave him 50.8 percent, as thin a lean as it makes, and the fans had him 11 to 7. Erosa dropped Lee and finished him at 1:31 of the first round, and took one of the night's four Performance of the Night bonuses with it. Szabová and Kropschot took two of the others.

The third was the one nobody saw.

Marshall vs Hope: one round, three judges

Every forecaster on the board had Francis Marshall. Dakota Hope beat him by split decision, and the scorecards show exactly where.

All three judges gave Marshall the first round, in which he landed his only takedown and held Hope for 1:27, even though Hope outlanded him 15 to 11. All three gave Hope the third, in which Hope landed 17 significant strikes to Marshall's 3 and spent 4:00 of the round in control. The second round decided the fight. Derek Cleary scored it for Marshall, 29-28 overall. Eric Colón and Sal D'Amato scored it for Hope, 29-28 the other way.

It was a close round by any count. Marshall landed 11 of 27 significant strikes and Hope 12 of 49; neither man completed a takedown, and Hope had 41 seconds of control to Marshall's 7. The preview's line was "Marshall wrestles, Hope throws." Over 15 minutes Marshall went 1 for 5 on takedowns, and Hope finished with 4:41 of control to Marshall's 1:34.

What the line did in the last week

Line movement is the market changing its mind, so it is fair to ask whether it changed its mind in the right direction. On this card it almost always did. Measured from the first sportsbook price on our board to the close, eight bouts moved two points of probability or more. Seven of them moved toward the eventual winner, including Cody Durden, who opened as a +106 underdog and closed a -117 favorite before winning 29-28 on all three cards, and Hope, who drifted from +211 to +188.

The exception was the main event, and it was also the biggest move on the card: about 6.8 points of probability, from Nolan to Moicano, in the wrong direction.

The fans, and the Fan Parlay

The fans never stood alone on this card. They agreed with the model on 10 of 12 fights and with the market on 11, and the two fights they missed (Moicano and Marshall) were misses they shared with somebody. Where the room was loudest, it was right: 20 of 21 on Reese, 20 of 22 on Luana Santos, 21 of 21 on Szabová. Reese won 29-27 on all three cards, Luana 29-28 on all three, and Szabová stopped Tainara Lisboa at 3:20 of the third round.

Saturday morning's Fan Parlay stacked the fans' eight most united picks into one ticket at 12.34x. Seven legs won: Szabová, Reese, Luana Santos, Jacobe Smith, Kropschot, Durden and Lebosnoyani. The eighth was Marshall. The ticket did not cash, and the leg that broke it was the only fight on the card where the model, the market and the fans all agreed and all lost. The Erosa vote that ticket left off, because it was already full, came in. None of this is betting advice; the ticket is a record of what the fans thought.

My own card

In the preview I picked all 12 fights and went 11 for 12, missing only Marshall. I had the method right on five of the 11 winners: decisions for Durden, Luana Santos and Rodriguez, and stoppages for Sezinando and Szabová. Nolan, Smith, Lebosnoyani, Erosa, Reese and Kropschot all won some other way than the one I named.

Before that reads like a victory lap, remember how those picks were made. I wrote them with all three numbers open. I took the model on the headliner and the market on Lebosnoyani, and that one fight is the entire difference between my 11 and the model's 10. That is not a fourth forecaster. It's the second-guessing of the other three, and last week at UFC 333 the same habit finished behind all of them.

What this card says about the forecasters themselves is less flattering to everybody. The model was right to ignore the market's flip on Nolan, and wrong to hold three prelim fights near 50 percent when the market was sure. The market read the undercard best and moved the wrong way on the one fight everybody watched. The fans, 86 of them, read it best of all. Every result is on the event page, and the pick'em standings for the card are on its leaderboard.

Four people picked Dakota Hope before the first bell. The model, 18 sportsbooks and I were not among them.