World Cup Recap: Day 8 (Round three, second classical game)

Certain World Chess Champions might have made a splash by poking fun at the Giri-Leko matchup and predicting nine straight draws, but Anish was having none of it. Giri instead won today, advancing to round four and earning a rest day. It was one of many decisive results in a bloody day of battle. All four players who lost yesterday failed to bounce back and were eliminated, and five other matches where game one was drawn saw a decisive victor today! Less than half the field, seven of 16 matches, will continue with tie breaks tomorrow.

Here are the nine players who have already punched their ticket to round four, and how much better “100%” is than their odds entering the day. Note that none of these nine actually improved their odds by more than 50%, meaning all of them entered today as favorites to eventually advance:

Seed Player Rating Rd 4 Odds Gain Player Eliminated
3  Fabiano Caruana (USA) 2798.7 1.5%  Anton Kovalyov (CAN)
4  Anish Giri (NED) 2793.5 25.0%  Peter Leko (HUN)
8  Ding Liren (CHN) 2782.0 16.5%  Gadir Guseinov (AZE)
10  Dmitry Jakovenko (RUS) 2743.0 42.1%  Vassily Ivanchuk (UKR)
11  Sergey Karjakin (RUS) 2766.6 6.3%  Yu Yangyi (CHN)
19  Shakhriyar Mamedyarov (AZE) 2737.2 4.3%  S.P. Sethuraman (IND)
20  Radoslaw Wojtaszek (POL) 2740.5 28.5%  Julio Granda (PER)
24  Wei Yi (CHN) 2735.8 30.2%  Alexander Areshchenko (UKR)
26  Pavel Eljanov (UKR) 2743.0 9.4%  Alexander Grischuk (RUS)

Meanwhile, here are the seven matches going to tie breaks tomorrow:

Favorite Rating R4 Odds Gain/Loss Underdog Rating
1.  Veselin Topalov (BUL) 2809.9 89.1% 4.4% 97.  Lu Shanglei (CHN) 2620.2
2.  Hikaru Nakamura (USA) 2810.6 76.8% -3.0% 34.  Ian Nepomniachtchi (RUS) 2705.6
5.  Wesley So (USA) 2767.5 68.3% 3.3% 37.  Le Quang Liem (VIE) 2702.6
6.  Vladimir Kramnik (RUS) 2779.2 64.9% -5.3% 27.  Dmitry Andreikin (RUS) 2727.8
21.  Maxime Vachier-Lagrave (FRA) 2762.2 54.3% 5.5% 12.  Evgeny Tomashevsky (RUS) 2747.8
15.  Michael Adams (ENG) 2739.6 53.1% 5.6% 18.  Leinier Dominguez Perez (CUB) 2729.8
17.  Teimour Radjabov (AZE) 2730.6 51.5% 5.7% 16.  Peter Svidler (RUS) 2726.0

It’s worth discussing here that one of the decisions we made in our model was to use classical ratings (not rapid or blitz ratings) to project tie break results. For the most part we stand by this decision as the best choice, because rapid ratings in particular tend to be based on unreliably small sample sizes, and we feel that most of the time a player’s classical rating is a strong reflection of their overall chess strength at any time control. However in this particular round there are two matches that may be exceptions.

Veselin Topalov has a classical rating of 2810, but a blitz rating of just 2647. Lu Shanglei has a classical rating of just 2620, but a blitz rating of 2780. In other words, if the blitz ratings are accurate Lu might be as big of a favorite in blitz as Topalov was in classical! Now for blitz games to actually be played, two rapid mini-matches would have to be drawn first. Both Topalov and Lu have rapid ratings that are similar (slightly lower) to their classical ratings, but also both are “inactive” on the rapid ratings list. Topalov hasn’t played a rated rapid event since 2013, and Lu’s only rated rapid event in the last three years came 15 months ago. It is entirely unclear whether classical ratings or blitz ratings are a better predictor of a player’s skill in rapid chess. We suspect, but cannot prove, that classical ratings probably are more useful in predicting G/25 strength, but blitz ratings might be better in G/10. Topalov is therefore probably the clear favorite in the first pair of games, but maybe at best equal in the second pair of games should it reach that point, and quite possibly an underdog in the third pair of games if it goes the distance. Overall none of these considerations are built into our model, but the point is that the number one seed is probably much less than the 89% favorite our model currently indicates.

Another player who may be less of a favorite than the model indicates is Hikaru Nakamura. This is perhaps surprising, because he’s well known as perhaps the best speed chess player ever (currently ranked #1 in the world in rapid and #2 in blitz behind Carlsen). How, then, could rapid and blitz tie breaks not benefit him? Well his opponent this round is also a speed specialist. Nepomniachtchti is just 39th in the world in classical chess, but is ranked #10 in rapid and #4 in blitz. Nakamura’s classical rating edge is over 100 points, but his rapid and blitz rating advantages are both around 60. Looking at the classical rating, our model makes Nakamura a 77% favorite in the tie breaks, but if we treat him as only 60 rating points higher his odds would drop to about 67%. So obviously a player as strong in speed chess as Nakamura is a favorite in tie breaks, but perhaps not favored by as much in this case as we would think.

