Dominion Strategy Forum

Please login or register.

Login with username, password and session length

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - terminalCopper

Filter to certain boards:

Pages: 1 2 [3] 4 5 ... 14
51
Dominion Articles / Re: What's stopping AI from mastering Dominion?
« on: March 02, 2018, 02:58:33 am »


Lack of training data
AlphaGo analyzed hundreds of thousands of professional games online before it reached a professional level. We have nowhere near that amount of data, not even taking into account the rating of the players. For one, there's really no such thing as a “professional dominion player”; there are a handful of highly-rated players who regularly place well in tournaments, plus stef himself (who doesn't participate in tournaments or rated games anymore). A neural network trained on data from all rated games would probably develop reasonably strong money-based strategies and beat lower-ranked players, but would lose to anyone who who frequented these forums. Furthermore, there's no good bot to play hundreds of thousands of games against in order to get a basis for good play. An AI that trains against Lord Rattington is going to end up marginally harder to beat than Lord Rattington, which isn't saying much.

2. Sheer number of starting game states
In games like chess, checkers, and go, there are a total of 2 possible starting game states for any given player: either you go first or you go second. In Dominion, that number is at least in the realm of quadrillions. An algorithm would need to analyze each game on the fly; it can't have a bank of games it's played with that same kingdom to fall back on. Even if a neural network could get to the point of basic kingdom analysis, small things like what the Bane is or what Obelisk is on can completely change a game. This type of intricacy is pretty common, but specific instances of it are rare enough that a neural network may not have a point of reference to determine whether, say, opening Young Witch is worth it in a given kingdom. It will also be bad at recognizing 2-card combos on its own. Something like Donate/Market Square May be obvious to a human, but how would a computer ever come up with that?

3. Randomness
Computers that play board games will often form a tree of possible moves when deciding what move to make next. The only variable they have to take into account is their opponent, and they can form a tree of moves for their opponent as easily as they can form one for themselves. However, it wouldn't be time-efficient for a computer to calculate each possible arrangement of its cards after a shuffle every time it plays or buys a card; it would need to develop some more abstract way of determining its choices. That's a level that AI as of yet tends to struggle with. This hypothetical dominion bot wouldn't be able to calculate its exact percentage chance of winning with each decision, so what would it go off of?

4. Number of small choices
For a while, Lord Rattington had a bug where if you played Vault until it had 0 cards in hand, it would freeze. This is an example of a bot not understanding a simple decision of whether or not to discard. Even after the bug was fixed, the bot continued to discard until it had 0 cards in hand. An AI would have to learn how to respond to every little prompt, and the difference between each one. It can't just choose not to discard to Militia, but it has to eventually choose not to discard to Vault. This is made harder by the fact that it's not necessarily train against an opponent who plays 5 Vaults per turn. It's going to have to somehow learn what the best response is to each little prompt, and it's going to have to do so quickly, as it may not get many other chances to go over that situation. This isn't how neural networks work; they need lots of reinforcement in order to be able to make good decisions.

5. Unique cards
An AI, if it got to this point, would likely be able to determine that Village and Worker's Village are pretty similar, and might group cards into categories much like humans do. But it could have more trouble with unique cards like Lurker or Villa that seem similar to other cards but play very differently. It could misjudge a card’s function based on a small number of games. I don't have a good idea of how this would play out; it's possible the AI could be better at this than I'd expect.


These are all good points, making a Dominion AI difficult. But I believe, the most important reason is missing:

6. By now, no one is willing to invest millions of dollars

I am pretty sure, if a couple of top players and a dozen brilliant guys from DeepMind would work together for a year, the resulting AI would beat us all.

52
Dominion Articles / Re: Tracker
« on: February 27, 2018, 01:24:03 pm »
A mere terminal Copper is something you never [...] ever want.

I hope my fiancee does.

53
Dominion Articles / Re: Shepherd
« on: February 22, 2018, 04:53:29 am »

I believe that the community is still far from deeply understanding this complex card. Anyway, I‘d like to contribute something I consider to be important:

Shepherd is an excellent opening, even if you don‘t plan to focus on a sheperd deck.

