AI card base value
Moderators: BAgate, drool66, Aswan jaguar, gmzombie, stassy, CCGHQ Admins
AI card base value
by stassy » 18 Nov 2012, 05:55
Current status of cards ranking by expansion
---
- | Open
Please do note that the objective of this ranking is first to make draft mode playable as fast as possible, so I will only fill cards with no AI card base value or with AI card base value of 0, there are many cases where this value seems wrong (like Darksteel Colossus set to 1) but please bear with it until the whole cards are covered, then we will starting to rank cards better.Rank 5 Filler cards, vanilla creatures with no abilities
Rank 9 Bugged currently unfixable card reduced as vanilla or unusable by AI
Rank 10 Lands and mana fixers
Rank 20 Average Creatures, with drawback or in need of heavy support to be efficient, color/type killer, heavy mana cost, situational efficiency
Rank 40 Medium quality Removal, temporary staller, land searcher, small damager
Rank 50 Above Average Creatures with at least one benefit and no drawback, cheap mana cost and giving support
Rank 60 Strong Spells, global range, multipes uses, token generator, creature retriever, counterspells, support planeswalker
Rank 70 Very strong Creatures but still in need of support or not game winner, global advantage/support
Rank 80 High Quality removal, no drawback, planeswalker with removal abilities
Rank 100 Bombs that potentially win the game on their own
I choose rank 9 because there is no 9 in others rank so I can search faster for them
---
How do we have to tell a card is "good"?Gargaroz wrote:About the AI's splitting, it's all linked to a field in Manalink.csv that tells the AI "how good" is a card. Since is leaved to "0" for a long list to card (it has to be inserted manually and it's a really boring work to do, if someone will volounteer, it wouls be great). I'll try to solve this though coding, but it will be still clumsy.
- stassy
- Moderator
- Posts: 5274
- Joined: 25 Feb 2009, 07:06
- Has thanked: 471 times
- Been thanked: 337 times
Re: Fact or Fiction
by Gargaroz » 18 Nov 2012, 15:46
It's completely up to you. We all know that Shivan Dragon is better than Brushwagg, but for other cards, it's really a matter of tastes.
----
- Current / medium term task: adjusting the code for making Misdirection and such usable
- Long term task: inserting all the good stuff I left out from the "Golden Years" mod
- Current / medium term task: adjusting the code for making Misdirection and such usable
- Long term task: inserting all the good stuff I left out from the "Golden Years" mod
- Gargaroz
- Programmer
- Posts: 7097
- Joined: 06 Nov 2009, 11:11
- Has thanked: 82 times
- Been thanked: 595 times
Re: Fact or Fiction
by Aswan jaguar » 18 Nov 2012, 16:09
What about the rating of users on Wizzards of the coast.1\2 to 5 stars?Although they have 4 stars for Savannah Lions. 
so forget about it.
Stassy this is guide by Jatill with values original cards had as example he wrote that on the wiki for coding cards:
AI Base Value determines how high the AI picks the card in limited, among other things. Use the following scale:
100-200 Bombs that potentially win the game on their own (Loxodon Warhammer, Meloku, Fireball)
80-99 High Quality removal (Terror, Swords to Plowshares)
70-79 Very strong Creatures (Air Elemental, Serra Angel)
60-69 Strong Spells (Tidings)
50-59 Above Average Creatures (Durkwood Boars, Prodigal Sorcerer)
40-49 Medium quality Removal (ex: Immolation)
20-39 Average Creatures
10-19 Lands and mana fixers
1-9 Filler cards (ex: Grizzly Bears)
I want to add that no original cards or just one used other numbers than *5.
