Congratulations to the 200th tournamentwin. Einstein forum
27 replies. Last post: 20070404
Reply to this topic Return to forum
hanfried_c at 20061205
Liebe Luise,
herzlichen Glückwunsch zu Deinem 200. Turniersieg auf Littlegolem. Mögest Du überall so erfolgreich sein wie hier.
Hanfried & Stefan

Jonny ★ at 20061206
Vor allen Dingen auch einen Glückwunsch zu einer erstaunlich konstanten Leistung, wie: Zu über 70 % gewonnener Spiele und über 50 % gewonnener Turniere. Nicht schlecht  im Gegenteil, sogar ausgesprochen gut… :)))
Olaf

Theo van der Storm at 20061206
My German vocabulary is rather limited and my German spelling is terrible.
These people have tolerated my English messages in a German forum, so in return let me for once give a translation of their text with which I concur.
Dear Luise,
My hearty congratulations with your 200th tournament victory on LG. May you be as successful everywhere.
Stefan
Sank you ferry mutch :D
Before anything else also congratulations with an astoundingly constant performance with over 70% won games and over 50% won tournaments. Not bad – quite the contrary – even exceptionally good.
Olaf
PS: Nobody look at Hanfried_c's winning ratio.

Jonny ★ at 20061207
Thankx for your translation Theo. My english is not very good (Lothar Matthäus ;)))
P.PS Nobody had played more than 1000 games.  Accept… And that is a good distance to see a constant performance.
But we will see… Some players are near 1000 games.
Olaf

hanfried_c at 20070314
It's incredible!
LuiseR won her 300. tournament and her performance (rating over 1800, nearly 70% won games and over 50% won tournaments) is still great!

Jörg Günther at 20070316
Hier am CebitStand merkte gestern jemand an, dass dieses Lob so wÃ¤re wie einen AlkoholabhÃ¤ngigen zu loben: “Herzlichen GlÃ¼ckwunsch, du hast gerade 50 KÃ¤sten Bier getrunken. GroÃŸartig!“…
;)
*duckundweg*

Rafael Marques at 20070316
Allerdings gibt es doch einen kleinen Unterschied: FÃ¼rs Bietrinken ist etwas weniger Hirnleistung (auÃŸer Geld zÃ¤hlen bei Kauf :)) nÃ¶tig als fÃ¼r EWN. ;))

Carroll ★ at 20070316
Liebe Luise,
Ja 70% is gut aber anderes sind besser :
Jonny 138 160 46%
Luiser 1753 758 70%
blackhat 54 21 72%
fraggle_c 123 49 72%
Theo van der Storm 89 32 74%
hanfried_c 131 45 74%
Telestes 66 34 66%
hubbie_c 100 50 67%
YHW 374 196 66%
Was denkst du von Jonny in erste rank mit 46% ?

Carroll ★ at 20070316
Mit tag ist besser :Jonny 138 160 46%Luiser 1753 758 70%blackhat 54 21 72%fraggle_c 123 49 72%Theo van der Storm 89 32 74%hanfried_c 131 45 74%Telestes 66 34 66%hubbie_c 100 50 67%YHW 374 196 66%Henrik SjÃ¸l 78 49 61%

Rafael Marques at 20070316
I try it in english (because we are on a international site :)):
Jonny has resigned many games after some stupid guys reproached (or blamed; I don't find the right synonym in my dictionary) him being a 'rating thief'. Correct me if I'm wrong, but I remember it so… ;)

Jonny ★ at 20070316
Hmm….., there are not so stupid. Because i played near 50 games and lost the bad games so fast as possible and played the other very slowly. Then i wanted to play all my winnig games, to climb to the Top.
So it ist possible to call me a “Rating Thief”. (See the Topic: Rating thiefs) Then i resigned all my winning games and take a time out for 2 Month.
ThatÂ´s it.
I said it again: Nobody had played more than 1000 games.  Accept… And that is a good distance to see a constant performance.
We others in the Top ten are far away from 1000 games.
May be in 1 or 2 Years…. ;))

Pawel Grabowski at 20070317
I played more than 1000 games (but i'm not in top ten nowadays :( ) Diamante has over 2000 games played recently

Jörg Günther at 20070317
> Ja 70% is gut aber anderes sind besser :
Carroll, nobody in the list has even 10% of luisers number of games played. Try posting a list of players with enough games to compare (like 1000).
And also consider with whom luiser played. She played every turnament some time ago (which leads to playing second and third rounds of MC with good opponents) and always played in champoinship level 1.

