deckstats.net

Author Topic: Suggestions for Improving Deckstats?  (Read 2279 times)

Nils

  • Administrator
  • Hero Member
  • *****
  • Posts: 901
  • Karma: 229
  • <admin>
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #30 on: August 05, 2019, 12:20:22 am »
Thank you for pointing this out. To be honest, I am not entirely sure how we should handle these types of cards. We could count all of them as "any" but then that is also not entirely correct. However, I think it's probably the simplest if I do that and then rely on people to use override <source=X> option if they think this is not appropriate.

crimsonking

  • Sr. Member
  • ****
  • Posts: 465
  • Karma: 112
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #31 on: August 05, 2019, 08:53:45 am »
I think if Exotic Orchard counts as an "any-color" land, Reflecting Pool should count as well. One would expect that.

Nils

  • Administrator
  • Hero Member
  • *****
  • Posts: 901
  • Karma: 229
  • <admin>
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #32 on: August 11, 2019, 04:43:15 pm »
I have just made a whole bunch of changes to our "mana source detector"! It should now handle the edge cases you mentioned properly and also recognize fetch lands. Please give it a try and let me know what you think.

crimsonking

  • Sr. Member
  • ****
  • Posts: 465
  • Karma: 112
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #33 on: August 12, 2019, 11:14:49 am »
So, here are my considerations:
  • Murmuring Bosk is considered as monogreen.
  • Crucible of the Spirit Dragon and Cascading Cataracts are considered as colorless lands.
  • Rhystic Cave and Meteor Crater are considered as non-mana-producing lands.
  • Lands that don't produce mana are listed as "unknown". You can put as many as you want, the statistics will only count 1 unknown source.
  • If you put for instance a Swamp and a Maze of Ith, the statistics will count 1 black mana source, but if you swap the Swamp for a Wastes it will count 1 colorless and 1 unknown. Of course, if you add Vesuva too, the unknown mana source will still be 1. I think this "unknown" category shouldn't exist at all.
  • Urza's Mine, Urza's Power Plant and Urza's Tower are considered to produce 3, 3 and 4 colorless mana respectively. Although I understand how the algorithm gets to this result, I think it is totally misleading.
  • Crystal Vein too is considered to produce 3 mana. Compare it with Ancient Tomb and you'll see how this is wrong.
  • Cards like Ebon Stronghold are considered to produce 3 mana. People tend to overlook colorless mana, but say you put an Ebon Stronghold and a Mountain, the statistics will say 75% of your mana base is black, which is clearly false.
  • Cards like Twilight Mire count as 7 mana: 3 black, 3 green and 1 colorless, although we all know it nets you only 1 mana.
  • Lake of the Dead is considered to produce a wooping 5 black mana! Although I wouldn't consider an error to list it as a 4-mana land, I think we could assume multiple mana come with drawbacks (that statistics are ignoring completely), so it would be more accurate to count lands at most once per color no matter what.
  • If you put 1 Crumbling Necropolis and 1 City of Brass, the statistics will say you have 25% of blue, black, red and any-color each. I think the correct result should rather be 50% any-color and 16.67% of blue, black and red.
  • According to the current system, Castle Sengir is better than Crumbling Necropolis, since it produce colorless mana too. Again, this is not strictly an error but it's weird to say the least.
« Last Edit: August 12, 2019, 08:35:40 pm by crimsonking »

crimsonking

  • Sr. Member
  • ****
  • Posts: 465
  • Karma: 112
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #34 on: August 12, 2019, 11:55:02 am »
As for the fetchlands, I actually think they should not count for mana statistics, at least not automatically.
I mean, I think the <source> tag is enough to allow the users to customize their mana-base evaluations.
Let me explain:
With the new algorithm, Verdant Catacombs counts as Golgari while Prismatic Vista counts as an any-color land.
At first look this could be acceptable, but in fact it's false.
The problem is the fetchlands can grab duals while Prismatic Vista only gets basic lands.
For instance, everybody knows that in a tri-color deck, any in-color fetchland can get any dual, so they effectively count as any-color.
But if we're playing 5 colors things get more complicated.
Take this deck as an example:
https://deckstats.net/decks/54477/1203990-domain-cascade
Here Verdant Catacombs can grab any dual, so it could count as an any-color.
The same is true for the other fetchlands aside from Scalding Tarn. In fact, Scalding Tarn can't get white mana.
No need to say here Prismatic Vista would be terrible since it could only grab a Forest or a Swamp.
Fetchlands are very dependant on the decklist structure, that's why I'd let them out of automatic calculations.
« Last Edit: August 12, 2019, 08:38:00 pm by crimsonking »

crimsonking

  • Sr. Member
  • ****
  • Posts: 465
  • Karma: 112
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #35 on: August 12, 2019, 12:45:36 pm »
Hi again,
Now that I was thinking of it, I'd have another suggestion for a quick improvement.
In the "Revisions" tab of the deck page, please put the "Compare Revisions" button on top.
I have many decks with 50 or more revisions and each time I want to see the changes I've made, I have to scroll down for a while only to reach the button.
After all, most of times you're interested in the last changes (so everything is already set up, you just need to click the button) or in the most recent ones (which are visible on top of the page anyway), so the scrolling down is unnecessary.
Moreover, why not just compare the last two last revisions straight away and leave the option to change them later (for instance through an expandable box)?
« Last Edit: August 12, 2019, 12:48:52 pm by crimsonking »

Nils

  • Administrator
  • Hero Member
  • *****
  • Posts: 901
  • Karma: 229
  • <admin>
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #36 on: August 12, 2019, 11:37:34 pm »
Thank you very much for your detailed analysis crimsonking! I have gone through and fixed some of the more obvious bugs you found, such as Urza's Mine counting as three mana (it's now counted as two), Murmuring Bosk only being counted as green, and mana being wrongly added together for cards like Ebon Stronghold and Crystal Vein.

