USA's round 5 opponents at the Olympiad: how Germany and Uzbekistan play

Short answer

In round 5 of the 46th Chess Olympiad, on 20 September 2026, the United States play Germany on board 2 of the open section and Uzbekistan on board 1 of the women's. Drawing on 949 classical games broadcast by Lichess over the last three years, this piece lays out each player's repertoire by colour, the lines where they score below their own average, where the engine finds the mistakes — with every quoted move re-measured at depth 24 — and what the verified online accounts show of their blitz and rapid play.

Event: 46th FIDE Chess Olympiad

How this was compiled: Compiled on 19 September 2026, with round 4 finished and round 5 not yet played: it starts on 20 September at 3pm Samarkand time (10:00 UTC). The pairings were checked on chess-results; the official PGN that names each team's line-up is only published when the round starts, so this dossier covers both full squads rather than four boards each. The repertoire uses over-the-board classical games from the last three years broadcast by Lichess, linked to each player by FIDE ID and excluding round 5 itself: 510 games across the five German players (80 to 131 each) and 439 across the five Uzbek players (65 to 116). Chess960 games were excluded from the classical sample starting this round. The error profile was measured with Stockfish 19 at depth 16 over each player's 12 most recent classical games, and every move quoted in the text was re-measured at depth 24 with a dedicated search. The blitz and rapid repertoire comes from Chess.com accounts verified one by one, over the last 90 days, and sits in a section of its own.

The essentials in a few lines

• In round 5, on 20 September 2026, the United States play Germany on board 2 of the open section and host nation Uzbekistan on board 1 of the women's. All four teams arrive on 8 match points.
• Vincent Keymer scores 26.9% in 13 games against 1.d4, with high confidence, but 77 of his 80 games in the sample come from closed elite events. The comparison that survives is with his own 36.2% baseline with Black, drawn from those same tournaments.
• Alexander Donchenko has the two most solid weak lines on the German side, both high confidence and both Indian: 46.7% in 15 games with White after 1.d4 Nf6 2.c4 e6, and 45.8% in 12 with Black in the King's Indian.
• Four of the five Uzbek players open 1.e4 in more than 89% of their White games. The exception, Gulrukhbegim Tokhirjonova, carries the section's sharpest edge: 30.0% in 15 games after 1.d4 Nf6.
• Frederik Svane's 55 centipawns in the endgame is not a collapse: re-measured at depth 24, the three moves the shallow pass flagged come from an already-lost position, and two of them are the engine's own first choice.
• Umida Omonova opened 1.e4 in 30 of her 31 classical White games and plays 1.b3 in the three openings she repeats most online — 53 blitz and rapid games in 90 days. The two sources are never merged into one number.

The round 5 picture

Both American teams arrive on 8 match points after beating Ukraine 3.5-0.5 and Spain 3.5-0.5 in round 4. Both opponents arrive on 8 as well, which is what the Swiss system does: it pairs the tied, not the similar. In the open section Germany leads on the game-point tiebreak, 12.5 to 12; in the women's the United States leads it, 15 to 14.

The women's match is board 1 of the entire section, and it is also the host nation's team. Uzbekistan entered two more squads in the women's event, seeded 43rd and 65th, and they are playing on boards 34 and 41 of this same round — the opponent here is the first team, seeded 13th.

The rating gap runs one way in the open section and the other in the women's. Germany's five average 2663, topped by Vincent Keymer at 2764; the American squad averages higher still. In the women's section the American five outrate the Uzbek five on every board. Neither fact settles anything: on this same day 24 boards separate these four teams from each other.

The sample column below is the part worth reading twice. It says how much of each dossier stands up, and it varies by a factor of two across the ten players — from Umida Omonova's 65 classical games to Alexander Donchenko's 131. The totals are 510 games for Germany and 439 for Uzbekistan, all played over the board in the last three years and broadcast publicly by Lichess.

Why Keymer's 36% is not the number it looks like

The lazy reading of a dossier is that a low percentage means a weakness. The percentages here are not adjusted for opposition strength, and for most players that barely matters, because most players face a mixed field. Keymer is the exception in this round, and his table row is the one that would mislead a reader who skipped the sample column. Norway Chess accounts for 17 of his games, Tata Steel for 13, the Grand Chess Tour Finals for 17 across semifinals and finals, the Sinquefield Cup and the Romania Super Chess Classic for nine each. Three games come from the current Olympiad and four from the Bundesliga. There is essentially no easy game in there.

