We’ve added (and are continuing to refine) a major new feature on CouncilRoom.com: Popular Buys.
By default, it provides static stats about Isotropic’s population as a whole; you can see, for instance, that Fishing Village is the most popular* card in the game — even more so than Province or Colony! In fact, Caravan and Nobles are both also more popular than any Victory or Treasure card. At the same time, poor Chancellor wins the prize as the least popular card in the game. Even Coppers and Curses are bought more often than Chancellor.
*To be precise, this ‘popularity’ is defined as the percentage of decks in which at least one copy of that card was gained. Technically, this is the number of players who bought or gained at least one Fishing Village in a given game divided by sum of the sizes of games where Fishing Village was available.
This “popularity” data is especially interesting when compared to “Buys”, which is a rough average of how many of each card each player buys. Unsurprisingly, non-terminals and Victory cards top the list; the highest terminals are Goons, King’s Court, and Bridge, which makes sense, since they all thrive when combined with each other.
In addition, “win rate given buy” provides a rough proxy of a card’s strength. Naturally, expensive cards tend to top the list, but so do many of the $5 attacks, as well as several Alchemy cards. Interestingly, Copper and Curse also do well, since if you’re buying them, you usually have a pretty good reason.
Most exciting, however, is the per-player data on that page. You can combine frequency of purchase, win rate given purchase, and win rate given availability to identify individual strengths and weaknesses; for instance, theory buys Lighthouse and Conspirator fairly often, but does quite poorly with them. rrenaud‘s long and futile fascination with Talisman reflects on his page; Obi Wan Bonogi, the top player on the leaderboard, shows a similar weakness for Quarry. Both, however, excel in sets with Vineyard, probably indicative of a love of Action-oriented games.
We hope you find all this data useful (and possibly educational)! If you have any suggestions, please let us know either in the comments or on the CouncilRoom uservoice page.
Dominion Strategy 
First of all thanks and then being a little bit pedantic: Nobles can’t be more popular than any victory card, since it is a victory card itself. 🙂
Second at least the FF3.6 doesn’t print “Search by player” in the Field where you should enter the name for the player to search, don’t know if you want to adress this, i want to update to FF4 anyway. Will see if it works there. But if I hadn’t read this article I wouldn’t have known that the side can give the stats by player…
Firefox4 works
That’s annoying. I guess we should change it, since apparently several browsers don’t support it.
Could you add definitions for Trashes and Returns to the page too?
I have them (and they are still in the data that gets sent to the browser), but they are sufficiently uninteresting and space filling that I deleted them from the table. People trash coppers/estates/curses and return coppers/estates, for the most part, just as you’d expect.
I’m just saying that if they’re in the table it would be nice to have a definition. (Is a “return” just for Ambassadors, eg?)
Couple of wishes for the popular buys page:
1. When searching for a player – include their base win rate as useful comparison. (Comparing win % rate normally vs. win rate given buy or avail is useful for analyzing strengths and weaknesses)
2. Allow a date filter or maybe only look at most recent N games or something. This would be helpful to test whether or not you’ve improved (or whether or not theory has decided to join other unnamed players in KC -> Goons -> Masq. locking.)
3. This might need to be a different page, but I’d love to see the same content, but with ALL players for a given card. (So I can see where I rank playing with Pirate Ship relative to other players, or I can see whether my games are shorter or longer on average than other players)
You can always look at their win rate for cards that are always available and that will be their base win rate.
I second all of these suggestions.
Some other things I would be interested in seeing:
1. I would like to see the average number of buys for a card not including the times it was not bought. This should be just is just “Buys”/”% +”, but I think this stat is more meaningful than “Buys”.
2. To further the thoughts on #1, (and venturing into the realm of data overload,) it would be interesting to see win percentage based on how many of a card were purchased. (How often do players win when they buy 1 smithy? …2? ect.) This stat borders on meaningless since the optimal number of a certain card is greatly effected by the rest of a set. It makes me want to snowball this suggestion into separating the stats for terminal actions into games where extra actions are available (e.g. villages) and games where they are not. This type of stuff would probably fit in well with separate ‘per card’ pages as with Captain_Frisk’s 3rd suggestion. It would be even cooler if it could be queried with a card, or set of cards (e.g. Smithy & Village, or even Smithy & (Village | Fishing Village).
3. This is more of a general request/inquiry. In my opinion, it would be more user friendly to express win rate as a win percentage. I assume it has been done in the current format because of games with more than two players; however, I am curious if this format has any significant benefits over treating 3+ game results as discrete 2 player games between each pairing. I know I am biased here because I almost exclusively play 2 player.
