chessgames.com
Members · Prefs · Laboratory · Collections · Openings · Endgames · Sacrifices · History · Search Kibitzing · Kibitzer's Café · Chessforums · Tournament Index · Players · Kibitzing

🏆 Grand Prix Riga (2019)

  PARTICIPANTS (sorted by highest achieved rating; click on name to see player's games)
Levon Aronian, Wesley So, Maxime Vachier-Lagrave, Veselin Topalov, Hikaru Nakamura, Shakhriyar Mamedyarov, Alexander Grischuk, Anish Giri, Sergey Karjakin, Peter Svidler, Pentala Harikrishna, Yu Yangyi, David Navara, Nikita Vitiugov, Jan-Krzysztof Duda, Daniil Dubov

Chessgames.com Chess Event Description
Grand Prix Riga (2019)

The Riga FIDE Grand Prix took place in the National Library of Latvia from July 12-24 2019. The 16-player knockout was the 2nd of four legs of the 22-player Grand Prix series that will determine two places in the 2020 Candidates Tournament. Players compete in 3 of the 4 tournaments, which each have a 130,000 euro prize fund, with 24,000 for 1st place. There are from 1 (quarterfinal loser) to 8 (winner) Grand Prix points available, plus an additional bonus point for each match win without tiebreaks. The overall series prize fund is 280,000, with 50,000 for 1st place.

Each round consists of two games of classical chess, with a time control of 90 minutes/40 moves + 30 min to the end of the game, with a 30-second increment from move 1. If the match is tied two 25+10 rapid games are played. If still tied, there are two 10+10 games, then two 5+3. Finally a single Armageddon game is played, where White has 5 minutes to Black’s 4 (with a 2-second increment from move 61) but Black wins the match with a draw. Leading partners of the series are PhosAgro and Kaspersky Lab.

Shakhriyar Mamedyarov beat Vachier-Lagrave in the Armageddon game of the final and collected 10 Grand Prix points.

Round 1 July 12-14 Quarterfinals July 15-17 Semifinals July 18-20 Final July 22-24

Karjakin ˝˝ 10 ˝˝ ˝˝ 1 5 Giri ˝˝ 01 ˝˝ ˝˝ 0 4 Karjakin ˝˝ ˝˝ ˝˝ 0˝ - 3˝ So ˝˝ ˝˝ ˝˝ 1˝ - 4˝ So ˝˝ 1˝ -- -- - 2˝ Harikrishna ˝˝ 0˝ -- -- - 1˝ Mamedyarov 1˝ -- -- -- - 1˝ So 0˝ -- -- -- - ˝ Svidler ˝˝ ˝0 -- -- - 1˝ Duda ˝˝ ˝1 -- -- - 2˝ Duda ˝˝ ˝0 -- -- - 1˝ Mamedyarov ˝˝ ˝1 -- -- - 2˝ Mamedyarov ˝1 -- -- -- - 1˝ Dubov ˝0 -- -- -- - ˝ Mamedyarov 10 ˝˝ ˝˝ 10 1 5 Vachier-Lagrave 01 ˝˝ ˝˝ 01 0 4 Vitiugov ˝˝ 00 -- -- - 1 Grischuk ˝˝ 11 -- -- - 3 Grischuk ˝˝ 1˝ -- -- - 2˝ Yu Yangyi ˝˝ 0˝ -- -- - 1˝ Aronian ˝˝ ˝˝ ˝˝ 10 0 4 Yu Yangyi ˝˝ ˝˝ ˝˝ 01 1 5 Vachier-Lagrave ˝1 -- -- -- - 1˝ Grischuk ˝0 -- -- -- - ˝ Nakamura ˝˝ ˝0 -- -- - 1˝ Topalov ˝˝ ˝1 -- -- - 2˝ Topalov 0˝ -- -- -- - ˝ Vachier-Lagrave 1˝ -- -- -- - 1˝ Vachier-Lagrave 1˝ -- -- -- - 1˝ Navara 0˝ -- -- -- - ˝

Grand Prix points:

Pts Bonus Tot Mamedyarov 8 2 10 Vachier-Lagrave 5 3 8 Grischuk 3 0 3 So 3 0 3 Duda 1 0 1 Karjakin 1 0 1 Yu Yangyi 1 0 1 Topalov 1 0 1

Official site: https://worldchess.com/news/2089
Chess.com: https://www.chess.com/news/view/mam...
ChessBase: https://en.chessbase.com/post/fide-...
Chess24: https://chess24.com/en/watch/live-t...
TWIC: https://theweekinchess.com/chessnew...

