Chris Martin
England
London
flag msg tools
mbmbmbmbmb
Prolegomenon
In 1921, the Czech playwright Karel Čapek released a play entitled "R.U.R.". In it is described how the mad scientist Rossum created an automaton which would work endlessly by following simple orders. These "roboti", meaning "hard workers", went on to destroy humanity, and so the robot was born. Little did Čapek know that 88 years later his fantastical tale would come true, in the unassuming guise of a badly-drawn man with a silly hat and a downright ridiculous beard.

Build-up
On Friday night, my flatmate Paul and I went round to our friend Richard's house to play some Dominion. Richard, Paul and I usually beat other people against whom we play Dominion, and for shorthand I will describe this state of affairs as "We are at least quite good at Dominion." I was excited to introduce to them the Envoy cards that I had recently received from BGG.
"Here's a tip, guys," I advised them, "I've found that these can be quite a good first turn buy." We played a number of games, had a good time, and late that night Paul and I caught the last Tube home. On the Tube I asked Paul what he'd thought of the Envoy.
"It's alright," he said, Kiwi accent filling the carriage, "but I think it plays quite robotically. Remember that last game we played with it, where I won pretty easily? I only had two Envoys and just bought money or VPs other than that."
"Ahh," I said, "so it's basically a version of my Smithy game." I had embarrassed Paul and Richard some time back by handily beating them with a deck that contained just two Smithies and sackfuls of Gold, against their more interesting but less effective "supercombos". What can I say - I'm a minimalist.
Paul nodded. "Yeah," he agreed, "but it's even more powerful than that. And I don't like that style of play. Honestly, where's the skill? A robot could do it." Paul loves Action cards. (I love big Treasure cards and Richard loves, well, Richard loves Chapels.) I hadn't picked up on his first use of the word 'robot', but I did this time.
"All right then," I countered, "let's create an Envoy robot using a few simple rules and we'll play against it. We'll let it buy up to two Envoys and set the others aside, and we'll play with ten random decks and try to beat it." The road to hell is paved with good intentions.

The algorithm
The algorithm followed a few simple rules. Firstly, it should always buy Provinces when it could. We could have tried for some subtlety about how many cards were in its deck, but we figured that we could save that tweak for later, if the robot seemed to need it. Secondly, it should buy Gold when it had 6 or 7 cash. Thirdly, it should buy a second Envoy if it didn't have one yet and it had 4 or 5 cash. Paul and I had a bit of a debate about the order of points one and two: Paul originally wanted it to pick up its second Envoy ASAP, even at the cost of a Gold, but I convinced him that as it didn't really need its second Envoy until later anyway, it should always buy Gold even as a priority over its second Envoy if it could. Fourthly, it should buy Silver if it had 3 cash, or if it had 4 or 5 cash and a second Envoy already. Fifthly, it should buy nothing if it had less than 3 cash. In between the first and second rules, it should play an Envoy card if it had one in hand and had an Action with which to use it. This was part of the principle of "Always buy a Province!"
After our first three games, we included three "kicks", points at which it would change its programming to reflect the approaching endgame. As soon as only six cards needed to be bought to trigger the end of the game (usually Provinces), it would start buying Duchies on 5 cash instead of a Silver. As soon as only four cards needed buying, it would start buying Duchies on 6 or 7 cash instead of a Gold. At two cards, it would simply buy the highest VP card it could afford.

As you can see, it was extremely simple. Even with my hedging, caveating and explanatory notes, it has been described in two fairly short paragraphs. There follows a more detailed treatment (including rules for the variant "equal turns" ending which we play as standard). It is a bit abstract and unless you are thinking of creating a computer program to play this strategy I suggest that you skip it. Obviously, a "Buy" command indicates a stop.