For #1, why is your measure better? It seems just more complicated to me. Certainly it measures something different, but avg buys is a pretty good measure of how useful multiple versions of a card are overall.
For #2, this gets toward what you want.
http://councilroom.com/win_rate_diff_accum.html?cards=Cost%3D%3D5%26%26Actions%3D%3D1
Of course, supporting filtering by restrictions on the kingdom set would be great to have there, it’s that scanning over hundreds of thousands of games is pretty damn slow, and so there are computational barriers.
#3. The reason for the win rate rather than win % is as you suggest. OTOH, division by 2 just isn’t hard, IMO.
To pile on to rrenaud’s hatred on #1, I believe that the average number of cards bought given buy should be [average cards bought] / [percent chance of buying]
So if your average fishing villages buys were 1.5, and you buy them in 50% of games, then in games in which you buy fishing villages, you must buy an average of 3 / game. I *think* that math holds out, but I’m too lazy to really think about it.
Good point on using the win% for cards that are always there as your base win%.
I feel that it is more meaningful to look at the average number of a card that was bought when it wasn’t ignored. If the presence of other cards makes a card irrelevant in a set, why should it’s average # of buys be reduced when it could have just as easily been a different card that didn’t get bought. In my opinion, the fact that .31 of an Adventurer is bought each time it is available doesn’t mean anything. That data alone only lets me know that it isn’t a very popular card, and the +% stat already tells me directly how popular a card is. I don’t suppose knowing that when they are bought, an average of 1.29 Adventurers are bought isn’t a terribly useful stat either, but it just seems more meaningful to me.
Interestingly, Copper and Curse also do well, since if you’re buying them, you usually have a pretty good reason.
There have been a few times when I’ve bought curses for strategic reasons, but much more often I’m buying them because it’s the last turn, I’m pretty far ahead, I have extra buys, and I feel like rubbing my lead in my opponent’s face. So there’s a bit of a jerk factor that you need to account for here, I think.
On the flip side, if I’ve lost but don’t want to resign for one reason or another, I will intentionally make nonsensical buys to mess with rrenaud.
Thanks for posting that…I’ve been wondering why anyone would buy a Curse. The only logical reason I could think of is that you have Watchtower in your hand and you want to trash the Curses. People sometimes talk about buying Curses to send them to other players with Ambassador, but in most cases you couldn’t guarantee getting the Curses and the Ambassador in your hand together, so that seems dangerous to me.
The usual reason that someone buys a curse is to deplete them as a 3rd pile. Usually a player will do this only if he is sure he is ahead in VPs, thus giving curse buyers a high win rate.
I am apparently not so good with Gardens. I wouldn’t have guessed.
I am apparently very bad with stash ; Part of this most likely is that it took me forever to figure out the deck-stacking interface (which is counterintuitive to me and doesn’t work well in safari to boot) but I also think part of it is that I haven’t thought or read anything about when it is right to buy it and where it is right to put it in the deck. This is me making a suggestion, by the way.
Well, I think Stash works well in “boring” sets (= no other power combos, no Colonies) and with proper support (anything that accelerates cycling: Chancellor, Warehouse, Golem w/ max 1 other action card …). If I go for Stashes I buy them relentlessly until I have 4 of them (i.e. also for 6 or 7 $) and then try to cycle my deck as fast as I can and buy Provinces (and maybe also Duchies – they don’t disturb you that much in Stash decks) whenever I can.
I’ve spent some time with the popular buys feature, and here’s what I’ve learned:
1. Many of the top players “cherry pick” their sets. I try to play mostly random, although I can see that my stats are skewed in the same direction as the general populace, most likely because people propose games to me with their settings. (Possession is my least common card, although I don’t think i’ve ever excluded it) I also played alot of my first games with 5 prosperity cards because I wanted to play with them. I won’t name names, but it appears that stronger players have a tendency to prefer colony games (longer games mean more control), and avoid some of the attack cards. Good players tend to play more from the 2p position than 1p, so attacks impact them more.
2. Buy comparing your “normal” win rate to your win rate given avail, and your win rate given buy, you can identify cards that you play better or worse than YOUR normal average.
For example:
One of my favorite cards is Hoard. I gain a Hoard in 80% of my games. When I buy it, my avg. win points are 1.19. The average for the population buying hoards is 1.02… so it looks like I play hoard better than average.
However, my Average Win Points Are 1.328. This means that I actually play Hoard poorly relative to my normal games, and thus I should probably ignore it more.