5,10,15,20,etc I don't know if that way are better understood-recognised by AI or was easier for Microprose Developers.
so forget about it.Stassy this is guide by Jatill with values original cards had as example he wrote that on the wiki for coding cards:
AI Base Value determines how high the AI picks the card in limited, among other things. Use the following scale:
100-200 Bombs that potentially win the game on their own (Loxodon Warhammer, Meloku, Fireball)
80-99 High Quality removal (Terror, Swords to Plowshares)
70-79 Very strong Creatures (Air Elemental, Serra Angel)
60-69 Strong Spells (Tidings)
50-59 Above Average Creatures (Durkwood Boars, Prodigal Sorcerer)
40-49 Medium quality Removal (ex: Immolation)
20-39 Average Creatures
10-19 Lands and mana fixers
1-9 Filler cards (ex: Grizzly Bears)
I want to add that no original cards or just one used other numbers than *5.
5,10,15,20,etc I don't know if that way are better understood-recognised by AI or was easier for Microprose Developers.
---
Trying to squash some bugs and playtesting.
Trying to squash some bugs and playtesting.
-
Aswan jaguar - Super Tester Elite
- Posts: 8078
- Joined: 13 May 2010, 12:17
- Has thanked: 730 times
- Been thanked: 458 times
Re: Fact or Fiction
by stassy » 29 Nov 2012, 04:52
So I had a peek at the Manalink.csv and I have severals questions before trying to fill it out :
- What editor must be used in order to open, modify and save Manalink.csv? I know that Microsoft Excel modify the csv structure so it might not a good idea, can I use LibreOffice?
- The column to modify is the one labelled BO | AI Base Value right?
- Should old cards tagged as 0 stay as it is or upped to 1-9 (Animate Wall for example)?
- What editor must be used in order to open, modify and save Manalink.csv? I know that Microsoft Excel modify the csv structure so it might not a good idea, can I use LibreOffice?
- The column to modify is the one labelled BO | AI Base Value right?
- Should old cards tagged as 0 stay as it is or upped to 1-9 (Animate Wall for example)?
- stassy
- Moderator
- Posts: 5274
- Joined: 25 Feb 2009, 07:06
- Has thanked: 471 times
- Been thanked: 337 times
Re: Fact or Fiction
by Aswan jaguar » 29 Nov 2012, 05:04
- I don't think it is a good idea to change values from original cards.As ranking is not fully understand.Fill only cards from new sets that don't have any values there.
-For me as I wrote in a post above original cards only used *5 numbers 5,10,15,20,.... maybe it is better to keep that,too.
-As for what program to use to change the values excell should be fine (Sonic can probably save or change it to what is needed)but you better ask Sonic for that via pm.
-For me as I wrote in a post above original cards only used *5 numbers 5,10,15,20,.... maybe it is better to keep that,too.
-As for what program to use to change the values excell should be fine (Sonic can probably save or change it to what is needed)but you better ask Sonic for that via pm.
---
Trying to squash some bugs and playtesting.
Trying to squash some bugs and playtesting.
-
Aswan jaguar - Super Tester Elite
- Posts: 8078
- Joined: 13 May 2010, 12:17
- Has thanked: 730 times
- Been thanked: 458 times
Re: Fact or Fiction
by stassy » 29 Nov 2012, 05:43
So that mean filler cards don't get subranking (as between 1-9 there is only 5)? How will I be able to rank Barbary Apes from Forest Bear or Gray Ogre then? 
- stassy
- Moderator
- Posts: 5274
- Joined: 25 Feb 2009, 07:06
- Has thanked: 471 times
- Been thanked: 337 times
Re: Fact or Fiction
by Gargaroz » 29 Nov 2012, 13:26
Shouln't we move this last part of the thread in the "development" forum ?
----
- Current / medium term task: adjusting the code for making Misdirection and such usable
- Long term task: inserting all the good stuff I left out from the "Golden Years" mod
- Current / medium term task: adjusting the code for making Misdirection and such usable
- Long term task: inserting all the good stuff I left out from the "Golden Years" mod
- Gargaroz
- Programmer
- Posts: 7097
- Joined: 06 Nov 2009, 11:11
- Has thanked: 82 times
- Been thanked: 595 times
Re: AI card base value
by stassy » 29 Nov 2012, 16:19
Done.