Where t is the amount of Treasure in hand, h is a binary function expressing the state of having an Envoy card in hand and an Action with which to play it and x is a count of how many Envoys are in the deck,
1. IF t ≥ 8 THEN Buy Province
ELSE GOTO 2
2. IF h = 1 THEN Play Envoy AND GOTO 1
ELSE GOTO 3
3. IF t ≥ 6 THEN Buy Gold
ELSE GOTO 4
4. IF t ≥ 4 AND x < 2 THEN Buy Envoy
ELSE GOTO 5
5. IF t ≥ 3 THEN Buy Silver
ELSE Pass


And with the kicks added, where p is a binary function expressing whether Provinces remain to be bought, d is a binary function expressing whether Duchies remain to be bought, e is a binary function expressing whether Estates remain to be bought and c is how many cards need to be bought for the game to end,
1. IF t ≥ 8 AND p = 1 THEN Buy Province
ELSE GOTO 2
2. IF t ≥ 8 AND d = 1 THEN Buy Duchy
ELSE GOTO 3
3. IF t ≥ 8 AND e = 1 THEN Buy Estate
ELSE GOTO 4
4. IF t ≥ 8 THEN Pass
ELSE GOTO 5
5. IF h = 1 THEN Play Envoy AND GOTO 1
ELSE GOTO 6
6. IF t ≥ 6 AND c > 4 THEN Buy Gold
ELSE GOTO 7
7. IF t ≥ 6 AND d = 1 THEN Buy Duchy
ELSE GOTO 8
8. IF t ≥ 6 AND c > 2 THEN Buy Gold
ELSE GOTO 9
9. IF t ≥ 6 AND e = 1 THEN Buy Estate
ELSE GOTO 10
10. IF t ≥ 6 THEN Buy Gold
ELSE GOTO 11
11. IF t ≥ 5 AND c > 6 AND x < 2 THEN Buy Envoy
ELSE GOTO 12
12. IF t ≥ 5 AND c > 6 THEN Buy Silver
ELSE GOTO 13
13. IF t ≥ 5 AND d = 1 THEN Buy Duchy
ELSE GOTO 14
14. IF t ≥ 5 AND c > 2 AND x < 2 THEN Buy Envoy
ELSE GOTO 15
15. IF t ≥ 5 AND c > 2 THEN Buy Silver
ELSE GOTO 16
16. IF t ≥ 5 AND e = 1 THEN Buy Estate
ELSE GOTO 17
17. IF t ≥ 5 THEN Buy Silver
ELSE GOTO 18
18. IF t ≥ 4 AND c > 2 AND x < 2 THEN Buy Envoy
ELSE GOTO 19
19. IF t ≥ 3 AND c > 2 THEN Buy Silver
ELSE GOTO 20
20. IF t ≥ 3 AND e = 1 THEN Buy Estate
ELSE GOTO 21
21. IF t ≥ 3 THEN Buy Silver
ELSE GOTO 22
22. IF t ≥ 2 AND e = 1 THEN Buy Estate
ELSE Pass


The algorithm is noticeably more complex, but a lot of that is just to deal with buying cheaper cards when the Province and/or Duchy decks have run out. And it is still only 22 steps long.

The trial and a request
We played a total of 9 games against the robot, 3 using the first ruleset and 6 using the second. Of these 9, I won 3 and the robot won 6. The presence or otherwise of any "supercards" did not seem to have an impact. (I should mention that we played the variant "everybody picks their opening split" rule, and that the robot always therefore bought an Envoy and a Silver.) I then took it to a big club day on Sunday and the robot smashed everybody (I came top of the human players when the robot played, and I won every human-only game of Dominion played that day but one).

I know that you're thinking, "This guy must be pretty rubbish at Dominion!" and I won't tell you you're wrong. Maybe I really am just rubbish at Dominion and should up my game. Maybe not only am I rubbish, but everyone I know is even worse. Maybe. Or maybe the robot really is that good.

So, my request. Copy out the rules (or learn them - they're not hard) and play some games against the robot. Set aside two Envoys for the robot to use and don't let anyone else take an Envoy. Follow the rules of the program exactly, and play a dozen or so games to allow any luck to even out. Then post your results in here. If you are beating the robot reliably, let me know how you're doing it. If you're not, then maybe we need to start building the First Law of Robotics into our boardgames.

