Charles Radtke is the pick at 73.4% against Stephen Thompson's 26.6%, and it is almost entirely an age and activity call rather than a rating one — Thompson is rated higher, 1962 to 1842, and the rating difference is listed as a factor favouring him. Against that, Thompson is 43 on fight night to Radtke's 36, has not fought since July 2025, and has lost four of his last five. The stylistic edge the model names is control: Radtke holds 229 seconds per 15 minutes against Thompson's 158, in a division averaging 189, and Thompson concedes 255 against the division's 248. Radtke is the lower-volume striker, 45.5 landed per 15 minutes against Thompson's 52.2, so this is not a case of the pick winning the striking exchanges. A decision is the likeliest ending at 49.5%, and the 72.9% is the largest gap between model and reputation on the card.

Thompson remains, on the numbers, a better striker than the man opposite him. He lands 52.2 significant strikes per 15 minutes against a welterweight average of 48.1, 41.7 of them at distance against 35.7, and absorbs 52.6 against the division's 54.3. The holes are where they have always been and they are now deep: 158 seconds of control per 15 minutes against the division's 189, 255 conceded against 248, 3.1 takedown attempts against 4.2, and 0.34 submission attempts against 0.49. He is 43 and 15.7 months out, and his last five are a split loss to Gabriel Bonfim, a third-round knockout loss to Joaquin Buckley, a submission loss to Shavkat Rakhmonov, the Kevin Holland knockout, and a decision loss to Belal Muhammad. The model is not reading his striking as declined; it is reading his grappling as a door and his age as a reason it opens more easily. Radtke is unspectacular and does the two things that exploit that. He controls 229 seconds per 15 minutes against the division's 189 — the clearest single edge in this fight — attempts 4.1 takedowns at 0.397 accuracy against division figures of 4.2 and 0.34, and attempts 0.54 submissions against 0.49, with a submission-finish share of 0.103 against 0.04. His striking is below average in volume at 45.5 landed and 32.6 at distance, but efficient at 0.463 accuracy, and his defence is the better of the two at 0.54 of strikes missed against Thompson's 0.508. He is 36, 6.7 months out, and has won three of his last five with a second-round knockout loss to Mike Malott among them. This is a wrestler against a striker with a measured takedown problem, and the model is backing the wrestler by a wide margin. The mechanism is specific rather than vague: 229 seconds of control against a man conceding 255 per 15 minutes, with Thompson attempting a below-average 3.1 takedowns of his own so there is no counter-grappling to speak of. The version where Thompson wins is the one where he keeps it standing, and his distance numbers say that version is comfortable for him. Here is where I think you should argue with the model and where I think it is right anyway: 73.4% on a man rated 120 points lower is a very strong statement, and almost all of it is seven years of age plus a 15.7-month layoff plus a control differential. Two of those three are measured well. The age term was checked across 4,314 walk-forward fights and if anything runs slightly weak, so the one number here I would not discount is the one that looks most like a prejudice.