Previous (and 1st) GP event: FIDE Grand Prix Moscow (2019). Next: Grand Prix Hamburg (2019)

 page 1 of 3; games 1-25 of 69  PGN Download
Game  ResultMoves YearEvent/LocaleOpening
1. M Vachier-Lagrave vs Navara 1-0192019Grand Prix RigaB11 Caro-Kann, Two Knights, 3...Bg4
2. Nakamura vs Topalov  ½-½392019Grand Prix RigaC84 Ruy Lopez, Closed
3. Aronian vs Yu Yangyi  ½-½292019Grand Prix RigaD10 Queen's Gambit Declined Slav
4. Vitiugov vs Grischuk ½-½452019Grand Prix RigaC58 Two Knights
5. Mamedyarov vs D Dubov  ½-½352019Grand Prix RigaD33 Queen's Gambit Declined, Tarrasch
6. Svidler vs J K Duda  ½-½422019Grand Prix RigaC24 Bishop's Opening
7. W So vs Harikrishna ½-½452019Grand Prix RigaC53 Giuoco Piano
8. Karjakin vs A Giri  ½-½162019Grand Prix RigaC65 Ruy Lopez, Berlin Defense
9. Topalov vs Nakamura  ½-½252019Grand Prix RigaC53 Giuoco Piano
10. Yu Yangyi vs Aronian  ½-½252019Grand Prix RigaE06 Catalan, Closed, 5.Nf3
11. Grischuk vs Vitiugov ½-½102019Grand Prix RigaE11 Bogo-Indian Defense
12. D Dubov vs Mamedyarov 0-1782019Grand Prix RigaA34 English, Symmetrical
13. J K Duda vs Svidler  ½-½422019Grand Prix RigaA15 English
14. Harikrishna vs W So  ½-½292019Grand Prix RigaE06 Catalan, Closed, 5.Nf3
15. A Giri vs Karjakin  ½-½302019Grand Prix RigaE01 Catalan, Closed
16. Navara vs M Vachier-Lagrave  ½-½462019Grand Prix RigaB51 Sicilian, Canal-Sokolsky (Rossolimo) Attack
17. Vitiugov vs Grischuk 0-1242019Grand Prix RigaD32 Queen's Gambit Declined, Tarrasch
18. Yu Yangyi vs Aronian  ½-½322019Grand Prix RigaC50 Giuoco Piano
19. Aronian vs Yu Yangyi  ½-½462019Grand Prix RigaD11 Queen's Gambit Declined Slav
20. Aronian vs Yu Yangyi  ½-½632019Grand Prix RigaD11 Queen's Gambit Declined Slav
21. Yu Yangyi vs Aronian  ½-½452019Grand Prix RigaE06 Catalan, Closed, 5.Nf3
22. Aronian vs Yu Yangyi 1-0572019Grand Prix RigaD11 Queen's Gambit Declined Slav
23. Yu Yangyi vs Aronian 1-0442019Grand Prix RigaE32 Nimzo-Indian, Classical
24. Aronian vs Yu Yangyi  ½-½532019Grand Prix RigaD11 Queen's Gambit Declined Slav
25. Nakamura vs Topalov  ½-½502019Grand Prix RigaA04 Reti Opening
 page 1 of 3; games 1-25 of 69  PGN Download
  REFINE SEARCH:   White wins (1-0) | Black wins (0-1) | Draws (1/2-1/2)  

Kibitzer's Corner
< Earlier Kibitzing  · PAGE 14 OF 14 ·  Later Kibitzing>
Jul-27-19
Premium Chessgames Member
  AylerKupp: <<devere> The form of qualification that is least subject to form is rating (not average rating!). Rating measures cumulative performance over a lifetime of chess playing.>

True, but remember that all that any event can do, including the WCC match, is to determine who the best player is during the time that the event was conducted. And that applies to any sport. For example, does anyone think that the New York Giants in American football was a better team than the New England Patriots who finished 16-0 during the season and 2-0 in the playoffs? But the New York Giants were the better team during Super Bowl XLII, winning 17-14, and that's all that matters.