Thank you for reading.
8 
 Thumb up
0.25
 tip
 Hide
  • [+] Dice rolls
Sam Lindsay-Levine
United States
St. Paul
Minnesota
flag msg tools
mbmbmbmbmb
Why don't you let anyone else take an Envoy...? Without that, it doesn't seem like a meaningful experiment. You might as well show that if you let a robot have a Chapel when nobody else can have one, it will have a big advantage, which I don't think will be a big surprise.

Surely if you want to test the robot, you have to make Envoy and 9 other random kingdom cards available for purchase.
7 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Daniel Winterhalter
United States
Broomfield
Colorado
flag msg tools
mbmbmbmbmb
chrisjwmartin wrote:

The algorithm is noticeably more complex, but a lot of that is just to deal with buying cheaper cards when the Province and/or Duchy decks have run out.


This might have been just a mistype on your part, but you were ending the game when the Province deck ran out, correct?

Regardless, I've found the Envoy/Smithy/Council Room + nothing but money deck to be fairly effective- unless there are Thieves about. I'll try and play a few against your robot and see how it goes this evening; it'll be entertaining!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Joe Rickard
United States
Los Angeles
California
flag msg tools
mbmbmbmbmb
SamLL wrote:
Why don't you let anyone else take an Envoy...? Without that, it doesn't seem like a meaningful experiment. You might as well show that if you let a robot have a Chapel when nobody else can have one, it will have a big advantage, which I don't think will be a big surprise.

Surely if you want to test the robot, you have to make Envoy and 9 other random kingdom cards available for purchase.


I'm guessing the point of the robot is to show whether or not when the envoy is in play are there better options than just "playing like a robot" and just buying two envoys coins and VPs. If you could buy envoys it wouldn't prove much other than the envoy is a good card. That is apparent from just reading the text on the card.

I don't think that a robot that was limited to only buying a chapel or two coins and VPs would do nearly as well. I think it could win sometimes, but not very consistently.

1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Chris Martin
England
London
flag msg tools
mbmbmbmbmb
Trekspert1 wrote:
you were ending the game when the Province deck ran out, correct?
Hi Daniel,

Like I noted, we were "including rules for the variant "equal turns" ending which we play as standard". I.e., if the start player buys the last Province, everyone else still gets a turn. It makes it a bit fairer as otherwise the start player has a big advantage. Because of this, we needed the extra steps in our algorithm. If you don't play that variant you won't need so many steps.

SamLL wrote:
Why don't you let anyone else take an Envoy...? Without that, it doesn't seem like a meaningful experiment. You might as well show that if you let a robot have a Chapel when nobody else can have one, it will have a big advantage, which I don't think will be a big surprise.

Surely if you want to test the robot, you have to make Envoy and 9 other random kingdom cards available for purchase.
Hi Sam,

Three points. Firstly, I removed the extra Envoys as a control. If you want to try using them in a more varied strategy, please go ahead. However we will still have demonstrated that the Envoy is very powerful, which I think is an interesting result.

Secondly, I can usually beat a Chapel deck, and I don't play Chapels. Perhaps my opponents can't play them very well, though. This leads into point three.

Thirdly, the Chapel deck requires skill. If you can program a robot to play a Chapel game with no outside assistance using just five steps, then you're a better man than I am. And this is the main point at which I was driving: this Envoy strategy takes next to no skill. A rank amateur with the memory of a goldfish could pick it up and suddenly be a serious competitor. To play a Chapel deck requires experience and finesse. What's more, everything that Thieves and Witches and Bureaucrats and Spies do to the Envoy deck, they seem to do much more so to the Chapel deck.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Sam Lindsay-Levine
United States
St. Paul
Minnesota
flag msg tools
mbmbmbmbmb
chrisjwmartin wrote:
Three points. Firstly, I removed the extra Envoys as a control. If you want to try using them in a more varied strategy, please go ahead. However we will still have demonstrated that the Envoy is very powerful, which I think is an interesting result.


