The crown jewel of the 2022 chess calendar is finally here. Eight Candidates will converge on Madrid and from the 17th of June until the 4th of July will, over 14 rounds, determine who earns the right to challenge Magnus Carlsen for the world championship.
Why are we so excited for this event? Of course the stakes are higher than anything other than a world championship match, and of course the talent level is through the roof. The field consists of players rated 2, 3, 4, 7, 8, 11, 15, and 18 in the world (per live ratings at the time of publishing this preview). In addition to all of that though, one of the most beautiful things about this tournament is how unpredictable it is. As critical as this tournament is to the world chess landscape, we go into this event with very little idea who will ultimately win.
Of course the cornerstone of our analysis, as always, is our computer model that simulates the event to translate the players’ Elo ratings into odds of winning first place. We will open this preview with the model’s projections, and then break down each participant in more detail. But we also have a lot of reason to be skeptical of exactly how accurate the model really is in this case, for a variety of reasons that are relatively unique to this particular tournament, and we will close this preview with a breakdown of all the extra reasons for uncertainty. Because to our mind, that only makes it more exciting to watch, as we have to wait and see what will ultimately happen!
But with no further ado, let’s take a look at our model.
These numbers are based pretty much exclusively on Elo ratings, so how much they tell us about an individual player’s chances is limited by the accuracy of that player’s rating – we’ll look closer at how accurate those ratings might be later on – but cumulatively there’s a lot to be gleaned from this chart. First of all: while there is an order of magnitude difference between first and eighth, it’s nevertheless true that everyone has a chance to win. This tournament does not have a pure sacrificial lamb whose chances can be entirely discounted, everyone starts out with a valid shot.
The second point is a corollary to the first. Nobody is positioned to run away with this. There is not an odds-on favorite whom the entire field is chasing from the beginning. And this is consistent with history – since this particular Candidates format was introduced in 2013 we have seen five tournaments and the pre-event rating favorite has only won once (that was Magnus in 2013, when he had a 62 point rating advantage on #2, yet only snuck through on tiebreaks). It certainly can’t be a bad thing to be the highest rated player in the field, but it’s no guarantee of anything in this format.
Our third takeaway is that collectively the top three ratings do impress our model. There is almost a 20 point rating gap between #3 and #4, and the odds really show how big of a gap that is if we assume the ratings truly reflect playing strengths. According to our model the three co-favorites have a combined 66% chance to win, while the rest of the field has just a 34% shot. Interestingly this does not align so well with recent history, as the #2 and #3 seeds have never won in this format. After Carlsen won as the #1 seed, the other four winners we’ve seen were seeded 4th, 7th, 5th, and 4th respectively.
So to summarize: everyone has a shot, the rating favorites are probably marginally more likely to win than those who are lower rated, but the uncertainty factor is through the roof here. Which as far as we’re concerned, uncertainty means excitement. This should be an amazing tournament to follow!
How did these eight players get here, how have they performed over the course of the qualifying cycle, and what do we need to consider about each of them beyond their rating, so that we can be more informed than if we just blindly followed the odds generated by the model? Let’s take a look.
First things first, some data. Here are performance ratings (and the sample size of games played in parenthesis) for each of the eight Candidates over three timeframes. Going back to January 1, 2020, as well as just the last year and just the last six months. This list is sorted from most to least active across the three time periods.
We will refer back to these numbers in each individual player’s writeup, so for now we won’t elaborate too much beyond this raw data, but it gives us a sense of both how active these players have been and how successful.
So with that said, let’s start at the bottom of our odds table and consider each player in turn (note: world rankings will be based on the live rating list as of this article being published, while ages are as of the tournament start date).
World Rank: #18
Qualifying method: Wild Card
Odds to win: 3.7%
Radjabov qualified for the 2020 Candidates tournament by winning the 2019 World Cup, but when he saw the perils of Covid on the horizon, he objected to holding the event, and after his objection was ignored, he withdrew. When that tournament was then postponed halfway through, his objections were validated, and FIDE ultimately committed to essentially compensate him by guaranteeing him a spot in the next Candidates. So now here we are in the odd situation where he essentially qualified for this event three years ago and has barely played chess since.
