The Art of Reducing Variance

hmmm spokone

Prends RUG, un des meilleurs decks du format, je pense qu'il a quasiment aucun deck qui gagne plus de 70% contre toi.
De plus, le mec en face ne va pas toujours avoir la god hand.
De base, tu gagnes déjà 10% de tes matchs car le mec va flood/death/mull à 4. Donc il y a pas de match impossible à gagner.
Dans vos exemples, vous partez toujours du fait que le mecs à les 7 bonnes cartes dans sa main de départ ce qui pas possible à 100%.

Tu peux venir avec ton RUG, je veux bien parier une boite de RTR que tu ne gagnes pas 30 matchs sur 100...

MatF a écrit:Prends RUG, un des meilleurs decks du format, je pense qu'il a quasiment aucun deck qui gagne plus de 70% contre toi.
De plus, le mec en face ne va pas toujours avoir la god hand.
De base, tu gagnes déjà 10% de tes matchs car le mec va flood/death/mull à 4. Donc il y a pas de match impossible à gagner.
Dans vos exemples, vous partez toujours du fait que le mecs à les 7 bonnes cartes dans sa main de départ ce qui pas possible à 100%.

Je me exprimerai en Anglais pour la suite - j'en ai pas du airco chez moi (contraire qu'au boulot) et c'est trop chaud pour faire des efforts languistiques...

1) You apparently use a different definition of matchup-percentages than me. Taking account mana screw and extreme mulligans seems futile, since the sample size of your games will never be high enough to filter out that variance on both sides. On a side note, though, you should definitely not always "go back to 7" while playtesting, since mulliganing is part of the game, and an essential part of your skillset ; also, aggresively mulliganing (or not) is part of strategy just like every other aspect of your deck which you're trying to test. But this "10%" of games shouldn't play a part in your determination on how two distinct decks match up against each other.

2) Using RUG Delver to demonstrate this point seems unfair, since that deck is by design the "50-50" deck in the format. The deck is specifically tailored to have an extreme amount of interaction with your opponent while presenting a steady clock. The amount of early-game interaction the deck packs, is why it will never have absolutely abysmal matchups across the entire field. However, there are plenty of viable, equally competitive (given the right metagame) decks which have more variable matchup-precentages. I've pointed to Maverick even though that deck does have a lot of play against a lot of opponents, just because I had this clear cut example from the GP, with a decent (understatement) player you're all familiar with to boot. Most linear combodecks, however, demonstrate how lopsided (both ways) soms matchups in Legacy tend to be.

Je confirme bien qu'il y a des matchs up (quasi) impossible. Pour ma part Tezzere Control vs RUG delver, il faudra demander à ceux qui jouent delver ce qu'ils en pensent mais cela leur laisse un gout relativement amer.

Moi par contre j'étais incapable de gérer Jitte sur un noble hierarch ou un simple bird. J'étais le bye de stoneblade (meme si je n'en ai pas joué)

Avant side, il y a toujours pas mal de match up impossible. Chalice à 1 en legacy, cela laisse tres peu de solution à l'adversaire.

La variance il me semble est vachement réduite dans ce format donc je n'en tiendrai pas forcement compte. Par contre je reconnais qu'en draft la variance est énorme. Un jour je gagne la draft et les 2 autres fois je fais 0-3 comme pas mal de personne. Moto reste sur 4 x 0-3 d'affilé alors qu'il remportent souvent celles-ci. Idem pour vincent

3) I will agree with spokone that RUG tends to have difficulties with Chalice-based decks, but I strongly contest his "70%" statement, even excluding insane mulligans or terrible mana issues. I would gladly take that bet, if I lived in Brussels. And that's given the fact that I'm pretty terrible at playing tempo-based aggro (probably the archetype I have the most trouble playing correctly with, in fact). I suggest MoTo gives it a go though, he'll gladly take you display I bet (btw, I'm assuming that 2/3 of the games are post-SB here, since that's how tournament magic tends to work...).

4) In Legacy, players have a lot of tools to their disposal to make their decks as consistent as possible (fetch-dual manabases to assure the right colors of mana are available, cantrips,...) but that doens't completely eliminate in-game variance - does he have the FoW/Daze/StP/whatever or not. Also, because of the vastness of the format and the immense strategic diversity, variance in your pairings is high than ever. I get what you're trying to say, but I do believe that variance in Eternal formats is equally present as is the case in other formats. It just lies in differents aspects of the game.