Maybe I'm confused as to what the hypothesis being tested here is. I thought the hypothesis you wanted to test was "a robot can play the Envoy deck as well as / better than a human", in which case the only variable you should change is the play algorithm. If instead you're testing "a player with access to Envoy does better than a player without", well, that isn't very meaningful to the actual game of Dominion since that situation does not arise.

What is your hypothesis that you are attempting to show?

chrisjwmartin wrote:
Secondly, I can usually beat a Chapel deck, and I don't play Chapels. Perhaps my opponents can't play them very well, though. This leads into point three.


Hmm, my group's consensus (and, I thought, the BGG consensus) is that the Chapel is the most game-warping card in that when it is present, it is profoundly difficult to beat without also going Chapel.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Chris Martin
England
London
flag msg tools
mbmbmbmbmb
SamLL wrote:
Maybe I'm confused as to what the hypothesis being tested here is. I thought the hypothesis you wanted to test was "a robot can play the Envoy deck as well as / better than a human", in which case the only variable you should change is the play algorithm. If instead you're testing "a player with access to Envoy does better than a player without", well, that isn't very meaningful to the actual game of Dominion since that situation does not arise.

What is your hypothesis that you are attempting to show?
Ah. This was a multi-faceted experiment! Yes, in part we were aiming to show that the Envoy could be played robotically. But for that to be meaningful, it had to be a strong strategy: after all, I could write a five-line algorithm using any card, but it's the fact that this algorithm can win and win routinely that is interesting.

I suppose the main point is that including the Envoy seems to bring an otherwise subtle and skilful game down to the level of a five-step algorithm. If I am faced with a set including an Envoy and I want to win, why would I not pick the boring algorithm that can beat me seven times out of ten rather than the clearly inferior strategies composed of playing other cards? But perhaps you're right and there are strategies to routinely beat the robot that involve using the Envoy card. In which case, try them out at home and come back and tell us your mixed Envoy strategy that beats the robot back into its place!

SamLL wrote:
Hmm, my group's consensus (and, I thought, the BGG consensus) is that the Chapel is the most game-warping card in that when it is present, it is profoundly difficult to beat without also going Chapel.
I think that this is a discussion for another day. Do you at least agree that playing a Chapel deck still takes a degree of skill? If so, then we agree on the main point: the Envoy robot is (probably) the strongest strategy available for that level of simplicity.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
David desJardins
United States
Burlingame
California
flag msg tools
mbmbmbmbmb
SamLL wrote:
If instead you're testing "a player with access to Envoy does better than a player without", well, that isn't very meaningful to the actual game of Dominion since that situation does not arise.


It's pretty interesting to me, as my current hypothesis is, "The Envoy is rarely worth buying." So if the statement were true then it would certainly shed light on this question.
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dave Kudzma
United States
Millsboro
Delaware
flag msg tools
Trumpy, you can do stupid things!
mbmbmbmbmb
DaviddesJ wrote:
SamLL wrote:
If instead you're testing "a player with access to Envoy does better than a player without", well, that isn't very meaningful to the actual game of Dominion since that situation does not arise.


It's pretty interesting to me, as my current hypothesis is, "The Envoy is rarely worth buying." So if the statement were true then it would certainly shed light on this question.


I've been using the Envoy since the day they had pics up for me to proxy. They often get used, and are VERY often beneficial. The trick is either getting high coin saturation or having a spare action to make the choice of which card to discard painful. They actually gain in power as the game progresses.