Our activity chart shows that Radjabov has played just 26 games in the past two and a half years, and has not impressed. After not playing at all in 2020, he played two events in 2021 and drew every single game (17 in total). Then just this month we finally saw him in action again, as he played dismally in the 2022 Norway Chess tournament, finishing in 9th place out of 10 and failing to win a classical game while losing three.
Importantly, we want to emphasize that struggling in one’s final tournament prior to the Candidates is most likely not a major concern. We will be mentioning such a struggle many more times as we work through this list of players. And it stands to reason that with the Candidates on the horizon, one might not go all-out in other events. Whether it’s a matter of conserving energy, keeping preparation hidden, or just not being fully focused on a “less important” tournament, it isn’t that odd that someone might underperform in any tournament they play after they’ve qualified for the Candidates. And especially as it draws closer.
So we strongly advise not to use the Norway results as a reason to dismiss Radjabov outright. But there are other reasons for concern as well. His inactivity runs deep and long, and it seems like a reasonable default to be highly skeptical of anyone who has been so inactive, until proven otherwise. So far, Radjabov has not proven anything (by say winning even one game) in the timeframe we’re looking at.
All of that said, Teimour was once ranked as high as #4 in the world, and even as recently as nine months ago was still ranked in the top ten. If this is the biggest underdog in the field, that drives home how strong of a field it really is. There may be many reasons for doubt here, but let’s be clear: he could win the whole thing. It’s a longshot, but we can’t rule it out completely.
Best case: If it turns out that time off hasn’t led to rust, the lack of recent wins has been circumstantial to low importance games, and old-school elite Radjabov shows up with years of accumulated preparation, he will almost certainly finish in the top half of the field and have serious chances to win the whole thing. We don’t think this scenario is likely, but it is possible.
Worst case: In an extremely strong field, Radjabov has been least active – and least successful when he did play – over the last couple years, and comes in with the lowest rating. Inevitably someone will end up at the bottom of the table, being the key target of everyone battling for first, and Radjabov is the player most likely to land in that victim role and suffer his way to a last place finish.
World Rank: #15
Qualifying Method: World Cup – 1st place
Odds to win: 5.5%
Duda is the second youngest player in the field, and the seventh youngest Candidate going back to 2013. He is still something of an up and comer, not yet having ever cracked the top ten of the world rankings in his career, but he has been excellent over the past year posting a 2783 performance rating despite a disappointment in his most recent tournament (Tata Steel 2022). As we said, disappointments in most recent tournaments will be a very common theme here, and not one we find concerning. Before that, the highlight is of course a 2832 performance rating at the World Cup to earn this spot (including famously being the player to knock Carlsen out of that event), but that was bracketed by a 2828 performance at the Prague Masters and a 2792 showing at the European Team Championships. In other words we know Duda can perform at a high level (although winning the Candidates has always required a performance of 2850+ in this format, and we haven’t quite seen him hit that high mark yet).
For Duda to win this tournament, it would take the greatest performance of his life. He has never before quite reached the heights it takes to win an event like this, but no doubt he has shown the potential to do so. Perhaps inexperience will be too great an obstacle to overcome, but like everyone in the field we also can’t count him out. It does seem like he’s on the cusp of a breakthrough and there’s no reason it can’t happen here on the largest stage. That’s why his odds sit at better than 1 in 20.
Best case: The Magnus-killer can clearly beat anyone when he’s on point, and if Duda shows up in top form he will be able to do very well even against a field as strong as this one. A performance like he had at the World Cup will likely be enough for a top three finish, and an outside shot at first place.
Worst case: Success at this stacked event will require a performance at the caliber we expect from a top ten player. Duda will be one someday, but isn’t one yet. If his time hasn’t yet come and his play here doesn’t live up to that standard, it won’t be pretty, and a bottom half finish becomes very likely.
