HyuN vs. Bogus

HyuN wins (60.47%):
      3-0: 17.22%
      3-1: 22.92%
      3-2: 20.34%

Bogus wins (39.53%):
      3-0: 8.73%
      3-1: 14.58%
      3-2: 16.22%

Most likely winner: HyuN (60.47%)
Predicted outcome: HyuN 3-2 Bogus (20.34%)

details, data source, code

Ryung vs. Sniper

Ryung wins (42.81%):
      3-0: 9.83%
      3-1: 15.88%
      3-2: 17.10%

Sniper wins (57.19%):
      3-0: 15.61%
      3-1: 21.62%
      3-2: 19.95%

Most likely winner: Sniper (57.19%)
Predicted outcome: Ryung 2-3 Sniper (19.95%)

details, data source, code

ES.Genie

Germany1370 Posts

November 26 2012 16:20 GMT

#48

On November 27 2012 00:55 TheBB wrote:
Sniper wins (57.19%):
      3-0: 15.61%
      3-1: 21.62%
      3-2: 19.95%

Most likely winner: Sniper (57.19%)
Predicted outcome: Ryung 2-3 Sniper (19.95%)

Is this a mistake or am I stupid?

Bluedraqy

Denmark496 Posts

November 26 2012 16:25 GMT

#49

On November 27 2012 01:20 ES.Genie wrote:

Show nested quote +

Is this a mistake or am I stupid?

The latter I guess?

aintz

Canada5624 Posts

November 26 2012 16:35 GMT

#50

i voted for hyun and sniper. zergs be zerging.

SnipedSoul

Canada2158 Posts

November 26 2012 17:24 GMT

#51

Terran power! Ryung and Bogus.

TheBB

Switzerland5133 Posts

November 26 2012 17:42 GMT

#52

On November 27 2012 01:20 ES.Genie wrote:

Show nested quote +

Is this a mistake or am I stupid?

"Predicted" outcome does not mean "most probable outcome."

For example, if a 0-8 victory for player A was 20% likely, and you had 10% each for 8-0, 8-1, 8-2, 8-3, 8-4, 8-5, 8-6, 8-7 victories for player B, would you say that an 8-0 whitewash victory for player A was a good predicted outcome? After all, 80% of the time, he loses.

My predicted outcome is the one that is least extreme, i.e. the middle of the bar. It takes into account that there is a natural order to the various outcomes. If there were not, we probably couldn't do better than picking the most likely one.

In the above (contrived) example, my method would give a predicted outcome of 8-4 or 8-5 to player B, which seems much more in line with the spread of possible results.

A 3-2 victory for Sniper is the least surprising outcome in the sense that it's the outcome that corresponds best to the current rating gap between Sniper and Ryung, or in other words, it's the outcome that would incur the smallest possible rating adjustments.

BathTubNZ

New Zealand2556 Posts

November 26 2012 20:28 GMT

#53

On November 26 2012 23:04 Shockk wrote:
How quickly are GSL vods uploaded these days? The finals take place at 3:00 AM CET and I'd like to watch the VODs during breakfast five or six hours later. Will they be up by then?

Most of the time a couple of hours, occassionally they can be a bit late. However I'm not sure how moving to Vegas will screw their usual time tables up.

Shellshock

United States97276 Posts

November 26 2012 20:36 GMT

#54

I made this prediction at the end of the ro8
+ Show Spoiler +

On November 22 2012 07:08 Shellshock1122 wrote:
Through the proven science of marinekingology i have discovered a way to predict this season's gsl champion:

In 2012, 5/8 GSL finalists defeated marineking at some point in the tournament
GSL 1: DRG ro16, Genius ro16
GSL 2: neither finalist
GSL 3: MC ro16
GSL 4: Mvp ro16, Life ro8

Ryung is the only remaining player in GSL 5 that defeated MarineKing (ro8)

Only 1 player has made the finals after losing to marineking (Genius) but he beat marineking later on the same day

the only final where it was a mkp victor vs a non mkp victor was MC losing to seed.

Taking this data i can hypothesize that HyuN and Sniper will not make the finals because they lost to marineking in this GSL. Ryung defeated MKP so his chances of making the final are higher. However he will lose to Bogus in the finals since MKP victors can only beat other MKP victors in the finals.

You cant argue with these totally unrelated coincidences!!
+ Show Spoiler +

not serious..... right?

