Our Roman Safiullin vs Andrey Rublev prediction for Friday, October 2, 2026 (10:00 PM ET) in Beijing starts with a market that trusts Rublev. Safiullin is priced at +120 to win outright, and on paper that number looks soft next to the week he's having. He's beaten a top-7 player in straight sets and climbed the rankings in real time.
But the model's read on this one is a PASS. That's an official no-action call, not a quiet lean dressed up as a pick. No side of this match gets backed today, and the reasoning below explains the tension in the matchup rather than a verdict the model is willing to stake anything on.
Why Safiullin's week looks better than his price
Safiullin arrives off a Beijing win over Flavio Cobolli, a top-7 player, 6-3 6-2, and the underlying numbers back it up: 80% of his first-serve points won, nine holds out of nine service games, and five of five break points saved. Across his four matches this stretch with full stats available, against a field that included Cobolli and a near-win over world No. 6 Daniil Medvedev, he won 52.8% of total points. That's a tougher field than the one Rublev just played through.
There's also a head-to-head data point worth flagging, carefully: Safiullin beat Rublev at Wimbledon three months ago, out-pointing him 176-164. That came on grass, a different surface from Friday's Beijing hard court, so it reads as a sign Safiullin can compete with Rublev rather than a blueprint for how Friday goes. It doesn't carry the pick on its own.
The case against: Rublev's return game finds the weak spot
Here's the problem the model can't get past. Rublev's hard-court game this year holds serve 85% of the time and wins 22% of his return games and 38% of return points, numbers built for grinding a match into a tight third set or tiebreak. Safiullin's second serve has been landing at just 55%, and he hit four double faults in his three-set win over Quentin Halys in Hangzhou. That's exactly the kind of crack Rublev's return game is built to find.
And when Safiullin's matches go the distance, his recent loss list reads like a warning label: deciding-set or tiebreak defeats to Grenier, Mpetshi Perricard twice, Galarneau and Medjedovic. If Friday's match tightens up the way Rublev's return numbers suggest it could, Safiullin has been on the wrong end of that scenario more than once lately.
Workload and form, nearly a wash
Fatigue doesn't clearly favor either guy enough to lean on. Rublev has played five singles matches worth about 9.6 hours plus a doubles loss in the last seven to eight days, with only one day off between his Hangzhou final and Beijing. Safiullin's five singles matches this stretch ran roughly 6.5 hours over eight days, with two days off built in, including today. Lighter load, slightly fresher legs.
But the point totals cut the other way. Rublev won 53.3% of total points this stretch against a field topped by No. 55 Tallon Griekspoor, a match he survived 109-107 in a final-set tiebreak. Safiullin's 52.8% came against the harder field described above. Those two numbers sit close enough to call it a wash rather than an edge for either side.
The verdict
Add it up and this is a genuinely tight read: Safiullin's form and schedule look better, Rublev's return game and Safiullin's deciding-set history say the match could tip the other way right when it matters most. That's exactly the kind of matchup where the model passes rather than guesses. This is not a bet the desk is making on Friday, at +120 or any other number, and that's true regardless of which side looks more appealing on the surface.
- Event · Beijing · ATP (2026) · Friday, October 2, 2026 · 10:00 PM ET
- Market · Roman Safiullin +120 — recorded when the pick was posted
- Model · T Money, Botcappers' AI Tennis model — how it is built and what it does not know
- Record · every pick is timestamped before the event and graded in public, win or lose
- Statistics · the model's own match/game database, current to the posting time above. Figures a reader cannot open here are internal to that database, not published third-party data.

