October 7th, 2003
|04:25 am - Wrap up warm|
Quick recommendation for Middlesbrough inhabitants: the Belle Vue Fisheries are breaking the rule that takeaways can only do one type of food well by serving the best takeaway curry in town as well as competent fish and chips. You can currently get a starter (eg chicken tikka plus salad), a sundry (eg cucumber raita), a curry (eg Chicken Rogan), a rice (eg vegetable pilau) and a sizeable bread (eg garlic naan) for £8. The five items together will provide meals for two and the quality is some of the best I've had outside Birmingham.
The instruction booklet with information about the types of puzzles to be solved at the World Puzzle Championship at the end of next week is now online. Looks like it would be a great deal of fun. Good luck to all, especially the UK team.
As I reported at the time, NAO Chess Club Paris won their final game 4-2; Peter Svidler, on a hot run of form, won as did Fressinet on the bottom board, with the other four holding draws to claim the championship. Let's hope that Mme. Ojjeh doesn't lose interest in chess now the novelty of her club winning the championship has gone; lest we forget, St. Petersburg beat Paris in an at-distance four-board 15-minute challenge match in May. (If the world can get the technology working properly, then perhaps the goal of a world chess league with radically reduced travel might eventually be feasible.)
Match two saw Polonia Plus GSM Warsaw overturn a possibly slightly unmotivated Ladya-Kazan-1000 by the same scoreline; Boris Gelfand had one of his bigger wins for a while over Viktor Bologan. (Viktor has gone WWWWLWL for the tournament and so deserves some sort of Japanese-style "fighting spirit" award for a draw-free week.) This improves Warsaw's position from last year's third to this year's second, with which they will be extremely pleased. Kazan finish fifth (also one place better than last year!) but this will slightly disappoint them. Kiseljak of Bosnia-Herzegovina won match three to finish third, a gnat's fiancetto behind second place only on board count for both they and Warsaw had six match wins and 29 game wins; however, Kiseljak's defeat came in round two and accordingly they faced relatively modest opposition in rounds three to seven to climb the leaderboard so high. These consequences of an early loss are sometimes referred to as the "Swiss Gambit".
Beer-Sheva's final-round defeat dropped them to seventh. Among the other teams I was following, Werder Bremen will be happy to have finished twelfth; last year's champions Bosna Sarajevo disappointed massively to finish sixteenth, with the same score as 20th-placed finishers Barbican Chess Club. Asker Schakklubb came 31st of 43. The last three places were taken by a Welsh team and two Irish teams. *buttons lips*
One of the biggest attractions to me of chess fandom, being the number-oriented, stats-freak traditional male sports fan that I am, is the ratings system. There are some reasonable background articles about it at the Wikipedia and about.com, with Sam Sloan providing his own take. Normally I would apply a bit of "mm" and "hmm" in warning here because Sam is, shall we say, a bit of a legend (the words "bags" and "spanners" come to mind) but this article of his is entirely safe for work and, frankly, quite an important and useful summary. It's also probably fair to say that the rating system as it has been implemented for chess is possibly the best-known worldwide rating system of its type in all sport.
The rating system attempts to measure who has demonstrated the best results over time. Everyone involved has a rating, the higher the better. To get into the FIDE ratings at all, you need to be rated at least 1800. (Used to be 2000 until recently. Getting a rating in the first place is another matter.) The highest-rated player is Garry Kasparov and he has been highest-rated almost continuously for almost twenty years. His current rating is 2830; in 1999, he peaked at about 2850, which is the highest rating anyone has ever achieved. It is a matter of some debate as to what extent you can compare ratings in any one year with those of other years and whether you can extrapolate this to "highest rating ever directly implies best player ever".
Essentially the rating point system is a zero-sum game. When two people play each other, the result of the game dictates how many points flow from one to the other. If the game had a decisive result, the loser of the game pays some rating points to the winner of the game. If the game is drawn, the player with the higher rating previously pays some rating points to the player with the lower rating previously. Players' rating changes are collected on a tournament-by-tournament basis and these changes are applied to the ratings every three months (used to be every six months until recently).
