Author: Pat King
Date: 09:57:22 03/21/99
Go up one level in this thread
On March 18, 1999 at 15:39:24, KarinsDad wrote: >On March 18, 1999 at 13:17:05, Charles Unruh wrote: > >>On March 18, 1999 at 10:44:46, KarinsDad wrote: >> >>>On March 18, 1999 at 10:07:01, Charles Unruh wrote: >>> >>>>I have played 155 tournament games in my career and i have never lost or drawn a >>>>single game against any person that was 150 rating points below me. >>> >>>Charles, >>> >>>Your playing history is an anomally. I have played players 400 points below me >>>and lost and I have played players 400 points above me and won. >> >>No it's not really an anomally, most players that i know, and i know hundreds, >>indeed are not as steady as myself, but as far as it comes to losing to players >>150 points lower than themselves though it happens it's usually less than 1 in >>30 or 40 games that it does happen. There is simply something critical that is >>being left out in the statistical calculation. > >After you read this message, I challenge you to find any open tournament in the >United States of 50 or more players and 4 or more rounds (standard playing >times, not blitz) in the last year where there was a ratio of 1 game won for >every 29 games lost (or a lower ratio) by players 150 or more points lower than >their opponents. You will not find such a tournament and any rebuttal you make >to me without data will only be rhetoric. > >You are wrong. Bottom line. You are basing your information off of your own >record, guessing, and memory which is biased. I can illustrate this. I went to >my local state chess web page and pulled a recent open tournament. I removed all >byes, all withdrawals, and all games played by unrated players (there were no >forfeits). > >There were 139 games played between people with established ratings. > >Of this, I took the statistics for all games played between players of a ratings >range of 150 to 250 points. There were 24 of these games. The record of the >higher rated players in these games were: +14=5-5, so they only won 58% of the >games. The lower rated players drew 21% of the games and won 21% of the games. >Also, as an aside, most of the draws and losses (8 of the 10) here by the higher >rated players were by class A through master level players (e.g. class A players >playing masters, class B players playing experts, and class C players playing >class A players, etc.). So since most of the higher rated players who lost or >drew to players 150 to 250 points below them were established good players >(class A and above), this seems to contradict your theory (the record was +8=4-4 >when I got out of the class A and higher rated players, half of the games were >drawn or lost). > >I also took some other statistics. For the games where the rating differential >was 250+ points, I kept track of the upsets by the lower rated player. Here are >the results (I rounded to the nearest 10 differential): > >550 draw >290 draw >350 draw >390 loss >390 loss > >700 loss This was a master who lost to a D rated player (not the same game I > mentioned earlier). > >420 draw >320 loss >290 loss >260 draw >590 draw >370 loss >410 loss > >600 loss These 3 wins were made by an up and coming player who's rating was >510 loss (before this tournament) a lot lower than it should have been. >560 loss > >350 draw >340 loss >390 loss >330 loss >310 loss >360 loss >370 loss >680 draw > >This is a result of 8 draws and 16 wins by players rated more than 250 points >lower than their opponents in 139 games. Cut me a break. You have not researched >this at all. > >I know most of the players in this tournament and about 1/4th of these upsets >were due to up and coming players, or elderly players who have difficulty >playing well for an entire game anymore. But realizing that this is part of the >overall data, the number of upsets by lower rated players by 150 or more points >in 139 games is 13 draws and 21 wins. This means that about 25% of the games >were upsets (not the 1 in 30 or 40 that you claim). The result may be about 18% >(not 2-3%, or the 0% in your case) if you pulled out the statistical anomalies >(such as elderly players and up and coming players). > >You have no clue of what you are talking about. > >>> >>>There are some players who play "similar" to computers in that for their level >>>of ability, they make very few tactical mistakes. You may be one of those types >>>of players. I am not one of those types of players. I can make a valid sac of a >>>knight against a player 600 points higher than me and worry the heck out of him >>>and 5 moves later, I can make a putzer move that 3 moves earlier, I mentally >>>said to myself could not be made in the game. >>> >>>My current rating is almost 400 points lower than yours and I would make a wager >>>that in a 10 game match of 40/2 G/60, I would