[Article] Too Much Information -- Seattle, Pittsburgh, Richmond, Boston Legacy Opens

[Mad Zur]

I think the link to this will go up on SCG's main page on Monday, but if you're going to Atlanta, you might want to see it now (http://www.starcitygames.com/magic/legacy/22734_Too_Much_Information_Seattle_Pittsburgh_Richmond_And_Boston_Legacy_Opens.html). It's got breakdowns of the last four Opens, matchup statistics for the most popular decks, and a graph of their popularity since New Phyrexia.

[Gheizen64]

Nice article, thanks, look like the format is finally maturing around some stable decks.

[Gui]

What's EV?

[Amon Amarth]

What's EV?

Expected Value. Something that is bad EV would be trying to break CYOS because, well, it's not very relevant.

[Gui]

Expected Value. Something that is bad EV would be trying to break CYOS because, well, it's not very relevant.

Hmmm... I see... how is it obtained?

[ajfennewald]

I think what they mean by it is expected performance of based on historical matchup records and the make up of the open.

[Gui]

I think what they mean by it is expected performance of based on historical matchup records and the make up of the open.
I got it, but these are specific %s there, how were they obtained?

[Obfuscate Freely]

We calculate the EV ("expected value") number by multiplying a given deck's aggregate matchup win percentage against another deck (right now, that would be the overall matchup result since New Phyrexia became legal) by that other deck's percentage of the field, then doing the same for every other matchup the given deck could have faced, and then adding the results together. The result is an average of a deck's win percentages, weighted by the breakdown of the tournament. It is the win percentage against a random matchup in that field.

We introduced this concept in our analysis of the Boston and Charlotte Opens back in May. Although the numbers are outdated, now, I'll quote the partial calculation, in case it makes things clearer.


For example, to calculate Merfolk's expected value for Boston, we multiply its matchup against the mirror over the course of the year (49.58%) with Merfolk's field presence in Boston (11.11%), which equals 5.51%, then add Merfolk's matchup against Bant (50.90%) times Bant's field presence (9.52%), which equals 4.85%, and then continue through the rest of the field, for a final value of 51.74%.

An EV of 51.74% is not exceptional, but it is positive, and it isn't far from Merfolk's actual performance in the event. If the numbers were further apart, we might seek explanation, such as matchups somehow changing for that event, or pairings that were not representative of the field.

I hope that makes sense. We aren't too sure how to interpret EV numbers, or the differences between them and actual performances, simply because it's difficult to define the significance of either. However, they are interesting, and we've decided over the course of writing these articles that win percentage and EV are close enough, often enough, that deviations are at least worth mentioning.

[Gui]

Ok, so let me see if I got it right:

You get the data from previous tourney, calculate EV using the aforementioned algorithm, and compare to the current data to check the deviation from EV to real winning % against the field?

Oh, and if this is so, could you post the next EV for these decks? :D

[Koby]

I think one thing stood out for me very oddly. The low amount of Maverick being played, which is no surprise to anyone in the USA. It's incredible how few people play it here in the US, while its numbers in Europe are abundant.

[Xantid Swarm]

Just want to repport 1 little mistake in the SCG Pittsburg chart. According to it, there was 1 "Hulk combo" deck in the field, that finished 131th with a 2-2 record. I was in fact playing "Hulk combo" and was probably the only one at Pittsburg, and my result was 7-1 + 2 I.D. for a 4th position. I understand that did not matter a lot, but still.

[Julian23]

Conclusion: Right now, play NO Rug if you're good and want to do well in a big tournament. If you're feeling a little lucky, go for Hive Mind which has greater variance.

[death]

Right now, RUG is popular and is a bad time for playing Hive Mind. U/W Stoneforge is a better choice since it has decent match ups against the tier decks.

[Julian23]

I'm considering NO Rug > UW because it just has the sickest sideboard options being UGr. Especially red with Ancient Grudge, Red Blasts, Grim Lavamancer and maindeck Bolts and Firce Ice destroy like half of the format right now.

[death]

I would like to point out that 4C Loam decks have been flying under the radar. With an EV of 54%+, it has outperformed Team America and BUG in the last 3 Opens, decks which were previously dominant post-MM.

[kiblast]

Conclusion: Right now, play NO Rug if you're good and want to do well in a big tournament. If you're feeling a little lucky, go for Hive Mind which has greater variance.

Or just play a blue deck with 4 Lavamancers maindeck and 4 Pyroblasts SB and smash both, as well as Merfolk.

[dontbiteitholmes]

Conclusion: Right now, play NO Rug if you're good and want to do well in a big tournament. If you're feeling a little lucky, go for Hive Mind which has greater variance.

Or if you're actually good you find a deck that has a favorable matchup vs. NO Rug, UW Mystic, and Merfolk and you have a favorable matchup vs. the most important 25%+ of the field.

[Mad Zur]

Just want to repport 1 little mistake in the SCG Pittsburg chart. According to it, there was 1 "Hulk combo" deck in the field, that finished 131th with a 2-2 record. I was in fact playing "Hulk combo" and was probably the only one at Pittsburg, and my result was 7-1 + 2 I.D. for a 4th position. I understand that did not matter a lot, but still.
It looks like your deck from Pittsburgh was listed as "NO Toolbox", though it was listed as "Hulk Combo" in Providence and Boston. There are a few other decks that have been called "Hulk Combo" (including the 131st place deck in Pittsburgh), but since I don't have the lists, I don't honestly know how similar those decks are to yours.

[Beatusnox]

i know for fact belcher went 4-5 in boston and won matches against burn, rdw, glimpse si, and non-existant opponent number 1. lol. By the way playing four turns of draw go against that si almost made me crap my pants.

[Gui]

It's a great article, keep up the good work! ^^

Could we have the EV for the next event?

[Mad Zur]

Only when we can get the field breakdown for the next event.

EV(x) = matchup%(x vs. deck1) * field%(deck1)
+ matchup%(x vs. deck2) * field%(deck2)
+ matchup%(x vs. deck3) * field%(deck3)
+ ...
+ matchup%(x vs. deckN) * field%(deckN)

So for every deck in the field, we need to know how likely you are to get paired against it and how likely you are to win if you do.

[Julian23]

Asking for the EV for the next event makes me giggle. It's like asking for predicting the future :-). You can however try to calculate it based on your metagame prediction. That won't be very accurate though...

[Gui]

That's why I asked in the previous post if the EV you were comparing to was from the previous event. As it's name is Expected Value, I'd figure you were using the EV calculated from the previous event to check if in the next event, the value was similar to the % of wins against the field. I reckon I didn't knew if it was from the current or previous event.