My worst card when its available is Woodcutter. My Win Rate Given Avail is 1.12. Even though I only gain one 20% of the time, when I do, my win rate is 0.85. This says I am most likely gaining woodcutters when I shouldn’t.
My best card when its available is Familiar. I win 1.47 points when it is available, and 1.51 points when I buy it (94% of the time, with an average of buying 2.17)
I also do well win Treasure maps are in play, but NOT when I play them. When I play them I win 1.31 points, but I win 1.46 when they are available. This indicates that I play them about about my average level (ie I play them when its generally a good idea), but that the general audience has a tendency to buy them when they shouldn’t, and I coast to an easy victory.
Cards that I don’t play well are:
Pirate Ship (0.22), Bureaucrat, Woodcutter, Outpost, Library… all of these cards I actually win < 1 point per game with. With pirate ship, I clearly don't know how to use it. With the others, I just don't know how to use them any better than the average isotropic user.
Cards that I might want to consider buying more often (cards with low popularity, but a very high win%) – Adventurer, Stash, Duke
Cards that I probably should buy less of like (high popularity, but a lower than expected win %) – Nobles, Festival, Caravan,Lookout,University,Throne Room,Haven,Salvager,Ghost Ship,Bishop,Remodel,Trade Route)
Some of those are a pretty big surprise to me (bishop and ghost ship).
Like everyone else, my favorite card is fishing village, apparently I ALWAYS buy one.
Interesting. Perhaps automating the break down of cards into different “quadrants” would be useful.
Very. I would love a graph that has your win rate given avail, win rate given buy, with lines for (average win% given buy, your average win %)
It would be great if the confidence intervals would have its own column.
As it is, I cannot paste it into excel and make my own calculations with any value that has a +/- in it.
You can paste it into a text editor, CTRL-F for +/- and replace it with a tab, and then paste it into Excel and the confidence intervals will have their own column.
A sloppy workaround, but at least an immediate solution :]
here is how I did it (not in a very elegant way, but it works):
*)open Word
*)set the page size to 50 x 50cm
*)ctrl-a-c-v the table into word
*)delete all stuff above and below the table itself
*)convert the table into text
*)replace all +/- with ^t
*)add tabulators into the head line after all columns which contain +/- to get the correct number of columns
*)ctrl-a-c-v into excel
FWIW, I might consider adding a csv output for the data in the future.
I love JSON, but it seems like the world loves CSV. Sad, sad world! The world has structure, let’s preserve it!
You can also add your own columns that use left(CellAddress,4) i think to get most of the base values. Not that hard.
I think it would also be interesting to see the standard deviation of “win % given buy” across the set of players. I think this would show the power and skill required to play a certain card. For example, the hypothetical card “you win the game” that costs $4 or less should have a stdev of 0 since it would reduce the game to nothing but shuffling luck and turn order bias.
One problem of this analysis is that the average win rate of a player may not be comparable to a specific card: some cards are intrinsically more swingy than others. For the hoard you mentioned, I believe the lower win rate comparing with your average comes at least a part from the fact that the game becomes more luck dependent when both players are buying them.
I think one of the easier way to identify individual strength/weakness is to look at the win rate/availability. While in principle this can also be affected by the randomness of the card, in practice it seems it is more directly related to how well the player plays the card. Also, if win/availability is higher than win/buy, that pretty much says the player still buys it too often.
To test this hypothesis, what other cards would you consider to be powerful and easy to play correctly such that they would tend to mute player skill?
Mountebank would be another example I can think of, without help from any other cards. In general, Cursing cards with reactions would tend to be swingy, which manifests itself in Mountebank.
The other kind of possibility are cards which facilitate a money deck. In this category though, it seems only hoard is strong enough to be in a consistently winning strategy. (Think of Beauracrat/explorer)
To second timchen’s comments, I don’t think you should conclude that you should ignore Hoard more often. I wouldn’t say that this data is an indication that Hoard is more swingy, but rather that it is likely it is a card that is powerful and easy to play. It doesn’t mean that you are playing worse when Hoard is in the set, but it does probably mean that it is more likely you will be playing a “mirror match” when it is in the set, which will pull your Win Rate closer to the coin flip result of 1.00.
Has anyone done hoard simulations? Should it be bought if there aren’t helper vp cards? In general I use it less in colony games, but I’m wondering if you should buy it with your first 6, or if you should get at least 1 gold first.
What makes me think that I overplay it is that that like B-Crat, it bloats your deck making combos harder… and a given hoard only appears a few times in a typical fast game… so is the delay on your economy by purchasing a hoard during the 2nd shuffle worth the fact that your deck won’t tank as quickly as you buy VPs.
