What The Numbers Say
What the Numbers Say
In the run‑up to the 2026 Candidates, every fan has an opinion. But what do the numbers say? A chess data blog, Chess by the Numbers, has crunched the performances of all 48 players who competed in the modern Candidates (2013–2022) to answer a deceptively simple question: if a player overperforms their rating in one Candidates, are they likely to do so again? The answer, it turns out, is complicated—but tantalisingly suggestive. With small sample sizes and plenty of caveats, the data hints that a few players may genuinely be “Candidates specialists”. And that could matter for how we view the 2026 field.
From 2013 to 2022, eight players competed in each of the six Candidates tournaments, yielding 48 individual performances. Of those, 11 players appeared only once and never returned, leaving 14 players with multiple appearances. The blog’s author was careful to note the statistical limitations: “Any attempt to analyze this question with such limited data is going to be ‘statistics’ in the sense of when baseball broadcasters tell you a players batting average with runners on scoring position – on Tuesdays – than it will be ‘statistics’ in the sense of the academic discipline.”
Nevertheless, they looked at three relationships: (1) first vs second appearance, (2) first vs all future appearances, and (3) average of prior performances vs final performance. The results were… interesting.
When comparing each player’s first Candidates to their second, the correlation was essentially nil. Half of the players improved; half did worse. The average change was just 5 rating‑performance points—statistically noise. “A player’s first Candidates appearance does *not* appear to predict their second result in any way,” the blog concluded. Good news for debutants like Sindarov, Wei Yi, Esipenko, and Blübaum: their first attempt offers no predictive signal for the second.
Looking at each player’s first appearance against the average of all their subsequent appearances, the trendline was flat—until one point was removed: Teimour Radjabov’s disastrous 2013 performance (‑146 relative to rating) followed by a strong 2022 (+75). With that outlier excluded, a slight upward slope appeared. Players who overperformed in their first Candidates tended to overperform later, and those who underperformed tended to underperform again. “The three biggest first‑event overperformances are players who overperformed their future events as well,” the blog noted.
For the 2026 field, this might suggest that Caruana (who overperformed in 2018 and has mixed results since) and Pragg (who had a neutral first outing in 2024) are less predictable, while debutants have no track record to lean on.
The most compelling finding came when the blog compared each player’s average overperformance across all prior Candidates to their final (as of 2022) performance. Again, Radjabov was an outlier. Removing him, the correlation became strong: players who had historically overperformed in the Candidates continued to do so, and underperformers continued to underperform.
The table of players with at least two prior appearances showed that the four players with the best prior average overperformance—Karjakin, Anand, Nepomniachtchi, and Caruana—all overperformed again in their most recent outing. For Caruana, the sample is complicated because his 2018 win was a huge overperformance (+80) while his 2020 and 2022 were slight underperformances (‑4 and ‑32). But overall, he has a net positive over his five appearances.
The blog’s author concluded: “Perhaps it is at least slightly valid to think that a select few players are legitimately and uniquely strong at this specific event.” They noted that Ian Nepomniachtchi’s prior average overperformance of 108 points (over two wins) is extreme, and that a simulation model might reasonably treat him as a 2780 player rather than his actual 2758. In the 2024 Candidates, this reasoning would have boosted Nepo’s odds—though in reality he finished second to Gukesh.
For 2026, the data suggest that among the veterans, Caruana (net +28 over four prior appearances) and Nakamura (net +15 over four) have positive histories, while Giri (‑14 over two) has underperformed. Among the debutants, we have no history to draw on—but that in itself may be an advantage: they have no track record of either success or failure in this unique format.
What does this mean for Cyprus? If the “Candidates specialist” effect is real, then Caruana, despite recent stumbles, remains a strong candidate because his prior performances (especially 2018) show he can peak in this format. Nakamura’s positive history (including a second‑place in 2024) suggests he can rise to the occasion. Giri, by contrast, has underperformed relative to his rating in both of his Candidates appearances—though his 2025 Grand Swiss win shows he has the capacity to break that pattern.
For the debutants, the numbers offer no guidance. Sindarov, Wei Yi, Esipenko, and Blübaum enter without a statistical profile in this event. They could be the next Gukesh (who had no prior record and won) or the next Abasov (who finished last). The data simply say: past performance in the Candidates is a weak but possibly real predictor—but only once you have a history to speak of.
The blog’s author was scrupulous about limitations: tiny sample sizes, the influence of outliers, the danger of over‑interpreting. “We will not be adjusting our simulation model to account for anything we found here,” they wrote. Still, the exercise is valuable because it reminds us that chess, even at the highest level, is not fully captured by rating alone. Form, format, and the peculiar pressure of the Candidates can amplify or suppress performance in ways that ratings miss.
For the 2026 edition, the numbers offer a lens—not a verdict. Caruana and Nakamura have shown they can handle this stage. Giri has struggled. The young guns are a blank slate. And whatever the statistics say, the tournament will be decided by fourteen rounds of hard‑fought chess, not by regression lines.
As the players prepare to meet in Cyprus, the numbers whisper a few quiet truths. Experience in the Candidates is not everything, but for those who have it, it may be a subtle edge. The true test, however, will not be found in spreadsheets. It will be found on the board, where the only statistic that matters is the final score.
Comments
Post a Comment