Dominion Strategy Forum
Dominion => Simulation => Topic started by: NoMoreFun on June 22, 2013, 05:41:19 am
-
Is there a strategy out there that utterly annihilates the standard big money bot to the point where it never even wins a a single game?
I'm thinking there'd be a bot out there that can end the game or set up a Masquerade pin (with Monument to guarantee victory after it stops pinning) so fast that the Big Money bot never wins a single game.
-
Here's an extremely simple one (Ironworks rushing Great Halls and Islands):
<player name="Ironworks Rush"
author="Anonymous"
description="No description available">
<type name="Bot"/>
<type name="UserCreated"/>
<type name="Province"/>
<type name="TwoPlayer"/>
<buy name="Ironworks"/>
<buy name="Great_Hall"/>
<buy name="Island"/>
</player>
-
What's the winrate? 100%?
Edit: Did it on the simulator. Big money still won a few games (it was a 99% winrate)
-
I added a Woodcutter (just one, prioritized over Silver) into the IGG/Gardens bot, and got a 100% win rate over 10,000 games. Then I ran it for 100,000 games and got 1 win for BMU, and 2 ties. I think that might be the best we can get. Without the Woodcutter it was also very good, but since there are no other actions in the deck there's no reason not to throw in a Woodcutter. I'm wondering if there is any other minor improvement that can be made to it?
Edit: With Nomad Camp instead of Woodcutter (but still getting a Woodcutter if it doesn't get $4 early), over 1 million games, I get 6 ties and 3 losses. I know the challenge is to find something that's guaranteed to beat BMU, but this is really really close. Maybe if there are just a few cases if awful shuffle luck vs. perfect shuffle luck it might be possible to track them down and force a win in another way just for those cases?
-
I'm pretty sure that a properly tuned Goons engine could win 100% against BMU. Unlike a rush, it doesn't depend on ending the game quickly, which means the only hope for BMU is to empty the Provinces ASAP, which isn't that fast even with perfect shuffle luck. Throw in a curser too and BMU is going to have a tough time.
I think a composition like Masquerade, Quarry, FV, Wharf, Witch, Goons, KC should do the trick. Maybe throw in some Worker's Villages too to rack up points and +buys. I can't test this now because my Java install is borked.
My submission to Simulation Tournament: Quints (http://forum.dominionstrategy.com/index.php?topic=2193.50) might do the trick, though Governors are not a great road to go if facing BMU with worst possible shuffle luck for you and best possible shuffle luck for BMU, because you'll tend to miss your Goons and cycle their deck.
-
If there's space, you could get Watchtower+Trader to not get bogged down by coppers you buy...
-
If there's space, you could get Watchtower+Trader to not get bogged down by coppers you buy...
You'll never need to buy coppers. By that point, you'll have more VP than exist as green cards. :)
-
How about a Bishop-Fortress Golden deck helped by Chapel?
BMU really takes a long, long time to clear out Provinces on his own, ~30 turns or something?
That's plenty of time for a deck like this to come together and get 12 VP per turn until BMU buys out every card but one in the kingdom (if it's programmed not to suicide). And then you buy the last card for the win.
-
Hermit/Market Square? Mega Poor House? Typically end the game on, like, turn 11-12. I expect there are lots of strategies with 100.000% win rates against BM.
-
I generally think that strategies that resolve around actually buying VP cards tend to lose at least 1 unlucky game against BM.
Not buying any VP cards, instead focusing on tokens, is better I think.
-
Well, if you only buy green cards on the turn you end the game, you aren't risking stalling.
-
Well, if you only buy green cards on the turn you end the game, you aren't risking stalling.
No, but you're risking BMU to get over half the VP before your final turn.
-
So using blueblimps code for the unluckiest path for BigMoney to 8 Provinces, I got the luckiest path for it and it's 16 turns.
CCCVV CCCCV
CCCVV SCCCC V
SSCCV SCCCC GVV
GCVVV SSSCC GGCCC
SSSCC GSCCC GGCCV VVV
VVVVV SSSCC GSCCC GGCCV
3 4 3 6 6 6 4 8 9 8 8 8 0 8 8 8
So that's the time you have when you always want to beat BM...