World Rank: #11
Qualifying Method: FIDE Grand Prix – 1st place
Odds to win: 7.5%
Until quite recently, Nakamura hadn’t played classical chess since November 2019 and was considered retired, having traded in his career as a tournament player for a spectacularly successful new career as a professional chess streamer instead. As a result he has played even fewer games than Radjabov since 2020, but a key difference is that the games he did play, he made the most of. After receiving a wild card invitation to the 2022 Grand Prix, Nakamura shed any concerns that he might be “just” a streamer, by winning the first leg with a 2842 performance rating, then following up with a 2812 performance and second place finish in the final leg. This secured first place in the overall Grand Prix standings, and thus a berth in the Candidates.
Nakamura’s inactivity is in some ways comparable to Radjabov’s. Both have played very little classical chess in recent years, and both have strong histories as top players when they were active. There are also some key differences though, that point to Nakamura having the better chances of the two – and perhaps better chances than our model gives him credit for. First of all his past peak is higher, as Hikaru was once the #2 player in the world. Additionally, it’s sort of misleading to keep calling him inactive because in his “retirement” he has actually done nothing but play chess. Of course it’s been online, mostly blitz, and framed largely to entertain his audience. Streaming chess is different than playing a classical tournament. But it is still chess, and he has been immersed in the game. And from what we saw at the Grand Prix, that perhaps translates very well back to the tournament environment.
Our model measures Naka’s chances based on his current rating of 2760. That is up 24 points from where he was at the beginning of the year; that’s how many points he gained while winning the Grand Prix. But the open question is whether 2760 really reflects his playing strength. It certainly falls far short of his 2830 performance rating. Now the Elo formula is built that way intentionally, 20 games is too small of a sample size to simply assume that performance rating reflects “true playing strength”, so rating increases are intentionally more gradual. In this particular case though, from what we’ve seen so far it’s certainly believable to think that maybe his current strength is better than his current rating. Perhaps the rating gain at the Grand Prix was too conservative. Perhaps he is back in similar form to when he was rated over 2800 earlier in his career. Or perhaps not. We’re not making any promises here, but we’ve already seen him play that way twice this year, the only two times he’s played. If his play at the Candidates is in any way comparable to his play at the Grand Prix, he will be a very serious contender to win the whole thing. Is that going to happen? We will just have to wait and see.
Best case: If the Grand Prix truly represented Nakamura’s current strength, that would mean he’s playing the best chess of his life. A peak like that would make him a serious threat to win, or possibly even the overall favorite, even if he also literally doesn’t care.
Worst case: If the Grand Prix was a fluke, running hot over a relatively small number of games, and regression to the mean catches up with him, then Nakamura will not be able to contend. And if early struggles lead to a loss of focus or interest, that could snowball into a bad finish, perhaps even last place.
World Rank: #8
Qualifying Method: FIDE Grand Prix – 2nd place
Odds to win: 8.5%
Nine months ago Rapport was in almost exactly the same position as we previously described Duda having been in prior to Tata Steel: an up and coming player in his mid-20s rated in the 2760s who hadn’t quite cracked the top 10 of the world rankings yet. The difference is that Rapport proceeded to take that next step. He posted a 2832 performance rating at Norway Chess 2021 and climbed to #10, and then in 2022 added a 2793 performance at Tata Steel, and most importantly a spectacular 2870 performance in the second leg of the Grand Prix, which punched his ticket to the Candidates. In the process his world rank climbed as high as #5.
He has dropped a little from that peak. He struggled in his most recent tournament – there’s that refrain once again – but like with everyone we’re not super concerned about his 12 rating points lost at the Grand Chess Tour event in Bucharest. That event actually had half of the Candidates field in it, and all four lost rating. Not really a red flag for any of them, as previously discussed. Overall Rapport’s performance ratings over the various timeframes we’re looking at all strongly support his current rating being accurate, but individual events (particularly that Grand Prix leg he won) show that he has a spectacular ceiling when he plays his best chess, and that ceiling would definitely put first place at this event within his reach.
