Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[rrenaud]

So this is inspired by wanting to make councilroom better for this purpose:

http://forum.dominionstrategy.com/index.php?topic=2301.0

If I had some training data, I could try to optimize the card (combo) ranking function.  The existing ranking is based on QVist's single card data.

This is sort of in the spirit of the QVist community rankings, except without anyone doing the hard work of aggregating and building a uniform ranking and providing commentary.  Of course, if someone wanted to take something akin to his role, I'd happily let this morph into that if someone wants to take the job.

Basically, list a card, followed by : +followed by cards that you think make it better, ?followed by cards that you think are mostly independent/non-interactive, - followed by cards that you think make it worse.

For example,

Mountebank: +King's Court, Scheme, Throne Room, Chapel, Village ? Caravan, Apprentice, -Counting House, Gardens, Ambassador, Trader

So I like Mountebank more in games with KC, Scheme, Throne Room, and Chapel, Village, I am mostly indifferent to the presence of Caravan or Apprentice, and I like it less in Counting House, Gardens, Ambassador, and Trader games.

This is just overall, in aggregate.  Don't try to stretch to hard to find some conceivable reason for the indifferent cards to matter.  For example, KC/Caravan/Mountebank might work better than just KC/Mountebank by itself, but mostly Caravan won't influence my decision for Mountebank, so I list it as indifferent.

Broad coverage is more useful than different people telling me 50x that KC and Bridge are awesome together.  It's also important to know about the more generic and less interesting combinations.  Eg, University/Bazaar is a fine and good combo, even if it's not exciting.

It's also useful for you to be unbiased.  Try to come up with your own thing without looking at others feedback or councilroom.

[Donald X.]

Mountebank: +King's Court, Scheme, Throne Room, Chapel, Village ? Caravan, Apprentice, -Counting House, Gardens, Ambassador, Trader

Probably you want to define some categories, and let people list those categories too. Villages, +Buys, etc.

[rrenaud]

Okay, fair enough.  It's also okay to list general classes of cards with the same 'syntax' as councilroom uses, enclosed in [].

Expansion=="Dominion" || Expansion=="Intrigue"

Cost==7 returns Bank, Expand, Forge, and King's Court
Cost == "P2" returns Apothecary, Scrying Pool, and University
(""+Cost).indexOf("P")>=0 returns all cards with Potion in cost

Actions >= 2 returns all cards that provide 2 or more Actions
Cards < 2 returns all cards that draw either 0 or 1 Card when played
Coins == "?" returns Bank, Philosopher's Stone, Pirate Ship, Salvager, Secret Chamber, Trade Route, Tribute, and Vault
Buys == 1 returns all cards that provide +1 Buy
VP >= 6 returns Province and Colony
Trash == 4 returns Chapel

Action && Victory returns Great Hall, Islands, and Nobles
Treasure && Victory returns Harem

Attack returns all Attack cards
Duration returns all Duration cards
Reaction returns all reaction cards

Library: +[Actions >= 2], Militia ? Spy, Talisman - Council Room, Caravan
Scout: +[Action && Victory], [Victory && Treasure] ? Festival - [Trash > 0]

[theory]

I assume we should feel no obligation to be comprehensive?

[rrenaud]

No, write as little, or as much as you want.  Even one reasonably thought out line for one card is fine.

Obviously, more stuff from better players is better .  More coverage of stuff that people haven't yet covered is also good.

And even doing the straight up boring work of parsing previous dominionstrategy.com articles Works with/Conflicts sections with could be a useful thing for someone to do that requires only knowledge of English, and no actual Dominion skill.

[blueblimp]

It's really difficult to think of the "?".

Torturer:
+ Village, Border Village, Hamlet, Fishing Village, Worker's Village
? Smithy (NOTE: there is asymmetry here: Torturer makes Smithy worse, but Smithy doesn't make Torturer worse!)
- Lighthouse, Sea Hag, Masquerade, Trading Post, Ambassador

[qmech]

Here are some possibly less popular ones.  I take full advantage of my freedom to not be comprehensive, particularly on the indifferent section.