There's a wonderful MtG card comparable, but perhaps still not as good called Fact or Fiction, with which you draw the top 5 cards of your deck, break them into 2 piles, and let an opponent choose which on you keep and which you discard.
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Chris Martin
England
London
flag msg tools
mbmbmbmbmb
quartersmostly wrote:
I don't think that a robot that was limited to only buying a chapel or two coins and VPs would do nearly as well. I think it could win sometimes, but not very consistently.
I agree, Joe: but more than that, I don't think that a robot could reliably play a Chapel deck. When you play Chapel there are often some tough early-midgame decisions to be made about when to keep trashing and when to start ratcheting up. The kind of programming to make those subtle decisions would be way more complex than anything here.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Chris Martin
England
London
flag msg tools
mbmbmbmbmb
DaviddesJ wrote:
SamLL wrote:
If instead you're testing "a player with access to Envoy does better than a player without", well, that isn't very meaningful to the actual game of Dominion since that situation does not arise.
It's pretty interesting to me, as my current hypothesis is, "The Envoy is rarely worth buying." So if the statement were true then it would certainly shed light on this question.
Then you're almost the perfect person to do the trial the original way! Let us know how it goes, is all I ask. Yes, even if it beats you...
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Chris Martin
England
London
flag msg tools
mbmbmbmbmb
locusshifter wrote:
I've been using the Envoy since the day they had pics up for me to proxy. They often get used, and are VERY often beneficial. The trick is either getting high coin saturation or having a spare action to make the choice of which card to discard painful. They actually gain in power as the game progresses.
We figured that a mixed Envoy deck would be better than a pure Envoy deck - until the players playing pure decks convincingly beat the players playing mixed decks. maybe that was just chance. But sure, it'd be interesting to see whether your mixed strategies could beat the pure strategy - remember to post up your findings!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
David desJardins
United States
Burlingame
California
flag msg tools
mbmbmbmbmb
chrisjwmartin wrote:
Then you're almost the perfect person to do the trial the original way! Let us know how it goes, is all I ask. Yes, even if it beats you...


Unfortunately, it seems like too much effort. I'll be interested to hear what others think, though.

I take it this strategy is supposed to be applied unaltered even if there are Witches in the game, is that right?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Chris Martin
England
London
flag msg tools
mbmbmbmbmb
DaviddesJ wrote:
Unfortunately, it seems like too much effort. I'll be interested to hear what others think, though.
It really isn't any effort, especially if you just do the basic one. Buy Provinces when you have 8+. Buy Gold when you have 6 or 7. Buy up to two Envoys when you have 4 or 5, otherwise when you have 3 to 5 buy Silver. If you have less, buy nothing. Simple! Once you've got that worked out you can start adding in the rules for buying Duchies and Estates.

DaviddesJ wrote:
I take it this strategy is supposed to be applied unaltered even if there are Witches in the game, is that right?
Yes, it is. In the games with Witches, we found that they slowed the Envoy down a bit but not enough.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Yaron Racah
Israel
Tel Aviv
flag msg tools
mbmbmbmb
Dominion: Envoy Promo Card » Forums » Strategy
Re: Rossum's Universal Envoys, OR The Success of an Algorithm Employing Just Two Envoy Cards
Why make 10 experiments when you can make 10,000?

Here's the thing: while I have no experience with the Envoy, I've played many games with and against a similar Robo-Smith strategy (just use Smithy insteand of Envoy). My impression, based on those games, is that this strategy presents a credible threat, but that in most setups, you can do better.

So, for me, the question is this:

Is Robo-Envoy much more powerful than Robo-Smith?

I tested this by letting my PC play solitaire games of Robo-Envoy and Robo-Smith, based on the parameters you outlined. For each strategy, I checked the average number of turns needed to get to 4 Provinces. Obviously, a lot depends on late Duchy and Estate purchases, but I assume that the faster Province-gobbler can get those just as well as the slower one.

The main "bug" in the strategy is grabbing a Province the first time you can, rather than getting some Gold first, but I did that for both Smithy and Envoy, so hopefully it doesn't skew the results.

I ran each strategy 10,000 times, so the 1st digit after the decimal point is accurate.

Results:

Pure money Robot: 16.8 turns

1-Smithy Robot: 14.9 turns
1-Envoy Robot: 14.6 turns

2-Smithy Robot: 14.9 turns
2-Envoy Robot: 15.0 turns


As you implicitly assumed, robots that went for 3-5 Smithy/Envoys did worse than their 1-2 counterparts. I won't bother copying the specific numbers.