Best case: A highly creative player, Rapport can beat anyone when he’s sharp. We saw his potential peak in the Grand Prix, which was absolutely a caliber of play sufficient to also win a Candidates Tournament, should he maintain it for 14 rounds. Rapport is a serious threat to win.
Worst case: Creativity can backfire. If he overreaches, Rapport could lose a couple games early, and at that point he might be fortunate even to finish in the middle of the pack, while facing risk of dropping to the bottom of the table.
World Rank: #7
Qualifying Method: World Championship Match
Odds to win: 9.1%
Nepomniachtchi won the previous Candidates Tournament in 2020/21, and earned the privilege of losing to Magnus Carlsen in a world championship match for his trouble. With that comes entry back into the Candidates to try his hand at earning a rematch. Obviously we know with certainty that Nepo can win a Candidates because we have seen him do it, albeit under somewhat odd circumstances as the event was split into two halves. Combining the two, we get a performance rating just over 2850, and know that Nepo has that kind of capacity. However we have not seen it since.
After winning the Candidates, Nepo has only played three events. First at Norway Chess 2021 he posted a 2718 performance. Then in the world championship his performance came in at 2722. Finally he was the second of the four players we mentioned had disappointing results in Bucharest on the Grand Chess Tour, with a performance there of 2717. For those who want to aggressively overreact and assume that Nepo is just a ~2720 level player now, let’s be clear: that is very unlikely. A couple disappointing results and a rating drop from 2792 to 2766 are a slightly worrying trend, but it’s just three tournaments. Nepo hasn’t really been active enough to have any certainty about whether anything has truly changed for him or not, and clearly the player who won last time around is still in there somewhere.
Nepo’s ascent to world championship challenger is intriguing and worth taking a look at. His path actually looked quite similar to (and in fact even slower than) what we’ve previously observed in looking at both Duda and Rapport. He didn’t actually crack the top 20 in the world rankings until he was 26 years old, and was 28 by the time he reached the top ten for the first time. Then he climbed as high as #5 while bringing his rating up to a new peak of 2776 and earning a spot in the first Candidates of his career, but he entered that event seeded 4th by rating, and with no history of proving he could compete at a level quite high enough to win such a prestigious event. That did not stop him, as we well know.
Perhaps Nepo’s story last time around bodes well for the chances of others who are in their first Candidates or who don’t have the most astronomical peak ratings yet in their career – it at least shows that those aren’t hard prerequisites to win this event. Alternatively, maybe it just says something about Nepo himself, and rather than extrapolating to others, the takeaway should be that he is a bigger threat to win than our model gives him credit for, for some unique reason. Of course since our core emphasis is uncertainty, we also have to recognize that if he does come in playing around a 2720 level like he did the last couple times he sat at a chess board, it could get bad for him fast in an event this tough. So there is a wide range of possible outcomes here, as there are for everyone, but certainly any former winner of this event is a threat to win it again.
Best case: He won the last one, and he could win this one too. Don’t let his struggles in the match against Magnus make you forget how strong Nepo is. He did go punch for punch with the world champ for 5 ½ games, and if he plays like that (or the way he played in the 2020/21 Candidates) it’s not clear how anyone can finish ahead of him here.
Worst case: There have long been concerns about Nepo’s stamina over the course of long events. He was perhaps fortunate that the last Candidates got split in half, allowing him to essentially play two strong 7-game tournaments rather than having to maintain his level for 14 straight games. After his strong start, we saw him collapse in the world championship match, losing some games in relatively shocking fashion. A collapse like that in the second half here could lead to a dismal bottom-half finish, even if he starts out strong.
World Rank: #4
Qualifying Method: FIDE Grand Swiss – 2nd place
Odds to win: 15.5%
For much of recent memory, Caruana has been a fixture in the #2 spot in the world rankings. He held that position on every rating list from May 2018 through September 2021. Carlsen was the world champion, ranked #1, and then there was Fabi behind him. He hasn’t been ranked outside of the top ten since 2013, when he was just 20 years old. This will be Caruana’s fourth consecutive Candidates appearance and in the prior three he tied for second, then won, then tied for third, and never scored worse than 7.5/14. So we are talking about a well established veteran of this format, with a long history of success both in the Candidates and overall. History suggests perhaps he ought to be seen as a favorite to win.