Other questions: If you draw < 5, should you buy an estate with hoard? If you draw < 8, should you buy an early duchy?
Heck, maybe I over-estimate Hoards strength. (My stats=> base win:1.25, available win: 1.24, bought: 67%, bought win: 1.73). Clearly I win a lot more when I play it than when I ignore it. I have no idea what conclusions can be drawn from that.
Your questions are excellent, but don’t have any data to support the theories I could come up with as answers. I think the standard deviation of “win % given buy” across the set of players that I suggested above would be a good measure of which cards mute player skill.
One more thing I would bring up regarding your Hoard example is that while your win points of 1.19 when buying Hoard are lower than your average win points overall, your win points when Hoard is available are 1.18. This means that when Hoard is available, you win more often when you buy it than when you do not. This could be interpreted to mean you should be buying Hoard more often. The top 6 players on isotropic all have reduced win points when Hoard is available, which points to it diluting skill, but I don’t think any of them are reduced quite as much as you are. If I had to guess, what I am seeing is pointing more towards you not playing it optimally, so I suppose that means you need some answers to the questions you posed.
calculating your personal [win_buy] minus [win_available] can tell you a lot about which cards you buy too often.
I cropped 9 cards with buy rate below 0.1 to avoid the weirdest results, and got as my worst buys:
Contraband,Workshop,Mint,Ghost Ship,Pirate Ship
Except for Ghost Ship that is according to what I already knew.
On the other hand the same statistic would advise me to buy more Herbalists, Chancellors and Smugglers among the top 5, which I am for sure not going to do. Statistics even tell me that if I buy Smugglers or Herbalists, I win far more than the average player buying those cards [personal win_buy] minus [overall win_buy]. Personally, I think that many Smuggler wins were through Gardens, which makes me want to exclude Garden games (games that ended with empty garden pile + 2 other empty piles) from all statistics.
Finally, how can you calculate what cards you correctly avoided? I went for [personal win_available]divided by[personal_buys], and got mostly the cards that I avoided in total. Relating the ranks of that value to the rank of [not bought] results in a confusing meta-ranking, in which smugglers, phil.stones and stashes somehow stick out.
Like everyone else, my favorite card is fishing village, apparently I ALWAYS buy one.
ISTR reading somewhere that Donald said he didn’t want any card to become “everyone buys this card until it runs out, then play the game”, but the last time I didn’t treat FV this way, I got completely stomped by the people who did. (Even in that game I bought *some*, it’s just that I also occasionally bought other things, so I ended up with 2 while the other players had 4 each, IIRC). It’s so good it’s game-distorting, even more so than the scary attacks (most of which are terminal, and need something like, say, fishing village to be REALLY powerful).
FV is very often purchased over silver because it’s not much worse in terms of producing money, but it has a big upside with getting rid of the risk of colliding terminals. It needs some powerful terminals or cards that like “disappearing money” like minion or menagerie be worth it over silver though. Of course, most kingdom sets have at least one useful terminal action, and this is why fishing village is very often purchased.
But that doesn’t mean it should really run out that much. You want useful $5 cost terminals to combo with it, not a ton of fishing villages by themselves.
And there are indeed a number of situations where you’ll prefer silver (some big venture decks, if there aren’t any terminals worth getting, possibly if you really don’t want more than one terminal total), just not that many.
availability:
least favourite card: possession (by a margin)
then all other potion cards (obviously, due to randomizing)
least favourite non-alchemy card: Saboteur (again by a margin)
favourite card (besides basic treasure/victory cards): Chapel (b.a.m.)
you can easily see what people like and what they don’t.
Wanna lose? Go for Thief!
alternatives: copper strategies (C-smith or C-House), Saboteur, Bureauscrat Talisman, Transmute
looking for the best actions?
Grand Market, Alchemist, Wharf, Scrying Pool, or any 5$/6$-attack card that gets you cards or virtual money.
what is trashed?
1.) your starting deck
2.) cards that trash themselves (or copies of themselves)
3.) cards that become obsolete over the course of the game, including Silver
4.) Curses (only nr. 10)
5.) Quarry, Potion and Gold
what makes games quick?
Caravan (b.a.m.), Chapel, Wharf, Fishy, Council Room, possibly Lab
(those 6 lead with or without colonies)
what puts on the brakes?
Cursethrowers, Discard, other attacks, Vineyard, Black Market (!?), Monument, Herbalist (?!)
I have a super record in games with Embargo, Smugglers and Workshop by not buying them 😉
On the other hand don’t buy and lose when Mine and Herbalist are there.