Theo van der Storm at 20070317
It's a pity I cannot go the CebitStand in Hannover to tell you that your suggestion is like: “Try posting a list of alcoholics.” :)
Perhaps more to the point:
I don't think anybody in the top 10 underestimates the “Powers of Luise”.

hanfried_c at 20070320
The question arised who is “better” – a player with a better ratio but less games or a player with worse ratio but more games. The first one could be just even luckier…
I reformulated the queston an did some mathematics.
Because of the randomness in the game it makes sense to compute for each player X a value x as high as possible such that the statement “The player X wins significantly more than x percent of his/her games.” is true.
I made such computations for a view players and came to the following results.
The two last columns are the numbers of won (W) an lost (L) games. The second column is the winning ratio. The first and third columns are similar with the following meaning:
The probability that a player who wins x_l/1000 of her games will win W or more out of W+L games is smaller than 2.5%.
The probability that a player who wins x_h/1000 of her games will win W or less out of W+L games is smaller than 2.5%.
That means with a level of significance of 97.5% the player is better than x_l/1000 and worse than x_h/1000, respectively. Now we can compare the players.
LuiseR had won significantly more than 68.0% of her games and this ratio is higher than for any other player.
x_l x_m x_h name w l
680 698 716 luiser 1761 761
675 744 803 hanfried_c 131 45
651 736 806 Theo van der Storm 89 32
640 712 775 fraggle_c 121 49
628 737 823 derob 56 20
623 664 703 JohnMarc 'Johnny' McGregor 350 177
614 654 692 YHW 376 199
610 720 809 blackhat 54 21
590 669 739 hubbie_c 101 50
582 727 837 david_artois 32 12
582 650 713 conillet 130 70
580 707 808 kitaktus 41 17
580 655 722 Erik 110 58
563 660 745 Telestes 66 34
561 591 621 Pawel Grabowski 604 418
558 683 787 quadrix 41 19
545 641 727 Carroll 66 37
543 724 853 aoba63 21 8
527 614 694 Henrik SjÃ¸l 78 49
524 650 758 collegestudent 39 21
522 614 698 ypercube 70 44
520 578 634 graff 163 119
519 632 576 Marius HalsÃ¸r 166 122
518 586 650 Kd Hoffmann 123 87
517 602 681 Trevor Green 80 53
515 625 723 test_c 50 30
510 569 627 szamba_ 156 118

Theo van der Storm at 20070320
Nice exercise, but the 97.5% is a bit arbitrary.
Let's take 85% and the table might look very different.
You know what they say about statistics :)

Theo van der Storm at 20070324
Using the same 97.5% I'm getting other numbers:
New x_l x_m x_h New R name
0.680323 680 698 716 0.715740 1 luiser
0.679282 675 744 803 0.801857 2 hanfried_c
0.656478 651 736 806 0.804251 3 Theo
0.643629 640 712 775 0.773167 4 fraggle_c
0.637444 628 737 823 0.820127 5 derob
0.623973 623 664 703 0.702559 6 JohnMarc
0.615222 614 654 692 0.691105 8 YHW
0.618627 610 720 809 0.806190 7 blackhat
0.594601 590 669 739 0.737115 10 hubbie_c
0.596619 582 727 837 0.832356 9 david_artois
0.584692 582 650 713 0.711216 12 conillet
0.591036 580 707 808 0.804613 11 kitaktus
0.583783 580 655 722 0.720748 13 Erik
0.568827 563 660 745 0.742706 14 Telestes
0.561127 561 591 621 0.620366 17 Pawel
0.567916 558 683 787 0.783130 15 quadrix
0.550249 545 641 727 0.724054 18 Carroll
0.564600 543 724 853 0.847154 16 aoba63
0.531727 527 614 694 0.691796 20 Henrik
0.533127 524 650 758 0.754059 19 collegestudent
0.527192 522 614 698 0.695539 21 ypercube
0.521606 520 578 634 0.632894 23 graff
0.520543 519 632 576 0.630769 26 Marius
0.520698 518 586 650 0.648497 24 Kd
0.520660 517 602 681 0.678243 25 Trevor
0.522387 515 625 723 0.719435 22 test_c
0.512075 510 569 627 0.625214 27 szamba_

hanfried_c at 20070327
I have to look at the data. Perhaps the reason is, that I simply used the normal(?) distribution and not the binomial distribution. That gives larger errors for players with only a small number of games.
Or you used newer datas?