However there is cause for concern in his recent results. Fabi has played more games in the past year than anyone else in the field, but hasn’t been quite so dominant as we’re used to seeing. At the 2021 Superbet Chess Classic he had an uncharacteristic negative score and a performance rating of just 2700. Then he was knocked out of the 2021 World Cup in just the second round he played. Then he struggled in the US Championships, only posting a 2733 performance even after a late surge. And in 2022 we’ve seen underperformances at Tata Steel and – like half the field – the Superbet Classic in Bucharest.
Don’t misunderstand. It hasn’t been all doom and gloom for Caruana. He has had positive results, and even some of the “struggles” we’ve listed would be solid results for most players, and are only disappointing because he has established such high standards for himself. A great reality check is that he is currently ranked #4 in the world and that has us asking “what’s wrong with him?” when only three of his opponents in the event have ever been ranked higher. But his overall performance ratings in the 2760 vicinity over the past year do make it look like perhaps he’s dropped slightly from definitively being the best player not named Magnus to being “just” a typical top ten caliber player. Of course even if that’s true, he would have chances to win this tournament, but it also might mean a risk of another disappointing event and perhaps the first bottom-half result we’ve ever seen from Candidate Fabi. Alternatively though, if he shows up in the vintage Caruana form from years past then he’ll have even better chances than what our model says. So he’s very likely to be one of the top players in the field, competing until the end, and he may be the “real” favorite despite what the numbers say, but we want to see him prove at the board that nothing is “wrong” with him before we buy in fully.
Best case: If experience is the deciding factor, we could see the veteran win this event once again. He’s had a plus score in three prior Candidates, and if he does it again he should easily finish in the top half, and probably have first place within his grasp through the final rounds, leaving conversion as the only question.
Worst case: There’s a nagging feeling that something about Fabi’s play has been “off” for the last couple years, and if there turns out to be truth to that and he stumbles at the wrong time, he does run the risk of quickly turning from favorite to also-ran, perhaps dropping into the bottom half of the standings, and failing to maintain his streak of plus scores.
World Rank: #3
Age: 18 (Turns 19 on day 2)
Qualifying Method: FIDE Grand Swiss – 1st place
Odds to win: 20.7%
Firouzja is perhaps the hardest player in the field to predict. Let’s start with the upshot. He comes in as the second highest rated player in the field, making him a favorite of our model, and on top of that there are a variety of reasons one could argue he might still be underrated. For one, he’s still a teenager, which is an age at which most chess players are still rapidly improving their game. Since qualifying for the Candidates last fall, he has been almost entirely off the grid. If he spent that time improving, and can avoid any issues with rustiness, he could play the Candidates at a higher level than we last saw from him. Which is a proposition that should terrify his opponents, because we last saw him at a nearly inhuman level.
Let’s talk about the fall of 2021. In September, Firouzja had a remarkable 2871 performance rating at Norway Chess, finishing second to Carlsen (and notably ahead of two of his opponents in this event – Rapport and Nepomniachtchi). From there he proceeded to win the Grand Swiss with a 2855 performance rating, locking up his spot in the Candidates. How did he celebrate? With an astonishing 3015 performance at the European Team Championship, scoring 8/9! Suddenly Firouzja was the youngest player ever rated 2800+ and ranked #2 in the world. A prodigy’s potential had been realized. In total, since the beginning of 2020, Firouzja has the second best performance rating in this field, and in a much larger sample of games (86) than the player who has performed better (Nakamura with just 20 games).
