Niemann Logged Multiple 100% Engine Games, Defying FIDE Review

The most incriminating evidence against Hans Niemann — Yosha - Echecs

Yosha - Echecs

A chronological analysis of Hans Niemann's classical tournaments reveals a sustained performance streak with a 1-in-77,000 probability of occurring naturally, contradicting FIDE expert Kenneth Regan's finding of normal variance.

AI-assisted summary of The most incriminating evidence against Hans Niemann — original source. Check the original for full context.

Source published:

Key ideas

  • A public dataset compiled by Gambitman catalogs Hans Niemann's 2019–2022 classical over-the-board games, tracking engine correlation scores generated by ChessBase's Let's Check tool with Stockfish 15. (0:00)
  • The analysis reveals multiple classical games where Niemann achieved a 100% engine correlation, surpassing the historical record of 98% set by Sébastien Feller during a confirmed cheating run in 2010. (9:23)
  • During a five-tournament stretch starting in July 2021, Niemann averaged over 73% correlation, exceeding the career-peak averages of Bobby Fischer, Garry Kasparov, and Magnus Carlsen. (14:34)
  • While Kenneth Regan found Niemann's ROI distribution normal in aggregate, evaluating the chronological sequence of six consecutive above-average tournaments yields a joint probability of just 1 in 77,000. (16:23)

Detailed notes

## Background This video was published on YouTube by the channel Yosha - Echecs on September 25, 2022. According to the description, the video presents a new set of compromising evidence against Grandmaster Hans Niemann following cheating accusations made by Magnus Carlsen, relying on data from an over-the-board spreadsheet compiled by Gambitman and statistical analyses from Kenneth Regan.

## TL;DR FIDE Master Yosha analyzes classical over-the-board games played by Hans Niemann between 2019 and 2022 using ChessBase's 'Let's Check' engine correlation tool and public data compiled by Gambitman. The analysis reveals that Niemann achieved an unprecedented series of games with a 100% engine correlation—surpassing the 98% mark set by convicted cheater Sébastien Feller—and maintained an average correlation above 73% across a five-tournament stretch in 2021. Yosha also scrutinizes FIDE statistician Kenneth Regan's Raw Outperformance Index (ROI) model, demonstrating that while individual tournament scores appear standard in isolation, Niemann's consecutive run of high-performing events represents a 1-in-77,000 joint statistical probability under clean conditions.

## Atmosphere & style The presentation carries the methodical, engrossing tone of an analytical chess detective laying out a forensic case file. Yosha delivers the findings with calm, measured articulation mixed with dry, understated disbelief, especially when pointing out repeated 100% engine correlation scores that eclipse known historical benchmarks. Rather than shouting accusations, the video adopts an intellectual, data-focused cadence that invites the audience to weigh statistical oddities and draw their own conclusions from the numbers.

## Bulletin - [00:00](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=0s) A public dataset compiled by Gambitman catalogs Hans Niemann's 2019–2022 classical over-the-board games, tracking engine correlation scores generated by ChessBase's Let's Check tool with Stockfish 15. - [09:23](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=563s) The analysis reveals multiple classical games where Niemann achieved a 100% engine correlation, surpassing the historical record of 98% set by Sébastien Feller during a confirmed cheating run in 2010. - [14:34](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=874s) During a five-tournament stretch starting in July 2021, Niemann averaged over 73% correlation, exceeding the career-peak averages of Bobby Fischer, Garry Kasparov, and Magnus Carlsen. - [16:23](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=983s) While Kenneth Regan found Niemann's ROI distribution normal in aggregate, evaluating the chronological sequence of six consecutive above-average tournaments yields a joint probability of just 1 in 77,000.

## Key takeaways - [01:18](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=78s) ChessBase's Let's Check engine correlation metric measures how frequently a player's moves align with top engine suggestions over the course of a game. - [01:18](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=78s) ChessBase's documentation cites Sébastien Feller's 98% correlation during his 2010 Paris Championship cheating run as the historical record benchmark. - [01:18](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=78s) Historical baselines show elite grandmasters averaging 62% to 67%, correspondence champions averaging 72% to 75%, and peak world champions averaging around 69% to 72%. - [04:47](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=287s) Standard over-the-board games by grandmasters, including several of Niemann's games in the 2022 Turkish Super League, typically fall between 45% and 70% correlation. - [04:47](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=287s) Magnus Carlsen's renowned positional masterclass against Ian Nepomniachtchi registered a 79% correlation, reflecting elite human play at its peak. - [06:57](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=417s) Niemann's widely scrutinized victory over Abhimanyu Mishra featured an unconventional knight retreat to h2, but registered a normal master-level correlation of 72%. - [09:23](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=563s) In his victory against GM Matthieu Cornette at Cap d'Agde 2020, Niemann registered a 100% ChessBase Let's Check correlation. - [12:21](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=741s) Niemann repeated the 100% correlation score across numerous cataloged classical games, including games of 20, 22, 27, 28, 31, 35, 37, 38, and 45 moves. - [14:34](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=874s) Over a five-tournament sequence starting at the 2021 Philadelphia International / World Open, Niemann recorded consecutive correlation rates of 69%, 70%, 79%, 79%, and 69%. - [16:23](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=983s) Kenneth Regan's anti-cheating model concluded Niemann's performance was clean by viewing his ROI scores as an aggregate normal distribution centered around 50. - [16:23](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=983s) Analyzing Regan's ROI metrics chronologically shows six consecutive tournaments above 50, creating a compound probability of 1 in 77,000 for a clean player. - [20:04](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=1204s) Under Bayesian probability with a conservative baseline assumption of cheater prevalence, an anomaly of 1 in 77,000 points toward an approximately 89% probability of cheating. - [20:04](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=1204s) In Niemann's head-to-head wins against Magnus Carlsen, Niemann scored a 67% correlation at the FTX Crypto Cup and a 71% correlation at the 2022 Sinquefield Cup.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=0s) Introduction and the Discovery of the Gambitman Dataset