This is exactly why the dossier compares a player to himself and not to a rating table. His baseline with Black — 36.2% over 40 games — carries the same bias as every line inside it, because it is drawn from the same tournaments. When one first move stands out below that baseline, the elite field has been held constant on both sides of the comparison. Against 1.d4 he scored 26.9% in 13 games, and the dossier marks the confidence as high. Against 1.e4, the most common choice against him with 18 games, he scored 41.7% — above his own baseline.

With White the picture reverses, and the sample is friendlier: 58.8% in 40 games. He is also the least predictable first-mover of the German squad, splitting his White games between 1.Nf3 in 11, 1.e4 in 11 and the rest elsewhere. After 1.e4 he scored 77.3%, so the reply that troubles him least is the one he sees most.

None of this makes 26.9% a plan on its own. It makes it a comparison worth carrying into the round, next to the fact that a 2764 has the White pieces half the time and that a match is decided on four boards, not one.

Germany: two brothers, two very different endgames

Matthias Bluebaum is the most predictable German at move one and the hardest to dent afterwards. He opened 1.d4 in 47 of 54 games with White, and the reply that costs him most is the most obvious one: after 1...Nf6 he scores 52.3% in 22 games, against a 59.3% baseline. With Black he is the only German whose two colours score alike, and the edge against him is thin but well sampled — 1.c4 held him to 50.0% in 12 games, with high confidence.

Alexander Donchenko brings the biggest sample of the round, 131 games, and it buys the two most solid weak lines in the German half of this dossier. With White, after 1.d4 Nf6 2.c4 e6 he scores 46.7% in 15 games against a 61.5% baseline. With Black, the King's Indian — 1.d4 Nf6 2.c4 g6 — gives him 45.8% in 12 against 60.6%. Both are marked high confidence, which in this series is rare enough to be worth naming twice.

Frederik Svane is the squad's widest gap between colours: 68.0% with White across 50 games, 44.3% with Black across 53. He opened 1.e4 in 39 of those 50 White games. Against him with White, the dossier points at 1.c4, which held him to 28.6% in seven games — medium confidence, and the largest drop from baseline in the German squad.

Rasmus Svane is the quiet one, and in preparation that is information rather than a gap. Across 88 games the dossier finds no first move that beats his own baseline with Black, 52.3%, and no weak line with more than low confidence. He opened 1.d4 in 25 of his 44 White games and 1.Nf3 in 13.

Uzbekistan: four of five open 1.e4, and almost always

A squad in which four of five players open the same move in more than 89% of their games is not a coincidence of taste; it is what a school looks like from the outside. It also makes the American preparation cheaper on four of the boards and more expensive on one, which is the opposite of how a scouting report usually reads.

Gulrukhbegim Tokhirjonova is the exception and the most exposed player in the section. She is the one Uzbek here who played for another federation — she represented the United States before returning to the Uzbek squad, and the official PGN lists her under Uzbekistan in all four rounds so far. Her 1.d4 gives up 30.0% in 15 games after 1...Nf6, against a 51.2% White baseline, with high confidence. That is a 21-point gap on a sample large enough to mean it.

Guldona Karimova has the largest sample of the Uzbek five, 116 games, and the flattest profile: 45.6% with White and 46.6% with Black, the only player in this round who scores better with the black pieces. Against her, 1.e4 held her to 38.9% in 36 games, high confidence — the best-sampled edge of the entire round on either side.

Umida Omonova and Nilufarkhon Imomkuzieva are the two whose weak lines are real but thinly sampled. Against Omonova's near-universal 1.e4, the Caro-Kann held her to 25.0% in four games; with White against her, 1.Nf3 gave 25.0% in six. Against Imomkuzieva, 1.c4 held her to 12.5% in four — the lowest number of the round, on the smallest sample of the round, and the two facts belong in the same sentence.

Where the engine finds the mistakes

The table gives average centipawn loss per move by phase. The opening column runs from 4 to 6 for all ten players and says more about preparation than about play. What follows the preparation is where these ten stop looking alike.

The German squad splits cleanly in two. Bluebaum and Donchenko lose almost nothing after the middlegame — 2 and 18 centipawns in the endgame — while Frederik Svane's 55 and Rasmus Svane's 36 are the two worst German figures. Frederik Svane's number invites an obvious story, and the obvious story is wrong. Three of his five worst moments come from one endgame, against Yuriy Kuzubov in the Linares league on 21 August 2026, and all three were re-measured at depth 24 before this sentence was written. At move 63 the position was already lost for him by six and a half pawns, and the move the shallow pass flagged — 63...Kxe3 — is the engine's own first choice. At move 68, the same. At move 69 he was facing mate in fifteen and chose a line that mates in fourteen.