Like few other cards, it is both a very good cycler and a card with good odds to hit five bucks - here‘s the scenarios where you spike the price point, with the according probabilities:

If sheperd doesnt miss shuffle (5/6),
4 coppers with shepherd, two estates miss shuffle  (6 choose 4 / 11 choose 4)* 3/7*2/6
3 coppers + pasture with shepherd, Sheperd draws silver (6 choose 3 / 11 choose 4) * 2/7
3 coppers + pasture with shepherd, Sheperd draws two copper (6 choose 3 / 11 choose 4) * 4/7 * 3/6
3 coppers + pasture with shepherd, Sheperd draws two estates (6 choose 3 / 11 choose 4) * 3/7 * 2/6
Sheperd comes with at least one estate or silver (1- (7 choose 4 / 11 choose 4))

If sheperd misses shuffle (1/6),
misses with estate: 3/11 *1
misses with copper/pasture, 5 treasures in one hand: 7/11 * 1/6
misses with copper/pasture, other treasures split „4-3“, silver in the hand with 4 treasures: 7/11 * 5/6 * 1/2
misses with silver, 5 copper/pasture in one hand 1/6*1/11*1/6

5/6*
((6 choose 4 / 11 choose 4)* 3/7*2/6)+ ((6 choose 3 / 11 choose 4) * 2/7) + ((6 choose 3 / 11 choose 4) * 4/7 * 3/6)
+ ((6 choose 3 / 11 choose 4) * 3/7 * 2/6) + (1- (7 choose 4 / 11 choose 4))

+1/6*
(3/11*1 +  7/11 * 1/6 + 7/11 * 5/6 * 1/2 + 1/6*1/11*1/6)

=

5/6 * 0.9437+ 1/6 * 0.6464 = 0.894

-> if you open Sheperd/Silver, you have an 89.4% chance to hit 5$ in T3/4.

54
Dominion Articles / Re: Witch, revised for 2018
« on: February 18, 2018, 02:12:35 pm »
I now propose: nobody ever lost Dominion by playing King's Court on a Mountebank.
Edge case: There are 1-3 Curses left and two empty piles and you're too far behind to catch up this turn.

Another edge case: Your opponent is playing Trader/Feodum and has Trader in hand.

Refuting an empirical claim with theoretical edge cases doesnˋt make sense.

55
Help! / Re: Posting losses in an attempt to get better.
« on: February 07, 2018, 09:35:42 am »
Hmm what about Gardens on this one? Gaining cards is enabled by Treasure Trove, and Delve for cheap Silver. To try and get a lot of Treasure Trove plays, tempted also to Mine Silver into Treasure Trove.

I don't think it can compete with the engine.

Yeah the engine seems really explosive here once you start the cost reduction train, though I'm not sure how fast Treasure Trove / Delve money might be when it comes to Province rushing.

Even without thinking, I would have chosen the engine path.  I totally agree with Mic that I want 2 Salvagers and multiple Archives here, but no Guardians. The only point where I disagree is that I wouldn't open Mine on 5/2; Bridge Troll looks stronger to me. I think it is more important to have better economy than our opponent right from the start than having a better shot at CQ's later. We should be able to get CQs by salvaging silver or salvaging salvager, also we don't need that many. Even a single CQ can provide enough actions to bring 4 Bridge Trolls in play, add four highways and pick 5 provs for free. No need to do better.

However, to be fair, it's not totally clear to me whether engine is best here. BM is really a powerful option in this kingdom, because there's a lot more synergy going on than Treasure Trove & Delve. Salvager helps milling provinces, it can also salvage Treasure Trove's Gold; Archive is great in huge decks, it smoothes out draws, and helps to bring Salvagers and their targets together. In general, Archive/Treasure Trove/Delve are good in BM, and the described synergies turn BM into a tough benchmark. If you open 2/5, an approximate plan might be

T1/2: Treasure Trove / Delve
T3/4: Salvager / Archive
T5-T9: 2 Provinces, Salvager, Delves, Archive
T10 - T12: Buy 2 more Provinces, mill a third Province
-> 4 provinces, with 3 provinces left

This performance is close to Gear-BM, which is one of the best BM-attempts in Dominion. And it has to be mentioned that the deck doesn't stall in T12, there's more Gold coming in, and milling will go on.