FIDE Master Yosha introduces the cheating allegations surrounding Grandmaster Hans Niemann following accusations by Magnus Carlsen. While acknowledging the impossibility of providing an absolute mathematical proof, Yosha frames the presentation as a breakdown of the most incriminating publicly accessible material to date: “Obviously I cannot promise that I will give mathematical proof that Hans cheated because it's impossible, but I can guarantee that I will give you the most incriminating evidence publicly available against Hans Niemann.”

The investigation began when an observant friend directed Yosha to an obscure Twitter response to a post by Peter Heine Nielsen that had gathered zero retweets and zero likes. The link led to a detailed Google Sheets spreadsheet compiled by Gambitman, cataloging classical over-the-board games played by Niemann between 2019 and 2022. While initially difficult to parse, the data tracked metric outputs derived directly from ChessBase’s analytical tools.

[01:18](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=78s) Understanding ChessBase's "Let's Check" Engine Game Correlation

To investigate the dataset, Yosha turns to the "Let's Check" tool in ChessBase, run with Stockfish 15. The feature computes an "engine game correlation" percentage for each player, measuring how frequently their played moves align with top engine suggestions throughout a contest.

```mermaid flowchart TD accTitle: ChessBase Let's Check Workflow accDescr: Diagram illustrating the steps to run Let's Check analysis and obtain engine correlation percentages. subgraph Execution["ChessBase Workflow"] A["Load Game Notation"] --> B["Select Analyze Menu"] B --> C["Run 'Let's Check'"] C --> D["Query Stockfish 15 / Cloud DB"] end subgraph Output["Statistical Result"] D --> E["Overall Move Correlation"] E --> F["White Correlation %"] E --> G["Black Correlation %"] end ```

Yosha reads directly from the official ChessBase documentation explaining the correlation metric. The built-in help text notes that high scores on single tactical games are not proof of cheating, while very low numbers serve to disprove engine usage. Notably, the ChessBase help documentation explicitly states that the historical record correlation was set on October 13, 2011, at 98% by Sébastien Feller at the 2010 Paris Championship—a tournament in which Feller was subsequently convicted of cheating using computer assistance to achieve a 2859 tournament performance rating.

Based on discussions with an analyst who processed thousands of games across the ChessBase engine network, long-term multi-game rolling averages yield recognizable thresholds across chess history:

| Player / Level Category | Multi-Game Average Correlation | |:--- |:--- | | **Sébastien Feller** (Paris Championship 2010 cheat run record) | **98%** (single game high cited by ChessBase) | | **Pre-Engine Correspondence Champions** | **72% – 75%** | | **Bobby Fischer** (Peak 20-game winning streak vs. Taimanov, Larsen, Petrosian) | **72%** | | **Magnus Carlsen** (Peak career play) | **70%** | | **Garry Kasparov** (Peak career play) | **69%** | | **Super Grandmasters** (Elite tier) | **62% – 67%** | | **Standard Grandmasters** (Normal GM baseline) | **57% – 62%** |

[04:47](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=287s) Baseline Controls: Normal Games and Magnus Carlsen’s Masterpiece

To demonstrate how the correlation works under standard conditions, Yosha examines a sample of classical games played by Niemann in the 2022 Turkish Super League: - A win against a GM rated 2641 produced a standard 49% correlation for White (Niemann) against 37% for Black. - A draw against Yuriy Kuzubov yielded 69% against 61%. - A draw against Maksim Chigaev scored 47% against 46%. - A loss against Sanan Sjugirov resulted in 51% for Niemann against 74% for Sjugirov.

As a reference point for elite human play, Yosha displays Magnus Carlsen's renowned victory against Ian Nepomniachtchi. In that game, Carlsen exerted long-term positional squeeze in an endgame with two bishops and a space advantage, took advantage of Nepomniachtchi's errors, and smoothly converted a two-pawn margin. Carlsen's clinical play earned a 79% correlation score—an exceptionally high mark for an individual game, perfectly reflective of a World Champion playing at his absolute best.

[06:57](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=417s) Niemann vs. Mishra: The "Suspicious" Knight Maneuver

Yosha next evaluates an encounter between Niemann (playing Black) and Abhimanyu Mishra, often cited by online commentators as one of Niemann's most unnatural and machine-like games. The game featured an extraordinary knight journey: Black pushed...g3, then rerouted the knight through g4 to h2, checked the white queen, harassed the rooks, and left the piece temporarily trapped on h2.

Niemann exchanged rooks and transitioned into an endgame via...Ra6, deliberately accepting an immobilized knight on h2 because White’s king was completely cut off from the action. Because an active king functions as an aggressive endgame piece superior to an immobilized knight, Niemann cleanly converted the technical advantage. Despite public suspicion surrounding these non-human maneuvers, ChessBase’s Let's Check score for Niemann’s side was 72%—a great game, but completely normal for an over-the-board master performance.

[09:23](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=563s) Niemann vs. Cornette: The 100% Correlation Game

The focus shifts to a game played at the 2020 Cap d'Agde tournament between Niemann (White) and French Grandmaster Matthieu Cornette, a personal friend of Yosha. Out of a symmetrical pawn structure, Niemann sacrificed a pawn, traded queens, and utilized the activity of two dominant rooks against Cornette's disconnected rooks alongside an opposite-colored bishop to regain material and force an early resignation.

When Yosha asked Cornette about the experience, Cornette recalled being outplayed completely, comparing it to an earlier drubbing he had suffered against GM Vladislav Tkachiev. Cornette—who has drawn games against elite competitors like Maxime Vachier-Lagrave and Magnus Carlsen—felt Niemann was simply operating at an overwhelmingly higher level. In the post-game post-mortem, Cornette asked whether his move...Ng4 had been the decisive mistake; Niemann dismissively and provocatively replied that it wasn't the knight move, but rather that Cornette’s entire opening was bad.

Running the Let's Check analysis on this game reveals an unexpected result: Niemann achieved an engine correlation of 100%. Yosha remarks with dark humor: “Someone should tell ChessBase, because they still pretend in the help menu that the world record is Sébastien Feller’s 98%—no, it's not, it's Hans Niemann.”

[12:21](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=741s) "Mister 100 Percent": Multiple Perfect Games Uncovered

While an isolated 100% game might theoretically happen by coincidence in a short miniature, Yosha demonstrates that Niemann repeated this feat across numerous games in the Gambitman dataset:

```mermaid flowchart LR accTitle: Hans Niemann 100 Percent Correlation Game Catalog accDescr: Sequence of opponents against whom Hans Niemann scored a 100 percent ChessBase engine correlation. subgraph PerfectWins["Cataloged 100% Engine Correlation Victories"] direction TB G1["vs. Christopher Yoo (22 moves, W)"] G2["vs. Miguel Soto (27 moves, B)"] G3["vs. Ostrovskiy (28 moves, B)"] G4["vs. Martin del Campo (20 moves, B)"] G5["vs. Unnamed Opponent (31 moves, W)"] G6["vs. Liam (35 moves, 2021)"] G7["vs. Unnamed Opponent (38 moves, W)"] G8["vs. Unnamed Opponent (45 moves, W)"] G9["vs. Strong Grandmaster (37 moves, W)"] end ```

- **vs. Christopher Yoo**: A 22-move demolition where Niemann registered a 100% correlation. - **vs. Miguel Soto**: A 27-move victory played as Black ending in 100% correlation. - **vs. Andrey Ostrovskiy**: A 28-move checkmating win playing Black, scoring 100%. - **vs. Martin del Campo**: A miniature lasting just 20 moves playing Black, registering 100%. - **Additional classical games**: Wins spanning 31 moves, 35 moves (against an opponent named Liam in 2021), 38 moves, 45 moves, and a 37-move victory over a strong Grandmaster—all registering a 100% correlation mark in the ChessBase system.

Noting the sheer absurdity of one player repeatedly matching engine lines across full-length classical games, Yosha prepares to cross-reference this cluster of perfect games back into the broader Gambitman database.

[14:34](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=874s) Gambitman's Dataset and Niemann's 5-Tournament Run

Yosha introduces a detailed spreadsheet compiling ChessBase "Let's Check" analysis for Hans Niemann covering his classical over-the-board games between 2019 and 2022. The document was put together not by Yosha personally, but by an anonymous collaborator going by the handle "Gambitman," with whom Yosha spent extensive time discussing the findings. A public link to the spreadsheet is provided so viewers can download the PGN files and verify the Let's Check evaluations directly within ChessBase.

Beyond isolated 100% engine-correlation games, the spreadsheet exposes an extraordinary five-tournament streak starting with the Philadelphia International / World Open in July 2021. Across these five consecutive tournaments, Niemann recorded engine correlation rates of 69%, 70%, 79%, 79%, and 69%. To contextualize those figures, Yosha reminds the audience that correspondence chess champions typically score around 72% to 75%, and Bobby Fischer during his legendary 20-game winning streak averaged 72%.

“And here for five tournaments in a row, Hans played at more than 73 percent on average.”

This sustained peak performance prompts an obvious question: how could FIDE’s leading anti-cheating expert and renowned mathematician, Professor Kenneth Regan, examine the data and find nothing suspicious?

[16:23](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=983s) Kenneth Regan's Statistical Model vs. Consecutive Probabilities

To investigate why FIDE detected no foul play, Yosha pulls up Kenneth Regan's publicly accessible data files covering Niemann from 2019 to 2022. Regan's methodology relies on an intrinsic performance metric called ROI (Raw Outperformance Index). Across Niemann’s games, Regan's ROI values range between 36.6 and 62.3. As Regan explained in a ChessBase interview, approximately 50% of Niemann's tournament ROI values fall above 50 and 50% fall below 50.