Firouzja also has the highest score against other players in the field, having scored 56% in his collective games against the 7 opponents he will face here. We want to take a moment to discredit a persistent but false narrative. Many discussions of Firouzja claim he got to 2800 by beating up on players rated in the 2600 range, but that he hasn’t proven himself yet against the 2700+ opposition he will face in this event. That is simply false. From October 2020 to November 2021, as he improved his rating from 2754 to 2804, Firouzja played 19 games against opponents rated between 2700 and 2799, winning 8 and losing just 1. Only two of those wins (and only four of the games) came against opponents under 2750, and the wins included five games against other Candidates: Nepomniachtchi, Rapport, and Duda (thrice). It can be said that Firouzja hasn’t yet proven himself against 2800+ opposition, but he’s not the only player with a Magnus problem and that probably can’t be considered a black mark for this tournament where Magnus isn’t competing.
So why isn’t Firouzja the clear and definitive favorite? Experience. For all the success we saw last fall, and as splashy as he’s been as a prodigy, we have not yet seen him play on a stage like this. We simply can’t be certain how he’ll hold up over 14 rounds, under the pressure a Candidates Tournament brings. And yes, this is an open question for many other players in the field as well, but it seems like an effect that could be magnified for a teenager. The Candidates Tournament is a relatively unique challenge in the chess world, where rating is less predictive than normal (more on that later) and other factors could ultimately determine who emerges on top. We don’t know – yet – how those factors will shake out for Alireza. Maybe very well, but maybe not.
Best case: He is the heir apparent. If he was only getting started last fall, and this teenager has spent the last six months continuing to improve, there’s a hidden chance that he’s truly on another level by now, with a chance to steamroll the field. Alireza may have the highest upside of anyone, and if everything clicks he might win first place in dominating fashion.
Worst case: It’s not his time yet. The Candidates is a grueling high pressure test, the likes of which Firouzja has never experienced. Raw talent may not be enough to carry him if he cracks under that pressure, and things could go very wrong very fast if he’s not ready for this challenge. Of the top three players, Alireza might be the one most likely to win by a large margin, but if his mental fortitude fails, he’s also probably the one most likely to finish in the bottom half of the standings.
World Rank: #2
Qualifying Method: Alternate by rating
Odds to win: 29.4%
This spot initially belonged to Sergey Karjakin who finished second in the World Cup, but he was then banned from competitive play for six months for violating FIDE’s code of ethics with his vitriolic comments supporting Russia’s invasion of Ukraine. With Karjakin unable to play, the rules called for the highest rated active player as of May 1st to replace him. Ding was the highest rated player not already in the field, but “active” status required 30 games and he had only played four in the designated timeframe. And so the Chinese Federation went to work on his behalf, rapidly organizing a series of tournaments allowing him to play an astounding 28 games in 30 days, meeting the activity criteria with just a few days to spare. Perhaps even more impressively, Ding managed to gain rating over the course of this marathon, as he posted a 2818 performance rating over the span and moved up to the #2 spot in the world rankings.
Ding is, along with Nakamura, one of only two players not entering this event after a poor showing in their last tournament. However if we’re looking for reasons not to anoint the highest rated player as the favorite, it’s worth considering that it’s not entirely clear how motivated anyone was – him or his opposition – during the 28 games that earned his spot here. The games just needed to be played, the results didn’t make a huge difference, so although the results were actually quite good it’s not certain that we should interpret that as him coming in with a high level of play. Aside from those games, we have almost no meaningful recent activity from him, and with that comes questions about what type of form to expect. So if – as a premise – we treat his recent games with skepticism, we could then logically raise some doubts as a result. It’s a shaky premise though; we generally are very reluctant to toss out data arbitrarily.
So all in all Ding is just quite simply a favorite. He has rating. He has the experience – this is his third consecutive Candidates appearance. He has the talent – he once went 100 games without a loss, and he has proven he can beat anyone. Unlike in 2020, all signs are that he had the proper opportunity to prepare for this event, and he is on his way to Madrid having successfully procured a visa (a challenge since Covid hit that largely explains his inactivity).
Best case: The favorite wins. There are no complex justifications needed for Ding, absolutely nobody should be remotely surprised if he wins this event. He may not be an odds-on favorite, but we have him as the most likely winner and we can’t imagine any way to analyze this event that doesn’t at least have him as a top three contender; he’s the only player we can say that for.