Very roughly, if you are playing someone with the same rating as you, beating your opponent will gain you five of their rating points, losing to an opponent will gain them five of your rating points and a draw will make no difference.
If you're rated 100 points higher than they are, you're assumed to have a 64-36 chance of winning (counting a draw as half a win for both sides). Accordingly, beating them will transfer 3.6 rating points from them to you, drawing with them will transfer 1.4 rating points from you to them and losing to them will transfer 6.4 rating points from you to them. (Likewise, if you're rated 100 points lower than they are, you're not expected to win, so you stand to lose 3.6 points, gain 1.4 or gain 6.4.)
If you're rated 200 points higher than they are, you're assumed to have a 76-24 chance of winning. Accordingly, beating them will transfer 2.4 rating points from them to you, drawing with them will transfer 2.6 rating points from you to them and losing to them will transfer 7.6 rating points from you to them. Rating changes is calculated to the nearest tenth of a rating point, but ratings are only published to the nearest point (until recently, to the nearest five points). A difference of fewer than 25 points or so is reckoned to be pretty incidental. The full chart of rating differences and associated expectanies is published at about.com and you can find the formula which generates it at the Wikipedia.
(Incidentally, if your FIDE rating is not particularly well-established, it will change 50% more quickly; technically you have a "K value" of 15 rather than of 10, but all the changes are expressed pro rata.)
Now one of the things that's useful about the system is that it can be used to compare two series of performances when they aren't against common opponents, to answer the age-old question of whether it's better to draw against really good players or to beat mediocre ones. You can take any string of performances and express it that, over the entirety of the tournament, you did exactly as well as you would have would a player with some other rating to perform - this some other rating is your performance for that tournament. Your current rating will then increase or decrease by some factor multiplied by the difference between your performance and your previous rating. (This factor is one-eightieth the number of games in the tournament - so, if your rating is 2500 and you get a 2600 performance over an 8-game tournament, your rating will go up to 2510.)
Assuming that all the participants in a tournament are of approximately equal rating, there's a quick and dirty way to estimate your tournament performance based on your overall score and the ratings of the other players. Take the average rating of the players you faced and subtract 400. Then express your score as a proportion (3 wins and 1 loss is 3/4, 3 wins, 2 draws and 1 loss is 4/6 and so on) and multiply that by 800. Add the two together and you've got a rough estimate of your performance for that tournament. If it's higher than your rating then you did better than you were expected to do and you can expect your rating to rise.
So a player who faces opponents with average rating 2500, wins five and loses three, has performed at 2500 - 400 + (5/8 * 800) = 2600. Note how this is calculated ignoring your own rating to begin with. This meshes exactly with the example two paragraphs above; if your rating had been 2500 before you gave that performance, you would expect your rating to increase to 2510 as a result. Your five wins at +5 each gain you 25 points and your three losses at -5 each lose 15 of them, so it's a 10-point increase either way.
This is essentially the way all ratings systems worldwide work. However, the chess system has never been particuarly good at explaining itself in simple terms like that. One fun dilemma which anyone who has ever played around with implementing a rating system has faced is working out what players' initial ratings should be. It's worth noting that the way this was solved first time was by defining someone who scored 50% in the US Open tournament to have had a 2000 rating (under a very slightly different rating system) then back-producing and forward-producing ratings from that basis.
Now we've got this framework in place, we can usefully use it to discuss performances at the European Clubs Cup, try to work out who shone and who stunk and so forth.
First off, despite ending with a loss and a draw, Garry Kasparov smote everyone in his path and took some pretty meaty scalps in the first five days; he ends up with a 2839 performance for the tournament, which increases his whopping 2830 rating still further. Vasily Yemelin played board six for St. Petersburg and won all five of his matches, the only player with a perfect record. Vladimir Akopian, an unlikely chap (who, lest we forget, made it to the final of the 1999 knockout) actually wins the prize for "best board one" for 5/6, helped by Alkaloid Skopje's weak opponents. NAO Paris and Polonia Plus Warsaw both have two of the top seven performances - Svidler and Lautier both shone as NAO boards 2 and 4 against tough opposition, while Ivanchuk and Macieja were a bulwark for Warsaw on boards 1 and 3.