Regan interprets this spread as a textbook normal distribution centered at 50, where standard deviations ($\sigma$) map to increments of 5 (e.g., $1\sigma = 55$, $2\sigma = 60$, $-1\sigma = 45$, $-2\sigma = 40$). Because Niemann recorded only two or three tournaments above 60 and a comparable number hovering around 42.2, Regan concluded that the entire dataset represents normal variance for a grandmaster.

However, Yosha argues that evaluating the distribution as an unordered aggregate hides a critical anomaly: the chronological clustering of these scores. Looking at Regan's own ROI values for that five-tournament run (and extending to a sixth tournament that also stayed above 50), the individual odds appear modest on their own, but become staggering when multiplied consecutively:

```mermaid flowchart LR accTitle: Cumulative Probabilities Across Consecutive Tournaments accDescr: Diagram tracing the cumulative probability of Niemann's consecutive high-performing tournaments according to Kenneth Regan's ROI metrics. subgraph Run["Consecutive Tournaments Above 50 ROI"] T1["Philadelphia Open<br/>ROI 57.9 (Odds: 1 in 18)"] T2["Tournament 2<br/>(Odds: 1 in 7)"] T3["Tournament 3<br/>(Odds: 1 in 8)"] T4["Tournaments 4 to 6<br/>(Odds: ~1 in 6 each)"] end subgraph Cumulative["Cumulative Joint Probability"] C2["After 2 Tournaments:<br/>P < 1%"] C3["After 3 Tournaments:<br/>1 in 1,000"] C6["After 6 Tournaments:<br/>1 in 77,000"] end T1 --> T2 --> C2 T2 --> T3 --> C3 T3 --> T4 --> C6 ```

- **Philadelphia Open:** ROI of 57.9 corresponds to individual odds of roughly 1 in 18. This is not an 18-to-1 probability of cheating; rather, in a clean player population, one out of every 18 tournament performances will reach this level by pure chance. - **Tournament 2 in a row:** Individual odds of 1 in 7, pushing cumulative joint probability under 1%. - **Tournament 3 in a row:** Individual odds of 1 in 8, dropping cumulative joint probability to roughly 1 in 1,000. - **Six tournaments in a row:** The continuous multiplication of above-average performances reaches an overall joint probability of 1 in 77,000.

“It's not at all proof of cheating or an indication of cheating, it's the probability of it happening in case Hans doesn't cheat.”

[20:04](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=1204s) Bayesian Priors and the Carlsen Showdowns

Addressing how one translates a 1-in-77,000 statistical anomaly into an actual cheating probability, Yosha points out that the result depends entirely on one's prior probability. If an observer assumes a world where nobody ever cheats, the posterior probability remains zero regardless of the data. Conversely, if one assumes a baseline prior where 1 in every 10,000 competitive players cheats:

- Out of a sample of 80,000 players, 8 would be cheaters. - In that same group of 80,000 clean players, approximately 1 would produce a 1-in-77,000 performance run by pure statistical coincidence. - Encountering a player with this profile yields 9 candidates total—8 actual cheaters and 1 legitimate outlier—meaning an 8-out-of-9 chance (roughly 89%) that the player is cheating.

Yosha distinguishes between French and English legal terminology, noting that in French the word *preuve* covers both, whereas in English there is a meaningful distinction between "proof" and "evidence." While this data does not represent irrefutable mathematical proof, Yosha considers it exceptionally heavy evidence.

Turning finally to Niemann's head-to-head encounters with Magnus Carlsen, Yosha highlights the ChessBase Let's Check numbers for their two most scrutinized clashes:

| Event / Game | Hans Niemann Score | Magnus Carlsen Score | Context | |:--- |:--- |:--- |:--- | | **FTX Crypto Cup (Game 1, chess24)** | 67% | 36% | Niemann wins as Black; produces the post-game quote: *"The chess speaks for itself."* | | **Sinquefield Cup 2022** | 71% | — | Niemann defeats Carlsen with Black in round 3, leading Carlsen to withdraw from the tournament. |

“So do what you want with this information, make up your own mind whether Hans has ever cheated OTB and whether he cheated against the world champion.”

## Notable quotes > [00:00](https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=0s) Obviously I cannot promise that I will give mathematical proof that Hans cheated because it's impossible, but I can guarantee that I will give you the most incriminating evidence publicly available against Hans Niemann. — Yosha

## Mentioned Hans Niemann, Magnus Carlsen, Yosha, Gambitman, Peter Heine Nielsen, ChessBase, Stockfish 15, Sébastien Feller, Bobby Fischer, Garry Kasparov, Yuriy Kuzubov, Maksim Chigaev, Sanan Sjugirov, Ian Nepomniachtchi, Abhimanyu Mishra, Matthieu Cornette, Vladislav Tkachiev, Maxime Vachier-Lagrave, Christopher Yoo, Miguel Soto, Andrey Ostrovskiy, Martin del Campo, Kenneth Regan, FIDE, Philadelphia International / World Open, FTX Crypto Cup, Sinquefield Cup 2022

## Video description I am providing you with a new set of extremely compromising evidence against Hans Niemann, the Grand Master accused of cheating by Magnus Carlsen. ▬▬▬▬▬▬▬▬▬▬▬ POUR ALLER PLUS LOIN ▬▬▬▬▬▬▬▬▬▬▬ ♔ Support the channel on Tipeee ➡️ https://fr.tipeee.com/yosha-echecs ♕ Take a private lesson with me ➡️ https://yoshachess.com/fr/cours-particuliers-echecs-paris-en-ligne/ Hans Niemann (OTB, Classical, 2019-22) by Gambitman ➡️https://docs.google.com/spreadsheets/d/127lwTsR-2Daz0JqN1TbZ8FgITX9Df9TyXT1gtjlZ5nk/edit?usp=sharing Kenneth Regan's analysis ➡️https://cse.buffalo.edu/~regan/chess/fidelity/data/Niemann/

▬▬▬▬▬▬▬▬▬▬ CHAPITRES DE CETTE VIDEO ▬▬▬▬▬▬▬▬▬▬▬ 0:00 - Intro 1:25 - Chessbase's Let's check analysis 9:30 - Niemann vs. Cornette 12:25 - Mister 100 percent 16:23 - Ken Regan's analysis 22:00 - Niemann vs. Carlsen and outro