We’ll find out tomorrow what happens in these, and the other, tie break matches. For the rest of our analysis we will treat the model’s use of only classical ratings as valid, but keep in mind that the other odds we mention are possibly off slightly in this case.

So what about reaching the finals, and earning a spot in the Candidates Tournament? Seven players improved their chances by at least half a percentage point, and 10 players saw their odds fall by at least that much. Interestingly only four of the nine players eliminated today entered the day with better than a half percent chance, so six of those “biggest drops” came for players who are still alive, just not as well positioned as before. Here’s everyone who saw significant change to their odds from yesterday to today:

Seed Player Rating New Odds of Reaching Finals Finals Gain
4  Anish Giri (NED) 2793.5 23.2% 5.3%
10  Dmitry Jakovenko (RUS) 2743.0 7.6% 3.7%
8  Ding Liren (CHN) 2782.0 17.7% 1.7%
26  Pavel Eljanov (UKR) 2743.0 7.6% 1.4%
11  Sergey Karjakin (RUS) 2766.6 11.7% 1.1%
20  Radoslaw Wojtaszek (POL) 2740.5 5.3% 0.8%
24  Wei Yi (CHN) 2735.8 4.7% 0.7%
56  Alexander Areshchenko (UKR) 2680.7 0.0% -0.5%
16  Peter Svidler (RUS) 2726.0 1.9% -0.9%
5  Wesley So (USA) 2767.5 7.9% -1.0%
7  Alexander Grischuk (RUS) 2751.9 0.0% -1.0%
6  Vladimir Kramnik (RUS) 2779.2 9.3% -1.0%
12  Evgeny Tomashevsky (RUS) 2747.8 3.5% -1.2%
29  Peter Leko (HUN) 2709.2 0.0% -1.3%
2  Hikaru Nakamura (USA) 2810.6 24.8% -1.6%
1  Veselin Topalov (BUL) 2809.9 27.2% -2.1%
23  Vassily Ivanchuk (UKR) 2726.6 0.0% -2.4%

It’s worth mentioning again that Nakamura and Topalov, listed here as the second and third greatest drops, are in that position despite perhaps being given more credit than they deserve. If we keep everything else the same, but assume that both players are only 70% to advance tomorrow (possibly still generous in both cases) then their odds of reaching the finals become 21.3% for Topalov and 22.6% for Nakamura, making their “drops” today 8.0 and 3.8 percentage points, respectively, and making them the two players who had the worst day overall (which is surprising, when they are both favored to advance and nine others were eliminated, but that’s what happens when you’re the top two favorites to win the whole event, and you face the high expectations that go along with that).

One other critical occurrence in today’s games also had a strong impact on potential qualification for the Candidates Tournament: Grischuk lost his game (in the process of being eliminated by Eljanov, who is now a spectacular 6/6 in classical games at this event and hasn’t yet even sniffed a possible tie break). This means Grischuk lost six more rating points, and we now officially are projecting Kramnik, not Grischuk, to be third in the average ratings list. We now have Topalov and Giri at a combined 50.3% chance of reaching the finals (though only around 40% combined if you prefer 70% as Topalov’s odds in tomorrow’s tie break), so there is a very significant chance of third place on the ratings list being important. We now project Kramnik to have a 38% chance of earning one of the seven non-wild-card spots in the Candidates Tournament, while offering Grischuk just a 24% chance.

Advertisements

11 thoughts on “World Cup Recap: Day 8 (Round three, second classical game)

  1. Lu Shanglei has done surprisingly well in this tournament so far, first eliminating Moiseenko (in rapid tiebreaks, not blitz, even) and then Wang Hao (without even needing a tiebreak!). Now he’s scored two draws in classical with the top seed? Assuming that these results are indeed because he’s been “on form” (as opposed to being simply a statistical anomaly), I’d say he’s possibly the favorite to win the match. We’ll see if I’m right tomorrow!

    Like

    • I understand your enthusiasm and nationalism. However with that “on form” adjective, I beg to disagree. AFAIK, Lu’s best performance was last year’s World Junior championship which is not enough to justify that he’s in the same form (i.e., level) as Topalov or the other elite GMs to make him the odds-on favourite. If instead of Lu we’d talk about Wei Yi, I won’t argue.

      Having said that and to reiterate previous remarks, it’s possible that Lu may pull an upset tomorrow bcoz the World Cup KO format and rapid tie breaks do not engrain the ratings. As I’ve said, expect a Top 10 to be shown the exit door in this Rd 3.