So the 55 is not a collapse; it is the arithmetic of playing on in a lost endgame. A depth-16 pass charges enormous sums for moves made in positions that are already decided, which is exactly why no worst moment reaches this text before being re-measured at depth 24. The column still stands — average centipawn loss caps each evaluation at 1000 — but what it describes there is a defeat, not a defect.

The limit cuts both ways. Bluebaum's 2 rests on only 68 endgame moves, because his games in this sample ended before the endgame more often than not. A number that low is a description of these twelve games, not a property of the player.

On the Uzbek side the shape is more uniform and generally higher: Imomkuzieva at 46 and 45 across middlegame and endgame, Tokhirjonova at 35 and 45, Karimova at 32 and 31. Khamdamova and Omonova are the two steadiest, both in the low twenties overall.

Blunder counts are deliberately absent from this table. The classification uses uncapped loss, so a winning player who picks a slower mate collects a blunder for a move that changes nothing. Average centipawn loss caps each evaluation at 1000 and does not suffer from that.

The online repertoire, and what it is worth

Eight of these ten players have a Chess.com account that meets the proof standard, and six have games in the last 90 days: Keymer (140), Bluebaum (300), Frederik Svane (299), Donchenko (239), Rasmus Svane (223), Khamdamova (47), Tokhirjonova (15) and Omonova (176). Imomkuzieva and Karimova have verified accounts with no games in the window, which says nothing about the account and everything about the window.

Omonova's split is the one worth staring at. A player who has opened 1.e4 in 97% of her classical games spends her online chess playing 1.b3. Both descriptions are true, and neither is evidence for the other. In a classical Olympiad game with 90 minutes and an increment, the repertoire that counts is the one from the board — which is why this dossier never merges the two sources into one number.

The German accounts add recency rather than surprise. Bluebaum's 300 games show the French Defence as his most played Black opening, 25 games in one variation and 12 in another; Rasmus Svane's 223 show the Caro-Kann in his top three Black openings, 13, 12 and 8 games. Both are consistent with the classical picture rather than at odds with it, and both are blitz and rapid, so they describe taste more than preparation.

Tokhirjonova's 15 games are included here for completeness and marked as what they are. Fifteen blitz games do not describe anyone's repertoire, and naming the sample size is the only way to publish a number like that honestly.

What to watch in round 5

It is worth separating what the dossier does from what it does not. It shows where an opponent has scored below their own recent average, with the sample size attached. It does not say what the American players are good at, does not measure anyone's form this week, and does not predict a score. The two only meet when the line where the opponent scores less is also a line your own side wants to play.

In the open section, the most quotable numbers belong to Keymer and the most usable ones to Donchenko. Keymer's 26.9% against 1.d4 is real but sits inside an elite-only sample, as the second section explains. Donchenko's two Indian lines are drawn from 131 games against a mixed field and are both marked high confidence, which is the combination that travels.

In the women's section, the shape of the preparation is unusual: four of the five Uzbek players open 1.e4 in almost every game, so the American preparation with Black converges on one first move. The single high-confidence edge sits with the fifth, Tokhirjonova, and requires her to have White and to play 1.d4 — 30.0% in 15 games after 1...Nf6.

One caution belongs at the end. Both these matches are top-table pairings against opponents on equal match points, and dossiers describe the recent past, not the board in front of anyone tomorrow. In round 4 the American open team won 3.5-0.5 against a Ukrainian side whose dossier offered plenty to work with, and the women's team beat Spain by the same score. Neither result was predicted by the numbers, because that is not what the numbers are for.

Frequently asked questions

Update log

• 2026-09-19 — Initial publication, with the round 5 pairings checked on chess-results and the round not yet played: repertoire from classical games broadcast by Lichess, error profile with Stockfish 19 at depth 16, quoted moves re-measured at depth 24 and, for the first time in this series, a blitz and rapid repertoire from verified online accounts. The dossiers cover both full squads; once the line-ups are out, the same file will show the four boards and the round's game.
• 2026-09-29 — The Olympiad ended on 27 September: Uzbekistan won the open and China the women's event. This coverage is no longer updated; the final result is in the standings story.

Sources

• 46th Chess Olympiad Samarkand 2026, Open — round 5 pairings — Chess-Results
• 46th Chess Olympiad Samarkand 2026, Women — round 5 pairings — Chess-Results
• Alexander Donchenko — broadcasts under FIDE ID 24603295 — Lichess
• Alexander Donchenko — Chess.com's own player page, which links the member account used for the online repertoire — Chess.com

Related guides

• How to build an opening repertoire: 1.c4 as White and the Sicilian as Black
• Corporate chess leagues: what changes when the point belongs to the team
• How to analyze a chess game with Stockfish without just copying the engine