However, there are still some more edge cases that will require some additional coding/thinking.

I actually used to calculate the mean numbers for each land, accounting for the different mana each land could produce. So for example Ebon Stronghold would be 1/2 + 2/2 = 1.5 black and Twilight Mire would be 1/7 colorless, 3/7 black and 3/7 green. However I ended up changing that because it also seemed to lead to even more confusion.

I do agree that the way we are counting things right now is also not 100% right though. Will need to look into this some more.

Nils

  • Administrator
  • Hero Member
  • *****
  • Posts: 901
  • Karma: 229
  • <admin>
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #37 on: August 12, 2019, 11:38:22 pm »
PS: I have just moved up the compare revisions button, good suggestion!

crimsonking

  • Sr. Member
  • ****
  • Posts: 465
  • Karma: 112
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #38 on: August 13, 2019, 02:06:48 am »
In my opinion, the algorithm should be more or less like this:
  • Each mana-producing land counts as 1.
    (This could be considered inaccurate for lands like Ancient Tomb, City of Traitors and Mishra's Workshop, but in the vast majority of cases I think it's a good approximation.
    Bounce lands too produce 2 mana, but this is compensated by the fact that they put you back one land drop.)
  • Non-mana-producing lands are ignored.
    (No more unknown mana sources in the statistics.)
  • Lands that say: "any color" or "any type" are considered any-color lands.
    (Any-color is a color, exactly as black, blue etc.).
  • Colorless is a color too.
  • If a land has n mana abilities, each of them will participate for 1/n of the total for that land.
    (For instance: Nimbus Maze will count for 1/3 as colorless, for 1/3 as white and for 1/3 as blue.)
  • If a mana ability says: "add {A} or {B}", it counts as two separated abilities: "add {A}" and "add {B}".
    (For instance: Underground River will count for 1/3 as colorless, 1/3 as blue and 1/3 as black.)
Some more examples to better explain my idea:
  • Murmuring Bosk will count for 1/3 as green, for 1/3 as white and for 1/3 as black.
  • Golgari Rot Farm will count for 1/2 as black for 1/2 as green (not ideal, but still acceptable).
  • Everglades will count for 1/2 as black for 1/2 as colorless (same as above).
  • Bottomless Vault will count as 1 black (namely, every storage land will count as 1 mana).
  • Moreover, Crucible of the Spirit Dragon will count for 1/2 as colorless and for 1/2 as any color.
  • Cascading Cataracts will count for 1/2 as colorless and for 1/2 as WUBRG (not 1 colorless and 5 of the rest).
  • Urza's Mine, Urza's Power Plant and Urza's Tower will count as 1 colorless (1/2 colorless plus 1/2 colorless; after all, they produce 1 mana most of the times).
  • Phyrexian Tower will count for 1/2 as colorless and for 1/2 as black (it could be 1/3 and 2/3 but I don't think it's that relevant).
  • Sulfur Vent will count for 1/2 as black and for 1/2 as blue-red, i.e. 1/4 as blue and 1/4 as red.
  • If I put 1 Ebon Stronghold and 1 Mountain, the result should be 50% black and 50% red.
  • Twilight Mire will count for 1/4 as colorless and for 3/4 as black, black-green and green, i.e. for 1/4 as colorless, for 3/8 as black and for 3/8 as green (actually, I think it should be 1/3, 1/3 and 1/3 but that's tricky to do).
  • Crumbling Necropolis will count as blue, black and red for 1/3 each, while Castle Sengir will count as colorless, blue, black and red for 1/4 each.
  • City of Brass will count as 1 any color, while Spire of Industry will count for 1/2 as any color and for 1/2 as colorless.
  • If I put 1 Crumbling Necropolis and 1 City of Brass, the result should be 50% any color, 16.7% blue, 16.7%black and , 16.7% red.
« Last Edit: August 13, 2019, 09:50:49 am by crimsonking »

Rikus

  • New Member
  • *
  • Posts: 1
  • Karma: 0
    • View Profile
  • Decks
Re: Suggestions for Improving Deckstats?
« Reply #39 on: August 16, 2019, 09:47:12 pm »
I was trying to filter by multiple sets, but perhaps I'm missing something.  Would it be possible to add multi-set filtering, and if not, could a set be created that included rotation safe sets?  Looks like there's a ton of set options to select, one called 2019 Rotation Safe Set, which include all cards from last four sets would be very nice.  And then when Throne of Eldraine is released a new 2020 Rotation set could be started, and updated as next three sets are released.  Or if not added as a set, perhaps as a format Standard-RS (Rotation Safe/Proof)?

Edit:  NVM, I found it.  Didn't notice a new box appeared after selecting (Enter Set Codes) form the set drop down list.
« Last Edit: August 16, 2019, 09:53:23 pm by Rikus »