Statistics against GSL style group format - Page 5

On December 21 2012 17:04 Fenrax wrote:
The OP is very stupid and the amount of people who don't immediately see why the OP is so stupid baffles me. And OP, if you don't understand enough about mathematics to see why your post is so bad you should make less data collecting and do some thinking instead.

The winner of the first BO3 already has a huge advantage. He can lose the next BO3 and still have a chance while the loser has to win two BO3s in a row. That gives the winner of the first BO3 a 75% chance of advancing and the loser only a 25% chance. That already was the advantage and it is a huge one, so any additional advantages on top of having a 3x higher chance of advancing would be unfair. The only exception is this:

(+ Show Spoiler +)

but it is only a corner case and really nothing to worry about because a) such groups are rare in general and b) those groups are even more rare in SC2 where usually everyone can lose to everyone because of BO losses / cheese. B/C will always have a shot at taking out A and they will also always have the risk to lose to D.

The system is basically flawless. You go 2-0 or 2-1 = you advance. You go 0-2 or 1-2 = you are out. Go GSL and OP you should feel bad.

On December 21 2012 19:40 Liquid`Nazgul wrote:
The mathematical goal of group/league play is not to finish above those you beat, it is to do better overall through playing multiple people/teams. If you win two bo3s and your opponent wins one bo3 it means you have done better. Simple math. The fact that you are 1-1 or at a losing record with him is mathematically entirely irrelevant.

Think of larger leagues which operate by the same principle you can lose to a football team twice in a single season, and finish #1 in your league while they finish #10. Nobody complains about that because mathematically the right person finishes ahead, not the one who beats the other. Group play is designed to have the right person advance mathematically. This is also why extended series don't make sense mathematically.

On December 21 2012 19:53 o)_Saurus wrote:
I think the main problem of this system is that it favours zerg quite heavily.
Not because of imbalances or shit but because zerg is supposed to be the more reactive race and therefore it is easier for the zerg to play a second time against a protoss or terran opponent then the other way round.

Just look at recent HSC:
ret loses to Hasu, then beats him.
Stephano loses to Feast then beats him.
DIMAGA loses to Bling then beats him.

I think that is the real issue and I would really like to see statistics about it. The "problem" in the OP is not really present in my oppinion.

On December 21 2012 11:43 lolmlg wrote:

You would have a very, very hard time convincing anyone that this is the case in SC2.

On December 21 2012 20:20 m0ck wrote:

But what are you arguing? In the GSL-format you actually finish in front of the people you beat, and behind the people you lose to (and in front/behind people you don't play).

As for the general discussion:

The problems with the GSL-system are:

1. Since race match-ups have a huge impact on the game, getting a lucky draw for the order of match-ups can have a big impact on the results. By meeting only two of three players in the group, you create a more random system than round robin.

2. It is possible to beat your opponent for second place in maps won, but still not advance. Outside of hand-waving about being "clutch" (which is what rolling a dice can look like from a first-person perspective), looking at it from a matter of probability, it is a problematic system that rates bo3 > bo5 (0 - 2; 2 - 1).

The GSL-system is a more random system than the round robin system. There is no way around it.

It is down to a matter of taste whether the lack of incentives in some situations in the RR-format is enough to prefer the GSL system. For my taste it is a no. I think you can work around the problems of incentives by having teammates always meet in the first match (which is uncontroversial in a RR format, but problematic in a GSL-system) and by incentivizing 3rd and 4th place (you win money/ranking).

I think having the better system for determining the best players should take precedence.

since D is already eliminated he may not be playing at his full potential and he can influence whether his opponent is qualified or not

On December 21 2012 20:56 DrakeFZX3 wrote:


The GSL-system is not more random than Round Robin. In fact, every Bo3 in the GSL-system has meaning whereas Round Robin will have meaningless games when top players have already guaranteed advancement. Why did you think the Naniwa-probe rush incident happened in the first place?

You're putting too much stock on who is playing who and that's entirely irrelevant. The fact of the matter is with the GSL format, you need to win two Bo3s. Plain and simple. Doesn't matter against who or what race.