-
Many ways to go here, but I think the safest one is Thief.
Ironworks, Chapel, Wishing Well, Throne Room, Thief, Monument, Smithy.
If you really want to overdo it, you could add Highway, Laboratory, Kings Court.
Edit: I played a bit around with the simulator and got it to win 100% over 100.000 games.
But oh boy is that bot stupid! I just can't get him to trash his coppers. It says 'Agressive trashing' on chapel but still he loves to keep some coppers around.
In the end I just allowed him to steal some golds and only then he rids himself of the coppers.
Starting out with some wishing well's really helps, mainly because the bot knows how to play them.
I removed the throne rooms, because of the opposite reason.
<player name="Steal da money"
author="Stef"
description="No description available">
<type name="UserCreated"/>
<type name="Bot"/>
<type name="TwoPlayer"/>
<type name="Province"/>
<board contents="King's Court, Village, Ironworks, Chapel, Laboratory, Smithy, Thief, Wishing Well, Throne Room, Witch" bane="null"/>
<buy name="Ironworks">
<condition>
<left type="countCardsInDeck" attribute="Ironworks"/>
<operator type="equalTo" />
<right type="constant" attribute="0.0"/>
</condition>
</buy>
<buy name="Chapel" strategy="aggressiveTrashing">
<condition>
<left type="countCardsInDeck" attribute="Chapel"/>
<operator type="equalTo" />
<right type="constant" attribute="0.0"/>
</condition>
</buy>
<buy name="Wishing_Well">
<condition>
<left type="countCardsInDeck" attribute="Wishing_Well"/>
<operator type="smallerThan" />
<right type="constant" attribute="3.0"/>
</condition>
</buy>
<buy name="Village">
<condition>
<left type="countCardsInDeck" attribute="Village"/>
<operator type="equalTo" />
<right type="constant" attribute="0.0"/>
</condition>
</buy>
<buy name="Smithy">
<condition>
<left type="countCardsInDeck" attribute="Smithy"/>
<operator type="smallerThan" />
<right type="countCardsInDeck" attribute="Village"/>
<extra_operation type="divideBy" attribute="3.0" />
</condition>
</buy>
<buy name="Thief">
<condition>
<left type="countCardsInDeck" attribute="Thief"/>
<operator type="equalTo" />
<right type="constant" attribute="0.0"/>
</condition>
</buy>
<buy name="Worker$s_Village">
<condition>
<left type="countCardsInDeck" attribute="Worker$s_Village"/>
<operator type="equalTo" />
<right type="constant" attribute="0.0"/>
</condition>
</buy>
<buy name="Village">
<condition>
<left type="countCardsInDeck" attribute="Village"/>
<operator type="smallerThan" />
<right type="constant" attribute="5.0"/>
</condition>
</buy>
<buy name="Thief">
<condition>
<left type="countCardsInDeck" attribute="Thief"/>
<operator type="smallerThan" />
<right type="constant" attribute="3.0"/>
</condition>
</buy>
<buy name="Monument">
<condition>
<left type="countCardsInDeck" attribute="Monument"/>
<operator type="smallerThan" />
<right type="constant" attribute="2.0"/>
</condition>
</buy>
<buy name="Laboratory">
<condition>
<left type="countCardsInDeck" attribute="Laboratory"/>
<operator type="smallerThan" />
<right type="constant" attribute="6.0"/>
</condition>
</buy>
<buy name="Village">
<condition>
<left type="countCardsInDeck" attribute="Village"/>
<operator type="smallerThan" />
<right type="constant" attribute="9.0"/>
</condition>
</buy>
<buy name="Gold">
<condition>
<left type="countCardsInDeck" attribute="Gold"/>
<operator type="smallerThan" />
<right type="constant" attribute="2.0"/>
</condition>
</buy>
<buy name="Silver">
<condition>
<left type="countCardsInDeck" attribute="Gold"/>
<operator type="equalTo" />
<right type="constant" attribute="0.0"/>
</condition>
<condition>
<left type="countCardsInDeck" attribute="Silver"/>
<operator type="equalTo" />
<right type="constant" attribute="0.0"/>
</condition>
</buy>
</player>
-
So that's the time you have when you always want to beat BM...