Worst case: Something unexpected could go wrong. Whether illness strikes, or whether it’s just a fluke, as solid as we typically expect him to be, Ding is of course not guaranteed to finish in the top three. He could drop lower in the standings, and even last place might be a possible result although of everyone in the field we would say he is the least likely to finish last.
We’ve gone over every player’s chances of winning based on our model, but how accurate is it as a predictor? To be entirely frank: not very. Bear with us please because we shared those odds for a reason and believe sharing them is valuable, but let’s first look at the reasons they aren’t everything. Our model is designed to be relatively simple. We believe we do the most to enhance the experience of following the tournament as a fan (our ultimate goal) if the model makes sense to our readers rather than being a black box. This simplicity comes at the expense of including every possible variable, even if doing so might possibly increase accuracy.
So what does our model do? Simply put the only inputs are the players’ ratings, a rough estimate of how much having the white pieces is worth, and a rough guess at the expected draw rate. Essentially the model is translating Elo to percentages. What is the purpose of this? Why not simply sort by rating? If a player is rated higher then of course a rating-based model will say they have a better chance to win, but we feel it’s not at all intuitive how much bigger. What is a 10 point rating edge worth? What about a 20 point edge? A 50 point edge? We use the model before the tournament begins to answer that question. Even though we know that the model is only accurate to the degree that those ratings accurately reflect the level of play we will see from each competitor.
Now most of the time this disclaimer isn’t necessary because Elo really is an excellent predictor of chess tournaments most of the time. But the Candidates is simply different. Players can often be rated slightly off from their level of play, and their form can vary from event to event, but those errors are generally relatively small. And most tournaments have relatively large rating gaps between the best and worst players, so those small errors get swallowed up and the odds come out relatively on point. The Candidates does not have those large rating gaps, and it’s a longer tournament (relative to the size of the field) than we often see, so that reduces variance, meaning that even small errors between a player’s rating and their “true strength” can throw off the model’s accuracy much more than normal.
And at the same time, those errors tend to be larger at the Candidates. This is an entirely unique event in that most players see it as the most important opportunity of their career, other than a world championship match should they reach one. Players will hold back opening novelties for years, not using them in other tournaments, just so they have a chance to use those ideas here. Players prepare much more rigorously for the Candidates than a “normal” event, and just approach these games differently. Which sometimes means that many players here don’t play in a manner that might be typical for them, and as a result their level of play is far more divorced from their rating than we usually expect.
Okay, enough words, you want numbers. This current Candidates format has been used five times dating back to 2013, so five events times eight players gives us 40 data points. Here is the graph of players’ ratings before each Candidates began versus their performance ratings in those events.
That is a random cloud of dots. While the trendline has a very minimal upward slope, the ultimate conclusion here is that rating simply doesn’t predict performance at the Candidates. So again, why are we building all our coverage of the event around a rating-based model? Well, to paraphrase Winston Churchill: Elo is the worst way to predict the Candidates except for all the other ways.
We searched for other variables that might do a better job, and were willing to consider tossing our regular model aside in order to build one more suited to the unique elements of the Candidates, but we couldn’t find a way to improve. Part of the problem is that 40 data points just isn’t that much to work with. So we’re rolling with the same model we’ve always used.
We want to emphasize that understanding this about our model’s odds is precisely the reason why we also investigated recent performances, and why we chose to highlight the things we did in our player writeups. We specifically wanted to add the context that is missing from the model’s output.
That being acknowledged, we also do believe the model still has a lot of value. It does still show, quite clearly, how much value (percentage-wise) can be found in various differences in playing strength (as measured in Elo terms). So even if you believe we have the players in the wrong order, the odds should still give a meaningful sense of how big of a favorite the top players are, whichever players those are. And, critically, we will be updating the model after every round. As games go into the books, each player’s chances to win the tournament will rapidly depend less on how they might play in the remaining games (hard to predict) and more on the known quantity that is completed results.