But in reality I'm thinking 3-2 for both HyuN and Sniper. both close series that could go either way. Then I'm going to say Sniper wins the GSL because it would be the most hilarious champion based on what was left (guy never gets any credit) + I think he's actually good enough to do it. The Kid is basically new DRG instead of mini DRG if he wins this GSL

Koerage

Netherlands1220 Posts

November 26 2012 20:39 GMT

#55

are the vods of this season also free? else i might stay up for the games

ThomasjServo

15244 Posts

November 26 2012 20:51 GMT

#56

HyuN and Sniper have been in such good shape lately, I am not saying Ryung and Innovation can't pull through but it is an uphill battle without a doubt.

opterown

Australia54784 Posts

November 26 2012 20:52 GMT

#57

cheering for sniper to take it all :D

c0sm0naut

United States1229 Posts

November 26 2012 20:57 GMT

#58

well
i feel that hyun is a brickwall when it comes to zvt and will beat bogus
but my god, him taking this series would be sick. because i know how it would happen.

Bogus goes up 1 game with some kind of timing and the next round does a proxy 2 rax. Hyun will drop 2 games and then turn it on in g3,4 by winning long, intense games. then in game 5 as fans we will see our lives flashing before our eyes, just as Bogus will, followed by two proxy raxes. The aftermath will be with heavy economic losses to both sides, followed up by a knockdown, drag out, 30 min game where both players play "whack-a-mole" with each others expansions, hyun using a lot of ling bane and aggressive creep spread, bogus doing a lot of pushes out to the middle tower and having drops poach around the side. Eventually bogus will come out on top, dropping hyuns expos while hyun moves out. bogus will barely hold a hive push and come out on top.

One can dream

Ryung vs Sniper

really difficult to say but recently sniper has not impressed me, whereas ryung has. i feel that ryung is one of the players most heavily affected by the map pool, he seems to win or lose based on positional errors or not being able to play comfortably because a map is too cramped. Ryung has good split second decision making but his problem is he often gets surprised by all ins or strange builds. If sniper plays a predictable style, i see Ryung absolutely crushing him, but i think if sniper plays the more flavor of the month fast mutalisk and early ling speed styles (which he has been doign lately in zvt), or even macro roach openings it will be super close

PVJ

Hungary5221 Posts

November 26 2012 21:17 GMT

#59

I need Bogus in the finals for maximum hype! Would like to see Ryung from the other bracket more than Sniper but then again a TvZ sounds better than a TvT :/

On November 27 2012 02:42 TheBB wrote:
My predicted outcome is the one that is least extreme, i.e. the middle of the bar. It takes into account that there is a natural order to the various outcomes. If there were not, we probably couldn't do better than picking the most likely one.

Sorry, just so I get it correctly: Are you basically saying that this score is your arithemtical mean? I am slightly confused what the bars mean, since I guess I've missed your explanation when you upgraded to this new look, so sorry if I'm asking something obvious.

Pretty professional and cool though, keep up the good work.

TheBB

Switzerland5133 Posts

November 26 2012 22:04 GMT

#60

On November 27 2012 06:17 PVJ wrote:

Show nested quote +

The bar is divided into six sections, each corresponds to an outcome, reading from left to right: 3-0, 3-1, 3-2, 2-3, 1-3, 0-3. So the leftmost bar (the reddest) corresponds to the biggest possible win for the guy on the left, the rightmost one (the bluest) corresponds to the biggest win for the guy on the right, and the ones in the middle correspond to less clear wins.

The width of each section relates to its probability. The whole bar is 100% in width. If an outcome was 50% likely it would take up exactly half of it.

My predicted outcome is whichever section contains the middle point of the bar.

You could take the arithmetic mean of all outcomes (weighted by probability), but then you'd get less than 3 for each player. For Hyun vs Bogus for example, you end up with an expected outcome of 2.28-1.76 in favour of Hyun. This means that Hyun is expected to win 2.28 games on average, and Bogus is expected to win 1.76, but it doesn't work so well as a prediction because at least one of them have to win 3 games. You can scale it so that one has 3 wins, which gives you 3-2.32 in this case.

Another thing you could do is to take the expected value of the difference between Hyun and Bogus' scores, and then adjust it so that one has three games. Then you get something like 3-2.39 in favour of Hyun.

So there you have it, three ideas. I kind of like the one I use because it's intended to be used as a prediction, and not an expectation (namely, it gives an outcome that is actually possible).

Input is always appreciated.