I don't think it makes sense to describe a similar buying plan for engine strategies, they are way more complex, flexible and situational. But my gut tells me that in a non-mirror against a good BM-player, our camp of engine believers would have to hurry up.


56
Help! / Re: Posting losses in an attempt to get better.
« on: February 01, 2018, 06:13:09 am »

This is probably the fifth time I look at this board, and finally, I feel like having an idea what to do :)

I'd open Ratcatcher/Bards, planning something like
2 Ratcatcher, 6 Bards + Lost Arts, 2 Libraries, Duplicate, gaining some treasures by Bards.

If I use Duplicate twice, this will be achieved until T10. My ratcatchers are slow, but being two, they should have trashed approx. 3 estates and 2 coppers, leaving me with a deck similar to

6 nonterminal Bards
2 Libraries
7 treasures.

In average, I hold 6/15 * 5 = two Bards per starting hand, and 2/15*5 = 0,66 Libraries.

This is a good starting point to achieve $11, some examples:
play one Bard, library draws 8$, bard gives +1$
play two Bards,  library draws 7$
play three Bards, have 3$ in hand, +extra card and extra coin

Something like this should work more often than not; buying 4 colonies by T16 seems possible to me.

Btw, I'd like to adress two interesting subthemes:

Monastery>Ratcatcher?

In most cases, yes. Definitely, I have Monastery on #3, it's bonkers! Compared to Ratcatcher,

a) it misses less shuffles,
b) it can trash multiple cards,
c) it's never drawn dead,
d) when trashing copper, copper still provides +1$.

Usually, this outperforms ratcatcher - but not in the situation I described. I don't really care about trashing faster (a) and b)), because my library wants to draw up to 7 treasures, and I don't want to spend time on replacing coppers by other treasures. c) doesn''t matter because libraries don't draw actions dead.

Compared to this, ratcatcher has three advantages,

a) It draws a card, which is important to get 4$/5$/6$
b) it trashes before library draws, which will draw one more card
c) it gets out of the way when it comes to greening.

Therefore, in this specific situation, I prefer Ratcatchers to Monastery.

Pirate Ship? Really? Maybe ...

I gave +1 to fausts posting because it is really a fascinating idea. If everything works well, Pirate Ship is really an option here.

However, it doesn't really scare me, because it's clumsy to achieve all three goals in a reasonable amount of time: trashing, gaining multiple Pirate Ships and spiking 6$ for lost arts. My gut tells me this will be slow, and in case that my Bard-Deck is threatened by this, the detour to replace lib's by some Poachers is not desirable, but possible.

But, to be fair, I'm not sure here. Maybe, really, this might be one of the rare decks where Pirate Ship is king, thanks to Lost Arts and Colonies.

57
Dominion General Discussion / Re: How thematic are card names?
« on: January 30, 2018, 05:05:44 am »
*:
Plan

I’d give Plan  *** or even ****.

Because whenever I buy Plan, I have a plan.

If I said this about my other purchases, it would be a lie.

58
Help! / Re: Posting losses in an attempt to get better.
« on: January 27, 2018, 11:47:16 am »
I believe in storyteller/market square, enabled by two amulets played every other turn. Ironically, I would open Courtier anyway, because it‘s best for spiking 5, and doesn’t collide with amulet. A short plan:

T3/4 Gain Gold, buy 2 amulets
T5/6/7 hopefully gain two storytellers, one market square, trash three cards -> good odds to already draw deck

After that: more storytellers, more market squares, more gold (with this priority), maybe add a village for mid-turn shenanigans with amulet/market square.
 
It should be easy to enable market squares and make the deck explode, because storytellers with cheap golds draw like hell.
.


59
Dominion General Discussion / Re: The Dominion Cards 2017 Edition: Boons
« on: January 27, 2018, 08:50:53 am »
Going for boons as copper trasher sounds to me like going for knights as an Alt-VP Strategy.