NAO will be faintly disappointed with Alexander Grischuk's relatively weak 50% record as board one, and they would have expected Michael Adams to kick considerably more "heiny" and concede far fewer draws than he actually did. Winning the tournament as top seed is never bad, but NAO should really have made it look easier than it actually was. Warsaw's second place was due to over-achievement all the way through the squad, with the exception of their occasional not-so-super sub Bartosz Socko who did not shine against people he might've been expected to put away. Bosna Sarajevo's well-below-par performance can largely be blamed on Evgeny Bareev, possibly heady after his rise to global #4 but hemorraging enough points here alone to put him back outside the top ten. Their lower order (boards 4-5 Kasimdzhanov and Georgiev) really didn't perform nearly ruthlessly enough against weak opposition, either.
Let's have a look at Asker Schakklubb, for I have been following them all the way. They only brought a six-player squad, so all six players played every round without breaks. 12-year-old Magnus Carlsen, who is apparently a really nice kid, faced some far older, far more experienced opposition and acquitted himself well; their board two, IM Berge Ostenstad, scored 5/7 to be the best-placed IM on second board in the competition - 9th of 45, all told. GM Jon Tisdall, who I used to work with, scored 50% on board three. You can find out a breakdown of each player's rating performance organised by team, organised by board or organised by performance.
Here's my all-Clubs-Cup MVP line-up:
Board 1: OK, he blundered badly once, but got to go with Garry Kasparov (Ladya-Kazan-1000) for being the top performer in the whole tournament.
Board 2: star performer on his team against classy opposition, Peter Svidler (NAO Chess Club Paris). Particular credit for beating Viktor Bologan on form.
Board 3: Macieja Bartlomiej (Polonia Warsaw).
Board 4: Joel Lautier (NAO Chess Club Paris) for 5/6.
Board 5: Alon Greenfeld (Beer-Sheva Chess Club). Now Alon actually played Board 4 for Beer-Sheva, so he should be an even better board 5.
Board 6: Vasily Yemelin (St. Petersburg), the only player with a perfect record - 5/5.
Team mascot: Gregor Fucka (Barcelona basketball team). At 7' tall, would you mess?
On a related subject, I'll also quickly explain how players win FIDE's titles. To become a Grandmaster - nay, an International Grandmaster, but the "international" is almost always dropped - you need to get a rating over 2500 and you also need to get either two or three "GM norm" tournament results within seven years. A "GM norm" is (roughly) won for a tournament performance over 2600 in a sufficiently challenging tournament - one with other grandmasters present and a wide mix of nationalities. Two GM norms will suffice if the tournaments had 24 games between them, otherwise three are required. There are some other qualifications (one tournament must be a round-robin, one tournament must have games of a certain length and so forth). To become an international master, the same thing applies but shifted down 100 points (2400 rating required, 2500 performance for an IM norm).
I'll check with Jon Tisdall about this, but I think Asker's top two boards both earned norms from the European Chess Clubs Cup. On board one, FM Magnus Carlsen scored 50% against 2501-rated opposition for a 2501 performance, which I think is enough for a IM norm. (He might have all his IM norms already, though - not sure.) On board two, IM Berge Ostenstad scored 5/7 against 2449 opposition for a 2601 performance, which I think might be a GM norm, though possibly not if he didn't face enough GMs along the way. It does seem that the decision to put the two of them ahead of Jon, the team's only GM, may have been the correct one.
Now once a title is earned, it is awarded for life; if your rating goes above 2500 and you have the GM norms, you can never lose the Grandmaster title even if your rating drops below 2500 again. One English GM remains active even though his rating has fallen to 2380. There are some tournaments which are allowed to break the rules and can automatically produce titles or GM and IM norms for their winner, regardless of the ratings; this can, as you'd expect, lead to unreliable results and accusations of some unexpectedly weak titleholders.
FIDE has some other titles as well: you earn the lightly-regarded tertiary title of FIDE Master simply by keeping your rating above 2300 for sufficiently long, with no need for norm requirements, and FIDE also recently introduced the quarternary title of Candidate Master for those who keep their rating above 2200 for long enough. Grandmaster, International Master, FIDE Master and Candidate Master all have their women's equivalents as well, all of which are roughly pitched 200 points lower. There are several Women's Grandmasters who are not even International Masters; only five Women's Grandmasters have earned the genderless Grandmaster title in their own right. Hey, I don't make the rules.