Contraband: + Apprentice, Salvager, Upgrade, Remodel, Expand, Remake, Colony, Platinum, Harem, Nobles ? Monylender, Spice Merchant - [Cost==5 && Cards >= 3], Wharf, Merchant Ship, Governor

Philosopher's Stone: + Ambassador, Platinum, Colony ? Coppersmith, Warehouse - [Cards>=1], [Trash>=1], [Duration], Governor

Explorer: + Duke, Gardens ? - [Cost==5 && Cards >= 3], Wharf, Merchant Ship, Governor

EDIT: inserted missing colons

[rrenaud]

If you are thinking a lot about ?, you are thinking too hard.  I expect it to be true of most card pairs.

[blueblimp]

If you are thinking a lot about ?, you are thinking too hard.  I expect it to be true of most card pairs.

I'm not sure this is true of most pairs. Any two terminals tend to conflict, especially if one is terminal draw. (So, for example, Torturer would tend to conflict with every terminal.) Villages tend to conflict with each other because you usually don't need more than one kind. Cantrips tend to work well together, because a deck that likes cantrips often wants to get as many as it can. etc.

[ftl]

Sure, I'll take a shot:

Conspirator:
+: [Cards >= 1 && Actions >= 1], [Actions >= 2], [Trash >= 2], [Buys >= 1], Warehouse, Cellar, Ironworks, Moneylender, Throne Room, King's Court, Pawn, Tactician, Loan, Spice Merchant, Stables
? [Cards >=2 && Actions == 0], Mine, Adventurer,
-: Witch, Young Witch, Sea Hag, Mountebank, Gardens, Trader, Swindler,  Duke, Torturer, Ghost Ship, Familiar, Rabble, Tournament, Fool's Gold, Tunnel, Jack of all Trades, Embassy, Jester, Ill-Gotten Gains, Governor

Am I doing this right?

I don't actually get what's supposed to go in the '?' category. I mean, by default everything that's not in + or - is ?, right? There's lots of thing which won't affect my decision to go conspirator or not - I wouldn't buy them for a conspirator engine, but I wouldn't stop my conspirator engine because of their presence, either...

[rrenaud]

Okay, take Torturer:

Monument, Merchant Ship, Baron, Herbalist, Haggler, Harvest, Fortune Teller, Explorer, Expand, Embassy, Duchess, Cutpurse, Counting House, Council Room, Conspirator, Mine, Mint, Moneylender, Workshop, Woodcutter, Treasure Map, Transmute, Bureaucrat, Rabble, Remodel, Pirate Ship, Black Market, Outpost, Oracle, Nomad Camp, Navigator, Adventurer

I think my decision to go for a torturer is basically independent of the presence of those terminals.

[Jorbles]

Ironworks:
+: Silk Road, Gardens, Duke, Vineyards, Caravan, Tournament, Ill-Gotten Gains, Island, Great Hall, Highway, Fishing Village, Lighthouse, Throne Room, Fool's Gold, Conspirator, Oasis
?: Most other things costing $4 or less
-: Basically everything else costing over $4

Did I forget anything major?

[rrenaud]

Squeezing it all on one line would be great, or at least definitely having the target and the interaction cards, cause eventually I am going to have to put these all together.

The purpose of the ? to be pushed towards "0" by the automatic ranker, and split between good and bad ones.  It's just like, there is no obvious connection between these cards.

[Jorbles]

How should we be handling counters? For example if Library and Militia are on the board should Library be a + for Militia or a -?

It all depends on who has the Militia in their deck whether Library is a + or -, and since that's the case I'd vote we just rank counters as ?s.

So for the above example
Militia:
?: Library, Watchtower, Jack of all Trades, Horse Traders

[rrenaud]

I think Militia would have a - for Library, Watchtower, Menagerie, etc, and Library, Watchtower, etc would have a plus for Militia. 

[Jorbles]

I think Militia would have a - for Library, Watchtower, Menagerie, etc, and Library, Watchtower, etc would have a plus for Militia. 