An early Smithy/Envoy can shave 2 turns off, compared to just money. 1 Envoy does a little better than 1 Smith, but I was surprised to see how small the difference is.

I'm also surprised that the 2nd Smithy is unhelpful, and the 2nd Envoy even detrimental. I suppose the very best strategy is a mix: buy the 1st Smithy/Envoy immediately, and pad your deck with some money before getting the 2nd. I don't think it would do much better than just buying the one Smithy/Envoy, though.

It actually does make sense that the 2nd Envoy is more problematic than the 2nd Smith. An Envoy is likelier to draw into its colleague (turning it into a dead draw) than a Smithy is.

Anyway, the main result is that Robo-Envoy isn't much more dangerous than Robo-Smith. Since I know I can handle the later, I'm not worried about the first. :-)
18 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Chris Martin
England
London
flag msg tools
mbmbmbmbmb
yaron wrote:
Why make 10 experiments when you can make 10,000?

Here's the thing: while I have no experience with the Envoy, I've played many games with and against a similar Robo-Smith strategy (just use Smithy insteand of Envoy). My impression, based on those games, is that this strategy presents a credible threat, but that in most setups, you can do better.

So, for me, the question is this:

Is Robo-Envoy much more powerful than Robo-Smith?

I tested this by letting my PC play solitaire games of Robo-Envoy and Robo-Smith, based on the parameters you outlined. For each strategy, I checked the average number of turns needed to get to 4 Provinces. Obviously, a lot depends on late Duchy and Estate purchases, but I assume that the faster Province-gobbler can get those just as well as the slower one.

The main "bug" in the strategy is grabbing a Province the first time you can, rather than getting some Gold first, but I did that for both Smithy and Envoy, so hopefully it doesn't skew the results.

I ran each strategy 10,000 times, so the 1st digit after the decimal point is accurate.

Results:

Pure money Robot: 16.8 turns

1-Smithy Robot: 14.9 turns
1-Envoy Robot: 14.6 turns

2-Smithy Robot: 14.9 turns
2-Envoy Robot: 15.0 turns


As you implicitly assumed, robots that went for 3-5 Smithy/Envoys did worse than their 1-2 counterparts. I won't bother copying the specific numbers.

An early Smithy/Envoy can shave 2 turns off, compared to just money. 1 Envoy does a little better than 1 Smith, but I was surprised to see how small the difference is.

I'm also surprised that the 2nd Smithy is unhelpful, and the 2nd Envoy even detrimental. I suppose the very best strategy is a mix: buy the 1st Smithy/Envoy immediately, and pad your deck with some money before getting the 2nd. I don't think it would do much better than just buying the one Smithy/Envoy, though.

It actually does make sense that the 2nd Envoy is more problematic than the 2nd Smith. An Envoy is likelier to draw into its colleague (turning it into a dead draw) than a Smithy is.

Anyway, the main result is that Robo-Envoy isn't much more dangerous than Robo-Smith. Since I know I can handle the later, I'm not worried about the first. :-)
Interesting! How did you set it up?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Yaron Racah
Israel
Tel Aviv
flag msg tools
mbmbmbmb
chrisjwmartin wrote:
Interesting! How did you set it up?


You mean in terms of the programming language? I used Matlab. I didn't simulate Dominion in any general way, just this specific scenario.
It's no more than 150 lines - if that kind of thing is up your alley, I can send you the file.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Branko K.
msg tools
mbmbmbmbmb
Yay, it's a Silver test all over again, now new and improved with Envoy spices!

But seriously, I'm always scared of these "direct" strategies. It's easy to get caught up in interactions between various Kingdom cards and forget that sometimes simple does it. In my earliest Dominion days I remember beating up experienced players with a deck containing one Smithy, one Council Room and a whole lot of Treasure.

