Forum
Advisor Blog
Reviews
Strategies
Ideas
grandgnu, olkainry38, 1st.skyler win 2011 TAZD*
Posted By: Vermont at 10:34 AM, Thursday April 12, 2012 EDT
Here are some additional scoring systems to the TAZD, including TAZD*, ASRm, and ASR. They are all preferable to base TAZD because they reward skill rather than just deviance from a defined middle point.
Specifically, a win (first place) could actually make your TAZD score go DOWN. In all the other systems it would make your score increase (or stay the same due to rounding.)
There are other issues with the TAZD that have been hilighted in the forum numerous times already.
Kudos to chloe for pulling all this together and publishing all her formulae and results. Due to the shortcomings of the plain TAZD, part of what goes on needs to be kept secret, but it's not the case with these. Security through obscurity is not necessary.
Here's Chloe's post, plus rankings:
Just say so if you'd like to join. If you have any questions about calculations or anything, just ask, and I'll happily answer. Suggestions are also always welcome and listened to. In fact! Monte recently suggested normalising the percentage profiles so everyone is on even footing- since not everyone's percentage profiles add to the same number. It's a small change, but it's an important one in my opinion.
An average of 35 games per month are required.
Unless someone specifically requests otherwise, I'll also include everyone in the yearly TAZD*. So far, I've included everyone with sufficient games that joined up to June. I'll add in the later entries soon. If you'd like to be removed, just say so, and I'll take you out no questions asked.
420 games minimum, 2400 games max bonus
Final standings, sorted by TAZD*:
ASRm ASR TAZD* Name
1.85 1096 28447 grandgnu
2.00 1639 23607 olkainry38
1.43 3465 22097 1st.skyler
1.46 2421 21374 Mercantile
1.31 2687 21307 kdiceplaya!
1.70 1719 19583 Mazaman
1.36 1653 19332 stakaboo
1.72 4303 19273 montecarlo
1.26 2474 18691 Lady Lite
1.72 0820 18559 masticore
1.40 8004 17820 Fonias
1.48 1136 17635 DrunkDaShiVa
1.26 3420 17021 Gurgi
1.29 3248 16598 dottir
1.22 1767 15354 cool g
0.89 2351 15331 chaiNblade
1.09 3709 14989 dirtyrolls
1.13 0534 14673 BAMMBI
1.38 2094 14563 Loobee
1.46 0493 14330 Xar
1.09 1609 14316 Az_Balu
1.26 1121 14285 ehervey
1.23 3678 13670 dasfury
1.23 1822 13631 charliedontsurf
1.14 1317 13372 Randomperfection
1.27 1219 13102 caesar-blue
0.63 -0030 12124 THRILLHO
1.18 0377 10911 Crazy Smurf
1.00 1082 10788 greekboi
1.07 0234 10059 ji-jo
1.13 0372 09926 TheBetterYodel
1.06 0396 09274 toad92
1.28 0089 08793 Kibble95
1.14 0068 07885 skrumgaer
1.01 -0205 07740 pooch723
1.02 0678 04874 Simon
0.93 -0041 04471 CCSKAOT
0.38 -0174 03302 joejoewhoa
http://bit.ly/v4mNPe
I got un-dumb and changed from manually summing products to using the SUMPRODUCT function in excel, but google docs apparently can't handle fixed references in its SUMPRODUCT calculation, hence the #VALUE! errors. Everything works fine in the original excel document, which I'd be happy to email to anyone who wants it.
Specifically, a win (first place) could actually make your TAZD score go DOWN. In all the other systems it would make your score increase (or stay the same due to rounding.)
There are other issues with the TAZD that have been hilighted in the forum numerous times already.
Kudos to chloe for pulling all this together and publishing all her formulae and results. Due to the shortcomings of the plain TAZD, part of what goes on needs to be kept secret, but it's not the case with these. Security through obscurity is not necessary.
Here's Chloe's post, plus rankings:
Just say so if you'd like to join. If you have any questions about calculations or anything, just ask, and I'll happily answer. Suggestions are also always welcome and listened to. In fact! Monte recently suggested normalising the percentage profiles so everyone is on even footing- since not everyone's percentage profiles add to the same number. It's a small change, but it's an important one in my opinion.
An average of 35 games per month are required.
Unless someone specifically requests otherwise, I'll also include everyone in the yearly TAZD*. So far, I've included everyone with sufficient games that joined up to June. I'll add in the later entries soon. If you'd like to be removed, just say so, and I'll take you out no questions asked.
420 games minimum, 2400 games max bonus
Final standings, sorted by TAZD*:
ASRm ASR TAZD* Name
1.85 1096 28447 grandgnu
2.00 1639 23607 olkainry38
1.43 3465 22097 1st.skyler
1.46 2421 21374 Mercantile
1.31 2687 21307 kdiceplaya!