60
I mostly agree, i. e., I have 15 of your top 20 in mine as well, and all of them would be at least, say, Top 40.

My least subtle disagreements:

3.) Wandering Ministrel > Ironmonger

WM is best for engines without trashing, and it‘s more reliable. Granted, Ironmonger is better with treasures. Treasures, meh.

2.) Torturer > Ghost Ship

The first Ghost Ship might be a notch better than the first torturer you play (I think the meaner attack is more important than 1 card). But the second play, it‘s faithful hound, whereas Torturer starts to take off. Torturer chains dominate boards way more often than ghost ship does.

1.) Tournament

... is Top 10 for me! To judge a cards power, consider games where one goes for it but the other doesn‘t. Skipping it is like saying „Here you are, some followers, a Steed and a princess for free - and hey, what about some gold and some duchies on top?“ This would lose many games, which is the reason why it  virtually never happens.

61
Some cards cost 7$ or more, because at 5$, they would be too swingy if one player opens 5$ and the other doesn‘t. This risk can be solved by debt. Example given, Forge could be priced 7 debt. Similarly, Prince would be more interesting as a debt card.


62
Help! / Re: Posting losses in an attempt to get better.
« on: January 20, 2018, 05:52:29 am »
Tournament is outstanding here, it's all about scoring 8$ ASAP. That makes salvager a strong opener. There's too many problems to set up an engine: no good copper trashing, (almost) no splitter, junking announced (followers), and margraves might attack you. That being said, despite the presence of wonderful payload cards like Groundskeeper/Wild Hunt, you should ignore these, because you'll never be able to chain them in a significant way.

Planning terminals is tricky. If you go for margrave, you don’t want to have more than three or four terminals, e.g. salvager, 2 margraves, maybe followers. If you go for vaults, you can pick some more. Personally, I‘d still give a small edge to margrave over vault, because the attack nerfs vault the hard way (which would have been pretty nice here, if there weren't handsize attacks). Just don't overdo the margrave purchases, e. g. once you have one, buy gold with 6$.

Concerning prices, get Trusty Steed first, then Followers. Also, don't auto-play Steed as Lost City, these silvers are nice here.  By the way, I'd get Bag of Gold over princess.

A second salvager can be helpful in the endgame, salvaging a colliding terminal or a (bag's) gold.

I think it's not going to be a very interesting game, if both players focus on 8$->province->tournament, luck will be the most important factor. Just don't get distracted by engine fantasies. It’s sad to see so many powerhouses starve, but in this scenario, Gold is simply better.

63
Help! / Re: 2 Bold Strategies to Choose From, Which One Would You Go For?
« on: January 19, 2018, 01:04:10 pm »


I’m with Gazbag here: Ports merit Pathfinding.

Whoever wins the Groundskeeper split should be in prime position to win so that is the first objective.  Sprawling Castle ... is worth 27 points.  Whomever gains this likely wins the game.

I think this is too slow:

I would go hard for the ports split, not for the groundskeeper split. The main reason is that it is much faster for two reasons:

- you merely need 4 buys,
- there will likely be some situations where we hit exactly 4$ (there is already a 5/6 chance in T1/2). If so, I believe that ports are the correct buy either way - even if you don’t plan to pathfind it.

@shark-bait: Assume that we both hit 4$ twice (buying ports), we both go hard for pathfinding (e. g., buy butcher, save coins for Pathfinding) and achieve it at the same time, the only difference being that

I spent four buys on eight pathfinding ports,
you spent these buys for four ports and two pathfinding Groundskeeper.

First, I am not sure if this will win you the groundskeeper split, because I am much closer to draw my deck. If you don’t win your split, you’re kinda lost.

Second, even if you win it, the VP/turn rate is not determined by the number of groundskeeper alone, but rather by the product ((number of Groundskeeper+VP/card) * number of gains). I believe that concerning the number of gains, I will have the upper hand because I draw my deck much earlier to support multiple butchers per turn.

However, it’s going to end on three piles, whether the third pile will be islands or estates depends on the situational tactics. In other words, I am almost sure we will never see sprawling castle.