So I think that what we are most interested in is who the best player is <currently> or at least for some time in the recent past; 2 years seem logical given that's the current period of the WCC cycle.

Consider using instead of ratings a form of TPRs. TPRs do not use player differentials like ratings calculations do, except when taking the average rating of a player's opponents in a tournament to calculate the final TPR. If we don't do this and use only the player's performance against other player's in a tournament (I'll call it Tournament Performance Score or TPS), then if we calculate each player TPS for, say, a 1-year or 2-year period, then each player starts out on an even basis with all the other Candidates Tournament contenders, and then player with the highest average TPS during that period would qualify for the Candidates Tournament.

We would have to add some additional constraints like a minimum number of games in, say, every 6-month period, to prevent any player on form at the beginning of the TPS calculation period from sitting on their initial TPS score and a minimum rating (say 2750, just like it was done for determining who qualified by rating to the 2018 Candidates Tournament).

Thoughts?

Jul-27-19
Premium Chessgames Member
  beatgiant: <AylerKupp>
<Besides, I don't think that the discussion above has anything to do with ratings themselves except tangentially.>

At the root of it is the statement by <devere> above that <rating measures cumulative performance over a lifetime of chess playing.>

If we think it accumulates linearly same as runs in a baseball game, then yes, it makes no sense to use the average of monthly ratings over a year. But the belief that it accumulates linearly is precisely where we differ, and that's a point about ratings math.

Jul-27-19  devere: < AylerKupp> Surprised to learn that you are also among the math-challenged. If players A & B start the year at 2700, and player A wins a game and gains 5 points in month 1, while player B wins 11 games and gains 55 points in month 12, Player A's average rating for the 12 months is 2705 while Player B's average rating for the 12 months is 2704.58.
Jul-27-19
Premium Chessgames Member
  beatgiant: <devere>
Your latest example does not satisfy the requirements to be eligible by rating for the Candidates: must play at least 30 rated games over the 12 months, of which at least 18 are in the last 6 months.
Jul-28-19
Premium Chessgames Member
  Sokrates: <AylerKupp> <...And others have also volunteered their forums for these discussions. But who would find these discussions in individual' forums if they want to participate in them? Just like no one would find them on this page (and others) unless they are truly determined to do so. And how many of us are willing to do that? ...>

I am sure you understand the point: Getting these mileslong posts and discussions about a peripheral issue away from the main CHESS forum and refer them to a place where they may be discussed infinitely - to the joy of those who are enthusiastic on the issue and to the equal joy of those who are not.

Jul-28-19  Absentee: <Sokrates: Hello, <Absentee>, You've certainly been an absentee for quite a while. Sort of missed you as a great contradictionee! :-) Hope you're all right ...>

Hey there! Thank you, everything's fine, I just haven't been following chess for a while. But I still take a peek to these pages now and then.

Jul-28-19
Premium Chessgames Member
  Sokrates: Hi, <Absentee>, Glad to hear you're fine. I know about on and offs on interests - had a very special one for almost twenty years, but then I suddenly had enough of it. It seems that I am permanently stuck with chess, though! :-)
Jul-28-19
Premium Chessgames Member
  beatgiant: <devere>
The distortion you are complaining about occurs because of the use of a previous month's rating in case of inactivity, which is why the activity rules matter.

Ideally, each player would play a lot of rated games every month, and that problem would go away. But that's not a realistic expectation.

In reality, the players who are affected by this criterion (e.g. Ding Liren, who is currently in the lead for the "top average rating" slot) are maintaining reasonably active schedules, just as we would expect for people who make their living by playing chess.

Jul-28-19  LameJokes: <<devere> All sporting events are affected by form… The current Grand Prix tournaments, whatever their format, clutter the chess schedule and prevent world championship aspirants from playing in other events they might prefer to participate in.>

Thanks for your response. I guess, you are right.

Jul-28-19  csmath: Where is Paris tournament?

MVL just beat Caruana in a fantastic rapid game. Great chess. And yet here nothing about this tournament for second day.

