You will be giving people what they want if you put work into a good tiering system; I don't really understand your aversion to it.
I have such an aversion to tiers that I spent a lot of my free time creating a big spreadsheet full of them...
There's no need to be sarcastic. I was just replying to your direct quote in which you said you didn't care much about creating a tiering system.
I'll make you a deal - stop reinterpreting what I say, and I'll stop being sarcastic. My words can stand on their own. And I already have a tiering system, in fact a very good one IMHO.
None of the deck databases online seem to have the paper PTQs. I don't think they are available, f anyone knows where to find them, please let me know.
You will be giving people what they want if you put work into a good tiering system; I don't really understand your aversion to it.
I have such an aversion to tiers that I spent a lot of my free time creating a big spreadsheet full of them...
There's no need to be sarcastic. I was just replying to your direct quote in which you said you didn't care much about creating a tiering system.
I'll make you a deal - stop reinterpreting what I say, and I'll stop being sarcastic. My words can stand on their own. And I already have a tiering system, in fact a very good one IMHO.
Just for reference, what is your current tiering system and how are tiers determined? I'm really not trying to be rude here and I don't want you to feel so defensive whenever I comment on that system. As someone who did this for years, implemented it on this site as a mod, did all the Nexus updates, and still cares about Modern tiering, I am contributing to help ensure the community gets the best possible product.
EDIT: To be clear, I know the tier % cutoffs from the sheet. I'm wondering what system resulted in their selection or calculation. For instance, why 4% and not 4.25% or 3.75%? How does that % change from month to month and what factors change it?
Archetype Definitions
"Death and Taxes" - The mono-white deck only, no B/W or G/W variants.
"B/W Eldrazi" - A distinct Eldrazi variant, sometimes called "Eldrazi and Taxes". Some similarity to the mono-white Death and Taxes but different enough to be separated.
Are you lumping the B/W Eldrazi version of Death and Taxes together with the BW Eldrazi midrange deck? I'm not sure what the consensus on that is, but I know from a player perspective Eldrazi and Taxes is a solid Death and Taxes variant, while the BW Eldrazi midrange deck is a separate archetype.
Archetype Definitions
"Death and Taxes" - The mono-white deck only, no B/W or G/W variants.
"B/W Eldrazi" - A distinct Eldrazi variant, sometimes called "Eldrazi and Taxes". Some similarity to the mono-white Death and Taxes but different enough to be separated.
Are you lumping the B/W Eldrazi version of Death and Taxes together with the BW Eldrazi midrange deck? I'm not sure what the consensus on that is, but I know from a player perspective Eldrazi and Taxes is a solid Death and Taxes variant, while the BW Eldrazi midrange deck is a separate archetype.
This guy's spot on. BW Eldrazi is a grindy midrange deck, while Eldrazi and Taxes is... well, a DnT deck. One of the key differences is that the midrange BW Eldrazi runs Lingering Souls mainboarded and no Vials.
Private Mod Note
():
Rollback Post to RevisionRollBack
Well, I can saw a woman in two, but you won't wanna look in the box when I'm through.
I'll take a look next time I add new data, probably Monday. I'm not that familiar with that deck, but I probably am lumping them together if it's B/W, and Death and Taxes is just mono-white.
Separating archetypes can be tricky business, and it's manual work. The naming is all over the map.
I tend to err toward merging decks because if you split every deck on every variant, then the decks share gets underrepresented. There's a fine line sometimes, and there isn't a perfect solution.
You see this most commonly with G/W, G/W Company, GW, GW Company, Bant, Bant Company, Counters Company, Counter Company, Knightfall, Bant Knightfall, G/W Hate Bears, GW Hatebears, etc etc ...
I'll start applying more scrutiny to the B/W decks.
You see this most commonly with G/W, G/W Company, GW, GW Company, Bant, Bant Company, Counters Company, Counter Company, Knightfall, Bant Knightfall, G/W Hate Bears, GW Hatebears, etc etc ...
This work is more detailed than it might seem at first, and the work is tedious to parse which Knightfall list has the Devoted Druid combo in it and which ones don't. Thanks for the effort!
Private Mod Note
():
Rollback Post to RevisionRollBack
Modern Decks
KnightfallGWUR
Azorius Control UW
Burn RBG
There are 17 (!) creatures in the E&T list that are not in the D&T list. To play Devil's Advocate to this argument, the decks are employing more or less the same strategy, they are just using different creatures to go about it, and the main deck still has about ~17 cards in common (Vials, Paths, Arbiters, Wisps, Splicers).