Theo van der Storm at 20070328
You should have taken the correct distribution, which is indeed binomial.
example for 89.0% significance:
x_l x_m x_h name w l
70.522% 74.432% 78.001% hanfried_c 131 45
68.832% 73.554% 77.808% Theo 89 32
68.714% 69.826% 70.916% luiser 1761 761
67.814% 73.684% 78.828% derob 56 20
67.068% 71.176% 74.979% fraggle_c 121 49
65.999% 72.000% 77.316% blackhat 54 21
65.121% 72.727% 79.186% david_artois 32 12
63.948% 66.414% 68.800% JohnMarc 350 177
63.863% 70.690% 76.700% kitaktus 41 17
63.254% 72.414% 79.939% aoba63 21 8
63.013% 65.391% 67.701% YHW 376 199
62.399% 66.887% 71.102% hubbie_c 101 50
61.501% 68.333% 74.462% quadrix 41 19
61.175% 65.476% 69.551% Erik 110 58
61.031% 65.000% 68.783% conillet 130 70
60.525% 66.000% 71.091% Telestes 66 34
58.628% 64.078% 69.199% Carroll 66 37
58.048% 65.000% 71.374% collegestudent 39 21
57.252% 59.100% 60.925% Pawel 604 418
56.425% 61.417% 66.191% Henrik 78 49
56.341% 62.500% 68.289% test_c 50 30
56.153% 61.404% 66.412% ypercube 70 44
55.246% 60.150% 64.868% Trevor 80 53
54.600% 58.571% 62.441% Kd 123 87
54.344% 57.801% 61.189% graff 163 119
54.215% 57.639% 60.996% Marius 166 122
53.423% 56.934% 60.382% szamba_ 156 118
This looks much nicer wouldn't you say :?

hanfried_c at 20070329
Since I'm a computer program it probably sounds strange, but I used the normal distribution to avoid the calculation of large bionomial coefficients.
Second it is not clear to me how to compute x_l and x_h (for binomial distr.) without approximation methods.
Yes, for you and me your table looks nicer. But nevertheless I prefer larger significance, i.e. larger “penalties” for players with a small number of games.
It's no problem to play the first 100 games in Level 7 tournament. This strongly affects the winning quota.

Hjallti at 20070330
How did you choose the players in your data? I would like to know how far I'm down in this list…

Theo van der Storm at 20070330
> Since I'm a computer program it probably sounds strange, but I used
> the normal distribution to avoid the calculation of large binomial
> coefficients.
Yes.
> Second it is not clear to me how to compute x_l and x_h (for
> binomial distr.) without approximation methods.
I hope this provides clarity:
Letâ€™s say we have a function that returns the smallest value for which the cumulative binomial distribution is greater than or equal to a criterion value. x_l: Use the bisection approximation method to determine the lowest value of p which exceeds the scored criterion.
x_h: Use the bisection approximation method to determine the highest value of p which falls short the scored criterion.
> Yes, for you and me your table looks nicer. But nevertheless
> I prefer larger significance, i.e. larger “penalties” for players
> with a small number of games.
Sure, no problem, you just want to contribute to the title of this thread.
> It's no problem to play the first 100 games in Level 7 tournament.
> This strongly affects the winning quota.
1. This is not how the high quota players got their high quota, so this is a suggestive and irrelevant remark.
2. From the start it was clear to me that the test (quoting you) “The player X wins significantly more than x percent of his/her games.” is not a proper test of playing strength. I commented on it to give some insight into statistics. Effectively you are saying your posting was just playing with numbers.

Theo van der Storm at 20070331
correction:
Use the bisection approximation method to determine the lowest value of x_l which COINCIDES with the scored criterion.

hanfried_c at 20070404
@Theo: Yes, you are right. It was playing with numbers. It was a respond on Jonny's remark who claimed that winning quotas of two players can only be compared if both have played “enough” games.
> It's no problem to play the first 100 games in Level 7 tournament.
> This strongly affects the winning quota.
> 1. This is not how the high quota players got their high quota, …
NOT? Surely this is not the only way to get high quotas. But it affects the data. Against players between 1400 and 1600 I lost only 6 of 58 games…
> 2. From the start it was clear to me that the test (quoting you) “The player X wins significantly more than x percent of his/her games.” is not a proper test of playing strength.
Yes. This test do not consider against whom someone played and “Among the blind the oneeyed is king.”
@Hjallti:
> How did you choose the players in your data?
I took the first 25 of the actual ranking, the first 25 of an all time ranking and the best 25 tournament players (or something similar, I don't know exactly anymore) getting a list of fiftysomething players. For these players I computed the values and posted all with x_l>50%.
Your scores: Won38, Lost32 leads to a x_l value (using Theo's method) of 43.3%. One reason is the “small” number of games you played.