take 1 to 2 points from you. Why? >>>Because I sometimes play at expert level and sometimes play at D level. It just >>>depends on how well I am concentrating at the time. >> >>Wager? I have played played a match(game in 60)last year with a player rated >>1584, i was rated 1812 at the time i beat him 10 straight games. I had beat him >>4 times previously that wasn't a part of the match. your wager is but >>conjecture, 400 points below me i doubt your wager. The odds say that i should >>beat Kasparov 1 in so many games, but he could play me down a pawn a 1000 games >>despite what the odds say, i doubt seriously that i'd win a game. > >Yes, this is conjecture. I am basing this guestimation on the fact that I have >beaten or drawn about a dozen games against class A players out of about 50 or >60 games that I have played against them. So, 1 in 4 or 1 in 5. I have also had >obviously winning games against experts (and one master) that I managed to blow, >so even though you are more consistent than I, there is a good chance of me >getting at least a draw (or more) out of you in 10 games. > >>> >>>> In a mere 6 >>>>game match yermo could be expected to take H7 out 6 0. >>> >>>Yermo will almost definitely not win 6-0. The reason is that if he is leading >>>the match (say 2-0), he will take a draw in a basically drawn position and not >>>push for a win that may not be there. This happens a lot at the GM level and I >>>cannot imagine that it will not happen in this match. >>> >>>> My rating is currently >>>>in the 1900s. I'm sure there are many players with similar records. >>> >>>I have drawn or beaten at G30 or slower times every 1900-1999 player in the 2 >>>main clubs I have played in within the last 3 years at one time or another and >>>my rating is currently in the high 1500s. I would predict that there are very >>>few players with your type of record. >> >>I have beaten people 200 points above me as well, some people have consistency >>some don't. >> >>> >>>> This leads >>>>me to believe that there is some error when one predicts that in a 6 game match >>>>that a 2500 player should score 2/6 against a 2630 player. I certainly hope H7 >>>>is much stronger than 2500, otherwise this will just increase speculation about >>>>the progs strength. >>> >>>Your assumption here is based on your previous assumption as to other players >>>having similar narrow banded win/draw/lose ratios like yourself. >> >>Other players do. Just look at kasparov Anand, and Kramnik, i don't have the >>statistic but it was placed in this group, and the predicted statistic for >>players 150-200+ points below them simply did not hold. Kasp played no one under >>2600. > >Yes, the higher you get up, the more difficult it is for a player 150 points to >beat a player higher than them. There are several reasons for this. One is that >higher rated players often play for draws if they see no way to win. Another is >that the mistakes made are smaller, so they are harder to take advantage of (and >if a GM is losing, you can bet that he will aggressively look for a way to >draw). But the other side of this coin is that even with a 150 point >differential, Hiarcs7 should be able to draw at least a few games against Yermo >just due to Yermo's tendency to go for a draw if he has the match won or if he >has a dead drawn position. > >Yermolinsky is rated almost 200 points below Kasparov and has a winning ratio >over Kasp. What does this say about your theory? > >> >>This is rarely >>>the case, so the second assumption here is probably also based on a false >>>premise. >> >>All that is rare is my perfect record, but as for the 1 in 30 or 40 games of >>most people that's not rare at all. > >This statistic is obviously wrong (see results above). > >KarinsDad No true stats, but true probabilities here, which fairly closely match Karin's actual results. This was an effort to predict outcome of Yermo-Hiarcs match, assuming an expected score of 0.67/round for Yermo and a GM draw rate of 30%. E(1) = 0.67 and P(0.5) = 0.3 -> P(1) = 0.52, P(0) = 0.18 for one game (note that this could account for Unruh's perception of the general lack of losses for higher rated players... the draw rate has a big effect on win/loss ratio). Extrapolating for a six round match, P(0.0) 0.0000 P(0.5) 0.0003 P(1.0) 0.0054 P(1.5) 0.0081 P(2.0) 0.0214 P(2.5) 0.0583 P(3.0) 0.1108 P(3.5) 0.1684 P(4.0) 0.2051 P(4.5) 0.1944 P(5.0) 0.1398 P(5.5) 0.0684 P(6.0) 0.0198 So Yermo will have to beat 50:1 odds to go 6 for 6. Bigger odds than I thought when I made the statement earlier in this thread that a 6.0 for Yermo wouldn't be significant, but even beating these odds only once isn't conclusive for me. Pat P.S. the Quattro Pro spreadsheet with these calcs is available at ftp://ftp.mindspring.com/~p_m_king/yermo.wb3
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.