Unless you have attacks. I think Chapel down to FV/Torturer would be a good start for the Poor House engine.
-
If you are starting player, you can get to more points than DStu's optimal BMU bot without attacks. Assuming worst shuffle luck here, i.e., Cha and IW always as late as possible and always collide.
In the beginning I use For just as a cantrip, so I don't display its plays
1/2 [Cha/IW]
3/4 [For]
5/6/7 IW(xFor),Cha (/CCC/)
8 IW(xFor),Cha (/CCC/)
9 IW(xBi),Cha (/CEE/)
At this point my deck is down to: Cha,IW,Bi,E,3xFor, so that I can play my whole deck every turn.
10 For,Bi(/For/),For,For,IW(xBi),Cha(/E/) VP:3
11 For,Bi(/Cha/),For,Bi(/For/),IW(xBi) [Ped] VP:8
12 Ped,For,Bi(/IW/),For,Bi(/For/),Bi(/For/) [For] VP:17
13 Ped,For,Bi(/For/),For,Bi(/For/),Bi(/For/) [Bi] VP:26
now the golden deck is completed, every turn I gain 12 VP, so that I have 62 VP by turn 16.
Probably this can be optimized further, so that one wins, when the BMU player starts as well.
-
Too bad Fortress isn't implemented in Dominiate, I really wanted to try fiddling with it.
And too bad there's never been an official update of Geronimoo's sim even though he's sneakily adding the cards for himself!
-
Most of the bots entered for the quints simulation tournament get 99%+ against big money.
This puzzle also reminds me of the 'lucky chancellor' problem where you had to design a deck that could beat the chancellor on best draws even though your deck had the worst possible draws. I think I came up with something like chapel/ironworks gaining workers villages and a remodel, switching up to goons, and then buying out the peddler pile. I doubt a simulator could do the card play for that one though.
-
What about a KC-KC-Goons-Masquerade pin? With Chapel trashing things, you might be able to get it set up quick enough. Although you can't get the simulator to trash everything, I've tried...
-
You have to guarantee setting it up quickly enough, in the worst case scenario.
-
I'm curious as to what the chances are that the simulator will ever actually run the worse-case scenario (and best-case scenario for BMU). Even across 100,000 games, I'm assuming that the number of possible games FAR excedes that. Even if you ran over a million games, I doubt the sim would ever put BMU's best possible game against the submitted solution's worst possible game.
-
Indeed; once the probabilities are 100% to a few decimal places, you need a new method. Exhaustive search is one, it doesn't have to be brute-force. Some people will throw about big factorial numbers but these are nonsense when most of your cards are not unique.
-
I have not thought it through completely, but I think you can guarantee to get your deck down to Chapel-Ironworks-Hamet-Scheme-Bridge in 9 turns when you start Chapel-Scheme, and from there it should take 5 turns for a Highway-Market Square megaturn into 13 Duchy-Dukes, which beats 8 Provinces, or one more for additional 10 Markets if you want to pile out.
-
I have not thought it through completely, but I think you can guarantee to get your deck down to Chapel-Ironworks-Hamet-Scheme-Bridge in 9 turns when you start Chapel-Scheme, and from there it should take 5 turns for a Highway-Market Square megaturn into 13 Duchy-Dukes, which beats 8 Provinces, or one more for additional 10 Markets if you want to pile out.
Remember that BMU will be buying some Duchies in the meantime though.
-
I have not thought it through completely, but I think you can guarantee to get your deck down to Chapel-Ironworks-Hamet-Scheme-Bridge in 9 turns when you start Chapel-Scheme, and from there it should take 5 turns for a Highway-Market Square megaturn into 13 Duchy-Dukes, which beats 8 Provinces, or one more for additional 10 Markets if you want to pile out.
Remember that BMU will be buying some Duchies in the meantime though.
3 4 3 6 6 6 4 8 9 8 8 8 0 8 8 8
BMU is busy buying Provinces....
-
I have not thought it through completely, but I think you can guarantee to get your deck down to Chapel-Ironworks-Hamet-Scheme-Bridge in 9 turns when you start Chapel-Scheme, and from there it should take 5 turns for a Highway-Market Square megaturn into 13 Duchy-Dukes, which beats 8 Provinces, or one more for additional 10 Markets if you want to pile out.
