Anastasia Zakharova vs Marie Bouzkova prediction time: the two meet in Ningbo on Wednesday, October 21, 2026, in a 1:00 AM ET start, and on paper this should not be close. Bouzkova is the No. 26 player in the world with a career high of 21; Zakharova is ranked around 110 and is only in this bracket because No. 2 seed Iga Swiatek withdrew with a left-thigh injury and handed her a lucky-loser slot, bye included.
The market has Bouzkova at -370, and the model's own numbers back that lean. But this is a PASS on our official card, no action, and the reasons for that sit alongside the reasons the model likes Bouzkova in the first place.
Why the model leans Bouzkova
Strip away the ranking gap and the recent form still points the same direction. Over the last three months on hard courts, Bouzkova is 9-6 against a notably tougher slate, that includes Jovic, Birrell, and tight three-setters against Swiatek and Zheng, while Zakharova has gone 4-7 over that same window and 2-4 in the last month. Her double faults have climbed from 4.4% of service points to 8.8%, a rate that has doubled as her form has slid.
The serve numbers follow the same pattern. Bouzkova's three-month hard-court line works out to roughly 60% of service points won, against Zakharova's 54%. Zoom out to 24 months and Bouzkova is 39-29 against a higher average opponent rating (p57) than Zakharova's 36-34 against p35 opposition. None of that is a blowout case, but it is a consistent one across every window the model checked.
The lucky-loser wrinkle
Zakharova did not win her way into this match through a normal draw. When Swiatek pulled out, Zakharova was elevated as a lucky loser and given a first-round bye, so she arrives at this round having played no actual main-draw tennis here. Her only competitive hard-court matches since the horizon came in qualifying: a clean win over Yuliia Starodubtseva (85% first serves in, 74% won behind them), followed immediately by a 3-6 3-6 loss to Polina Kudermetova in which she served 62% first-in, was broken five times on 12 break points, and committed seven double faults.
That is a wide range of outcomes from one week, and it is the only current form the model has on her. The clean version of Zakharova and the version that just got broken five times are both real possibilities walking into Wednesday.
The case against
Here is the part that keeps this off the card. Zakharova's best documented hard-court split is specifically against mid-level servers who look like Bouzkova: 75% hold, 39% return games won, an 8-3 record. The mechanism is her first serve. When it lands at the rate she showed in that Starodubtseva qualifying match, 85% in, 74% won behind it, it keeps her own shaky second serve out of play and drags return games into long rallies.
That style finds a real opening against Bouzkova, whose own second serve has won just 42% of points over the last month. It is also the level of opponent behind Zakharova's two best 2026 wins, over Pliskova and Kalinskaya, so the pattern is not a one-off. None of this flips the pick. It is simply enough uncertainty, layered on top of a lucky loser with almost no reliable current-event form and a price that already has Bouzkova as a heavy favorite, that the model is not turning this into an official play.
The verdict
To be clear: this is a PASS, not a bet we are making, and that is a statement about where the numbers landed, not a knock on the case for Bouzkova. At -370, the market is already pricing in most of what the model likes about her, fresher form, a tougher recent schedule, and a serve edge over the last three months. What keeps this off the card is not stated here, only that the combination of factors did not clear the bar for an official action. If you are tracking this one anyway, kickoff is 1:00 AM ET Wednesday, October 21, 2026, and Bouzkova is the side the model would lean toward if it were playing it.
- Event · Ningbo · WTA (2026) · Wednesday, October 21, 2026 · 1:00 AM ET
- Market · Marie Bouzkova -370 — 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.