Jul-28-19  Absentee: <csmath: Where is Paris tournament?>

In Paris, I should think.

Jul-28-19
Premium Chessgames Member
  Sokrates: <csmath: Where is Paris tournament? MVL just beat Caruana in a fantastic rapid game. Great chess. And yet here nothing about this tournament for second day.>

CG is probably on vacation - again. Why skip a holiday for a world class tournament?

MVL is playing great in these rapids. The others seem to win and lose every other day. Yes, highly entertaining games all over.

Jul-28-19  john barleycorn: try here GCT Paris Rapid & Blitz (2019)
Jul-28-19
Premium Chessgames Member
  AylerKupp: <<beatgiant> At the root of it is the statement by <devere> above that <rating measures cumulative performance over a lifetime of chess playing.>

It's a true statement. Dr. Elo's and FIDE's ratings calculation is essentially a recursive filter with R(t) = R(t-1) + RC(t) where R(t) = rating for the current period, R(t-1) = rating for the previous period, and RC(t) = ratings change for the current period. So the a player's current rating is the result of a lifetime of chess playing, although the more recent games have much more impact on a player's current rating than his much earlier games. And, given the accumulation of roundoff errors over time by rounding R(t-1) to an integer, a player's rating becomes more inaccurate over time. So the longer that a player has been playing, the more inaccurate his rating is.

Then again, with all the simplifications in ratings calculations done by Dr. Elo and FIDE over the years, it probably makes no sense to talk about "accurate ratings".

And I couldn't find any posts by <devere> that indicated that he thought that ratings accumulate linearly, which couldn't happen in any case because of the non-linearity of the Cumulative Distribution Function (CDF) used to calculate the ratings gain or loss based on the games' results and the rating differential between the players. It might have looked linear in the example he gave but that was just to make a point and not based on any ratings calculations.

Jul-28-19
Premium Chessgames Member
  AylerKupp: <<devere> Surprised to learn that you are also among the math-challenged.>

Sorry. I'm not among the math-challenged, at least not in this case, but I am definitely among the reading comprehension-challenged. I read your statement of "Player B beats eleven opponents with the same rating as Player A's January opponent during December 2019." as "Player B beats eleven opponents with the same rating as Player A's January opponent <each month> during December 2019." That's why my calculations are different than yours.

You would even have made your claim of a 12X advantage clearer by indicating that in this scenario Player B gains 60 rating points in month 12 rather than 55. In that case both Player A and Player B would have finished the year with a rating of 2705.00 but Player A only needed to gain 5 points in the first month while Player B needed to gain 60 points in the last month to achieve the same average rating, hence the 60/5 = 12X advantage. Although a 12X <potential> advantage is probably a more accurate statement, since it depends on Player A gaining all his points in the first month and Player B gaining all his points in the last month. This scenario, while theoretically possible, would be highly unlikely.

The requirement to play at least 30 games during the year and at least 18 games during the last 6 months could be satisfied by both players playing in tournaments every month, with Player A gaining all his points during his first tournament and Player B gaining all his points during his last tournament. But this scenario is even more unlikely.

Jul-28-19
Premium Chessgames Member
  AylerKupp: <<beatgiant> <devere> The distortion you are complaining about occurs because of the use of a previous month's rating in case of inactivity, which is why the activity rules matter.>

I think that we all assumed initially, myself included, that the reason that the players in <devere>'s example did not gain any rating points because they didn't play any games. And that's not necessarily a valid assumption. As I indicated above, each player could have been playing in tournaments all year (thus satisfying the minimum games played criteria to qualify for the Candidates Tournament via average ratings), but Player A just happened to gain all his ratings points in his first tournament and Player B just happened to gain all his ratings points in his last tournament. So the unchanged ratings may not be due to inactivity at all. As I also said, this scenario is highly unlikely, but there were are.

It would be "nice" if the Elo rating system rewarded activity and penalized inactivity as the Chessmetrics system claims to do, but the Elo rating system doesn't. Kasparov's standard (classic) rating is the same today as it was when he retired in 2005 although I don't think many would dispute that his playing strength has likely deteriorated after 14 years of inactivity in rated tournament participation.

