Monday, November 21, 2011

EC Gamewin & Timeout Statistics

Since I was asked about statistical data on past VtES European Championships, regarding the distribution of gamewins and timeouts, I dug up the Archons (ExcelSheets) from the ECs in Palma de Mallorca (2009), Paris (2010) and Warsaw (2011). After adding up the results from all the different tournaments (Legendary Vampire, Last Chance Qualifier, Day 1, Day 2 and First Chance Qualifier), the summary for these tournaments look like this:

GW5 GW4 GW3 GW2 No GW Sum Timeout
EC 2011 Sum 18 51 83 39 81 272 102
EC 2011 Pct 6,62% 18,75% 30,51% 14,34% 29,78%
EC 2010 Sum 19 63
EC 2010 Pct 6,33% 21,00% 29,33% 13,67% 29,67%
EC 2009 Sum 15 36 52 29 55 187 64
EC 2009 Pct 8,02% 19,25% 27,81% 15,51% 29,41%

As you can see, despite some suggestions given on, there's no trend towards less gamewins. It's even interesting to notice how similar the distribution of GWs are over the past three EC tournaments. At most, you can see see a very slight trend towards more timeouts (last column in the table above), but this could also very well to be within the normal statistical deviation. After all, the data proves only how individual impressions can lead to false conclusions which are not founded on the actual data.

Two remarks regarding the base data:
  1. The data from Prague is missing the Legendary Tournament (the format was introduced in 2010 in Paris).
  2. I am only counting the five player games in order to have more comparable results.

Update: On November 28th, I have added the missing data from the Paris FCQ tournament.


MS said...

Very nice work up.
The consistency of the numbers is quite remarkable in my opinion.
Without doing the math i would guess that all differences are well within the statistical margin of error.

Anonymous said...

Great Job!

Izaak said...

Just quick:

Let's assume that the 2009, pre-Villein, number is the accepted chance for timeout (34,22%).

H(0): p=0,3422
H(1): p>0,3422

The probabilty value of 2011 is then P(X>=102) = 1-P(X<=101) = 1 -Binominalcdf(n=272, k=101, p=0,3422)= 0,1411 which is well within any accepted statistical variance. One could speak of an upwards trend in case pv would have been less than 0,1 but this isn't the case either.

Therefore based on the data from the recent ECs, there is absolutely no reason to suggest that Villein makes game more likely to time.

This isn't a very surprising conclusion, since every card that is a Villein now was likely a Minion Tap in 2009 and decks just generally don't contained as much masters as they do now. Since masters have an average poolcost of more than 0, the reduced amount of minion cards (and thus ousting power) is compensated by people spending more pool on masters.

I'd do some more analysis, but I should probably put it in a more visible spot than Ralf's comments :)

Anonymous said...

I'm totally flabbergasted by the consistency!


Juggernaut1981 said...

I do find the increase in Timeouts curious. On samples of 200-300 a deviation of 2-3% is not insignificant. Especially since this also includes the greater availability of some cards because of the 'staples' printed in HttB Starters.

I'd be interested to see if it would be possible to spread this project a little further and include all of the Continental Championships. The increased sample size would increase the chances of having any significant trends becoming more obvious.

This consistency may also be a result of a "static" meta-game in Europe, which from my own conversations, seems to be focussed on minimising deck size to reach a goal of 'efficiency'. (I dispute the use of minimisation as a measure of efficiency, but since deck minisation seems to be labelled 'efficiency' lets stick with that).

I'd be curious to see if such a static trend occurred across the NACs for the last few years and if again, there was a confirmation of consistency in the NAC metagame.

extrala said...

@Juggernaut: Unfortunately I only have the Archons of the EC 2009 to 2011.

And I don't know how informative the data from the NAC are. The tournaments are usually only one third in size, so the variance would be much higher.

Juggernaut1981 said...

Oh I was thinking of adding those figures into these, to create a greater portrait of "timeouts" in the more competitive VTES scenes. The greater number of data points should improve consistency of the overall data, even if the data that was added was relatively less consistent.