So while our pre-tournament odds answer the relatively nebulous question of how valuable 20 Elo points might be, as the tournament progresses our updated odds will answer more concrete questions like “how valuable is a 1 point lead over the field” or “how hard is it to come from behind with this many rounds remaining?” We believe that seeing how the odds shift as the tournament progresses can bring tremendous added value to the fan experience, even if there are valid doubts about whether the pre-tournament odds started out perfectly accurate. And in our round by round updates we will walk you through the key takeaways from whatever shifts we saw in the odds, and help identify which upcoming games are the most critical to the overall results.
So to close out this preview, let’s look at a more concrete question now. Exactly how valuable or damaging is it to win or lose in the first round of a long tournament like this?
We are now going to look at all four games in the first round, and see what would happen to each player’s odds to win the tournament, if that game ends in either a white win, a draw, or a black win.
|Adjusted Odds if…|
|Player||Initial Odds||White wins (20%)||Draw (66%)||Black wins (14%)|
Duda and Rapport are two of the model’s relative underdogs, so this game grades as the least important of the round. However we do see that each player has a chance to roughly double their odds of winning the tournament if they can get off to a 1-0 start, with the counterpoint that a loss for either player would be extremely costly, cutting their already slim odds more than in half. Clearly this creates a challenging risk/reward question should a situation arise where a player has the opportunity to take a safe draw, or to seek complications. How the players handle this question in the first of 14 rounds will be a very interesting question to consider as the game unfolds.
|Adjusted Odds if…|
|Player||Initial Odds||White wins (14%)||Draw (66%)||Black wins (20%)|
Alireza opens with black, yet our model favors him to win, as his rating edge exceeds white’s advantage. If he did win, his odds of first place would improve to a much more favorable 32% – possibly enough to make him the new favorite if Ding didn’t also win – but again this comes with a serious risk as his odds could be cut in half should he lose. So is the youngster prepared to play for a win with black? He’s going up against ostensibly the weakest player in the field, and it’s possible that wins over Radjabov may prove a deciding factor at the top of the final standings, but we also haven’t actually seen confirmation yet that Radjabov really is a weak opponent that can be picked on at this event. We talked before about how volatile Firouzja’s range of possible outcomes is, and how aggressively he approaches this game (and how that decision works out) may set the tone for his entire tournament.
|Adjusted Odds if…|
|Player||Initial Odds||White wins (29%)||Draw (59%)||Black wins (13%)|
The two Americans come in both relatively middle of the pack in the odds, but with Fabi’s chances about twice those of Naka – according to the model. Those numbers would almost exactly flip, though, if Nakamura can win this game and serve notice to the field that he’s the same threat here that he was to the opponents he shredded in the Grand Prix. Fabi, on the other hand, has the comfort of the white pieces, and according to the model a 29% chance to win the game, and if he can do so it would be his own opportunity to make an assertion: don’t worry about a couple stumbles I may have had in previous events, I’m back to my true self and ready to win.
|Adjusted Odds if…|
|Player||Initial Odds||White wins (33%)||Draw (55%)||Black wins (12%)|
In the game our model deems most important (measured in terms of how much both player’s odds could potentially swing) our favorite rolls in with a chance to start off on his best foot with the white pieces, and cement his status as the most likely winner. However if he slips up and loses, he would drop to essentially the average odds to win (if all eight players had equal chances, that would put each of them at 12.5%) and in that scenario suddenly Nepo would be better than a one in five shot to win the Candidates for the second time in a row.
We can see that this tournament has a chance to start off with a bang. Even in round one, and even in such a long event, a first round win can have a very positive impact on your chances to win the event, while an opening loss can be extremely damaging. Does the former point mean players will come out swinging? Or will the latter point mean that this round sees some cautious starts as the players feel each other out? The stakes are astronomical from day one, and now you know as much as we do about all the implications. We have a number of possible storylines that we’re ready to watch for, and are prepared to spot others as they emerge. We have acknowledged that we really have almost no idea what has happened, and embraced that as a cause for excitement. So there’s only one thing left to do. Wait impatiently for the games to begin!