The question, of course, is "how much". I haven't been able to find any articles (I didn't look very hard) that calibrate the Chessmetrics predicted reduction in a player's rating as a result of inactivity such as Kamsky's from 1999 to 2004, Tarjan's from 1984 to 2014, Fischer's from 1972 to 1992, and many others. So I'm not sure how to properly address ratings change as a result of inactivity in a verifiable way.

Jul-28-19
Premium Chessgames Member
  beatgiant: <AylerKupp>

<And I couldn't find any posts by <devere> that indicated that he thought that ratings accumulate linearly>

He compared taking the average of the monthly ratings with taking the average of the number of runs after each inning in baseball.

A much closer analogy, in my opinion, is taking the average of a person's monthly weight. Yes, your weight is the cumulative result of a lifetime's eating and exercise, but it does still make sense to take the average of your monthly weight. The extra birthday cake you had when you were 5 years old has little impact now.

But what if you told your doctor that you weighed yourself once in January and then just assumed that your weight in February, March and April was the same? That's the actual problem we are talking about.

Aug-01-19
Premium Chessgames Member
  AylerKupp: <<beatgiant> He compared taking the average of the monthly ratings with taking the average of the number of runs after each inning in baseball.>

OK, I missed that. But I don't think that the important problem in terms of "fairness" is whether ratings accumulate linearly or non-linearly. It's the fact that the player(s) with the initially higher rating or who accumulate rating gains early in the average rating calculation period have a possibly unfair advantage because the effect of the higher initial or early rating is propagated disproportionably throughout the average rating calculation.

An initially higher rating is certainly well earned. But that higher rating was gained outside the average rating calculation period so it's questionable whether having a higher initial rating should influence the average rating calculation or whether rating gains early in the average ratings calculation period should have a proportional higher impact on the final average rating than rating gains later in the average rating calculation period.

Player ratings fluctuate during a period of time. If players ratings reflect the player's relative strength compared to other players, then that early higher rating may represent a peak in the player's rating and not an average indication of his playing strength. But when choosing the best players to participate in the Candidates Tournament, wouldn't we want to select the player(s) whose rating is increasing towards the end of the average ratings calculation period rather than the player(s) whose rating is decreasing towards the end of the period? That would help assure that the player that is selected to play in the Candidates Tournament is as close to his likely actual form when the Candidates Tournament starts.

Aug-01-19
Premium Chessgames Member
  beatgiant: <AylerKupp>
Interesting points, but it's not the argument that <devere> is making. You should look for a post where he explains his baseball analogy.
Aug-03-19
Premium Chessgames Member
  AylerKupp: <<beatgiant> Interesting points, but it's not the argument that <devere> is making.>

True, and I had looked at the post he made about his baseball analogy. I may be wrong but I don't think that it really makes a difference what analogy one makes. The basic problem is that the players who either have a higher average when the year starts or who gain the majority of their rating points early in the year have an advantage over those players who either had a lower average when the year starts or who gain the majority of their rating points later in the year whenever average ratings are calculated, even though the players in the second category may finish the year with higher rating than the players in the second category. This, of course, is because the ratings early in the year have a greater effect in the calculation of the final average rating than the ratings later in the year.

So whether the ratings accumulate linearly or non-linearly is besides the point. If we average the players' ratings, the "early risers" will always have an advantage over the "late bloomers".

Now that got me thinking, always a bad thing. Now, if instead of averaging the players' <ratings> during the year you would either average or simply add the player's <rating changes> during the year, would that be fairer to the "late bloomers"? It would if Player A and Player B had equal ratings at the beginning of the year, each played 4 tournaments during the year a month apart, and each gained the same number of rating points in each tournament. Their initial ratings would not be a factor and both Player A and Player B would gain the same number of rating points during the year, 20 total with an average of 5 rating points/month. Adding their rating gains (or losses) would be easier because if we took their averages then we would need to take into account when a player played in a tournament but did not gain or lose any rating points; that would affect the player's average rating even though it would not affect their total rating gains or losses during the year.

But this is an unrealistic situation because a player's ratings gains and losses are based on the rating differential between a player and his opponents and ratings are recalculated each month. And because of the way the ratings are calculated, this would make it unfair to the <higher> rated player at the beginning of the year because it's harder to win rating points (and easier to lose them) if you are a higher rated player than if you are a lower rated player, even if both players performed the same..

