Chess.com Fair Play Report: Estimating Cheating Rates in Titled Tuesday
In our first report on cheating in Titled Tuesday, we compared upset rates in over the board (OTB) blitz to upset rates in Titled Tuesday and found that cheating appears to be far more limited in Titled Tuesday than many believed. We also recognized that cheating does happen.
In this second report, we examine the issue of cheating in greater detail by digging into player performances in Titled Tuesday. We approach this by developing a statistical model that predicts the outcomes of Titled Tuesday games, and using it to simulate hundreds of thousands of historical Titled Tuesday games. It is unsurprising to find that every Titled Tuesday has players who overperform relative to expectations. Most overperformances do not indicate cheating and are explained by natural variance. To evaluate the prevalence of cheating, the key question is how many overperformances are expected compared to how many actually occur. We term the difference between these two the number of surplus overperformances – and it is at the heart of this report. Our key takeaway, visualized in the graph below, is that the estimated cheating rate (ECR) – which is the number of surplus overperformances divided by the total participants in a Titled Tuesday – is almost always below 2%, but generally greater than zero.
In other words, we can estimate that a large majority of Titled Tuesday events have one or more players cheating in them, but these players regularly make up a small fraction of Titled Tuesday participants. Importantly, in the one-year period from March 20, 2023 to March 19, 2024, the ECRs closely align with the actions taken by our Fair Play team, who closed 1.1% of Titled Tuesday participants for cheating that happened in Titled Tuesday in that same one-year period.
While not every overperformance is cheating, by calculating surplus overperformance we can provide the chess community with our best estimate of the extent of cheating that is occurring in Titled Tuesday.