5) Finally, citing a few drafts seems a pretty terrible way to demonstrate this issue. Determining the magnitude of variance is reliant on using large sample-sizes (=basic statistics). Drawing conclusions from small-sample-size observations seems iffy to me. Also, what kind of variance are you trying to demonstrate from your example?

Kind regards,

Tom

Nirves, si je dis cela c'est parce que l'on a vraiment testé longtemps avec Moto ce deck contre Delver et il faut bien le reconnaitre que dans son format classique (le side classique) RUg n'a vraiment pas beaucoup de chance du tout de l'emporter.

A gand j'ai perdu un RUG qui avait encore un life from the loam en side et qui s'est meulé jusqu'à trouver son seul ancient grudge alors que je n'avais a ce moment qu'un chalice à 1 et un ensnaring bridge sur table le tour d'après je posais le second bridge et un nouveau chalice via le tezzeret.

Je réitère que je veux bien prendre le pari sur base d'un RUG qui ne prévoit pas un side uniquement contre ce deck

1) Honestly, I think that if you does not consider mana screw/death/mull, you are making stat about another game than MtG. A game where people choose the 7 cards they start with. Those bad starting hand are part of the game and cannot completely not consider them. Yes Belcher and Elf will win 100% against you deck with perfect hand, but how often do you have the perfect hand? 90% of the time? 80%? If it is 90%, good for you, it is probably a good deck and you should play it.

2) RUG is not the only one, there are plenty of other deck that can interact and be able to play against everybody (UW, BUG landstill, ...) and combo deck can also by essence be able to win against any deck. Some deck worst opponent are themselves, not the deck in front of them (Belcher, ...). Honestly, most good deck in Legacy are able to win whatever the guy is playing in front of them.

3) Chalice is the best anti RUG deck, but it is a pretty poor deck against the rest of the field. I did not see many Chalice decks winning big event recently. You prefer to play a deck with 60% win against 90% of the deck and that lose to 10% of the field, or a deck that win 100% against 10% of the field and less than 50% against the rest. That why nearly nobody play chalice deck, because it is good only against some match up. A good deck is good against most match up.

NiRVeS a écrit:Ca nous porte un peu hors sujet, mais...

le HS dans le HS !

spokone: I'll repeat: I do not claim that your matchup will be bad - I'll wholeheartly agree that your deck has a positive matchup against RUG. And it seems pretty clear that both decks should be stock: I assume most RUG decks will have a singleton Ancient Grudge, maybe a Krosan Grip, maybe a Life from the Loam and probably a few REB/PBlast to side in against you, but nothing more than that. Whereas your deck is pretty much designed to crush RUG-style decks. However, a 70-30 matchups seems unlikely to me, just by the nature of how RUG is built and by the natural inconsistency stax-type decks tend to posess.

MatF:
1) I didn't want to imply that screw/death/mull shouldn't be considered from testing a deck, since these are all part of the game, as you state, and inherent features of the deck you're trying to construct or optimize. I'm stating that, when testing a matchup in, for example, a 10 game match (4 matches pre-side, with each player on the play twice, followed by 6 sideboarded game where each player starts three times), counting games where one player mulls to 4 or keeps a decent one-lander and never draws a second land within the first 5 turns (I'll assume brainstorm/ponder misses here, since I can't imagine other reasons to keep a onelander in most decks) will greatly dillute the quality of your testing. Mulliganing down to 6 or 5 cards is, as you state, part of the game and part of the skillset of the players. These are not "exceptional circomstances" in my book and can/should be taken into account, of course.

2) I think we actually agree on that one. Every viable deck has at least the possibility to defeat any opponent. Brainstorm-based decks tend to have a wider range of decks against which their matchups will be closer than others, while combo-decks tend to have a "wider spread" - more blowouts on both sides, less "close" matchups. The point I was trying to make, is that even though there are decks with very lopsided matchups (70-30 or even worse), RUG seemed to be a bad example since it tends to have very flat matchups throughout the entire field.

3) Valid arguments against the Chalice-decks. That's why it's been 3 years since I last touched my Chalices to play at events, even though I've been tinkering with UB Stax for some time now during our testing-evenings.

So yeah, I think we're actually on the same base here .

Variance et Mana vus par Conley Woods: Mo’ Mana, Mo’ Problems