Ah so, the goal being that if you have a card in your deck then is it a + or a - for the other cards to be on the board (as opposed to in your deck/in your opponent's deck)?

[theory]

Yeah, I think rrenaud's goal is:

I want to talk about card X on this board.  It is better than usual because card Y is also on the board.  It is worse than usual because card Z is also on the board.  My opinion of it is unaffected because Q is on the board.

X: +Y, -Z, ? Q

Militia: it is better when Chapel is on the board.  It is worse when Library is on the board.  It is unaffected when Transmute is on the board.

[Jorbles]

Okay then with the above in mind:
Venture:
+: Bishop, Mine, Moneylender, Upgrade, Loan, Masquerade, Lookout, Trade Route, Rabble, Spy, Oracle, Tunnel, Steward, Trading Post, Remake, Chapel, Forge, Scrying Pool, Witch, Familiar, Sea Hag, Duke, Cartographer, Scout, Fortune Teller
-: Thief, Noble Brigand, Ambassador, Contraband, Pirate Ship, Mountebank, Horn of Plenty, King's Court, Remodel, Governor, Cache, Counting House, Apothecary, Coppersmith, Chancellor, Gardens

Edit: Removed -: Bank.

[Jorbles]

Rabble:
+: [Actions >= 2], [Victory && !Action && !Treasure], Throne Room, King's Court, Scheme, Witch, Sea Hag, Mountebank, Familiar, Stash
-: Venture, Adventurer, Farming Village, Scout, Oasis, Warehouse, Cellar, Crossroads, Lighthouse, Cartographer, Swindler

[blueblimp]

Okay, take Torturer:

Monument, Merchant Ship, Baron, Herbalist, Haggler, Harvest, Fortune Teller, Explorer, Expand, Embassy, Duchess, Cutpurse, Counting House, Council Room, Conspirator, Mine, Mint, Moneylender, Workshop, Woodcutter, Treasure Map, Transmute, Bureaucrat, Rabble, Remodel, Pirate Ship, Black Market, Outpost, Oracle, Nomad Camp, Navigator, Adventurer

I think my decision to go for a torturer is basically independent of the presence of those terminals.

Good point, it's the more powerful card that makes the less powerful card less attractive, not the other way around. Although with these terminals, if there's no village, I might prefer going something like 1 Monument 1 Torturer over 2 Torturers, so Monument does make Torturer less attractive in that situation.

[michaeljb]

Okay then with the above in mind:
Venture:
-: Bank, Remodel, Governor

Wait, what? Why?

I don't at all see the reasoning for Bank...as for Remodel and Governor, is it just because of turning Gold into Province?

[WanderingWinder]

I can do a lot of these over time, but my real question is what the point of the neutral rating is. Is it to bias the rankings toward neutral, in case someone else thinks they AREN'T neutral? In that case, I would probably put a lot of nombos there. If not, I don't really see the point.

[Obi Wan Bonogi]

If your goal is "enjoyable games" why would you put KC and Mountebank together?  That two card combo is probably responsible for more early game resignations(at least for myself) than all other two card combos combined.

[rrenaud]

The reason to have the non-interactive cards is to guide the combo ranking optimization to find better solutions.  Putting the really good combos above the really good counters is easier to do and less informative than putting the really good combos over the nombos over the good counters.

The direct goal is to be able to estimate the quality of combos better.  How to use that information to make interesting kingdom sets is another question, and avoiding super gross combos (KC + mountebank) might need to be a component of that part.  Certainly, better machine understanding of the quality of combos makes the "generate fun/interesting" kingdom sets problem easier.  Getting a machine to understand "fun" OTOH, could be a big problem .

I am not sure if KC + mountebank is counterable in 2p, but keep in mind that these sets will be selected to have things like watchtower, gardens, trader, counting house, etc, to try to mitigate the combo.

[blueblimp]

Getting a machine to understand "fun" OTOH, could be a big problem .

Maybe harvest kingdoms from the Game Reports forum?