Dominion Strategy Forum
Dominion => Dominion General Discussion => Topic started by: Limetime on June 01, 2016, 09:55:15 pm
-
For me I think i am best at governor. Love the card and all the interesting decisions. When watching streams of govorner games I am like no, don't get a gold remodel that silver into a govorner and buy another one.
-
Transmute
-
Bishop might be one. I often will get Bishops later than my opponent and end up with more tokens from them.
-
Miser. Also I feel like I win more than not in Cultist and Urchin mirrors.
Edit: this is my 222nd post, and that's my lucky number!
-
Probably Scout.
-
Probably Scout.
Really? You?
I feel like I'm the only one who plays Scout well, but if I mamaged to convince everyone in the world as to my views of Scout, they would probably be able to utilize my strategy better than me.
-
I win almost every Herald board since I'm one of the few people who acknowledge its true power. 8)
Unfortunately, for every Herald there are at least four cards I completely suck at. Black Market, Rebuild, Possession, Tactician...
-
Last time I checked the actual statistics, it was Bridge, but it's been a while since then.
-
Somehow I think I might be OK with Possession (will probably lose next 666 Possession games and then die in pain screaming for mercy).
-
For a long time I was very good with duke, I think its power was not well known.
Now I'm not bad with harvest or farmland.
I don't know why, but seems like I'm very bad with butcher. But that's another topic I guess.
-
Black Market. I would play every game with that card in the kingdom if I could
-
Highway! Or maybe Gear.
-
Scout- I always know exactly when to buy it.
Never!
-
I am really getting the hand of the entire page line, I feel.
-
Golem, Lookout, Procession, and for reasons unexplainable to me, Merchant Guild.
-
Hoard
-
Are there any statistics easily available for this for someone like me who has only been around for Making Fun?
My guess would be Beggar. I seem to win most of the time when I buy that card, which I also believe I do significantly more often than most other people.
Other cards I suspect I do well with: Urchin, Ill-Gotten Gains, Explorer
Cards I suspect I do poorly with: Bishop, Tunnel, Market Square, Highway
I am very bad at determining when gaining Gold in a weird way via Tunnel or Market Square is good.
-
I just ran my stats across all professional games recorded by gokosalvager.
Here's the google doc: https://docs.google.com/spreadsheets/d/1fEm6TtcqLWrqq5ksaOdEXnXEn9FfDWaShpwtnuxLvWA/edit?usp=sharing
Note that games present is games where its in the supply; the "discrepancy" with Gear isn't a discrepancy: I lost a game where I bought it out of the Black Market deck; but haven't lost with it otherwise.
By looking at the base cards, I have a score % ((1 * wins + 0.5 * draws) / games played)) of 65%. The non-Adventures card I have the highest score percentage with is Great Hall, at 85%. Weird. I tend to buy it in those games as well.
As for actually buying the cards, I've apparently won every game (extremely small sample size) in which I've bought Stash, Harvest, Coppersmith, Duchess, and Bureaucrat (again excluding Adventures, my win rates are very high with those cards). For cards that I actually have played > 10 games with, buying Embargo, Herbalist, and Scrying Pool all lead to 90+ win rates for me in rated games. Interesting.
On the flip side, my win rates when buying Count, Native Village, Trader, Horse Traders, Advisor, and Workshop are all <40%.
By the way, my extremely simple methodology of computing this means that Knights aren't calculated correctly for being bought, but they are for being in the supply.
Also, no, I can't easily extend this to anyone who asks.
-
How do you run the stats for others?
-
How do you run the stats for others?
Well, first you download all your game logs from gokosalvager, then you parse them, noting for each game what cards appeared and what the result of the game was. Then you aggregate all that information, and generate stats. Part of this process for me was manual, because I was lazy and didn't feel like parsing the list of game logs off the log search in the same script that I used to then download them all. Instead, I just ran some javascript in my browser's developer console that got me the list of game log files, which I then copy pasted out into a file, which my fetcher script then read and downloaded all the logs. Then I used another script to parse them and generate the csv, which I imported into google docs.
Creating a general use tool wouldn't be that difficult, if people wanted it (and I could find a place to host it).
-
How do you run the stats for others?
Well, first you download all your game logs from gokosalvager, then you parse them, noting for each game what cards appeared and what the result of the game was. Then you aggregate all that information, and generate stats. Part of this process for me was manual, because I was lazy and didn't feel like parsing the list of game logs off the log search in the same script that I used to then download them all. Instead, I just ran some javascript in my browser's developer console that got me the list of game log files, which I then copy pasted out into a file, which my fetcher script then read and downloaded all the logs. Then I used another script to parse them and generate the csv, which I imported into google docs.
Creating a general use tool wouldn't be that difficult, if people wanted it (and I could find a place to host it).
I think a General use Tool would be really nice and much appreciated if you have the time and will to do it :)
-
Back when TheExpressicist's tool was available for grabbing player stats from the Goko implementation, I looked at my win % with various cards (but not looking at the cards I almost never bought) and it looked a lot like what I got if I took my overall win % and just did random simulations of sets of 200 games or whatever. I didn't do anything particularly rigorous, but I'm not convinced that the higher or lower win percentages with a particular card are anything but just the noise from our small sample sizes.
-
I'm more curious about what cards I'm not good at and why I'm not good at them.
-
Back when TheExpressicist's tool was available for grabbing player stats from the Goko implementation, I looked at my win % with various cards (but not looking at the cards I almost never bought) and it looked a lot like what I got if I took my overall win % and just did random simulations of sets of 200 games or whatever. I didn't do anything particularly rigorous, but I'm not convinced that the higher or lower win percentages with a particular card are anything but just the noise from our small sample sizes.
