Everyone knows that good luck and good game are such insincere terms that any man who does not connect his right hook with the offender's jaw on the very utterance of such a phrase is no man I would consider as such.
Everyone knows that good luck and good game are such insincere terms that any man who does not connect his right hook with the offender's jaw on the very utterance of such a phrase is no man I would consider as such.
Under RUg Storm Pyro, your "Most unusual decklist" is a stock Hive Mind deck.
It makes me wonder what other archetypes you missed.
Well! I'll think how to deal with it.
I would change up the colors. I can't tell if Pod is 30% of the meta, or if Other is.
I've never used google charts before, and those charts were made last night in 30 minutes So I'll surely make normal coloring, but don't know how fast - maybe in a day or two...
What criteria does it use to define each Archtype? And how does it handle overlap?
Hm... Initially it presumes there is no any archetypes. Then for each given deck check, if it seems "like a deck from the archetype" for each archetype. If it not, create new archetype.
After all archetypes created, go once more and make statistics on using any colors and some "corner" cards - and "humanify" names from "Archetype 105" to "RU Storm". If there is overlap, or a deck belongs to several arhcetypes there could be many things... New archetype, or, as in case with "Hive Mind", erroneous archetype. I'm working on the algorythm, so if you see any errors in deck archetypes - please write here
Could we get an additional pie chart showing only 1st place finishes by archetype, and then a bar chart showing first place finishes as a percentage of representation for the top 10 decks or so?
For example, Jund was represented 137 times, and has 24 1st place finishes. (17.5%)
Affinity, however, was represented 105 times, and has 34 1st place finishes. (33%)
Could we get an additional pie chart showing only 1st place finishes by archetype, and then a bar chart showing first place finishes as a percentage of representation for the top 10 decks or so?
For example, Jund was represented 137 times, and has 24 1st place finishes. (17.5%)
Affinity, however, was represented 105 times, and has 34 1st place finishes. (33%)
And, should it be sorted by percentage of 4-0's, or by total number of 4-0's?
1). Another pie chart showing raw number of 4-0's by archetype
2). A bar graph showing the percentage of 4-0's (compared to representation) for the top 10 most played archetypes.
Very cool. Archetype detection is definitely a tough nut to crack though.
Possible approaches to mull over:
a) Crowd sourcing - Give some interface where people can identify what archetype a deck is.
b) Headliners - A translation list from single card to what deck. These take some work to identify, but with the right choices, it should filter down a lot of the decks to the right places. This works best for identifying combo decks as they usually are using pieces which don't overlap elsewhere.
Maybe with some percentage to tell how likely of a tell having that card is? The combo cards can be 100% in most cases. So you check for those first, then descend in probability until you get a hit.
Example:
Hive Mind => Hive Mind Combo 100%
Living End => Living End Combo 100%
Ad Nauseum => Ad Nauseum Combo 100%
Splinter Twin => Splinter Twin Combo 100%
Pyromancer's Swath => Storm Combo 100%
Bloodbraid Elf => Jund 50%
Bloodbraid Elf => Zoo 50%
Loam Lion => Zoo 30%
Loam Lion => GWB Junk 30%
Blightning => Jund 30%
Blightning => Grixis Delver
So we take a decklist, find out which card in the deck has the highest rating, and see which decks that card is in. If its only one, we can stop and we know our archetype. If there are multiple decks, we find the next highest ranked card and see which decks overlap. We continue that until we only have one possible archetype.
Take a stock jund list with the above definitions for example:
Bloodbraid elf is the highest ranked card and so the possible archetypes are zoo and jund.
Its next highest card by that list would be Blightning, which is played in Jund and Grixis Delver. Jund is in common on both lists, so we know its a jund deck.
It is a really complicated system, but it should be very accurate once you work out the rules.
Private Mod Note
():
Rollback Post to RevisionRollBack
I support WotC's goal of shaping Modern in favor of diversity.
I ran a thought experiment on my blog Modern in a Nuclear Wasteland
of an extreme case of banning 20 more cards to make sure they get everything, then scaling back where appropriate. WotC seems to be on a slowly build up approach. Both ways probably reach similar end points.
The post Gatecrash metagame is proving to be closer to the endpoint than I estimated, so its very possible that few (if any) more cards need to be banned.