[Source](https://www.youtube.com/watch?v=jfPzUgzrOcQ)

中文简报

尼曼多次下出100%引擎重合度对局，与国际棋联审查结论相悖 对汉斯·尼曼慢棋赛事的按时间顺序分析显示，其持续的高水平表现自然发生的概率仅为7.7万分之一，这与国际棋联专家肯尼斯·里根得出的正常波动结论相矛盾。 国际棋联大师约莎（Yosha）利用ChessBase的“Let's Check”引擎关联工具以及Gambitman整理的公开数据，分析了汉斯·尼曼（Hans Niemann）在2019年至2022年间进行的慢棋实盘对局。分析显示，尼曼下出了一系列史无前例的引擎重合度达100%的对局——超过了已被定性作弊的塞巴斯蒂安·费勒（Sébastien Feller）创下的98%纪录——并在2021年连续五站赛事中保持了超过73%的平均重合度。约莎还审视了国际棋联统计学家肯尼斯·里根（Kenneth Regan）的原始超常表现指数（ROI）模型，指出虽然单站赛事的成绩孤立来看似乎正常，但尼曼在正常情况下连续多站打出高水平赛事的联合统计概率仅为7.7万分之一。 Gambitman整理的一份公开数据集收录了汉斯·尼曼2019年至2022年的慢棋实盘对局，并追踪了ChessBase的Let's Check工具使用Stockfish 15生成的引擎重合度评分。 (0:00) 分析显示，尼曼在多盘慢棋对局中达到了100%的引擎重合度，超过了塞巴斯蒂安·费勒在2010年被证实的作弊期间创下的98%的历史纪录。 (9:23) 从2021年7月开始的连续五站赛事中，尼曼的平均重合度超过73%，高于鲍比·菲舍尔、加里·卡斯帕罗夫和马格努斯·卡尔森职业生涯巅峰期的平均水平。 (14:34) 虽然肯尼斯·里根在汇总层面认为尼曼的ROI分布正常，但按时间顺序评估其连续六站高于平均水平的赛事，得出的联合概率仅为7.7万分之一。 (16:23)

Source quotes

[0:00] Yosha: Obviously I cannot promise that I will give mathematical proof that Hans cheated because it's impossible, but I can guarantee that I will give you the most incriminating evidence publicly available against Hans Niemann.

Original transcript

[0:00] hi everyone and welcome to this video [0:02] where I'm going to talk about the [0:04] shooting allegations against Hans [0:06] Niemann if you don't know me I'm feeding [0:09] Master Josh from friends and nice to [0:11] meet you [0:12] so obviously I cannot promise that I [0:15] will give mathematical proof that Hansa [0:18] cheated because it's impossible but I [0:21] can guarantee that I will give a user [0:25] most incriminating evidence [0:28] publicly available against and sneeman [0:32] so it all started when a friend of mine [0:35] were is even more honored than me asked [0:38] me whether I saw this trade and back [0:40] then it had [0:42] zero chords zero like so I was like [0:45] come on girl no only you have seen this [0:49] street in response to a barrister books [0:52] to it from two weeks ago [0:54] but okay I clicked the link and it [0:58] brought us to um to this Google Sheets [1:02] full of numbers that we had a struggle [1:06] to make sense of at first and then we [1:11] understood what it was all about [1:14] and it brought us to chess Bays so let's [1:18] go there there's this thing in chess Bay [1:21] that I wasn't aware [1:22] um of before yesterday because let's [1:25] check analysis basically you go to [1:28] analyze then you click let's check it [1:31] will bring the default engine for me [1:34] stock h15 and then you click once more [1:38] to analyze let's check and after a while [1:42] you will get these numbers [1:44] the world game analyzed and this [1:48] correlation engine game correlation for [1:50] white 49 for black 30 7 percent [1:56] and it's a bit remindful of the famous [1:59] chesscom number that you'll see [2:03] everywhere [2:04] but what does chess base [2:07] um say about this engine game [2:09] correlation let's see what does engine [2:11] game correlation mean at the top of the [2:13] notation after the let's check analysis [2:15] this value shows a relation between the [2:17] moves made in the game and those [2:19] suggesting by the engines Pro this [2:22] combination isn't a sign of computer [2:24] cheating because strong players can [2:26] reach high values in tactical simple [2:28] games there are historic games in which [2:31] the correlation is above 70 percent only [2:34] low values say anything because are [2:37] sufficient to disprove the illegal use [2:38] of computers in a game among the top 10 [2:41] gunmaster it is usual to find the win [2:44] their games with the correlation value [2:47] of more than 50 percent even if [2:49] different chess programs agree in [2:51] suggesting the same variation for [2:53] position it does it does not mean that [2:56] this must be the the best moves sorry [2:58] the current record for the highest [3:00] correlation October 13th 2011 is 98 [3:06] percent in the game fellas at Raman [3:08] Palace championship 2010. [3:10] this Precision is implementing various [3:13] other games in this tournament and [3:15] results in an early performance of 28 [3:19] 59 that made him the clear winner by the [3:23] way in case you don't know in 2010 [3:25] Federer was convicted after having [3:29] cheated with the help of the engine so [3:32] I've spoken with someone who has [3:34] analyzed thousands of games using this [3:37] let's check analysis by a chess base as [3:40] I told me the following if you take [3:43] um a few dozen a few dozens of [3:46] consecutive game by someone you will get [3:48] approximately these results so 98 is [3:53] Sebastian fellas record in Paris 2010 [3:55] according to chess base [3:57] 72 to 75 percent [4:01] um its correspondentral champion before [4:03] the engines 72 percent is Bobby Fischer [4:07] at his Peak doing is a 20 consecutive [4:10] winning straight like again time enough [4:12] against Ben Classen [4:15] 70 percent is minus Carlsen at his best [4:18] 69 percent Gary Casper at his best 62 to [4:22] 67 Super Grandmaster and 57 to 62 [4:28] percent normal GM [4:31] so it's for many games you know like the [4:34] average after [4:36] um if you take a few dozens of games [4:38] because obviously for only one game [4:42] basically more or less anything can [4:44] happen so now let's Supply this analysis [4:47] to Hans Neiman it's a game that Hans [4:49] played two months ago in the Turkish [4:52] Super League 2022 as you can see he won [4:55] his game against her Grandmaster rated [4:58] 26 41 and he had a correlation of 49 [5:03] percent against such a seven percent so [5:07] very normal game [5:10] that he won in this game he drew against [5:13] kuzu buff with a 69 percent correlation [5:16] against 61 percent [5:18] no more game [5:21] in this game uh he drew against Mexico [5:24] perform with a correlation of 47 against [5:27] 46 percent normal game [5:31] in this game he lost against sanan with [5:34] a correlation of 51 percent against 74 [5:37] percentile so obviously a great game by [5:41] the Indian player but obviously he [5:45] doesn't play at this level [5:47] um for many games in a row [5:50] so what is a like perfect modern game by [5:54] a Grandmaster [5:58] [Music] [6:00] if you don't remember it was like minus [6:05] Cloud sense the best player ever doing [6:07] what he does best which is pressing in [6:10] um end game with a small Advantage so [6:13] two Bishops a space advantage [6:15] and then after a mistake by uh yeah [6:19] minus one upon [6:21] here [6:23] A5 was fast and minus one upon and then [6:27] he pressed pressed and [6:30] finally won a second Pawn here [6:33] and that's it [6:36] the animation resigned so perfect game [6:39] by Magnus Carlsen [6:41] 79 percent congratulations minus so [6:45] obviously uh [6:47] my new skeleton doesn't play uh on [6:50] average at this level but for one game [6:52] one Victory especially [6:54] it's normal for minor skeleton [6:57] so now let's analyze this game [6:59] um my handsome man which was considered [7:01] by many as very suspicious [7:04] if you don't know this game it was [7:07] Anthony Manuel's black and he plays this [7:10] incredible maneuver like never seen [7:12] before like very creative very [7:15] computer-like and obviously the best [7:17] move according to the computer which was [7:19] to buy G3 then Knight to G4 okay you [7:23] cannot take but what's the point of the [7:25] G4 it is to go to H2 Queen history [7:28] attack the world because the work moves [7:30] and then white folks white work moves [7:34] and the Knight is taken Edge too and [7:36] it's very strange to back like this but [7:38] towards the point of all this well we'll [7:42] see [7:43] black [7:44] now we wins an exchange but [7:47] they have to give it back [7:50] and now we enter this end game hook to [7:54] H6 [7:55] you might think it's to take as opponent [7:58] D6 but no because what what text this [8:02] text sorry would lose actually for black [8:04] because after Knight G3 and G3 work G3 [8:08] the light and H2 is lost and a white [8:11] will emerge upper Pawn [8:14] as a point was to play work to A6 to [8:19] exchange other work and in this end game [8:22] and black basically play without a [8:25] knight because it's stuck but white is [8:28] playing without a king and the king is [8:32] um a better piece a better attacking [8:34] Pace In the End games on the night so [8:36] like [8:37] it was it was a world point of this [8:40] incredible United to maneuver and as the [8:44] man conversion is an advantage in [8:47] perfect with a perfect technique like [8:51] great game [8:53] like and many for many people it was the [8:56] most uh suspicious game by hansenmann [9:01] and okay so poor Mishra [9:04] and where is by somebody in case you [9:06] don't know the youngest Grandmaster ever [9:07] support Mission here [9:11] so what is the correlation [9:13] for black 72 percent [9:16] so [9:17] great great game by your hands but [9:20] nothing too suspicious [9:23] so now let's see these games that Hans [9:25] played against Metro corner at a very [9:27] strong French gun master and a friend of [9:29] mine it was played as a suggest a [9:33] suggestion band 2020 [9:35] symmetrical Palm structure and hence [9:37] basically [9:39] outplayed his opponent in perfect [9:41] fashion [9:43] white sucked upon and here there is a [9:47] queen exchange but