Waiting for Sonic answers then I will begin the value filling starting from latest to earlier expansion.
Edit: I will have to finish to test en 2nd part of RTR since recent expansions cards are much more subtle than their predecessors in ranking term
Waiting for Sonic answers then I will begin the value filling starting from latest to earlier expansion.
Edit: I will have to finish to test en 2nd part of RTR since recent expansions cards are much more subtle than their predecessors in ranking term
- stassy
- Moderator
- Posts: 5274
- Joined: 25 Feb 2009, 07:06
- Has thanked: 471 times
- Been thanked: 337 times
Re: AI card base value
by stassy » 04 Dec 2012, 05:45
Ok so just one question before beginning : Do we rate cards for their potential or their current state in ML3?
A typical example is Dryad Militant which is a potential sideboard 50-59 rank medium above average creature because it can shutdown many decks but in ML3 as its ability is not working it's downgraded to 1-9 vanilla creature...so rank 50 or rank 1 ?
A typical example is Dryad Militant which is a potential sideboard 50-59 rank medium above average creature because it can shutdown many decks but in ML3 as its ability is not working it's downgraded to 1-9 vanilla creature...so rank 50 or rank 1 ?
- stassy
- Moderator
- Posts: 5274
- Joined: 25 Feb 2009, 07:06
- Has thanked: 471 times
- Been thanked: 337 times
Re: AI card base value
by Aswan jaguar » 04 Dec 2012, 14:58
For me if it acts like a vanilla creature should be valued as vanilla creature as the AI has to evaluate it's true value not the potential.
Of course,it would be nice if you could make a list of such cards in a text editor maybe, to have such cards listed in case they get coded properly so they are easy to track and revalued.
Of course,it would be nice if you could make a list of such cards in a text editor maybe, to have such cards listed in case they get coded properly so they are easy to track and revalued.
---
Trying to squash some bugs and playtesting.
Trying to squash some bugs and playtesting.
-
Aswan jaguar - Super Tester Elite
- Posts: 8078
- Joined: 13 May 2010, 12:17
- Has thanked: 730 times
- Been thanked: 458 times
Re: AI card base value
by stassy » 06 Dec 2012, 12:15
What about confirmed cards and old confirmed non fixed cards yet?
Currently I am tagging unfixable/unusable with red background, shall I do the same for the above in yellow for example?
Anyway here is my RTR cards ranking, attached is an xls file with sorting respectively by rank and by name, the following 2 tabs are text spoiler and rank explanation which I will show here:
Feel free to suggest modifications or modify the xls file, I am not a mtg expert and the rank only reflect what I think the AI should rank cards.
Once the ranking is finalized, I plan to use an uber macro of my own to insert nearly automatically this ranking in the Manalink.csv, unless skymagic is still useable but I doubt so.
If this goes well, I would like to test the AI behavior and the others not so known areas in the csv with for example Fact or Fiction, does the csv to dat compiling tool in the magic_updater folder is enough to do it?
Currently I am tagging unfixable/unusable with red background, shall I do the same for the above in yellow for example?
Anyway here is my RTR cards ranking, attached is an xls file with sorting respectively by rank and by name, the following 2 tabs are text spoiler and rank explanation which I will show here:
Rank 5 Filler cards, vanilla creatures with no abilities
Rank 9 Bugged currently unfixable card reduced as vanilla or unusable by AI
Rank 10 Lands and mana fixers
Rank 20 Average Creatures, with drawback or in need of heavy support to be efficient, heavy mana cost
Rank 40 Medium quality Removal, temporary staller, land searcher, small damager
Rank 50 Above Average Creatures with at least one benefit and no drawback, cheap mana cost and giving support
Rank 60 Strong Spells, global range, multipes uses, token generator, creature retriever, counterspells, color/type killer, support planeswalker
Rank 70 Very strong Creatures but still in need of support or not game winner, global advantage/support
Rank 80 High Quality removal, no drawback, planeswalker with removal abilities
Rank 100 Bombs that potentially win the game on their own
I choose rank 9 because there is no 9 in others rank so I can search faster for them
- Rank 5 | Open
- Rank 9 | Open
- Rank 10 | Open
- Rank 20 | Open
- Rank 40 | Open
- Rank 50 | Open
- Rank 60 | Open
- Rank 70 | Open
- Rank 80 | Open
- Rank 100 | Open
Feel free to suggest modifications or modify the xls file, I am not a mtg expert and the rank only reflect what I think the AI should rank cards.
