EPT
About an hour ago I was watching the 3set of Canata-JD (fun game, btw) and saw MBCGame's usual "predict the winner" poll pop on to the screen (it read 77%-23% in favor of Jaedong). I began to wonder whether MBC's voters were smarter than TL's Liquibetting population, and then if Korean StarCraft fans in general were smarter than TL's (predominantly non-Korean) memberbase. The relative massiveness of Korea's SC fanbase must have two competing effects on its collective ability to predict game winners: on one hand, BW's popularity means Korean fans have easier access to interviews, statistics, professional-quality game analysis, and VODs; on the other, BW's mainstream-ness means the average Korean fan is probably less zealous than the average non-Korean Liquibetter.
After some poking around I was delighted to find that YGosu, TL's Korean sister site of sorts, has a voting system (which I think is called Lotto, based on its url) sort of similar to Liquibet (most importantly for my purpose, both systems incentivize "unbiased" voting by recognizing — or boosting the reputation of — accurate predictors).
I decided to pit TeamLiquid's Liquibetters against YGosu's Lottoers by comparing each group's collective predictive capability over the entire Bacchus OSL 2009 Ro16. The results of this little competition substantiate some interesting theories about the Korean and foreign SC fanbases and their perceptions of various OSL players — you can read those below, under "Conclusion".
Results
Here is the grand Liquibet vs. Lotto showdown's results (out of curiosity I decided to include my somewhat sorry LB history, in that third column on the right):
![[image loading]](/staff/JWD/bacchus_osl_s2/lb_ygosu.jpg)
+ Show Spoiler [Graph Description] +
Columns on the right display the percent of the voting population (Liquibetters, Lotto players, me) that correctly predicted the result of the set listed at left. To make the graph easier to read, I've highlighted these cells to indicate prediction accuracy: the more green a cell is, the more accurate (closer to 1) its prediction was; the more red a cell is, the more inaccurate (closer to 0) its prediction was. The sets are ranked according to the success of TL's prediction of their winners.
As the above graph shows, TL just barely edges out YGosu in predictive capability over the OSL Ro16 — but neither site's voting population demonstrated itself to be significantly more intelligent than a monkey (perfectly random voter). Uhhh...let's preserve our dignity by chalking that up to this Ro16's many upsets.
Oh yes, and, of course, I spanked both TL and YGosu with a whopping 54% accuracy. FINGER BOOM!
Conclusion
Liquibets are consistently more one-sided than Lotto votes. I think this is due to a combination of two factors:
- Because TL is much smaller than YGosu, Liquibet is both easier to win and its winners probably receive more (local) prestige. Thus, Liquibetters are less likely to vote along fan lines. Less voting along fan lines means more uneven votes, because underdogs' total votes are less inflated by devoted (stupid) fans.
- When Liquibetters do vote along fan lines, they are much more likely to do so in unison than are Lottoers. I think our (foreign fans') relatively narrow exposure to the Korean BW scene has made us much more monolithic than Korean SC fans. Because we only have a few sources of independent news analysis (and these sources are deeply involved in, and thus influenced by, the foreign BW community) and we only have access to in-depth interviews and coverage of certain players, we tend to both favor mega-stars (favorites) and ignore new and seemingly personality-less players (underdogs) more than Korean fans.
Yarnc. Two of TL's three most inaccurate predictions are Yarnc wins. YGosu didn't call these games right, either — but at least they were closer. This one makes sense, given Yarnc's recent results: we grossly underestimated him in Group A. (Honorable mention for this category goes to overestimating Kwanro, who Liquibetters incorrectly predicted to win both vs. hero and Canata.)Lottoers' weakness: underestimating
fantasy. YGosu just isn't as enamored with Fantasy as we are here at TL, and it hurt them here. Stork had a 46% chance against Fantasy's TvP? BackHo 26%? Come on.Liquibetters' strength:
go.go hate. Hot damn do we hate go.go...Liquibetters overwhelmingly picked him to lose in each of his three games, and hit the jackpot when he actually did (vs. ZerO, one of two games that TL called right but YGosu did not).Lottoers' strength:
type-b love. Holy crap, Lottoers actually picked Type-b to win two Ro16 games — and they were right both times. What did Korean fans know about Type-b that we didn't??Fun facts:
- According to Liquibetters, the OSL Ro16's biggest upset was Yarnc > Jaedong. According to Lottoers, it was go.go > Bisu.
- According to Liquibetters, the OSL Ro16's most lopsided win was Fantasy > Type-b. According to Lottoers, it was Fantasy > BackHo. (So: according to both, Fantasy is a monster.)
- Despite having a slightly worse overall prediction record, Lottoers actually called one more Ro16 game correctly (14/24) than Liquibetters (13/24) did. (TL's strength seems to be more overwhelmingly sticking to heavy favorites.)
- Everyone overrated Hwasin. I think Zero > Hwasin was a pretty easy call, and was really surprised to see both TL and YGosu about 70% in favor of the Red Sniper. I guess people put too much stock in his TvZ?
but I suppose if YGosu scored worse than us, we look alright.