obviously only white [9:51] can be better thanks to the activity of [9:55] the pieces like [9:57] through perfect works against two [9:59] disconnected works if another black can [10:02] still Castle [10:03] and opposite corrupt Bishop [10:05] which faberize activity and yeah so [10:11] right we gained the prawn and then [10:16] after a few moves poor Metro cannot [10:19] resigned and [10:21] I asked I'm not sure about this game and [10:23] they told me well I've been dominated I [10:27] like rally but it happened to me before [10:30] for instance against blood ketchup and [10:32] we're talking about a stronger master [10:34] who drew against MBL who drew against [10:36] minus calcium [10:38] some you know he's capable of playing at [10:41] the same level at least for one game as [10:44] the best player in the world but he told [10:46] me okay I felt that Hans was a much [10:50] stronger player than me but otherwise [10:52] nothing too suspicious [10:54] even though after the game [10:57] mature asked his opponent [11:01] whether his mistake was this move net G4 [11:05] which it is you adds a bad move and [11:08] hence uh answered in his typical [11:12] provocative way that no it's not about [11:15] Niger it's your opening is bad and [11:18] mature was like come on you just watch [11:20] me like don't be such sure whatever [11:24] but okay according to Metro nothing [11:28] suspicious about this game [11:31] so what is the score maybe 79 percent [11:34] like a manuscripts and guts are for His [11:37] Brilliant Victory against yeah or [11:40] something like that [11:41] well [11:42] 100 percent [11:45] wow perfect score even better than [11:47] Sebastian Federer when Sebastian was [11:50] cheating [11:52] and by the way someone should um [11:55] should tell chess base because they [11:57] still pretend in The Help menu that the [12:01] workout is a selection figures 98 no [12:05] it's not it's Hans Neiman [12:08] but okay it's not a proof of anything [12:11] because [12:12] for one game anything can happen [12:16] as it happened two times well let's see [12:21] hence game against Christopher you [12:25] um us protection [12:27] as you can see 100 let's just click [12:30] through the game [12:33] and just crashed his opponent obviously [12:39] and yeah [12:41] um very quick Victory 22 moves [12:45] 100 correlation with uh according to [12:48] chess base [12:50] and as it happened more than twice yeah [12:54] sometimes as we can see it's a victory [12:58] in 27 moves with blood by handsome man [13:01] against Miguel Soto [13:04] a first time [13:06] and victory against Astra whiskey a [13:10] victory in 28 moves with blight [13:13] checkmate [13:16] another time with black still Victory [13:20] against [13:21] um Martin Duke I don't know [13:25] um 20 20 moves victory [13:28] with black [13:30] another time a victory with white [13:34] um [13:35] in 31 moves congratulations to Hans [13:41] another time [13:43] against liban in 2021 [13:48] victory in 30 farmers [13:52] magnificent another time right 100 [13:56] correlation [13:58] a victory [14:00] in 38 moves congratulations [14:05] another time yeah Victory with his wife [14:09] 100 correlation [14:11] victory in 45 months [14:14] congratulations maybe to his opponent I [14:17] guess to have resisted that many moves [14:22] and another time against a strong [14:25] Grandmaster a victory with white [14:28] in 37 moves wow [14:32] so now let's go back to this Google [14:34] Sheets which was not compiled by myself [14:36] but by government man a person with whom [14:39] I've spoken a lot yesterday and we want [14:41] to remain anonymous [14:43] so are quite discussed to them and of [14:46] course uh I post a link in the [14:49] description so you can check by yourself [14:52] and you can even download the games to [14:56] um to see the let's check analysis [14:59] um they have done unjust base if you [15:02] have a chess base [15:05] so as you can see it's a compilation of [15:08] all these let's check analysis for Hans [15:11] Newman between 2019 to 2022 [15:16] so there are this shoe 100 [15:20] results and another another [15:24] thing that is very strange it's five [15:27] tournaments in a row and from the [15:30] Philadelphia open in July 2021 [15:35] um Hans paid at incredible at an [15:40] incredible level 69 percent 70 percent [15:43] 79 79 69 percent [15:47] so if you remember [15:49] 72 to 75 percent is correspondence for [15:53] Champion 72 percent it's Bobby Fischer [15:55] doing his 20 consecutive winning strike [15:58] and here for five tournaments in a row [16:01] and spayed at more than 73 percent uh on [16:06] average [16:07] so it's simply incredible and it made me [16:09] wonder but how it's possible that um Can [16:12] Wigan will with fitted top anti-cheating [16:16] expat and a great mathematician [16:20] has seen nothing and so I took a look at [16:23] um wigan's results [16:26] were publicly available on his website [16:31] so and of course I put the link in the [16:34] description so what matters is the