or here's a thought experiment: consider a monkey who plays kdice. he clicks randomly, plays like a... well, like a monkey, doesnt connect, rolls 2v5s, and lots of times just clicks the sit-out button. his final distribution is:

0% 0% 0% 0% 13% 40% 47%

his tazd would probably beat any human, even the most "skilled" (according to tazd).

i understand that only computing the top-100 ranked-by-points kdicers should help you skew your sample away from these types of players, but the tazd really really really makes me want to create an alt, and sit as many 0 tables as possible, and leave them without playing. every day for a month until there are 2 days left. then play for real, win a tourney maybe, and get in top 100. then laugh at everyone as my tazd destroys, literally destroys them, like they read the forum post and their heads explode.

i know these circumstances seem nonsensical. but my point is that, with chloes ASR, you cant come up with a loophole that awards bad play. whereas with TAZD, its quite obvious what the loophole is.

and it seems obvious how you can close the loophole. just make positive deviations from 1st to 3rd be positive, and negative deviations from 5th to 7th be positive. forget 4th.

or perhaps close the loophole with Ryan's system of placement weightings. you know, where 3rd is always +10 pts. just choose the point distribution from a 0 table or something. for every 1st you get +120, 2nd is +45 (i forget what the numbers are actualy), 3rd +10, 4th -8, 5th -27, 6th -54, 7th -72.

the strong point of tazd is the zero datum. i like the idea that youre comparing people to the average stats of the worst players here. think it would be interesting to see chloe use your zero datum along with a sensical reward/punishment scale. and since chloe actually listens to our ideas and implements them, ummm... hey chloe, what do you think?

monte, there actually is a way to game the ASR but it's a pain in the ass to do. You can have a negative ppg and a negative multiplier in order to achieve a positive (and possibly 'good') ASR. But, as I said, it's not easy to do. Your percentage profile would have to be like the monkey one you posted in order to do that.

Also, about that alt account: If you were willing to do that for a whole year, the annual tazd doesn't require you to be top 100 so you wouldn't have to actually try the last two days of the month. :)

To back up monte's argument with the computations: The monkey's percentage profile, with 540 games (the minimum to compete in the tazd this year) earns a tazd of 19738, which places the monkey in fourth for the year based on my data set, which differs slightly from skrum's because I substituted montecarlo's december for the full monte's december (and snow leopard for greekboi in november). Some people also asked to be included in the 2010 ASR but did not ask to be inclued in the TAZD.
The average number of games for the TAZDers is 1520. If the monkey plays this many games, his TAZD is 33115, which blows everyone else out of the water.

From my data set, these are the results of the 2010 tazd:

22535 fatman_x
21798 olkainry38
20865 the full monte
18416 Fonias
17542 Xar
17194 trendz
16858 cool g
14953 leeeroy jenkins
14727 ZIGIBOOM
14698 ProxyCheater
14460 greekboi
14084 snmlmz
13950 caesar-blue
13015 noamlang1
12306 chaiNblade
11929 Troy11
11362 dasfury
10919 Kdice_CPR
09785 yellowfin
07448 speciale528
06686 kendawg
06614 pooch723
06350 AlexBallDrop

The standings, of course, are a bit different if you don't weight by the number of games played. Without that weighting, there is no reward for playing a large number of games and no penalty for playing a smaller number of games. The biggest movers are noamlang1 and Fonias, who both dropped 7 places. These two played the largest number of games among the competitors. snmlmz, who played 558 games, jumps up 9 places, but he did not opt to compete in the TAZD.

729 olkainry38
671 the full monte
596 snmlmz
558 fatman_x
530 chaiNblade
523 leeeroy jenkins
452 trendz
433 Xar
428 ZIGIBOOM
411 ProxyCheater
389 dasfury
364 Fonias
361 cool g
311 greekboi
299 caesar-blue
286 yellowfin
285 Kdice_CPR
255 speciale528
246 AlexBallDrop
244 Troy11
231 noamlang1
175 pooch723
158 kendawg


As for using the zero datum to create another stat, I'll work on a formula and get back to you monte. Are you thinking similar to the asr, with table points weighted, just straight percentages, or something weighted by games (or with a small games bonus)?

monte,

If you leave without playing, you would have 100% sevenths unless someone else flags first, which would be possible when that player gets fewer lands. If you use the same alt all the time, people might get wise to your strategy and adjust their play accordingly; so it would be better to use a different alt each time and keep records of your place of finish. Or, instead of sitting yourself, watch a number of games and keep a record of players who lag out without playing and their places of finish. You can do this at any level of table. This datum would not have the problem of including "good" players who are temporarily at zero because of bad luck.

