Kayla Harrison is the pick at 61.1% against Amanda Nunes's 38.9%, on ratings of 1954 and 1738. The rating difference is the largest listed factor and the layoff is the largest one against her: Nunes has been out 40.8 months and Harrison 16.9, and the model applies a two-year return penalty to Nunes that it does not apply to Harrison. On the tape the two are strikingly similar — 54.4 and 54.6 significant strikes landed per 15 minutes, 55.5 and 56.1 absorbed, striking defence of 0.544 and 0.543 — so almost nothing in this pick comes from style. What separates them in the model is grappling volume and recency: Harrison holds 275 seconds of control per 15 minutes against Nunes's 249, in a division averaging 229. A decision is the likeliest ending at 48.4%, and the honest caveat is that Nunes's profile describes a fighter from June 2023.

Harrison's profile is a grappler's with a competent striker's numbers attached. Her 275 seconds of control per 15 minutes is well above the division's 229, but it carries an uncertainty of 299 — larger than the value — because three recorded fights cannot pin it down, so read it as a direction and not a quantity. Her striking is slightly above average in volume and clearly above average in defence: 54.4 landed against 52.2, and 0.544 of incoming strikes missed against 0.52. She concedes almost nothing in knockdowns, 0.14 per 15 minutes against a division average of 0.16, and her submission rate of 0.124 is twice the division's 0.06, which is where the Peña and Holm finishes come from. She is 36 and has not fought since June 2025. Nunes's numbers are a champion's and they are three and a half years old. Her striking volume and accuracy match Harrison's almost exactly, she knocks people down more often — 0.81 per 15 minutes against Harrison's 0.69 and the division's 0.72 — and her takedown accuracy of 0.415 is the best figure either woman has, against a division average of 0.35. She also controls well, 249 seconds per 15 minutes, and concedes only 170 against the division's 212. The problem is not the profile, it is its date: 40.8 months out, and her last three fights were all five-round decisions, two of them wins over Peña either side of the submission loss. The model has no mechanism for what three and a half years does beyond the layoff step, and it says so. Two near-identical strikers who both prefer to grapple makes a long fight, and a 48.4% chance of a decision with only a 36.4% chance of a knockout follows. The live question is the clinch and the floor, where Harrison's 13.0 clinch strikes and 275 seconds of control per 15 minutes meet Nunes's 8.0 and 249 — a real but uncertain edge to Harrison, and the one place the fight is likely to be decided rather than contested. Nunes's route is the knockdown rate: she scores more often than Harrison and Harrison's striking defence, good as it is, has never been tested by someone with Nunes's accuracy. What should make you uncomfortable about the 61.1% is how much of it is a layoff term rather than a read: strip the two-year penalty out and this is close to a coin flip between two profiles the model cannot tell apart.