As an example, take 2 players, A and B, rated 2800 and 2700 respectively at the start of the rating period. As before they each play in 4 tournaments a month apart each consisting of 10 players and 9 rounds. Player A plays his 4 tournaments at the beginning of the year and Player B plays his 4 tournaments at the end of the year. The average rating of their opponents in each tournament is 2750 and both Player A and Player B win 5 of their games and draw 4.

Again using http://www.kosteniuk.com/EloCalc/el... and a K-factor = 10, in this scenario Player A's rating in months 1, 3, 5, and 7 would be 2799, 2798, 2797, and 2796 for a total rating loss of 4 points. Player B's ratings in months 6, 8, 10, and 12 would be 2711, 2721, 2730, and 2738 for a total rating gain of 38 points. Leaving aside the likely desirability to have a 2796 player participate in the Candidates tournament rather than a 2738 player, Player B would be selected to qualify for the Candidates Tournament over Player A.

So it seems to me that using average ratings or rating differentials (whether averaged or summed) is not fair to either the "early risers" or the "late bloomers" in terms of having a player qualify for the Candidates Tournament, assuming that one of the objectives of the qualification process is "fairness". A better scheme should be found.

Aug-03-19
Premium Chessgames Member
  beatgiant: <AylerKupp>
The rating criterion is not designed to reward "rating gain" but rather "stable high rating."

Whether it should or not is a different question. Whether "fair means memoryless" is a question we've debated ad infinitum. I think that depends on whether we want the world title to be more of a "lifetime achievement" kind of award or a "player of the year" kind of award.

Aug-04-19
Premium Chessgames Member
  AylerKupp: <<beatgiant> The rating criterion is not designed to reward "rating gain" but rather "stable high rating." >

Well, if the rating criterion is intended to reward "stable high rating" then it seems to me that it's even more important to look at monthly rating gains <and loses> rather than overall average ratings. After all, it's the rating gains and losses that, when combined with a player's rating prior to the average rating calculation period, determines whether the player achieved a "stable high rating" during that period. Besides, if a player has the highest rating than all the other contenders trying to qualify via rating for the Candidates Tournament, then that rating was achieved by that player's performance <outside> the average rating calculation period. So that doesn't seem consistent to me.

For example, the attempt to qualify for the 2018 Candidates Tournament via rating was a 3-horse race between Caruana, Kramnik, and So. Of the 3, Caruana started with by far the highest in Jan-2017, 2827, compared with So's 2808 and Kramnik's 2811. All 3 lost rating points during the year and Caruana experience the greatest drop in ratings, -28 points, vs. Kramnik's -24 points and So's 20 points. Yet it was Caruana and So, not Caruana and Kramnik, who qualified for the 2018 Candidates Tournament via ratings.

And if it's stability that is being rewarded, So and Kramnik had the same standard deviation with respect to their overall ratings change, 9.15, vs. Caruana's 10.63. So if "stable high ratings" is the criteria, then So and Kramnik should have qualified for the 2018 Candidates Tournament and not Caruana.

The main reason that So outperformed Kramnik in the average rating calculation race is that he had the best performance of the 3 players by far early in the year, gaining 14 rating points in Feb-2017 while Kramnik did not play in any tournaments during the first 6 months of 2017, possibly due to illness. Therefore the disproportionate influence of ratings in the early part of the average ratings calculation period benefited So immensely, and punished Kramnik.

<Whether "fair means memoryless" is a question we've debated ad infinitum.>

I don't think that we've debated that at all.. I think that we've discussed "fairness" but that's not the same thing. "Fairness" has to do with equality of opportunity to achieve a goal. If a player has a relatively low rating compared with the other WCC title contenders and has a 2-year hot streak where he has a superior record against all the top players, including the defending WCC, than that player will become the new WCC regardless of how he performed prior to the WCC cycle.

Or, as Fischer put it far more succinctly than I ever could when he was young and began to make amazing progress, "All of a sudden I just got good."