Once the ranking is finalized, I plan to use an uber macro of my own to insert nearly automatically this ranking in the Manalink.csv, unless skymagic is still useable but I doubt so.
If this goes well, I would like to test the AI behavior and the others not so known areas in the csv with for example Fact or Fiction, does the csv to dat compiling tool in the magic_updater folder is enough to do it?
- Attachments
-
R2R.xls- (162 KiB) Downloaded 681 times
- stassy
- Moderator
- Posts: 5274
- Joined: 25 Feb 2009, 07:06
- Has thanked: 471 times
- Been thanked: 337 times
Re: AI card base value
by Aswan jaguar » 06 Jan 2013, 13:21
For the AI value don't have high expectations for AI evaluation on how to play the cards as mostly is used by Draft and Sealed mode (as Mok and Jatill mentioned) but in draft YES expect AI to select cards according to your evaluations.
I just regret I didn't mention that to you because cards that have really high value in certain decks and combined with other cards are not so valuable or just rubbish when they are drafted,like theme decks-cards like tribal cards-boosters.
Thank you very much for your work,stassy it is greatly appreciated.
I just regret I didn't mention that to you because cards that have really high value in certain decks and combined with other cards are not so valuable or just rubbish when they are drafted,like theme decks-cards like tribal cards-boosters.
Thank you very much for your work,stassy it is greatly appreciated.
---
Trying to squash some bugs and playtesting.
Trying to squash some bugs and playtesting.
-
Aswan jaguar - Super Tester Elite
- Posts: 8078
- Joined: 13 May 2010, 12:17
- Has thanked: 730 times
- Been thanked: 458 times
Re: AI card base value
by Aswan jaguar » 13 Jan 2013, 19:16
Stassy, as I discovered the AI card base value is essential for draft play as AI won't put cards with 0 value to his decks(AI values them as no good) and as a result AI fills the decks it makes with too many lands and so the decks can't be challenging.I believe AI certainly will appreciate your work and all of us who love draft,too.
---
Trying to squash some bugs and playtesting.
Trying to squash some bugs and playtesting.
-
Aswan jaguar - Super Tester Elite
- Posts: 8078
- Joined: 13 May 2010, 12:17
- Has thanked: 730 times
- Been thanked: 458 times
Re: AI card base value
by stassy » 23 Jan 2013, 17:01
Sadly after a few tests with a debugged Fact or Fiction I couldn't conclude of any lead on how the AI select defined cards, because the card effect itself does not show if AI choose some cards more than others (seems like it random select between the 5 cards and show 2 then 3 of them regardless of their AI base value.
Anyway here is the current status of the ranked cards by expansion :
Anyway here is the current status of the ranked cards by expansion :
Ranked
Innistrad
Dark Ascension
Avacyn Restored
Return to Ravnica
Duel Decks: Izzet vs Golgari
From the Vaults: Realms
Magic 2013
Ravnica: City of Guilds
Edit: Nevermind, it seems to be that if the AI cast Fact or Fiction and if player clearly make pile with "good ranked cards" and "bad ranked cards" the AI will always choose the "good ranked cards" pile, guess at least it's a sign of progress...In progress
Magic 2010
Magic 2011
Magic 2012
- stassy
- Moderator
- Posts: 5274
- Joined: 25 Feb 2009, 07:06
- Has thanked: 471 times
- Been thanked: 337 times
Who is online
Users browsing this forum: No registered users and 22 guests