ROI [16:37] which for [16:40] um niman between 2019 and 2022 it's [16:44] between [16:45] 62.3 and 36.6 [16:49] and as we can explain in his chest based [16:53] interview 50 of their have Hans results [16:57] are above 50 and 50 are under 50. which [17:03] is exactly what you would expect because [17:06] this Roi follows a normal distribution [17:10] so here we are on Wikipedia [17:13] it's a normal distribution centered on [17:16] zero with plus one Sigma plus two Sigma [17:19] minus one Sigma minus two Sigma and for [17:24] um regions Roi it's not zero it's 50 and [17:29] it's not plus one Sigma it's a 55 it's [17:32] 60 65 70 45 and 40 35 30. [17:39] and [17:40] according to Reagan's [17:43] um according to Reagan and results are [17:46] just normal like as you can see there [17:49] are three tournaments above 60 which is [17:52] more or less what you would expect like [17:54] two or three tournaments like this above [17:57] above 60 and also [18:01] Mountain amount above 40 but [18:04] two tournaments at precisely 42.2 so [18:09] normal results according to Reagan but [18:13] if we take a look at the ri according to [18:16] Reagan for this five tournaments and [18:19] even the sixth one because it's also [18:21] above 50. [18:23] and you see a different picture for [18:25] these tournaments because for the [18:28] Philadelphia it's 57.9 [18:32] um [18:33] Ry which [18:35] correspond to another of one in 18 so [18:41] it's not of course it's not a [18:43] probability of cheating it's [18:46] the opposite actually if you consider a [18:49] population of [18:52] players who are non-chitters [18:54] and one in every 18 tournament you will [18:58] see such a tournament has a hands had in [19:01] Philadelphia [19:03] so you see it's not at all proof of [19:06] cheating or an indication of cheating [19:08] it's [19:10] um [19:11] the probability of it happening [19:15] in case Hans doesn't cheat [19:19] so when in 18 is just it happens but the [19:24] second tournament in a row is one in [19:27] seven and the third one one in eighth [19:29] and one in sixth so as you can see [19:32] because you have to multiply the [19:34] probabilities uh it's getting stranger [19:38] and stranger the cumulative [19:39] probabilities are less than one percent [19:43] after two total amounts in a row one in [19:46] one thousand after three tournaments in [19:49] a row and after this six uh tournaments [19:54] in a row it's one in 70 [19:58] um [19:58] 77 [20:00] 000. [20:01] so very strange thing [20:04] so you might be wondering what is the [20:06] probability uh that hands has cheated [20:09] knowing this well we don't know it all [20:13] depends on your prior probability for [20:16] instance if you believe that you live in [20:18] a perfect world where nobody cheats then [20:21] the probability of hands are cheating [20:24] knowing this according to vegan is still [20:27] zero because well nobody cheats if you [20:30] believe we live in a world where one in [20:34] every 10 000 players cheats and that [20:39] Hansen man he's a an average player then [20:44] if you if we take like 18 [20:47] 80 000 players eight of them will cheat [20:50] and one of them will have this kind of [20:53] results without cheating [20:55] so we have nine players with these [20:57] results eight of them are cheaters one [21:00] of them is non-shitter so if you you [21:03] should take this one in ten thousand [21:05] prior probability then you will get [21:08] eight [21:10] um a chance out of nine that hence is a [21:13] cheater [21:14] so all this uh that are and all this [21:19] information that I gave you thanks [21:21] mainly to this combat man [21:24] um it's not a mathematical proof and [21:27] Alpha hands having cheated [21:29] but just let's say um very [21:34] not a proof but maybe a very big [21:37] evidence yeah I think in English you [21:40] have this difference between proof and [21:42] evidence that we don't have in French [21:43] for us it's just five but um in English [21:47] I would say that the evidence is high [21:52] but yeah not not a proof so do what you [21:56] want to visit [21:57] but you might be wondering what about uh [22:00] his game against minus Carlsen let's see [22:02] in his game from the F6 crypto cap as [22:06] that fat played on the chess 24 [22:09] that has won the famous chess picks for [22:13] itself game Hans had a [22:16] let's check analysis score of 67 percent [22:19] against 36 percent for the world [22:23] champion and in the even more famous [22:26] game as a single cup after which uh [22:30] minus quit the tournament Hans had a 71 [22:34] percent and let's check analysis results [22:37] so do what you want with this [22:39] information uh make your your own mind [22:43] make up your own mind [22:46] um whether Hans has ever cheated uh OTB [22:50] and whether he cheating uh he cheated [22:53] against the world champion [22:56] um I hope you have found this [22:59] new piece of evidence [23:01] uh interesting and thank you for [23:05] watching have a nice day ciao