1.70 1719 19583 Mazaman
1.36 1653 19332 stakaboo
1.72 4303 19273 montecarlo
1.26 2474 18691 Lady Lite
1.72 0820 18559 masticore
1.40 8004 17820 Fonias
1.48 1136 17635 DrunkDaShiVa
1.26 3420 17021 Gurgi
1.29 3248 16598 dottir
1.22 1767 15354 cool g
0.89 2351 15331 chaiNblade
1.09 3709 14989 dirtyrolls
1.13 0534 14673 BAMMBI
1.38 2094 14563 Loobee
1.46 0493 14330 Xar
1.09 1609 14316 Az_Balu
1.26 1121 14285 ehervey
1.23 3678 13670 dasfury
1.23 1822 13631 charliedontsurf
1.14 1317 13372 Randomperfection
1.27 1219 13102 caesar-blue
0.63 -0030 12124 THRILLHO
1.18 0377 10911 Crazy Smurf
1.00 1082 10788 greekboi
1.07 0234 10059 ji-jo
1.13 0372 09926 TheBetterYodel
1.06 0396 09274 toad92
1.28 0089 08793 Kibble95
1.14 0068 07885 skrumgaer
1.01 -0205 07740 pooch723
1.02 0678 04874 Simon
0.93 -0041 04471 CCSKAOT
0.38 -0174 03302 joejoewhoa
http://bit.ly/v4mNPe
I got un-dumb and changed from manually summing products to using the SUMPRODUCT function in excel, but google docs apparently can't handle fixed references in its SUMPRODUCT calculation, hence the #VALUE! errors. Everything works fine in the original excel document, which I'd be happy to email to anyone who wants it.
|
Vermont wrote
at 7:54 PM, Wednesday December 21, 2011 EST Chloe, what parts of your excel spreadsheet do you have to hide so people can't game your statistics?
|
|
skrumgaer wrote
at 8:04 PM, Wednesday December 21, 2011 EST I hide all of my excel spreadsheets because I don't want people to know where they reside. "Security through obscurity" is a good credo to follow in regard to what you want the world to know about your real life. You don't know who might be looking.
|
|
superxchloe wrote
at 10:20 PM, Wednesday December 21, 2011 EST In what universe does publishing your excel documents require revealing anything about your real life?
|
|
|
skrumgaer wrote
at 11:26 PM, Wednesday December 21, 2011 EST A universe where my knowledge of how to publish anonymously is nil. I learned my code when information was stored on punched cards. I will leave it to you young pups to publish stuff anonymously. You know how to do it. I will stick to what I know, which is statistics.
|
|
|
superxchloe wrote
at 12:10 AM, Thursday December 22, 2011 EST Step one: create a google account- call it whatever you want. Give google a fake name if you want. It really doesn't matter.
Step two: Go to docs.google.com Step three: Press the upload button, which is right next to the "Create" button in the top left- it's the button with disk drive with an up arrow on it. Step four: upload your document (i.e. select the document you want to upload and confirm the upload settings) Step five: Open the document (click on the title from the list of uploaded documents in the center). Step six: click the share button at the top right and change the setting to "public" or "anyone with link," whichever you prefer. Step seven: Copy and paste the "Link to share" that it gives you. I shorten the links using bit.ly because they're long. |
|
|
Vermont wrote
at 8:17 AM, Thursday December 22, 2011 EST My references to "security through obscurity" are due to the following statement you made:
"I have child-proofed the yearly TAZD, but I will not say in what way I have done it." The stat as a metric of skill is so flawed that you have to secretly manipulate the data to hide its shortcomings. At the end of the day, the fact someone can get a win (1st) and have their TAZD go down is obviously, patently, irreconcilably stupid. |
|
grandgnu wrote
at 9:11 AM, Thursday December 22, 2011 EST Sorry, I'm still furiously fapping after seeing my name at the top again.
|
|
|
skrumgaer wrote
at 9:59 AM, Thursday December 22, 2011 EST Verms:
The measure of skill is a function of the player's entire profile, not just the percentage of firsts. You haven't been able to explain why players with different profiles can still be good players. The TAZD is designed to reward good players with different kinds of profile. The profiles that I have designated with animal names, such as the "horse" for Sam, are real results of real good players. Who are you to set yourself up as judge as to which good players are good and which good players are not good? That a system can be gamed has nothing to do with whether it is flawed or not. Taking countermeasures against gaming has nothing to do with whether the system is flawed or not. Don't display your faulty logic for all to see. |
|
|
skrumgaer wrote
at 10:23 AM, Thursday December 22, 2011 EST Chloe:
Step one: Consider a probability distribution such as the calculation for the difference of the expected probability and the observed probability squared, divided by the expected probability, for first place. Step two: For repeated observations, the distribution of the mean of this calculation will tend toward a normal distribution. This result is independent of the number of degrees of freedom of the chi square of which the first place calculation is a part. Step three: The variance of this distribution will go as the inverse of the number of observtions. Step four: The standard deviation of this distribution will go as the inverse of the square root of the number of observations. Step five: To make a determination as to whether the means of the distribution for first place for two players is significantly different, when the number of observations for the two players is the same, you take the difference of the means and divide by the standard deviation. Step six: If you are comparing two players with each other who have different number of observations, you compare each with their appropriate norms and compare the results with each other. Since the standard deviations are different, the difference from norm will have to be scaled appropriately for each player by multiplying by their standard deviation. Step seven: Since the number of observations is the number of games, the appropriate scale for the calculation is the square root of the number of games. |
|
|
skrumgaer wrote
at 10:42 AM, Thursday December 22, 2011 EST An example for Step six:
If players A and B have played 100 games and players C and D have played 400 games and the difference between the means for A and B is 10, the difference between the means for players C and D would have to be 20 for them to have the same significnce of difference of skill as players A and B. |
