On August 01 2009 10:08 Roxen000 wrote:
This is pretty interesting. I heard somewhere that TL had the most users out of any Starcraft website Korean or not. Guess not?

TL may have more users than YGosu, but I am pretty sure (based on the YGosu Lotto numbers) that YGosu's betting population is larger than TL's.

(Keep in mind that size/members is not the same as voters.)

konadora

Singapore66364 Posts

August 01 2009 01:40 GMT

#42

Wow that was interesting haha

Glad to see 0.570 for type-b > Stork though (although it's for YGosu)

tonight

United States11130 Posts

August 01 2009 01:49 GMT

#43

This is actually really cool to see. I wonder how big the YGosu community is though?

Altar

United States577 Posts

August 01 2009 01:49 GMT

#44

I would love to see more of this type of stuff, It is beyond interesting to see who everyone picks.

GoSu

Korea (South)1773 Posts

August 01 2009 02:18 GMT

#45

YGoSu, My gf love read this website XD

eshlow

United States5210 Posts

August 01 2009 03:28 GMT

#46

On August 01 2009 01:14 Athos wrote:

Show nested quote +

This just shows how upset prone pro-gaming actually is. We're betting only .07% better than a coin flip right now.

It's .507... which means it's 50.7%... which means it's .7% not .07%.

Although still bad. :p

Seraphim

United States4467 Posts

August 01 2009 04:02 GMT

#47

Nice work dude

zazen

Brazil695 Posts

August 01 2009 06:33 GMT

#48

Hahah I don't hate go.go, I think he has a "role" in the progamer scene: out of so many mannered guys, one HAD to be the asshole hahaha...

But I do think he's a pretty shitty player. I still can't get over his win vs. Bisu.

Railz

United States1449 Posts

August 01 2009 08:12 GMT

#49

On August 01 2009 02:34 JWD wrote:

Show nested quote +

Oh? Do you remember anything else about the interview? I tried searching for "boxer type-b interview" but didn't turn up anything useful.

I'm sorry to say I searched 10 minutes to find the interview again. All I can remember is the context in which the interview was given was it was about a month or two before Boxer's service was over and they were asking about what he thought about the current status of the PL. The picture that was in the thread, unassociated with the interview was boxer holding a "I Love TL.net" sign. =/ u_u Sorry I couldn't be of more use finding it.

JWD

United States12607 Posts

August 01 2009 08:16 GMT

#50

On August 01 2009 17:12 Railz wrote:

Show nested quote +

I think you could be referring to this? It's in the (heavily underrated) Articles forum. However, that interview is pretty old and in it Boxer names Savior and Sea as good up-and-comers...

Thanks for looking, anyway!

TheTeamLiquidTiger

United States627 Posts

August 01 2009 09:05 GMT

#51

wow man, what a write-up
and nice graph too!

7mk

Germany10157 Posts

August 01 2009 17:46 GMT

#52

Hmmm reading this entire thread and actually enjoying it made me realise how big a nerd I am
^^

alffla

Hong Kong20321 Posts

August 01 2009 17:54 GMT

#53

On August 01 2009 17:16 JWD wrote:

Show nested quote +

o wow boxer called stork and savior and he was fuckin rightttttt O___O

upset prone progaming hmm..
sometimes i wonder why i follow sc progaming xD

myrmidon2537

Philippines2188 Posts

August 01 2009 18:49 GMT

#54

Didn't Type-B have a really decent ZvP before? Its pretty much the only reason I LB'd him over stork (either he did or I had a premonition =O)

Type-B over Leta though... O_o

Geo.Rion

7377 Posts

August 01 2009 19:14 GMT

#55

i was thiniking exactly the same when i saw the results:

neither site's voting population demonstrated itself to be significantly more intelligent than a monkey (perfectly random voter)

foeffa

Belgium2115 Posts

August 01 2009 19:17 GMT

#56

Great article, very interesting to read.

Dagobert

Netherlands1858 Posts

August 02 2009 02:45 GMT

#57

Indeed, a very entertaining article.

However...

monkey (perfectly random voter)

this is incorrect. A monkey's actions are anything but random. I don't have the article here, but in a study of the infamous "give x monkeys typewriters and infinite time... and eventually they'll reproduce Shakespeare's works" misconception the monkeys produced lots of written text... consisting of nothing but exactly the same letter over and over again.

Coin flips on the other hand are usually pretty close to randomness.

Then again, humans are monkeys.

And here is why.

Ironically, seeing the voting data... I would have to tentatively conclude that monkeys *are* perfectly random voters. It's just that I think it's due to a floor effect (only 2 choices, even if you click on the same button forever you'll probably get 50%, can you do a check on how often option 1/2 won?)

igotmyown

United States4291 Posts

August 02 2009 03:38 GMT

#58

There would be much less variance if you analyzed a lot more games, regardless of the population size. There's no way you can say TL is more accurate when changing one game could potentially flip the results by 2*(difference)*1/16 = .125*difference, assuming everyone votes. Let's say effort beat Jaedong like in GOM, then the difference is .819-.635=.184, and now YGosu goes up by .025.

Let's assume TL has godlike knowledge of the odds, but not the outcome (since God plays dice).
summing p[ i ]*(1-p[ i ]) gives us the total variance of 2.9, or a standard deviation of 1.7. They will average 70% success rate with a 10% standard deviation.

And of course we usually accept results within 2 standard deviations (in fact this comes close to proving within 95% certainty that TL does not have godlike knowledge). Finally if we don't assume TL has godlike knowledge and merely comparing accuracy, the results of one series of 16 will have to differ by ~.3 to be meaningful.

And YG's results:
+ Show Spoiler +

would average .58 correct with an s.d. of 1.2. YG's odds have more variance, thereby exhibiting a less extreme betting pattern.

JWD

United States12607 Posts

August 02 2009 03:45 GMT

#59

On August 02 2009 12:38 igotmyown wrote:
There would be much less variance if you analyzed a lot more games, regardless of the population size. There's no way you can say TL is more accurate when changing one game could potentially flip the results by 2*(difference)*1/16 = .125*difference, assuming everyone votes.

Of course it's blatantly obvious that I cannot conclude, based on a sample size of 24 games, that the .01% difference in accuracy between TL and YGosu voters is statistically significant. Are you suggesting that somewhere I said "TL is more accurate"?...I'm not sure what you're getting at.