Forum
Advisor Blog
Reviews
Strategies
Ideas
Fatman_x takes the 2010 TAZD; computation of complete standings in progress.
Posted By: skrumgaer at 6:51 PM, Saturday January 1, 2011 EST
Here are Fatman_x's final stats:
1631 22% 16% 13% 12% 10% 11% 11% 22535 Fatman_x
Olkainry38 did not play any additional games after Dec 21.
Here were the two leaders on Dec 21:
1591 22% 16% 13% 13% 10% 11% 11% 21975 Fatman_x
0894 22% 20% 16% 13% 11% 08% 07% 21798 olkainry38
The November scores are below. They were recomputed when I found I had been using the 2009 datum instead of the 2010 datum.
The Test Against Zero Datum (TAZD) is a weighted sum of the squares of the differences between a player's percentage profile and the profile of a typical player with a zero score, adjusted according to the square root of the number of games played. If you would like to enter the cumulative TAZD competition for 2010, reply to this thread in the account that you want to enter.
In 2010, the cumulative TAZD began in April, because some January, February, and March profiles had corrupted percentages.
A minimum of sixty regular games per month was required.
Entries show number of games, percentages, and player name.
A minimum of 540 regular games played was required to remain in the standings as of December 31.
Here are the end of November standings with the new datum.
0815 22% 20% 16% 13% 11% 09% 07% 20192 olkainry38
1324 22% 16% 13% 13% 10% 13% 11% 19645 Fatman_x
0886 21% 18% 18% 13% 11% 08% 07% 19382 the full monte
1612 15% 19% 16% 14% 14% 11% 07% 17412 Xar
2297 18% 15% 14% 15% 13% 13% 10% 17190 Fonias
1282 19% 15% 14% 14% 13% 13% 09% 14698 ProxyCheater
1089 19% 16% 12% 15% 13% 12% 09% 14359 ZIGIBOOM
1817 16% 18% 12% 10% 11% 13% 16% 14038 cool g
0731 18% 22% 12% 10% 08% 11% 16% 14010 leeroy jenkins
1810 19% 14% 12% 12% 13% 15% 12% 13270 caesar-blue
2990 13% 11% 11% 08% 09% 10% 34% 12928 noamlang1
0519 21% 15% 18% 11% 12% 08% 12% 12080 chaiNblade
1597 19% 14% 09% 09% 10% 13% 24% 11916 greekboi
0769 22% 14% 09% 13% 13% 13% 14% 11355 dasfury
1054 18% 14% 13% 13% 12% 11% 15% 09627 yellowfin
0698 17% 15% 10% 13% 14% 14% 13% 06707 speciale528
0662 17% 14% 12% 12% 13% 13% 15% 06340 AlexBallDrop
1122 11% 15% 15% 14% 11% 12% 18% 05785 pooch723
1625 16% 12% 11% 10% 12% 15% 22% 06512 kendawg
0613 14% 11% 17% 11% 11% 15% 16% 04945 vIRGI
1631 22% 16% 13% 12% 10% 11% 11% 22535 Fatman_x
Olkainry38 did not play any additional games after Dec 21.
Here were the two leaders on Dec 21:
1591 22% 16% 13% 13% 10% 11% 11% 21975 Fatman_x
0894 22% 20% 16% 13% 11% 08% 07% 21798 olkainry38
The November scores are below. They were recomputed when I found I had been using the 2009 datum instead of the 2010 datum.
The Test Against Zero Datum (TAZD) is a weighted sum of the squares of the differences between a player's percentage profile and the profile of a typical player with a zero score, adjusted according to the square root of the number of games played. If you would like to enter the cumulative TAZD competition for 2010, reply to this thread in the account that you want to enter.
In 2010, the cumulative TAZD began in April, because some January, February, and March profiles had corrupted percentages.
A minimum of sixty regular games per month was required.
Entries show number of games, percentages, and player name.
A minimum of 540 regular games played was required to remain in the standings as of December 31.
Here are the end of November standings with the new datum.
0815 22% 20% 16% 13% 11% 09% 07% 20192 olkainry38
1324 22% 16% 13% 13% 10% 13% 11% 19645 Fatman_x
0886 21% 18% 18% 13% 11% 08% 07% 19382 the full monte
1612 15% 19% 16% 14% 14% 11% 07% 17412 Xar
2297 18% 15% 14% 15% 13% 13% 10% 17190 Fonias
1282 19% 15% 14% 14% 13% 13% 09% 14698 ProxyCheater
1089 19% 16% 12% 15% 13% 12% 09% 14359 ZIGIBOOM
1817 16% 18% 12% 10% 11% 13% 16% 14038 cool g
0731 18% 22% 12% 10% 08% 11% 16% 14010 leeroy jenkins
1810 19% 14% 12% 12% 13% 15% 12% 13270 caesar-blue
2990 13% 11% 11% 08% 09% 10% 34% 12928 noamlang1
0519 21% 15% 18% 11% 12% 08% 12% 12080 chaiNblade
1597 19% 14% 09% 09% 10% 13% 24% 11916 greekboi
0769 22% 14% 09% 13% 13% 13% 14% 11355 dasfury
1054 18% 14% 13% 13% 12% 11% 15% 09627 yellowfin
0698 17% 15% 10% 13% 14% 14% 13% 06707 speciale528
0662 17% 14% 12% 12% 13% 13% 15% 06340 AlexBallDrop
1122 11% 15% 15% 14% 11% 12% 18% 05785 pooch723
1625 16% 12% 11% 10% 12% 15% 22% 06512 kendawg
0613 14% 11% 17% 11% 11% 15% 16% 04945 vIRGI
|
skrumgaer wrote
at 11:17 AM, Wednesday August 4, 2010 EDT Vermont:
Also, since the TAZD is a stronger measure of positive skill, it is a weaker measure of negative skill. But still, it measures negative skill. The chi-square used here is a two-tailed test. You can complain that negative skill should not be rewarded, but that is like claiming that lowball poker should not be played. But there is nothing wrong with the computation. |
|
Vermont wrote
at 1:42 PM, Wednesday August 4, 2010 EDT If the tazd were ACTUALLY a measure of skill, there is no way noam would be in first. Anyone can look at that and see that obviously it is not actually accomplishing it. Sure, maybe he won't end up on the very top in the long run, but it should be clear to everyone that if this were truly a measure of skill that he would be last out of the ten people listed.