Remember that BMU will be buying some Duchies in the meantime though.
3 4 3 6 6 6 4 8 9 8 8 8 0 8 8 8
BMU is busy buying Provinces....
Ah right, it's "best-case" BMU, so no Duchies.
-
I have not thought it through completely, but I think you can guarantee to get your deck down to Chapel-Ironworks-Hamet-Scheme-Bridge in 9 turns when you start Chapel-Scheme, and from there it should take 5 turns for a Highway-Market Square megaturn into 13 Duchy-Dukes, which beats 8 Provinces, or one more for additional 10 Markets if you want to pile out.
Remember that BMU will be buying some Duchies in the meantime though.
3 4 3 6 6 6 4 8 9 8 8 8 0 8 8 8
BMU is busy buying Provinces....
Ah right, it's "best-case" BMU, so no Duchies.
Well, if you're going for a Duke-Duchy strategy, then the best case might not be the one which always hits $8. The challenge is to find an algorithm that guaranteed beats BMU, which means that it has to beat it in every possible case, not just in some pre-determined "best case". So the "best case" might vary depending on what strategy it is competing against.
-
Best case BMU doesn't deal with attacks though.
Adding even a single Militia may severely hamper it.
It will start buying Duchies and slow down to a crawl until (Estates and) Duchies are gone and only then it starts buying some economy again.
-
Best case BMU doesn't deal with attacks though.
Adding even a single Militia may severely hamper it.
It will start buying Duchies and slow down to a crawl until (Estates and) Duchies are gone and only then it starts buying some economy again.
It shouldn't; BMU doesn't buy Duchies until at least four Provinces are gone.
-
Well, eventually it will get to 4 Provinces, but have trouble buying more and it drops down to Duchies and keeps buying them until he has all 8. At that point he will start with his regular Gold and Silver again, but it will take a lot longer to get the game to completion this way.
Enough time, even in best case perhaps, for a VP-token based strategy to overcome it with a worst case scenario.
-
Well, eventually it will get to 4 Provinces, but have trouble buying more and it drops down to Duchies and keeps buying them until he has all 8. At that point he will start with his regular Gold and Silver again, but it will take a lot longer to get the game to completion this way.
I think a strategy that hits it with Militia two out of three turns (or more) is going to delay Provinces for a very long while; the BMU bot has to be holding GGSXX or better to grab a Province. That sort of density will take a long time. GSSXX just gets it another Gold.
-
Yes, and we're talking about a simple Militia here.
You could even Minion-attack first, then play Militia and it has to get lucky to have a 4-card hand of GGSX instead of GGSXX.
-
Yes, and we're talking about a simple Militia here.
You could even Minion-attack first, then play Militia and it has to get lucky to have a 4-card hand of GGSX instead of GGSXX.
But requiring luck is not a problem if you assume perfect luck for the opponent. I think Minion is a terrible card to use, since it will only discard hands full of provinces, estates, and maybe coppers.
-
Well, if we assume perfect luck, Thief always hits 2 VP cards, so that's out as well.
-
Well, if we assume perfect luck, Thief always hits 2 VP cards, so that's out as well.
Unless you're playing more than one Thief per reshuffle.
-
Well, if we assume perfect luck, Thief always hits 2 VP cards, so that's out as well.
Unless you're playing more than one Thief per reshuffle.
Still I think you could model it that it's pretty bad even when played multiple times.
Remember, we not only have to account for perfect luck for our opponent, but for the worst possible luck for ourselves. So Throne Rooms never find other action cards and terminals always collide.
-
Well, if we assume perfect luck, Thief always hits 2 VP cards, so that's out as well.
Unless you're playing more than one Thief per reshuffle.
Still I think you could model it that it's pretty bad even when played multiple times.
Remember, we not only have to account for perfect luck for our opponent, but for the worst possible luck for ourselves. So Throne Rooms never find other action cards and terminals always collide.
See, and the problem with this sort of exercise is that the non-BMU player's luck and the BMU bot's luck are entangled. Rearranging the bot's deck to account for the player's deck changes what is the "worst" for the player's deck, forcing that to be rearranged, which means the bot's deck has to be rearranged, and so on.
