Loss to cheater: 1297-1290

A three-week stretch from October 23rd to November 14th saw my Chess.com Rapid rating nosedive from an all-time high of 1310, reached on October 8th, to 1234, a level not seen since August 24th. Devastated by the sharp reversal of fortune, of which I believe overconfidence and impatience to have been the principal causes, I spent the next three weeks sedulously clawing my way out of the abyss. Between November 14th and December 3rd I built up an eight-game win streak, and on December 10th the stage was set for me to reclaim the 1300 mark.

Chess.com matched me with a 1328-rated player by the name of "aarghmaj" and assigned me the white pieces. This report examines the game itself, which ended with me resigning after twenty-three moves, and also explores the phenomenon alluded to in the title. What clues led me to suspect my opponent of cheating? What happened when I reached out to Chess.com and shared my suspicions with them? What does it feel like to be the victim of a cheater? Given that my suspicions were confirmed and the cheater apprehended, why doesn't the title say "Loss to cheater: 1297-1297"?

---

1. d4 e6
2. e4 f5
3. Nc3 g5

As is my wont when playing White, I opened with 1. d4, anticipating either 1...d5, after which I'd continue 2. c4, or 1...e4, after which I'd continue 2. dxe4. Aarghmaj, however, played 1...e6, and already I was in unfamiliar territory. What I should have continued with was 2. e4, seizing the opportunity to dominate the center with two pawns; but habit got the better of me and I continued 2. c4. Then Black played 2...f5, and I was thoroughly baffled: Who, other than Hikaru Nakamura and Ben Hunt, moves their f-pawn on move 2?

Daniel Naroditsky, referencing Nakamura's penchant for unorthodox openings, recently spoke of the "psychological tilt effect, because [...] someone who doesn't punish that can feel really frustrated."¹ I had a clear sense my opponent was playing an unsound opening, and I felt the onus was on me to prove it -- but it is exactly that pressure that leads to rash decisions and gives the advantage to the opponent. To wit, I knew 2...f5 was a garbage move, but I didn't know how to expose it as such and wasn't about to spend valuable time calculating what my best third move was. I therefore continued with the natural 3. Nc3, to which Black replied 3...g5.

---

4. Nf3 g4
5. Bg5 Be7
6. Bxe7 Qxe7

Still unsure of how to proceed, I played the natural-looking 4. Nf3. Black rejoined by attacking my knight with 4...g4, and here the frustration started getting to me. 5. Ng5 and 5. Nh4 were out because of queen takes, 5. d2 would block in my dark-squared bishop, 5. Ng1 looked silly, and 5. Ne5 would invite 5...d6 or 5...Bg7. What the engine wanted was for me to rewind the previous move and play 5. Ng1.

I, however, was determined to retain the initiative, and so my move was 5. Bg5 -- a pointless attack on Black's queen. Black counterattacked with 5...Be7; and for lack of anything better, I captured with 6. Bxe7. Black recaptured with 6...Qxe7, and after all that I was still left with my knight under attack. As indicated by the evaluation bar, Black was not objectively better, but it felt to me like he was in control of the game.

---

7. Ne5 d6
8. Nd3 Nf6
9. g3 Nc6

Where to move my knight this time? Between retreating the knight to g1 or d2 and advancing it to e5, the latter struck me as being slightly more aggressive. I continued 7. Ne5, whereupon Black, predictably, attacked the knight again with 7...d6. I retreated the knight to safety with 8. Nd3, and Black developed his kingside knight with 8...Nf6.

The light-squared bishop being my last remaining undeveloped minor piece, there were two options for involving it in the game: 9. e3 and 9. g2. While fianchettoing on the side where I intend to castle is seldom my preference, I opted for 9. g2 as the bishop would have more scope on the long diagonal. Black replied 9...Nc6, deploying his queenside knight with a threat of capturing my undefended pawn on d4.

---

10. e3 Bd7
11. Bg2 0-0-0
12. b4 e5

What to do about my pawn on d4? One option was to advance it to d5, attacking Black's knight on c6 and his pawn on e6 but weakening my control over the central e5 square. The move I played, which the engine agrees was best, was 10. e3, defending d4 and creating a pawn chain. Black replied 10...Bd7, deploying his last remaining undeveloped minor piece to the only square available to it.

Although I now had both g2 and e2 available for my light-squared bishop, I proceeded with the fianchetto plan and played 11. Bg2. Stockfish insists 11. h4 was a better move, perhaps to lock up the kingside or, in the event of en-passant, open up the h-file for my rook. Considering the mess Black's kingside was in, he made the logical choice on his turn of castling queenside with 11...0-0-0. With Black's king castled, I had a choice to make: castle immediately or delay castling and advance on the queenside. I chose the latter and proceeded 12. b4, which the engine says was best. Then began the crucial sequence of the game: Black struck at the center with 12...e5.

---

13. b5 Na5
14. Nd5 Nxd5
15. Bxd5 exd4

Black was attacking my pawn on d4. If allowed to capture, 13. exd4 would not be possible due to an absolute pin on my e-pawn by Black's queen. Moreover, my knight on c3 would be under attack and forced to move. Rather than resolve the tension between e5 and d4, I responded with an attack of my own: 13. b5. The engine agrees with my choice, but what the computer has that I don't is a perfect memory: In shifting my attention to Black's knight on c6, I soon forgot about my loose pawn on d4.