As much as I hate further fragmenting meta share, I'm leaning toward option 1 and splitting into three lists. E&T does feel like a different deck. BUT the problem that could develop, is that you can see decks that fit somewhere in the middle of E&T and D&T, or are further hybridizations between E&T and B/W Eldrazi.
Any thoughts are welcome. I'm not doing anything before Monday, so anyone who wants to chime in with their opinion before then would be helpful.
I like options 1 and 2. I see what you mean about E&T fitting between the two archetypes. It's implementing both a D&T strategy AND an Eldrazi strategy. I can tell you from personal experience playing the deck that it still feels like a D&T deck when playing it, but I appreciate why sites like mtggoldfish lean toward option 1. There is value to be had in splitting it off if it behaves differently from the rest of D&T on a statistical analysis level.
I like options 1 and 2. I see what you mean about E&T fitting between the two archetypes. It's implementing both a D&T strategy AND an Eldrazi strategy. I can tell you from personal experience playing the deck that it still feels like a D&T deck when playing it, but I appreciate why sites like mtggoldfish lean toward option 1. There is value to be had in splitting it off if it behaves differently from the rest of D&T on a statistical analysis level.
It looks to me like mtggoldfish doesn't touch the name and just takes it as provided by their data source. As I was alluding to earlier, you can see Knightfall lists as G/W Company, Naya Company, Bant Knightfall, etc. It's hard to glean the meta share of Knightfall from the mtggoldfish list.
Also, look at some of the names that show up there for E&T:
Eldrazi & Taxes
B/W Eldrazi Taxes
W/B Eldrazi Taxes
etc
Obviously these should be considered the same deck with one consistent name for meta analysis purposes. Right now I'm doing this manually, but eventually it would be cool to do it with machine learning.
You could probably do it programmatically! At least for this case. Any deck playing thalia, arbiter, vial, wisp, and eldrazi temple is E&T. That will also pick up the rare mono W E&T list.
You could probably do it programmatically! At least for this case. Any deck playing thalia, arbiter, vial, wisp, and eldrazi temple is E&T. That will also pick up the rare mono W E&T list.
Sure that's possible, but it would be more work, and I'm not quite ready to take on pulling in deck lists. Now I'm just getting results lists. The sources I'm getting data from pass HTML over the wire, not formatted data like JSON, so custom web page scraping/parsing code would be needed. I'm not there yet, but I've really just gotten started on it... I'll incrementally improve it for as long as I remain interested in the project. The long term goal would be to have something that's 100% automated.
Gotcha. Is there any consideration for open sourcing the project? If you did we could help build it out as a community
Yeah, I may eventually do that. The code isn't bad, but honestly it isn't in a state where I'd want anyone else to look it right now. If you or anyone else are seriously interested in collaborating, maybe take it to PM and we can talk.
The direction it needs to go next is automating generation of the tournament results list, which is currently ingested by my code as csv.
That part could be written independent of anything else. It's not trivial, though. I've done website scraping before, and not only is it messy and error-prone, it's fragile because web pages change over time and when they do, the parser breaks. That's a messy enough problem that I'm not super interested in taking it on considering that I'm not getting paid to do it.
It would be great if there was direct access to data without having to go through a web server and parsing HTML. If anyone knows of such a thing, please let me know. I'd gladly do lots of cool stuff if I could get structured data as JSON, csv, etc. Access to structured data is the barrier now.
Gotcha. Is there any consideration for open sourcing the project? If you did we could help build it out as a community
Yeah, I may eventually do that. The code isn't bad, but honestly it isn't in a state where I'd want anyone else to look it right now. If you or anyone else are seriously interested in collaborating, maybe take it to PM and we can talk.
The direction it needs to go next is automating generation of the tournament results list, which is currently ingested by my code as csv.
That part could be written independent of anything else. It's not trivial, though. I've done website scraping before, and not only is it messy and error-prone, it's fragile because web pages change over time and when they do, the parser breaks. That's a messy enough problem that I'm not super interested in taking it on considering that I'm not getting paid to do it.
It would be great if there was direct access to data without having to go through a web server and parsing HTML. If anyone knows of such a thing, please let me know. I'd gladly do lots of cool stuff if I could get structured data as JSON, csv, etc. Access to structured data is the barrier now.