Skrum, you miss monte's point. One can flag out immediately if they so choose. I'm pretty sure the ridiculous 7ths percentage was the point. An account with 100% 7th places and a decent number of games would do extremely well in the TAZD. People don't keep track of who plays and flags out all the time at 0 tables. No one cares that much. Flagging out immediately doesn't really qualify as a 'strategy'. Pretty sure that if anything, people would notice that you take 7th always and thus would rather sit with you because they're competing against 5 other players instead of 6 other players.

It should be noted that if that account earned its points from tourneys at the end of the month, it would also do fairly well in the ASR. Multiplier of -2, with -50 ppg and say 500 games? since you're sitting and leaving immediately, I don't think that's an unreasonable number. This earns an ASR of 510, which would have been 12th in December. Of course, the TAZD for the same stats would be 75284. This is literally more than five times greater than the top December TAZD.

I did not miss monte's point. You would have to teach a monkey to flag out. That means it's a skill. Merely teaching him to sit in a game takes much less work. So the TAZD for sitting and leaving would be less than the TAZD for sitting then flagging out. Would that make sense for a a statistic that measures skill?

I thought of a source of data for the average percentages for someone who sits and leaves. The leaderboard has the attack/defend stats (which are pretty much useless, since they are ratios, not raw numbers). Find all zero score players who have zero attack percentages, and determine the distribution of places of finish.

I still have the data from 900-odd players that I calculated the zero datum from. When I get back to my spreadsheet, I will sort the players by attack percentage and see what patterns I get.

"the tazd really really really makes me want to create an alt, and sit as many 0 tables as possible, and leave them without playing." is not the same as his point about the monkey. you missed it.

Clicking a box to quit a game without playing doesn't really qualify as a kdice skill imo. I would expect that it takes as much effort to train a monkey to click the check box as it does to train him to click the sit button.

Sitting and NOT playing just 35 games in a month gets you a TAZD of almost 20000. This is significantly greater than the best TAZD of real players, who put forth an effort to earn points and do well in games last month. Does this make sense for a statistic that tries to measure skill?

A monkey that always flags out seventh needs more training than one who merely sits out. (Remember, you have to have the fastest finger. Sixths or fifths just won't cut it). So a monkey that always finishes seventh deserves a bigger TAZD.

Looking at my zero datum data set (real players playing real games), I find 42 players that appear to have sat out and not flag. Their percentages of finishes were 00, 00, 00, 04, 23, 25, 48. A player with that profile would have a TAZD of 3708 for 35 games.

Based on these numbers, I would tend not to suspect a player of trying to game the TAZD unless he has more than 48% finishes. Even noam did not have that high a number over all his months. To prevent the TAZD from being lowballed, all I would have to do is to put a 48% cap on the number of sevenths.

If I start seeing alts named therealmonkey, monkeycarlo, or superxmonkey, I know that people are trying to game the TAZD. Then I will set up a competition called the TABAM (Test Against Bad Ass Monkey).

The fact that skrum took the monkey analogy literally rather than actually recognizing TAZD is a measure of deviation rather than skill shows a mind-boggling level of obtuseness. Arguing about how much skill a monkey would need is just ridiculous.

And the fact that you have to have a special test to exclude people shows your measure is flawed. You don't see any other competitions having to bump people out because they are doing well by trying to do poorly. The fact that that sentence makes little sense shows how bankrupt your logic is.

You're so unwilling to admit you might be wrong that you're twisting things around terribly.

Dear Red Sox,

we noticed that you have made it into the playoffs by losing many of your games. We apologize that our ranking system shows you in the top five, but we will now have to exclude you.

In the future, please try to play slightly less poorly so that your ranking decreases. Keep in mind that if you get too good it will start going up again.

We have considered moving to a system that accurately reflects facts and shows better teams ahead of lesser teams, but are busy discussing how much skill a monkey would actually need to lose games on purpose. (Not that we are implying you are a bunch of monkeys; that would be the Braves.)

Sincerely,
the commish