      Like

      • Well, my confidence is partially due to Lu’s good results in this tournament so far. He’s shown Moiseenko and Wang Hao–neither of whom are weak–the door already, and that without even getting far enough to engage his opponents in blitz, which is his specialty. That’s what I meant by “on form”–not his previous result in the World Junior Championship, though that was impressive too. In addition to that, Lu’s specialty, as I already mentioned, is blitz chess–his rating in blitz is a whopping 2780. Add to that the fact that Topalov is weaker in blitz than in classical, and I think Lu has good chances.

        Also, nationalism? Heh, despite my surname, I’m 100% American-born and raised. Nakamura and Caruana are my favorites here. I just happen to like the “Cinderella” aspect of Lu Shanglei’s performance so far.

        Like

      • It’s now moot and academic: Topalov is just too classy for Lu. Another Top 10 elite GM was axed, but it was Kramnik. To reiterate what I’ve said before the start of the World Cup, I expect Topalov (and the other Top 4 seeds) to breeze his group, enter the Quarterfinals but not necessarily win the Cup.

        Like

  2. Also, I’ve wanted to bring this up for a while now, but I kept forgetting to do so. Wouldn’t it be better to display the increase/decrease of someone’s chances to get to the finals as a percentage rather than as a flat, additive change? For instance, Grischuk’s 1.0% decrease definitely hurts him more than Topalov’s 2.4%, but since the latter dropped by a higher absolute amount, you have him listed below Grischuk. On the other hand, if the changes were expressed in percentages instead, you could say, “Grischuk’s odds are 0.0% of what they were yesterday, and Topalov’s odds are 91.9% of what they were.” This seems to better express the actual impact of the changes on the player’s chances, rather than simply displaying a flat increase or decrease.

    Like

    • I see your point, but don’t really agree. I think Grischuk going from 1% to 0% really is less impactful than Topalov going from 29% to 27%. Grischuk’s odds entering the day of just 1% essentially meant he was not actually in contention to win. Being eliminated only makes that status official, it doesn’t necessarily make the status much worse. On the other hand Topalov is a pretty strong contender, he really is quite likely to reach the final. When his odds shift multiple percentage points in a day, that has a major effect on our understanding of who will eventually win. To me that effect is more significant than a player who was already practically eliminated just officially sealing the deal with a loss. It’s scratching a player off the list who wasn’t really on the list to begin with.

      Of course Grischuk’s first loss yesterday, which dropped him from 7.7% down to a mere 1.0%, and which could be seen as the game that “actually” eliminated him more than his loss today, was huge. And when he lost THAT game his 6.7 percentage point drop did show up at the bottom of our table.

      Today, though, didn’t really change much. Grischuk showed up as a dead man walking, whose loss yesterday had basically ruined his chances. What does the last one percent matter, really? At least that’s how I see it. I agree there’s something compelling about the definitiveness and absolutism of an elimination, but I see everyone’s odds as a form of equity in the final, and Topalov lost more equity than Grischuk did today.

      Like

      • Mm, fair enough. I’m not really willing to let the point slide quite yet, though. I think I just selected a bad example with Grischuk vs. Topalov. What about Tomashevsky’s decrease, versus Topalov’s decrease? Tomashevsky lost less percentage points in an absolute sense, but I feel that there is a sense in which Tomashevsky’s decrease can be said to be more significant than Topalov’s. After all, for every 11 times Topalov reaches the finals using yesterday’s odds, he can still do so 10 times with today’s. On the other hand, if Tomashevsky would have reached the finals in 11 trials yesterday, today he can only manage ~8. Or for an even more striking example, Peter Svidler: his chances to reach the finals were nearly halved by today’s result!

        My point is this: the less probability you have originally to make the finals, the more significant each result becomes for you, because the change will represent a larger shift relative to your odds. For me, at least, Oladapu Ada from Round 1 (seed no. 128) experiencing even a half-percent increase for his odds would be a huge result!

        Liked by 1 person

      • Yeah, I can agree with your point as rephrased. However by clarifying the phrasing it now only applies in some specific cases not universally, and I think it would be awkward to go back and forth between the two presentations in my articles.

        I think perhaps I’ve offered the information you’re looking for in a slightly different manner in the spots where I’ve looked at “change in odds of reaching the next round” (which is much more sensitive to immediate results, and to eliminations). Saying a player who went from 40% to advance down to 0% to advance, and was one of the biggest losers of the round in terms of odds of reaching the next round, and then omitting that player from the other list because he was never in contention and his odds of reaching the finals only dropped from 0.3% to 0.0% seems like perhaps the best approach to cover both ideas?

        Certainly if Adu had gotten his odds up to half a percent that would have been huge… because without factoring in ratings increases from the wins it would have taken to get there, he could have REACHED THE SEMIFINALS and still only had a 0.1% chance of reaching the finals (although of course if that had happened it would have suggested a flaw in the model, as you have to question things the second or third, much less sixth, time that a player listed as a 99.9% to 0.1% underdog wins, lol).

        Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s