Black got his knight out of danger with 13...Na5, and the engine says I should have pursued it with 14. Qa4, which would have prompted 14...b6 because the knight would have nowhere to go. Instead, I went after Black's queen with 14. Nd5. This, it turned out, was not only a futile attack but a waste of a valuable tempo. After Black captured my knight with 14...Nxd5 and I recaptured with 15. Bxd5, Black proceeded to capture my pawn with 15...exd4. Black was up a pawn in material, and the advantage had swung in his favor.

---

16. 0-0 dxe3
17. Re1 exf2+
18. Kxf2 Qg7

Scrambling to get my king off the e-file, I castled short with 16. 0-0. The engine says 16. Qd2 was a better move, but my position was already crumbling. Black helped himself to another of my pawns with 16...dxe3. Recapturing with 17. fxe3 would have invited Black's queen to penetrate my third rank with 17...Qxe3, so I tried 17. Re2, pinning Black's pawn to the queen.

But the Black pawn continued its feast on my pawns unabated, openly mocking me as it did so: Because 17...exf2+ came with check, my pin on the e-file was useless. 18. Kxf2 recovered a lost pawn and activated my rook's attack on Black's queen, but the queen had only to sidestep two squares with 18...Qg7 and my attack was over. Stockfish's evaluation of the position after move 18 as -2.2 feels generous. I was teetering like a tired boxer waiting for the knockout punch.

---

19. Qa4 b6
20. Rac1 Qd4+
21. Re3 Rhe8

It was on move 19 that I played what the engine said I should have played on move 14, attacking Black's knight with 19. Qa4. Naturally, Black defended the knight with 19...b6. If I had any hope of breaking through Black's defenses, it was on the c-file, e.g. 20. c5 dxc5 21. Nxc5 bxc5 22. Qxa5 followed by 23. Qb7+.

It's generally advisable to position rooks on open files or files that one expects to become open. It's also generally advisable to line rooks up with the enemy king. Hence, in preparation for my queenside invasion, I played 20. Rc1. Black, in arrant contempt for my plan, went ahead with his: 21...Qd4+. I interposed with 21. Re3, but Black's 21...Rhe8 sealed my pinned rook's fate -- and mine.

The game continued:

22. Re1 Rxe3
23. Rxe3 Re8

And I resigned.

---

Dr. Jekyll and Mr. Hyde

It wasn't until the endgame that I gave serious consideration to the possibility my opponent was cheating. What I couldn't reconcile were the two wildly disparate personas it appeared I was playing against. There are players who don't care about their rating and open with random pawn moves purely to troll. That would explain the opening 1...e6 2...f5 3...g5. Then there are players who execute plans with the surgical precision of an elite assassin. That would explain much of the game's latter half.

This and the speed with which my opponent had played led to me examining aarghmaj's history. Lo and behold, he was losing his Blitz games against weak players but winning his Rapid games against strong players. It's true that the older you get, the slower you play, which is why seniors often compete in separate tournament brackets at chess events. Perhaps aarghmaj was getting up in years. But it's also true that Stockfish needs time to calculate the best moves, and time is in shorter supply in Blitz than it is in Rapid. Add to that an astounding 28-game win streak that boosted aarghmaj's Rapid rating from 1165 to 1416, and my mind was made up.

Report

There is no record of when exactly I filed my report concerning aarghmaj, but I believe it was in the evening the day after our game. Whenever the next time was that I logged into Chess.com, their reply was there in my inbox: "We have detected that one or more of your recent opponents has violated our Fair Play Policy."

I also believe I was the first of aarghmaj's victims to report him. It's even possible I was the only one to do so. The basis of my conviction is this: virtually no other player initiated a computer analysis of their game after losing to aarghmaj. Had more of them done so, they'd have seen aarghmaj was playing at levels of accuracy unfeasible for a human and reported him themselves.

One can only speculate how I'd have played on that December 10th against an opponent who didn't cheat. The two games in October (5th, 8th) that elevated my rating to 1310 had been won by strokes of luck, not brilliance; it was clear to me I hadn't earned the points through improved chess play. The fact that my Rapid rating collapsed both after hitting 1310 and after losing to aarghmaj means 1300 was and remains a strong resistance level for me.

That's my rational assessment of the situation, made with the clarity and composure of hindsight. At the time, though, the loss stung. Every loss stings, but this one stung more sharply for two reasons. One, I was three points away from 1300 before the loss but ten points, i.e. two wins, away after it. Another loss and I'd be a daunting three wins away. Two, every chess game is an intensely focused experience. For ten minutes and seven seconds, all my attention was funneled into this game; and when the game concluded, it left me feeling helpless and empty, like all my efforts were for naught.

The loss to aarghmaj was followed by another rating slide. Along the way, on December 12th, I lost a game but something unexpected happened: my rating did not change. It turns out that rather than compensate me for the loss to aarghmaj by adjusting my rating retroactively immediately upon determining that aarghmaj had cheated, Chess.com held onto my points as credit and only awarded them to me simultaneously with the December 12th loss.