64
Thx for this opportunity :)
What I would really like to have is the possibility to play against an AI that is able to build engines. It doesn‘t have to be supergood, but I would appreciate if it feels like playing against a human.


 1. An offline client, allowing you to play campaigns against bots
- Yes
2. A bot with a setting for different play styles Not necessarily
3. A strong bot. - Definitely, yes!
4. Options for timed games - No
5. Options for undo settings - Yes, sounds good
 6. Online mini-tournaments; the option to easily create one for you & your friends. - No
7. Option to invite specific players to your table Yes! That would be very nice.
8. More translations (to what language?) - No
9. Better animations when cards are gained/bought/played/... Not necessary
10. Better visualization of split piles, Archive/Crypt contents, Prince targets, ... It would be nice, but not important
10. Option to hide the log - No
11. Android app (or just improved compatibility?) - No
12. iPhone app (or just improved compatibility?) - Yes
13. Downloadable game logs Yes.
14. Improved moderation (dealing with chat abuse, slowplayers, ...) No, it‘s fine as is.
15. Tutorial on how to play Dominion - No
16. Tutorial on how to play dominion online. No

65
I really like your post, but on top, I‘d like to emphasize that your statement

play five Groundskeepers and Estates become Provinces.   

even underestimates Groundskeeper. The truth is, Estate+5VP is way better than a province, because you can trash that estate and keep your deck thin.

66
Fortune‘s Rating is probably based on how powerful the card is once you have it, but not its overall impact on the game. A typical situation is that you have built a reliable engine, hit $8, buy fortune, then pay debt / buy a province, then buy two provinces twice. It takes you 4 turns to get five provinces, which is one turn faster than buying them one by one without fortune.

That makes it worth going for, but the game didn‘t change by that much. You merely need one turn less - nice to have, but in terms of impact, it‘s far away from meriting place 5, at least it shouldn’t be ahead of potentially gamechanging cards like e.g. Lost Arts, City Quarter or Inheritance.

67



When I saw Sentry the first time, it seemed overpowered to me. But that‘s not true, for three reasons:

1.) Sentry doesn‘t trash much more than one card per turn on average.
I can‘t prove this with stats, and there‘s a lot to factor in, so let‘s try to look at a somewhat average case for the sake of simplicity and assume, that

• you play Sentry in T5, T7, T8,
• You gain one good card per turn.

When you play Sentry in T5, you have bought three (other) good cards and ten starting cards, so you will trash 2*(10/(10+3)) = 1.5. cards in average. From there, the ratio bad cards/good cards will decrease, e. g.    in T8 you might have 7 bad cards left compared to 7 good cards, which means that the number of expected trashed cards is close to 1.

In total, in the first three plays I expect Sentry to trash about 1.2 cards per turn, and less afterwards.


2.) Sentry trashes fewer estates than e. g. Jung Dealer.

A big downside is that you can‘t trash cards from your hand, and among three trashed cards, a typical number of trashed estates is probably one. I guess Junk Dealer has good chances to trash one more estate.

3.) The bonus (reordering) helps rarely in the beginning.

You don‘t want to discard good cards in the beginning, and the only situation reordering might help you is if you find two good cards, and put the convenient one on top for cards like Vassal. It is way more likely that you can make use of Junk Dealers +1$  when he trashes an estate, or Upgrades ability to gain a 3$-cost.

Summed up, Sentry trashes merely a little faster than upgrade or Junk Dealer, but suffers from catching less estates and providing fewer bonusses.

Of course, this doesn‘t make Sentry a weak card. It‘s a cantrip trasher, that‘s fine. It‘s just not OP like it looks at first glance.



68
The first article I wrote claimed that Junk Dealer was underrated, it was on place 32/61 in 2013. Now I think it went up too high ... I mean, yeah, it's powerful, so if you skip it, you often fall behind. But if you skip Cultist or Mountebank, in many games you have basically resigned.

69
Seems like many people haven‘t understood yet how to handle Crypt. Including me.
Once we get a better grasp, I think it will be a big winner next year.