<I think that depends on whether we want the world title to be more of a "lifetime achievement" kind of award or a "player of the year" kind of award.>

I've stated before, all that a sporting event proves is who the best player was at the time that the event was held. So it makes no sense to consider the world title to be more of a "lifetime achievement" award than an award for being the best player in the world during a very limited period of time. If we wanted to reward "lifetime achievement", there are far better ways of doing that than focusing on a narrow window of time.

Aug-04-19  kappertjes: So if we don't want to reward early vs late or vice versa and we want reward the consistently best player over the period of interest (so preferably not using their rating at the beginning), it seems to me a TPR for each player could work. Use the rating for the opponents as they were played and treat the whole 2-year period for performance calculation as in a tournament.

Possible abuse would ofc be to stop playing after having achieved 2900 TPR or w/e, so some kind of activity rule, like currently in place, will be needed to guarantee minimum shenanigans.

Aug-05-19
Premium Chessgames Member
  AylerKupp: <Qualifying for Candidates Tournament via TPRs> (part 1 of 2)

<<kapertjes> ...it seems to me a TPR for each player could work. Use the rating for the opponents as they were played and treat the whole 2-year period for performance calculation as in a tournament.>

Yes, I suggested something like that (World Cup (2017) (kibitz #2350)) and others suggested that before I did. We all suggested a 1-year period and you suggested 2-years, and either one is reasonable. But I think that the most practical one, given that the WCC match has recently been held in November and the Candidates Tournament has been held late March – early April, is (for the 2020 Candidates Tournament) to select the start of the period to be the Jan-2019 FIDE rating list (which reflects the games played in Dec-2017, the month after the WCC match completed) through the Jan-2019 FIDE rating list which reflects the reasonable requirement that whichever player is select by TPR or equivalent needs an early-2020 announcement so that he can make the proper preparations. For 2020 FIDE indicated that the participant to be selected to participate in the 2020 Candidates Tournament via rating include the results through the Jan-2020 FIDE rating list. I think that this might be pushing it but we'll see what happens during this WCC cycle.

That leaves us with about a 12 – 13 month period to determine the Candidates Tournament participant to be selected by this method. If we want a longer period we could start the period to coincide with the first month following the completion of the Candidates Tournament instead of the WCC match; with the Candidates Tournament finishing in April that would make it, say, FIDE rating lists from May-2020 through Dec-2023, a 19-month period. Either one would probably be fine.

A few other things need to be resolved. FIDE defines the calculation of the Performance Rating in section 1.48 of https://www.fide.com/fide/handbook.... by

RP = RC + D(P) where:

RP = Tournament Performance Rating (TPR)

RC = The average rating of the opponents

P = Number of points scored / Number of games played

RD(P) = The rating difference indicated by P from the FIDE Performance Rating Table 8.1a (Section 1.49) as given in the FIDE Handbook above.

Aug-05-19
Premium Chessgames Member
  AylerKupp: <Qualifying for Candidates Tournament via TPRs> (part 2 of 2)

A possible issue is that P in Table 8.1a has only 2 digits of significance and a given P (say P = 0.58) results in RD(P) = 57 while P = 0.59 results in RD(P) = 65 and P = 0.57 results in RD(P) = 50. What value do we use for RD(P) if P is between 0.59 and 0.58 or between 0.58 and 0.57, particularly if P is the result of a 8/12 score = 0.6666...? Section 1.49 indicates that any P should be rounded to the nearest 2 digits, with P = 0.005 rounded up. As long as one is consistent this should be OK, but much higher precision is possible with not much extra effort, given that relatively cheap computers and software are readily available..

FWIW I use Excel to calculate TPRs and I use a table ranging from a rating difference of -1000 to + 1000. Each entry represents 1 rating point and RD(P) is calculated with Excel's NORMDIST to Excel's default 15 digits of precision, using a SD = 200*SQRT(2) ~ 282.8247 instead of Dr. Elo's and FIDE's simplification of SD = 2000/7 ~ 285.7143. When there is no exact match to a given P (as happens most of the time), I calculate which P-value in the table is closest to the given P and use the RD(P) corresponding to that value. That's probably the best that can be done.

