Three tournaments added, OP rankings updated:
July 21, Magic Bazar Paris Monthly @ Paris, 33 players: 1.) Esperblade 2.) Humans 3.) Grixis Death's Shadow 4.) Dragon Stompy 5.) Grixis Control 6.) OmniTell 7.) RUG Delver 8.) Tempo Zoo
July 28, Staples @ Collector Legion (Gardena, California), 46 players: 1.) Grixis Control 2.) Eldrazi 3.) Dredge 4.) Sneak & Show 5.) Esperblade 6.) Sneak & Show 7.) Sneak & Show 8.) ANT
July 28, Millennium Games Legacy Series @ Rochester NY, 44 players: 1.) Lands 2.) TES 3.) UWr Stoneblade 4.) OmniTell 5.) Grixis Control 6.) UWr Stoneblade 7.) Lands 8.) Punishing Maverick
One tournament added, OP rankings updated:
July 29, MTGO Challenge, 7 rounds: 1.) Grixis Control 2.) Aluren 3.) RUG Delver 4.) Miracles 5.) Miracles 6.) RUG Delver 7.) Grixis Control 8.) Miracles 9.) Soldier Stompy 10.) Sneak & Show 11.) B/R Reanimator 12.) Sneak & Show 13.) Death & Taxes 14.) Grixis Control 15.) Steel Stompy 16.) ANT
Last edited by Scott; 07-30-2018 at 09:23 PM.
So, which deck is good against Miracles, Show and Tell, Stoneblade, Death & Taxes, Grixis Control, RUG Delver, and Chalice? Or most of it.
Next is Show and Tell or those impossibly cheap creatures like Delver and Angler. I mean Grixis can play 20 counters, mana denial and drop 5/5 with no drawback. Curiously they are the top dogs.
This is the base of 80% of top decks
4 Daze
4 Fow
2 Pierce
4 Ponder
4 Brainstorm
3/4 Preordain
Then the finisher can be either SFM, Angler TNN, Delver, YP, Mentor, Clique (super cheap and effective creatures that don’t distract your mana too much) or SnT/Sneak Attack/Reanimate
The results depends on the variance of draws in the end. The problem? It has been like this for 10 years. The only variants are Eldrazi, BR Rea and...?
Last edited by Poron; 07-31-2018 at 09:20 AM.
D&T, Lands, moon stompy, elves, …
Funny, i thought we'd see more maverick..
Why? Control decks have Terminus/K.Return and black has Toxic Deluge
Maverick forte has been historically the lands are everyone is playing mana denial
Sort of related tangentially, is anyone a programmer on these boards with experience programming recurrent neural networks? What would happen if you feed such a network all the top decks from the past year of tournements - what decklists would the program give out? (Related, because I'm assuming if the network sees X certain blue cards over and over, it will create a deck list with those x blue cards). I'm guessing the feed into the network will have to be curtailed (unless we want random elves/goblins etc showing up) to "blue" decks, but would be curious what blue decks the network would give out (and would also be curious on the overall input/output).
All Spells Primer under construction: https://docs.google.com/document/d/e...Tl7utWpLo0/pub
PM me if you want to contribute!
Somewhat. I've had something of an ongoing project for a couple years now teaching a computer to play MTG. The projects currently on hold but I was able to identify a few successful cards and deck configurations with what I had built.
It gets really, really complex but I'm always happy to talk about it if people are interested. Suffice it to say, I found a system which I believe works, but isn't time efficient enough to handle a big pool of cards.
My current approach is actually a bit different, and ranks cards vs cards given criteria to create pass/fail conditions, and then uses a bit of social network analysis to create groups of core cards. The trick to this, is that it requires an absolutely immense amount of data, but should in theory be possible because it just has to parse oracle text properly. Though, such a system is only good for aggro, midrange, and control. It will not identify combos.
@Brael: that's really interesting, but I was primarily thinking that the only variables would be the card names and numbers, no need to parse oracle texts. In effect instead of a network figuring out the best deck of a certain type for a certain metagame (what I believe you're trying to do), I wanted to know if the network could figure out the most common cards and numbers based on decks that have been played within a certain time frame.
That's easy, you don't even need a neural network for it. A database and charting the results is enough. I don't think the results are particularly helpful though. If I understand you correctly, all you're really trying to do is correlate the prevalence of specific cards to the success or failure of particular decks.
The real issue with that has nothing to do with technology and more to do with the amount of data available. Magic is an incredibly variable game, and it's pretty hard to get statistically significant amounts of information like this out of the tournaments that get reported. There's enough data available to make an approximation as to what decks are generally successful, but not enough to go into details for that success.
This is very important when you're trying to look at single cards because minor deck changes may or may not have any impact what so ever in outcomes. For example, I've run simulations before where over 10 million games, a 24 land deck made it's first 4 land drops many more times than a 26 land deck (50 vs 45%) and where a 28 land deck drew 0 lands in more opening hands than a 22 land deck (1% vs <0.1%). A handful of cards, in a handful of decks, over a very small sample size isn't really guaranteed to give you any accurate information.
Two tournaments added, OP rankings updated:
July 28, The Last Sun 2018予選 - 晴れる屋トーナメントセンター#2, 72 players: 1.) Grixis Control 2.) Grixis Control 3.) Eldrazi & Taxes 4.) Dark Depths 5.) ANT 6.) BUG Death's Shadow 7.) Miracles 8.) Sneak & Show
July 28, LML Sexta etapa 22/7 @ TCGeek, 35 players: 1.) Elves 2.) Miracles 3.) Goblins 4.) ? 5.) Turbo Depths 6.) Dredge 7.) Esperblade 8.) Merfolk
Three tournaments added, OP rankings updated:
July 28, 37º Alpha Legacy @ Magic Store (Brasil), 40 players: 1.) U/W Helm 2.) U/W Stoneblade 3.) Steel Stompy 4.) Shardless BUG 5.) OmniTell w/ Sneak 6.) B/R Reanimator 7.) BUG Lands 8.) Jund
July 29, レガシー神トライアル, 83 players: 1.) Miracles 2.) U/B Death's Shadow 3.) Grixis Control 4.) Colorless 12-Post 5.) Grixis Control 6.) Aluren 7.) Aggro Loam 8.) Pox
July 29, Monthly July @ London, 42 players: 1.) Miracles 2.) Miracles 3.) Elves 4.) Lands 5.) Aggro Loam 6.) Goblins 7.) Goblins 8.) Lands
Early one morning while making the round,
I took a shot of cocaine and I shot my woman down;
I went right home and I went to bed,
I stuck that lovin' .44 beneath my head.
There are currently 1 users browsing this thread. (0 members and 1 guests)