Sadly, when I did the metagame updates and articles, there was also no better way to gather information than manually. What sites are you using for your dataset? I'm happy to share more sources if you need additional paper tournaments in the sample.
Sadly, when I did the metagame updates and articles, there was also no better way to gather information than manually. What sites are you using for your dataset? I'm happy to share more sources if you need additional paper tournaments in the sample.
I'm using the SCG site for all their stuff, and mtggoldfish for the GPs and MTGO. And that part is manual - copypasta into spreadsheet, do some cleanup and archetype assignment, export as csv (tab-delimited, actually). It doesn't take long though, then my code does all the real work; I just have to set some date ranges in a config for the different analyses.
I don't currently have a source for paper PTQs if you know of one.
Do you all think B/G should be considered a separate archetype? I had been considering "Jund" to be basically "Non-Abzan G/Bx Midrange". I could just change Jund to G/Bx Midrange, or I could separate them into Jund and G/B Midrange.
I definitely feel like Abzan is its own thing because white is rather important in the build, whereas with Jund red is usually a minor splash.
And all the above are distinct from the Shadow variants, which will stay separate since that's how folks seem to want to reckon them.
Do you all think B/G should be considered a separate archetype? I had been considering "Jund" to be basically "Non-Abzan G/Bx Midrange". I could just change Jund to G/Bx Midrange, or I could separate them into Jund and G/B Midrange.
I definitely feel like Abzan is its own thing because white is rather important in the build, whereas with Jund red is usually a minor splash.
And all the above are distinct from the Shadow variants, which will stay separate since that's how folks seem to want to reckon them.
I'd separate them. It's important to know whether you are more likely to see Kommand vs. Souls, Path vs. Bolt, etc. It's also important to know which is better positioned.
Anyone following this probably noticed, but I put the project on hold. I'm going to make some changes to how I was splitting archetypes, and I also have some other changes I want to do in conjunction with that to make it easier to see what's going on in the meta. I'm busy right now with a work deadline coming up, but probably next week I'll have time to look at this again.
Private Mod Note
():
Rollback Post to RevisionRollBack
To post a comment, please login or register a new account.
I have such an aversion to tiers that I spent a lot of my free time creating a big spreadsheet full of them...
There's no need to be sarcastic. I was just replying to your direct quote in which you said you didn't care much about creating a tiering system.
I'll make you a deal - stop reinterpreting what I say, and I'll stop being sarcastic. My words can stand on their own. And I already have a tiering system, in fact a very good one IMHO.
https://docs.google.com/spreadsheets/d/1ZRLhtRToCI2VkJ-xYVwwvYuYtFmCLyMyRH1pbjCdBg8
None of the deck databases online seem to have the paper PTQs. I don't think they are available, f anyone knows where to find them, please let me know.
Just for reference, what is your current tiering system and how are tiers determined? I'm really not trying to be rude here and I don't want you to feel so defensive whenever I comment on that system. As someone who did this for years, implemented it on this site as a mod, did all the Nexus updates, and still cares about Modern tiering, I am contributing to help ensure the community gets the best possible product.
EDIT: To be clear, I know the tier % cutoffs from the sheet. I'm wondering what system resulted in their selection or calculation. For instance, why 4% and not 4.25% or 3.75%? How does that % change from month to month and what factors change it?
Are you lumping the B/W Eldrazi version of Death and Taxes together with the BW Eldrazi midrange deck? I'm not sure what the consensus on that is, but I know from a player perspective Eldrazi and Taxes is a solid Death and Taxes variant, while the BW Eldrazi midrange deck is a separate archetype.
This guy's spot on. BW Eldrazi is a grindy midrange deck, while Eldrazi and Taxes is... well, a DnT deck. One of the key differences is that the midrange BW Eldrazi runs Lingering Souls mainboarded and no Vials.
Separating archetypes can be tricky business, and it's manual work. The naming is all over the map.
I tend to err toward merging decks because if you split every deck on every variant, then the decks share gets underrepresented. There's a fine line sometimes, and there isn't a perfect solution.
You see this most commonly with G/W, G/W Company, GW, GW Company, Bant, Bant Company, Counters Company, Counter Company, Knightfall, Bant Knightfall, G/W Hate Bears, GW Hatebears, etc etc ...
I'll start applying more scrutiny to the B/W decks.