Buying stacks of Laboratory, Caravan and Nobles seems to cost me a lot of games.
Fishys and Minions on the other hand I often bought successfully, as well as Kings, Golems and Stewards.
My Market games last only 15 Turns on average
I never fail to buy a Chapel
Speaking of black market, a question that came to mind when i was looking at these stats is how black market cards are dealt with ; does a card that is in the black market deck count as available? If not, how do cards purchased by black market counted?
Personally, i’d want to exclude all black market games, because they skew the meaningfulness of the results ; Especially in a black market game where (taking from recent painful memory) an especially powerful card becomes present in an otherwise slow game, that card would overvalue the black market of the person lucky enough to get it, while devaluing the strategy that would otherwise have won the board, if the lucky black marketeer hadn’t pulled the power card in the 2nd shuffle.
It’s a hack. There is no really great way to deal with black market, IMO.
Excluding it altogether is one reasonable option.
This is what I implemented. Cards from black market count as available if and only if they were purchased from the black market. This means they are in some weak sense ‘available’ but not counted when they are in the black market but not purchased.
I think you need to be very careful how you read these statistics. My most successful card appeared to be the bureaucrat. I suspect this is because I only buy them rarely and only when they’re highly likely to work. I doubt it is because I’m experienced in using bureaucrats or because of the power of the card.
It does show you are better at knowing when to deploy it. Also you can see how often you buy a card as well, to put together a good picture of why you’re successful with it.
If / when rrenaud ever makes player data available, my plan is to search for players who are effective with b-crats, and figure out what they do to actually make them effective.
Well, it’s fifth lowest in win rate given buy, so apparently it isn’t that effective. My win rate with it is 1.33, but I can’t think of any strategy I’ve used. It’s pretty good when your opponent has a clogged deck (Gardens strategy, etc.) or when you have a ton of Curses and need as much money in the mix as possible. It’s obviously better in a Province game than it is in a Colony game. But usually it’s not a good card to buy.
My most successful card appears to be Curse, followed by Thief.
I think I’ve bought Thief a grand total of once. And that was basically more or less as a lark, on a bad draw near the end of a 3p Village / Smithy / Hoard game which I was already winning.
In other words, it’s a fluke of small sample sizes.
Any way to calculate % buy when player had enough cash? “When available” just means it’s in the supply right? But I’d be interested to know, for $5+ cards, how often players turn those cards down…
Hmm, my win rate isn’t very good when I buy Familiar. I wonder why? I always buy them if they’re available. Maybe because my opponents always buy them too? On the other hand, I need to buy more Stewards…seems I win with them a lot.
Apparently the Vineyard and the Scrying Pool are great cards, but I haven’t had much luck with either one. I look forward to the writeups on those two cards.
The bottom 10 cards in “win rate given buy” are all cards I usually avoid, but I’m a bit surprised Remodel and Chancellor don’t win more often.
I was also interested in the cherry-picking idea. I’ve come across a very few who seem to obviously do it (which I subsequently confirmed on councilroom as having card selections well outside 3 sigma). But I think a lot of the skew has to do with the preponderance of people who propose Kumbaya card sets without attacks. I’m not sure what the randomization criteria are for Colony games or Alchemy, so I can’t commend on that.
I will admit that I sometimes reject games that look uninteresting and painful (a la Ghost Ship and no trashing, with nothing else going on). At some point playing a 35 turn game against a new player who takes 20 minutes each time to decide which cards to return just isn’t worth it.
I’d be curious to see charts separated by number of players, so that one chart includes only 2 player games and another includes data from only 3-4 player games. For example, I would think the win rate for mountebank and witch would increase in the 3-4 player games.
The win rate for Thief and Pirate Ship would certainly be higher in 3-4 player games. That might be the case for Saboteur as well.
Are there *any* attacks that are stronger with 2 players? I can’t think of one.
Wait, that’s not true, never mind. Minion is better with 2, because you really need 5 or 6 of them to get the full effect.
The vast majority of attacks are stronger in multiplayer, though.
chapel buy 0.92 %+ 77%
only 77% will get at least one chapel when it is available and i think i can understand why
but buy = 0.92 mean that some people will buy more than one chapel when it is available, i really not understand why some people will do so.
Some people play very poorly 😛
It’s important, of course, to note that these are /correlations/, not causations. As a previous poster pointed out, It could be that you have a very high win rate with card X because you are very judicious in its use, and only buy them when it is a dominant strategy. That does not mean that in general you should buy it more (though it doesn’t mean you shouldn’t, either!) because “buying X causes you to win”.