The really cool thing in current version is that I really dont need anything to write about archetype names, except some rare names ("zoo", "jund"...). My algorythm is something like you offered, but more complicated (closer to regression methods). Cool idea about crowd sourcing... but I've never written any web services before, so it'll take a long time to create now I'm trying to rewrite my C# code on python and let it all work on a shellmix.com server... kinda tough
p.s. I'm not outputting last info because of fails in some archetypes. Approximate deviation is got, the precise answers will be after crowd help will be here
Its MTGO results. He's got one of these up for block. Of course block archetypes are a lot easier to determine.
Private Mod Note
():
Rollback Post to RevisionRollBack
Out of the blackness and stench of the engulfing swamp emerged a shimmering figure. Only the splattered armor and ichor-stained sword hinted at the unfathomable evil the knight had just laid waste.
To post a comment, please login or register a new account.
Notice that all names and colors are evalueted automatically, so don't be scared by "Human Wizard" decks
p.s. ANY critics and suggestions are applicatable
- To my youngest sister when she was 6.
- To my youngest sister when she was 6.
It makes me wonder what other archetypes you missed.
Well! I'll think how to deal with it.
I've never used google charts before, and those charts were made last night in 30 minutes So I'll surely make normal coloring, but don't know how fast - maybe in a day or two...
Hm... Initially it presumes there is no any archetypes. Then for each given deck check, if it seems "like a deck from the archetype" for each archetype. If it not, create new archetype.
After all archetypes created, go once more and make statistics on using any colors and some "corner" cards - and "humanify" names from "Archetype 105" to "RU Storm". If there is overlap, or a deck belongs to several arhcetypes there could be many things... New archetype, or, as in case with "Hive Mind", erroneous archetype. I'm working on the algorythm, so if you see any errors in deck archetypes - please write here
For example, Jund was represented 137 times, and has 24 1st place finishes. (17.5%)
Affinity, however, was represented 105 times, and has 34 1st place finishes. (33%)
If your answer is "No," then your morality does not come from God's commandments.
If your answer is "Yes," then please, please reconsider.
And, should it be sorted by percentage of 4-0's, or by total number of 4-0's?
p.s. I'll add it tomorrow...
1). Another pie chart showing raw number of 4-0's by archetype
2). A bar graph showing the percentage of 4-0's (compared to representation) for the top 10 most played archetypes.
If your answer is "No," then your morality does not come from God's commandments.
If your answer is "Yes," then please, please reconsider.
Possible approaches to mull over:
a) Crowd sourcing - Give some interface where people can identify what archetype a deck is.
b) Headliners - A translation list from single card to what deck. These take some work to identify, but with the right choices, it should filter down a lot of the decks to the right places. This works best for identifying combo decks as they usually are using pieces which don't overlap elsewhere.
Maybe with some percentage to tell how likely of a tell having that card is? The combo cards can be 100% in most cases. So you check for those first, then descend in probability until you get a hit.
Example:
Hive Mind => Hive Mind Combo 100%
Living End => Living End Combo 100%
Ad Nauseum => Ad Nauseum Combo 100%
Splinter Twin => Splinter Twin Combo 100%
Pyromancer's Swath => Storm Combo 100%
Bloodbraid Elf => Jund 50%
Bloodbraid Elf => Zoo 50%
Loam Lion => Zoo 30%
Loam Lion => GWB Junk 30%
Blightning => Jund 30%
Blightning => Grixis Delver
So we take a decklist, find out which card in the deck has the highest rating, and see which decks that card is in. If its only one, we can stop and we know our archetype. If there are multiple decks, we find the next highest ranked card and see which decks overlap. We continue that until we only have one possible archetype.
Take a stock jund list with the above definitions for example:
Bloodbraid elf is the highest ranked card and so the possible archetypes are zoo and jund.
Its next highest card by that list would be Blightning, which is played in Jund and Grixis Delver. Jund is in common on both lists, so we know its a jund deck.
It is a really complicated system, but it should be very accurate once you work out the rules.
I ran a thought experiment on my blog
Modern in a Nuclear Wasteland
of an extreme case of banning 20 more cards to make sure they get everything, then scaling back where appropriate. WotC seems to be on a slowly build up approach. Both ways probably reach similar end points.
The post Gatecrash metagame is proving to be closer to the endpoint than I estimated, so its very possible that few (if any) more cards need to be banned.
p.s. I'm not outputting last info because of fails in some archetypes. Approximate deviation is got, the precise answers will be after crowd help will be here
It's got from http://www.wizards.com/Magic/Digital/MagicOnline.aspx?x=mtg/digital/magiconline/whatshappening