Anyways, thanks for the robot algorithm, I think I'll try it out on a few mano-a-roboto..umm..'s mano plays. My hunch is that these "direct strategies" are efficient in solitairy plays but are pretty susceptible to attacks and this I'd like to test a bit. Also, I'm still unsure on how strong the Envoy is anyways, this will be a nice eye-opener.

Cheers!
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Werner Bär
Germany
Karlsruhe
Baden
flag msg tools
mbmbmbmbmb
yaron wrote:
The main "bug" in the strategy is grabbing a Province the first time you can, rather than getting some Gold first, but I did that for both Smithy and Envoy, so hopefully it doesn't skew the results.

I think buying a procinve with your first 8 is ok if you already have some (3?) gold in your deck.
In one of my envoy games, i had 8 money in turn 3. I bought a gold.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Chris Martin
England
London
flag msg tools
mbmbmbmbmb
Werbaer wrote:
yaron wrote:
The main "bug" in the strategy is grabbing a Province the first time you can, rather than getting some Gold first, but I did that for both Smithy and Envoy, so hopefully it doesn't skew the results.
I think buying a province with your first 8 is ok if you already have some (3?) gold in your deck.
In one of my envoy games, i had 8 money in turn 3. I bought a gold.
Yeah, the robot can buy a province on Turn 3 if two of its Estates are the 11th and 12th card of its deck and its Envoy is in the first five. But adding rules to stop it from doing so would make the program much more complicated. The point is to have a simple program.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Chris Martin
England
London
flag msg tools
mbmbmbmbmb
yaron wrote:
You mean in terms of the programming language? I used Matlab. I didn't simulate Dominion in any general way, just this specific scenario.
It's no more than 150 lines - if that kind of thing is up your alley, I can send you the file.
It would be up my alley, but I don't know Matlab. I'm not a programmer - I was going to create an Excel. To be honest, I mostly wanted to check to see whether you'd made any mistakes.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Greg Filpus
United States
Bellevue
Washington
flag msg tools
mbmbmbmbmb
The way I see it with "robot" strategies, the primary skill involved is assessing the setup to figure out if it's worth doing or if it's beatable. I've won games with the one Smithy + one Council Room + money strategy because I was the one to realize there was nothing better on the table. If multiple people go for it, it comes down to a little luck and a little of the finesse of the strategy- when do you buy the second Envoy, do you grab a Province or a Gold if you get 8 coins really early, when do you start going for Duchies and estate...

In the interest of realistic scenario, I'd want to see the Envoy as one of the 10 action cards. If the human players go for the straight Envoy strategy and win, or if the Envoys get bought out before the robot can get enough, those are outliers worth noting, but those can happen in real games. In particular, Envoy works really interestingly in a chain strategy- when you Envoy with an action left afterwards, it creates tough decisions like discarding the Gold or the Militia.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
David desJardins
United States
Burlingame
California
flag msg tools
mbmbmbmbmb
So, I'm not interested in "adding extra players" to my games, but I did try simply playing a two-Envoy-and-money strategy myself. Worked pretty well, I ended up tied for first (won on tiebreak under our house rules). It's certainly one of many viable plans, but hardly a gamebreaker. I wasn't trying to follow a precise algorithm.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Tom Chappelear
United States
Kensington
California
flag msg tools
mbmbmb
In our recent games, the feeling has been that if the Envoy is on the table, everyone is going to buy it in the first two turns. The second one is optional, but so far no-one has won a game with the Envoy on the table without buying it before the first shuffle.

Only about 12 games like this, but it does feel like we're in a bit of a rut here...
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
David desJardins
United States
Burlingame
California
flag msg tools
mbmbmbmbmb
tomchaps wrote:
In our recent games, the feeling has been that if the Envoy is on the table, everyone is going to buy it in the first two turns.


If everyone buys an Envoy then the winner will definitely be someone who buys an Envoy. I don't think that says anything about the card itself. Multiple players buying Envoys also help the other Envoy players by creating a game pace that favors the Envoys. There are other cards and strategies like that, the more players who follow them the less likely you are to win by doing something else.

2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
1 , 2  Next »   | 
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.