-
The best strategy might involve using Bureaucrat or Cutpurse to see the opponent's hand, so you see wetehr he has a hand you want to discard with Minion. This is kinda twisting my mind.
-
I have not thought it through completely, but I think you can guarantee to get your deck down to Chapel-Ironworks-Hamet-Scheme-Bridge in 9 turns when you start Chapel-Scheme, and from there it should take 5 turns for a Highway-Market Square megaturn into 13 Duchy-Dukes, which beats 8 Provinces, or one more for additional 10 Markets if you want to pile out.
Remember that BMU will be buying some Duchies in the meantime though.
3 4 3 6 6 6 4 8 9 8 8 8 0 8 8 8
BMU is busy buying Provinces....
Ah right, it's "best-case" BMU, so no Duchies.
Well, if you're going for a Duke-Duchy strategy, then the best case might not be the one which always hits $8. The challenge is to find an algorithm that guaranteed beats BMU, which means that it has to beat it in every possible case, not just in some pre-determined "best case". So the "best case" might vary depending on what strategy it is competing against.
I need one more turn to pile Provinces, so I doubt that taking Duchy>Provinces will help BMU much. We could also take Fairgrounds in the Kingdom, which will be easily worth 6VP when I have 13buys.
-
The best strategy might involve using Bureaucrat or Cutpurse to see the opponent's hand, so you see wetehr he has a hand you want to discard with Minion. This is kinda twisting my mind.
Is the best hand against Cutpurse + Minion a hand with a Copper plus enough money to buy whatever he wants, or a hand without copper?
-
For everyone who thinks there is a problem with this challenge, because "worst possible luck" is ill-defined, there's no problem with that, mathematically speaking. The challenge is "just" to give an strategy for every possible way you and your opponent can shuffle. Many cases will be easy (if you are "lucky"), and there will be some cases which are hard (if you have "bad luck"). There won't be one unique way both players can shuffle which can be called the "most difficult case", so the term "worst possible luck" is indeed not well-defined. There will be different ways to which the players can shuffle, which each bring their own set of complications.
Unfortunately, handling every possible case is quite hard. Giving a separate strategy for every different case is unfeasible. You can group similar shuffles together and give a strategy which works for all of these shuffles. Usually you will lose some efficiency with this, but this might work.
By the way, I'm convinced that you can win 100% of the times against any big money strategy, if the board is strong enough. It should be possible to guarantee to build up a masquerade pin by turn 15 (and it's impossible for big money to get to 8 provinces earlier). If that's not possible, then it should be possible a few turns later, but attacking the big money player enough that he can't get to 8 provinces before that time. It will be very annoying to show that such a strategy will always work, though.
-
http://forum.dominionstrategy.com/index.php?topic=462.msg7678#msg7678
-
http://forum.dominionstrategy.com/index.php?topic=462.msg7678#msg7678
Well then. Next question!
-
I thought the point of this thread was to consider strategies that existing simulators can play. (To be fair, in principle you can tweak a Dominiate AI to do anything.)
-
I thought the point of this thread was to consider strategies that existing simulators can play. (To be fair, in principle you can tweak a Dominiate AI to do anything.)
It's the "never wins a single game" that forces us into non-bot territory, as people have already demonstrated effective 100% win rates. However, I'm sure the anti-Lucky Chancellor not is readily programmable. Since it's a guaranteed win against LC, which will always be faster than BMU, it's a guaranteed 100% win vs BMU too.
-
I thought the point of this thread was to consider strategies that existing simulators can play. (To be fair, in principle you can tweak a Dominiate AI to do anything.)
It's the "never wins a single game" that forces us into non-bot territory, as people have already demonstrated effective 100% win rates. However, I'm sure the anti-Lucky Chancellor not is readily programmable. Since it's a guaranteed win against LC, which will always be faster than BMU, it's a guaranteed 100% win vs BMU too.
Well, one assumption made was that 3 piling was not a worry because the LC bought almost nothing but Provinces. With a less lucky Chancellor, it at least needs to be addressed that Duchies and Estates may be purchased sooner. I doubt that will matter much though.