For the last five years or so, there have been three players who have been clearly better than the rest - Garry Kasparov at number one, challenged (sometimes closely) by India's Vishwanathan Anand and Russia's Vladimir Kramnik. Nobody has been able to make a dominant #4 position stick; Evgeny Bareev is #4 this quarter, Peter Leko held it last time, Veselin Topalov at the start of the year with Michael Adams and Alexander Morozevich duelling between 2000 and 2002. (My tip for the top in January: Peter Svidler.) Alexei Shirov and Vassily Ivanchuk tended to squabble over the position in the late 1990s; the first half of the '90s was definite Karpov territory. You can see the current top 100 and look for other people's ratings, if you like.
Incidentally, you'll need to dig deep to find the ratings of the English team for the European Team Championship. The team lists for that have been published and it's pretty ugly. Russia will win because they're sending five of their top ten. Most of the other usual suspects - Ukraine, France, Germany, Spain and even Poland - are sending most, but not all, of their best players. Hungary are definitely fielding a weakened side, but not nearly to the disrespectful extent by which we are fielding roughly our fourth team. I shall not go on about the European Team Championship to nearly this extent, except possibly to go "Ooh, look, England lost to Azerbaijan today, just like we lost to Sweden yesterday and to Austria the day before that." Half-way would be a very good result.
So the big three of Kasparov, Kramnik and Anand have been dominant recently, though that's not to say that they've always been World Champion. The topic of the chess world championship is a matter for another day... no, it's a matter for someone else's web site, because it's a really horrible mess. Fun to point and laugh at, but really pretty horrible.
Oh, crikey. Now the chess is over, what next? Well, I've just realised the 3rd Mind Sports Olympiad in Prague actually started last Saturday and is rolling on until Sunday 12th. Tuesday 7th sees - as well as longsunday's birthday, which I hope is a good one! - tournaments in Oware and Lost Cities, Duplicate Scrabble (played in Czech, naturally) and 9x9 Lightning Go. It's going to be another great week of mind sports - here we go again! ;-)
Current Mood: complete
Current Music: Super Mario Kart Trippin' On Rainbows (acid jazz thing)
|Date:||October 6th, 2003 11:15 pm (UTC)|| |
Oxford's favourite sons
Re: Oxford's favourite sons
Keble's a reasonably big chess college, as it happens. There's a pretty good article about it
in the Brick, the Keble students-and-alumni mag. I overlapped with Rowson by one year, but don't think I ever met him. (Ruth Sheldon, who is something like England's #4-ish, #5-ish lady, also studied at Keble. I did meet her when the MSO covered an Oxford - Cambridge chess match on the Internet. ealuscerwen
looks fairly similar to her.
A random ratings fact for you: Magic the Gathering uses the same ratings system you describe above - I never realised it was how the Chess ratings were calculated ! (Back when I played chess ratings were worked out by averaging game ratings over a season of at least 12 games. Your game rating for a game was 50 points higher than your opponent if you won, 50 points lower if you lost, or the same as theirs for a draw. Worked quite well except that it discouraged good players from taking on much weaker opponents.)
Ah, British chess gradings, yes. :-) Scrabble is graded in (almost?) exactly the same way. I have a vague suspicion that there are some limits in place in both systems to ensure that your grading can't actually go down by beating an opponent over 50 points worse than yourself, nor go up by losing to an opponent over 50 points better than yourself.
Likewise, I think that you don't actually qualify for a rating if you have a 100% or a 0% record, on the grounds that you were clearly playing people who were so different in you to skill that your experience against them does not permit the grader to comment about your skill. If that isn't how it works, then that's how it ought to work, at least.
Kasparov sat out of match 2 last week for just the same reason because the opposing team only managed to score a point and be matched up against them through the full-point, 6-0 bye. He would've ended up playing against someone rated about 2150, I think.
Hee! Sankyu~ dear! *hugs all round*