KnightfallGWUR
Azorius Control UW
Burn RBG
Thanks. I was just looking at decklists. E&T kind of splits the difference between D&T and B/W Eldrazi, but it does lean more toward D&T.
There are three options:
1. Separate into 3 decks:
Death and Taxes
Eldrazi and Taxes
B/W Eldrazi
2. Merge Taxes decks:
Death and Taxes (includes Eldrazi and Taxes)
B/W Eldrazi
3. Merge Eldrazi decks:
Death and Taxes
B/W Eldrazi (includes Eldrazi and Taxes)
So yeah, #3 is what I did, which in retrospect is probably my least preferred of the three options.
Looking at lists, I think a strong argument could be made for #1.
Check out this E&T list from Syracuse Open (14th place):
http://sales.starcitygames.com//deckdatabase/displaydeck.php?DeckID=115360
And the winning D&T list from Roanoke:
http://sales.starcitygames.com//deckdatabase/displaydeck.php?DeckID=114531
There are 17 (!) creatures in the E&T list that are not in the D&T list. To play Devil's Advocate to this argument, the decks are employing more or less the same strategy, they are just using different creatures to go about it, and the main deck still has about ~17 cards in common (Vials, Paths, Arbiters, Wisps, Splicers).
As much as I hate further fragmenting meta share, I'm leaning toward option 1 and splitting into three lists. E&T does feel like a different deck. BUT the problem that could develop, is that you can see decks that fit somewhere in the middle of E&T and D&T, or are further hybridizations between E&T and B/W Eldrazi.
Any thoughts are welcome. I'm not doing anything before Monday, so anyone who wants to chime in with their opinion before then would be helpful.
-Adam
It looks to me like mtggoldfish doesn't touch the name and just takes it as provided by their data source. As I was alluding to earlier, you can see Knightfall lists as G/W Company, Naya Company, Bant Knightfall, etc. It's hard to glean the meta share of Knightfall from the mtggoldfish list.
Also, look at some of the names that show up there for E&T:
Eldrazi & Taxes
B/W Eldrazi Taxes
W/B Eldrazi Taxes
etc
Obviously these should be considered the same deck with one consistent name for meta analysis purposes. Right now I'm doing this manually, but eventually it would be cool to do it with machine learning.
Sure that's possible, but it would be more work, and I'm not quite ready to take on pulling in deck lists. Now I'm just getting results lists. The sources I'm getting data from pass HTML over the wire, not formatted data like JSON, so custom web page scraping/parsing code would be needed. I'm not there yet, but I've really just gotten started on it... I'll incrementally improve it for as long as I remain interested in the project. The long term goal would be to have something that's 100% automated.
Yeah, I may eventually do that. The code isn't bad, but honestly it isn't in a state where I'd want anyone else to look it right now. If you or anyone else are seriously interested in collaborating, maybe take it to PM and we can talk.
The direction it needs to go next is automating generation of the tournament results list, which is currently ingested by my code as csv.
That part could be written independent of anything else. It's not trivial, though. I've done website scraping before, and not only is it messy and error-prone, it's fragile because web pages change over time and when they do, the parser breaks. That's a messy enough problem that I'm not super interested in taking it on considering that I'm not getting paid to do it.
It would be great if there was direct access to data without having to go through a web server and parsing HTML. If anyone knows of such a thing, please let me know. I'd gladly do lots of cool stuff if I could get structured data as JSON, csv, etc. Access to structured data is the barrier now.
Sadly, when I did the metagame updates and articles, there was also no better way to gather information than manually. What sites are you using for your dataset? I'm happy to share more sources if you need additional paper tournaments in the sample.
I'm using the SCG site for all their stuff, and mtggoldfish for the GPs and MTGO. And that part is manual - copypasta into spreadsheet, do some cleanup and archetype assignment, export as csv (tab-delimited, actually). It doesn't take long though, then my code does all the real work; I just have to set some date ranges in a config for the different analyses.
I don't currently have a source for paper PTQs if you know of one.
I definitely feel like Abzan is its own thing because white is rather important in the build, whereas with Jund red is usually a minor splash.
And all the above are distinct from the Shadow variants, which will stay separate since that's how folks seem to want to reckon them.
I'd separate them. It's important to know whether you are more likely to see Kommand vs. Souls, Path vs. Bolt, etc. It's also important to know which is better positioned.