70
Skulk on 52, but Quarry on 17 doesn‘t match.

Skulk is better if you need +buy, with tfb, and the gold works with treasures and victory cards as well.
Quarry is better if you don‘t want the skulk, and if you buy multiple actions.

But the most important thing is that with both cards, you can open with (kind of) a gold to get expensive actions early.

Therefore I believe that Skulk and Quarry should be approximately at the same level.

71
Game Reports / Re: Figuring out the golden deck here
« on: January 10, 2018, 02:17:59 pm »


 I strongly believe you want a catapult here. I’m not sure about the optimal realisation; the best version I can spot is

T1 buy catapult
T2 donate, keep 5 coppers
T3 catapult copper, pay debt (granted, there‘s a 1/6 risk to miss catapult, but I think the extra coin is worth trying).
T4 catapult copper,  buy silver
T5 catapult copper, buy silver
T6 catapult copper, buy goons
T7 play goons, buy artificer

In the next turns, we get diplomats/artificers/goons while catapulting the remaining copper and the silvers.

I think that this beats any non-catapult-strategy handily which will have a hard time to get a T6 goons under permanent attack, and still has some leftover coppers on top.

In practice, it will play out quite differently if both players get catapults, but that‘s the whole point: you need a catapult here.

72
Help! / Re: Double Tac, Villa, Bridge, Storyteller, HoP
« on: January 09, 2018, 01:07:19 pm »
I think tac is weaker than storyteller, temple is too slow. It is better to open with bridge or dismantle.

The latter looks interesting - here is a plan i expect to be clearly faster than T17:

If everything works out, we can

open dismantle/silver
T3/4: Buy silver, silver; dismantle estate into copper, gold -> 17$, 15 cards
T5,6,7: Buy storyteller, storyteller, bridge; dismantle one more estate -> 22$, 17 cards (not counting Storyteller)
T8: draw your deck with two storytellers spending 12$ -> play bridge, buy villa, buy two more storytellers
T9 - T11: explode with bridges/villas/HOP/storyteller

T8 involves some luck, and we might also have bad luck in the beginning, or dud during T9-T11. But I don't see a reason to expect losing more than two turns on average, we might even get lucky with 5$ in T3/4.
Therefore, a megaturn in  T13 or similar seems realistic to me.

 

73
Another placement that seems weird is Guardian. That card is basically a delayed Copper unless there are nasty attacks, so how is it so good?

Nasty attacks are often the best cards in the Kingdom.
Also, a delayed copper is better than a copper.

74
Help! / Re: Battlefield and Distant Lands but no draw
« on: January 08, 2018, 03:15:37 pm »
One thing I really like about Butcher is the way it can alternate between turning Estates into $3 cards and Coppers into $3 cards. That's especially true on this board where there are a couple of nice $3s to get.

Butchering copper into a 3$ is rarely a good choice.

Instead of +1$ (copper) you have to invest a coin, so you play a terminal 5$ that kinda says  -2, trash a copper, gain a card costing 3$. It's like self-enchantering an activated conspirator to pay 5$ for a weakened workshop just to trash a copper. There might be some edge cases where  it is worth the effort ...

... but definitely, not here: having more 3$'s than you get by dud turns and transformed estates might even be bad (oasis, shanty and trade route all risk to harm your drawing power), you don't need to get rid of copper at all price (because of apothecary) and even if you want to, upgrade does this by far better.

75
Help! / Re: Battlefield and Distant Lands but no draw
« on: January 08, 2018, 07:54:19 am »
Is Upgrade really a lot better than Butcher on the first $5?

Yes it is:

- in general, +action +card is more often useful than +coin, compare lost city to baker.
- with shanty in hand, upgrade leads to three more cards than butcher.
- after playing apothecary, upgrade can draw whatever apothecary leaves on top.
- moreover, in this specific kingdom, draw is somewhat unreliable.

Butcher will be king when greening, but I want to trash all the estates and a few coppers as well before, this is why i think that 2 is the best number of upgrades to get.

I agree they will be butchered in the final phase.

Pages: 1 2 [3] 4 5 ... 14

Page created in 0.056 seconds with 18 queries.