We can, as you said, consider all the games played during the period as one big tournament or calculate the TPRs separately for each tournament and average the results. I think that these will be equivalent but I need to verify that. On advantage of making the calculations as though all the games were part of one big tournament is that the Normal distribution used in calculating RD(P) is appropriate, while with the smaller number of games when the individual TPRs are calculated per tournament it is not.

Then there are other considerations to ensure the best calculations such a specifying a minimum number of games and tournaments to be played and when a player participates in those tournaments. For a 1 year TPR calculation period a minimum of 30 games/yr as is done now seems reasonable and if the TPRs are calculated separately for each tournament then a minimum of 4 tournaments in that year with a minimum of 2 the first 6 months of the year and a minimum of 2 the second 6 months of the year. Maybe it would be beneficial that only a player's top 4 results are considered in the calculations; I don't know.

Different possible player strategies need to be considered. Some players might prefer competing in closed tournaments when the expected P will be relatively low but RC would be relatively high. Other players may prefer competing in open tournaments when the expected P will be relatively high since the best player will likely play much lower rated players in the first few rounds but the expected RC will be relatively low because of the much greater numbers of lower rated players in the tournament. I'm not sure if these two strategies are in balance or not; that also needs to be checked.

But basically, I and others agree with you that using either 1-year TPRs which I have dubbed YPRs (Yearly Performance Rating) or an average of the TPRs obtained in various tournaments during the year seems to be a "fairer" way to provide for Candidates Tournament qualification than using an average of the players' ratings.

Jump to page #    (enter # from 1 to 14)
search thread:   
< Earlier Kibitzing  · PAGE 14 OF 14 ·  Later Kibitzing>
NOTE: You need to pick a username and password to post a reply. Getting your account takes less than a minute, is totally anonymous, and 100% free—plus, it entitles you to features otherwise unavailable. Pick your username now and join the chessgames community!
If you already have an account, you should login now.
Please observe our posting guidelines:
  1. No obscene, racist, sexist, profane, raunchy, or disgusting language.
  2. No spamming, advertising, duplicate or nonsense posts.
  3. No malicious personal attacks, including cyber stalking, systematic antagonism, or gratuitous name-calling of any member Iincludinfgall Admin and Owners or any of their family, friends, associates, or business interests. If you think someone is an idiot, then provide evidence that their reasoning is invalid and/or idiotic, instead of just calling them an idiot. It's a subtle but important distinction, even in political discussions.
  4. Nothing in violation of United States law.
  5. No malicious posting of or linking to personal, private, and/or negative information (aka "doxing" or "doxxing") about any member, (including all Admin and Owners) or any of their family, friends, associates, or business interests. This includes all media: text, images, video, audio, or otherwise. Such actions will result in severe sanctions for any violators.
  6. NO TROLLING. Admin and Owners know it when they see it, and sanctions for any trolls will be significant.
  7. Any off-topic posts which distract from the primary topic of discussion are subject to removal.
  8. The use of "sock puppet" accounts to circumvent disciplinary action taken by Moderators is expressly prohibited.
  9. The use of "sock puppet" accounts in an attempt to undermine any side of a debate—or to create a false impression of consensus or support—is prohibited.
  10. All decisions with respect to deleting posts, and any subsequent discipline, are final, and occur at the sole discretion of the Moderators, Admin, and Owners.
  11. Please try to maintain a semblance of civility at all times.
Blow the Whistle See something that violates our rules? Blow the whistle and inform a Moderator.

NOTE: Keep all discussion on the topic of this page. This forum is for this specific tournament and nothing else. If you want to discuss chess in general, or this site, visit the Kibitzer's Café.

Messages posted by Chessgames members do not necessarily represent the views of Chessgames.com, its employees, or sponsors. All Moderator actions taken are at the sole discretion of the Admin and Owners—who will strive to act fairly and consistently at all times.
Spot an error? Please suggest your correction and help us eliminate database mistakes!


home | about | login | logout | F.A.Q. | your profile | preferences | Premium Membership | Kibitzer's Café | Biographer's Bistro | new kibitzing | chessforums | Tournament Index | Player Directory | Notable Games | World Chess Championships | Opening Explorer | Guess the Move | Game Collections | ChessBookie Game | Chessgames Challenge | Store | privacy notice | contact us


Copyright 2001-2019, Chessgames Services LLC