Chloe stated it much better: "...the tazd isn't a test of skill, it's a test of deviation from a specific data set." Her assessment is far more consistent with the results we're seeing here. |
|
|
Vermont wrote
at 2:00 PM, Wednesday August 4, 2010 EDT In thinking a little bit more about this, for it to work (more) properly, it would need to take into account if it is deviating in each position in a positive or negative direction.
So a 20% first place would be +6 while a 10% first place would be -4. So a 20% seventh place would be -6 while a 10% seventh place would be +4. Then the result would hinge not just on deviation from the norm, but positive deviation from the norm. Some players would end up with negative modified TAZDs, which would be accurate since they did worse than the "average" score. Does that make sense? I don't have the exact equation you're using so I may be off base. |
|
superxchloe wrote
at 6:58 PM, Wednesday August 4, 2010 EDT http://en.wikipedia.org/wiki/Goodness_of_fit
Basically, the tazd is a test that shows how well your actual data 'fits' with your expected data.. It truly is a measure of deviation from some 'expected' value, in this case the zero datum. This is usually achieved by earning a large number of 1sts/2nds and a small number of 6ths/7ths. Take for example noam's stats: 12% 11% 10% 07% 07% 10% 39% versus the zero datum: 10% 10% 10.3% 12.9% 15.5% 17.2% 19.4% (for the tapl, these numbers are all 14) The difference between the actual first percentage and 'expected' first percentage (12-10=2) is taken, squared (4), and divided by the expected percentage (4/10=.4). Do this for all seven values, add those numbers together, then mutliply by the square root of the number of games then multiply by 1000 to get a number that doesn't seem silly. Because the difference is squared, having 8% firsts is equivalent to having 12% firsts. The squaring is important to the test though- it means that a high first % and a low seventh % both help. Normally, people achieve a lower-than-expected seventh percentage because the zero datum set seventh percentage is so high (19.4%) but because his statistics are so extreme, noam's tazd is quite large (and his tapl is larger, as skrum noted). Hope that helps Vermont. I don't believe that flagging first round qualifies as 'negative skill'. It's far easier to achieve sevenths than it is firsts, so there isn't skill involved. If there were such a thing as 'negative skill' I would agree that noam should be rewarded for it, but there simply isn't. Lowball poker still requires knowledge of the game and its rules, but getting a seventh in dice doesn't require that you know anything about the game at all. |
|
|
Vermont wrote
at 8:27 PM, Wednesday August 4, 2010 EDT That makes perfect sense, and is exactly what I was thinking....this is no way measures skill. We can certainly come up with a more useful calculation.
|
|
|
skrumgaer wrote
at 9:54 PM, Wednesday August 4, 2010 EDT Chloe:
My check of noam's stats right now shows two sevenths in the last ten games, one of 2 rounds and one of three rounds. Instead of making inferences from small samples, perhaps we can hear from people who have observed noam play or from noam himself. Vermont: I have another measure that has negatives for lower places: the convolution integral. I haven't trotted it out for a while, but it is described on my Wall. Also, the baseball-style standings penalize the lower places. There is a rough correlation between people who finish high in the monthly TAZD and those who do well in the baseball-style standings, but we have sampling error because only the top 100 are drawn. |
|
|
Vermont wrote
at 11:24 PM, Wednesday August 4, 2010 EDT I agree that any of the other algorithms you use track actual performance better than this, which just tracks deviation.
|
|
greekboi wrote
at 2:26 PM, Thursday August 5, 2010 EDT 23% 23% 10% 10% 15% 8% 8%
l33t so far this month |
|
noamlang1 wrote
at 3:58 PM, Thursday August 5, 2010 EDT one of the reasons I have so many 7ths is I start every month on the 0 tables, If I see I cant get 3rd and above and I have 0 points. 4=5=6=7 meaning I dont care taking 7th and just getting out of there
The other reason is I am not that good, plus take my stats on may, and it takes it somewhat out of proportion my stats just get better every month |
|
|
superxchloe wrote
at 10:01 PM, Thursday August 5, 2010 EDT Vermont: I recommend checking out moondust's alternative skill ranking: http://kdice.com/discussion/topics/44798227?
which I feel is a much better skill ranking than tapl/tazd personally, though I'm unsure about the weights of the multipliers. The other plus is that the alternative skill ranking doesn't take into account number of games, so if you've played 100 games in a month you're not at a huge disadvantage to someone who has played 1000. |