I find it likely that it's not just noise about your overall win %. A calculation by rrenaud in 2012 (http://forum.dominionstrategy.com/index.php?topic=2798.msg47781#msg47781) on how much certain cards favor more skilled players finds that some cards favor more skilled players more than others.
Since you're one of the best players, you're more skilled than your typical opponent by quite a bit. Therefore, we'd expect to find that you win more often with the cards at the top of the list in a calculation like rrenaud's. Edit: If we grouped you and fellow players with similar ratings and did card win %s for the group, we'd have much bigger sample sizes and could probably tell that the card win %s are different for different cards. But we couldn't necessarily tell to what extent you differ from Stef who differs from SCSN, etc.
-
I'm pretty good with Provinces, especially when I have most of them.
Honestly though, maybe Scrying Pool? I think I'm pretty good at recognizing when it's skippable. (But watch me go lose against it a bunch now...)
-
Back when TheExpressicist's tool was available for grabbing player stats from the Goko implementation, I looked at my win % with various cards (but not looking at the cards I almost never bought) and it looked a lot like what I got if I took my overall win % and just did random simulations of sets of 200 games or whatever. I didn't do anything particularly rigorous, but I'm not convinced that the higher or lower win percentages with a particular card are anything but just the noise from our small sample sizes.
I find it likely that it's not just noise about your overall win %. A calculation by rrenard in 2012 (http://forum.dominionstrategy.com/index.php?topic=2798.msg47781#msg47781) on how much certain cards favor more skilled players finds that some cards favor more skilled players more than others.
Since you're one of the best players, you're more skilled than your typical opponent by quite a bit. Therefore, we'd expect to find the same general pattern as would be found in a calculation like rrenard's. However, we may not have the sample size to distinguish any cards that you are particularly good at, compared to the cards that other players of similar skill level tend to win with more often.
I can't see how rrenaud's results have anything to do with what I'm talking about. I doubt if there's any correlation at all between rrenaud's results and the card win % of top players. It's a measure that looked at 2.5 million games across all iso levels. The fact that in that set Trueskill apparently is right more often with certain cards doesn't seem to say anything about whether a sample of a few hundred or thousand games can tell us anything useful about which cards a player is best with.
I don't understand your second paragraph at all.
-
I can't see how rrenaud's results have anything to do with what I'm talking about. I doubt if there's any correlation at all between rrenaud's results and the card win % of top players. It's a measure that looked at 2.5 million games across all iso levels. The fact that in that set Trueskill apparently is right more often with certain cards doesn't seem to say anything about whether a sample of a few hundred or thousand games can tell us anything useful about which cards a player is best with.
I don't understand your second paragraph at all.
I'd expect that the cards at the top of rrenaud's list are correlated with the card win % of top players. Indeed, the top four are Goons, Colony/Platinum, Bishop, and Ambassador.
The second paragraph is attempting to say that if we grouped you and fellow players with similar ratings and did card win %s for the group, we'd have much bigger sample sizes and could tell that the card win %s are different for different cards. But we couldn't necessarily tell to what extent you differ from Stef who differs from SCSN, etc.
-
Here's a figure of win % for the top 20 isotropish players (data here (https://docs.google.com/spreadsheets/d/13mQ1humtQbPLY9nbKscR65dV7hbGPdI3AQkNjMHZpeM/pubhtml?gid=495443102&single=true), from Goko stats circa early 2015) with card in the kingdom vs that card's value on rrenaud's list.
(http://i.imgur.com/cN0qnoB.png)
Sad guy at the bottom is Walled Village, just too few games.
Your predicted trend would be a line from top left to bottom right (roughly). It's kind of there. The two axes are of course not from the same set of data or time period. It may be that if we had more data from the top players the correlation would be better, but that's my point really!
This is all pretty tangential to my original post, which I still don't understand how it has anything to do with the rrenaud data. When people look at their sets of 100 or 1000 games, of course there are some cards which they have won more or less with, but the deviations from the average win % as far as I can tell are not distinguishable from random. But maybe someone can show that they are. As it is, I don't consider the deviations any sort of evidence for special understanding of a card.
-
Coppersmith for sure.
-
Silver is my worst card. Every game this card is in I seem to lose.
-
Silver is my worst card. Every game this card is in I seem to lose.
Wrong thread. This is the thread called "What's the card your best at?", and not the thread called "What's the card your worst at?", where such an appropriate statement could then be made.
-
Silver is my worst card. Every game this card is in I seem to lose.
Wrong thread. This is the thread called "What's the card your best at?", and not the thread called "What's the card your worst at?", where such an appropriate statement could then be made.
I am best at losing with that card.
-
Estate.
-
Nobody says Curse, because Curse is clichè. Jsh already won the entire draft by saying Transmute, a terrible card that nobody likes. What card is in second place though? Hrm.
-
Maybe Horn of Plenty?
-
I'm pretty good at games where I have Followers and Trusty Steed. Village is a close 2nd.
-
I'm pretty good at games where I have Followers and Trusty Steed. Village is a close 2nd.
To be fair, if you have both Followers AND Trusty Steed, you're pretty damn ahead.
-
I'd find a way to lose those games
-
The title of this thread is wrong: it should be you're not your.
-
The title of this thread is wrong: it should be you're not your.
I was wondering when someone would mention that. I didn't want to be that person.
-
The title of this thread is wrong: it should be you're not your.
Screw phrasing and Grammer.
-
The title of this thread is wrong: it should be you're not your.
Screw phrasing and Grammer.
No thanks, I'm not interested in a threesome.
-
The title of this thread is wrong: it should be you're not your.
Screw phrasing and Grammer.
No thanks, I'm not interested in a threesome.
I mean are we talking Kelsey Grammer? Then sure.