Dominion Strategy Forum
Archive => Archive => Innovation General Discussion => Topic started by: rspeer on March 25, 2013, 03:53:50 am
-
I'm looking into doing stuff with the Innovation logs. Many interesting statistics are going to require parsing the game logs and reconstructing the game state, such as how likely you are to win based on your number of achievements, current tech difference, particular cards melded, and so on. This is of course daunting. Has anyone made a start on it? Is your code available?
On the other hand, I can get something interesting by just parsing the last line of each log: a breakdown of the reasons for winning (or at least for the game ending).
I've retrieved 27736 games. Of those:
- 15412 (55.6%) of them were won by achievements.
- 4586 (15.5%) were won by attrition (because all but one player resigned).
- 2567 (9.3%) were won by score after trying to draw an 11.
- 2252 (8.1%) ended with all players resigning. I guess that might tend to happen in solitaire.
- 741 (2.7%) were won by Bioengineering.
- 657 (2.4%) were won by Globalization.
- 629 (2.3%) were won by Self Service.
- 545 (2.0%) were won by Empiricism.
- 124 (0.44%) were won by Collaboration.
- 84 (0.30%) were won by AI.
- Only 11 games (0.040%) were exact ties on score and achievements.
I skipped over Echoes-specific win conditions in that list to make it clearer and minimize the "what is this I don't even" factor, but here are those:
- 67 wins by Social Networking
- 22 wins by Puzzle Cube
- 16 wins by Paper beating Rock
- 11 wins by Human Genome
- 6 wins by Saxophone
- 5 wins by Radio Telescope
- 2 wins by Rock beating Scissors
- Nobody has ever won by Sudoku.
-
16 wins by Paper beating Rock <---
5 wins by Radio Telescope <---
2 wins by Rock beating Scissors <---
Yay one of each of these are done by me :P
15412 (55.6%) of them were won by achievements.
2567 (9.3%) were won by score after trying to draw an 11.
So basically we have 6 out of 7 games ended by achievements. Great piece of information, which means that Math-teching to 11 is not that feasible. :P
-
Excluding all the resignations (since it's impossible to determine the reason), ties and echoes cards, we have:
Achievements: 74,2% (almost 3 of 4)
Score: 12,4%
Card-specific: 13,4%, namely:
Bioengineering: 3,6%
Globalization: 3,2%
Self Service: 3,0%
Empiricism: 2,6%
Collaboration: 0,6%
A.I.: 0,4%
-
as far as echoes game go,i won 2 by social network,1 by puzzle cube and 1 by human genome
i came really close to winning a game by sudoku but unfortunately the card that would have give me the ninth different bonus value covered up sudoku thus covering sudoku 11 bonus
-
Nice piece of information
Assuming all solitaire game is excluded...... most of the time, achievement is the main way to win
p.s. I have a new aim now..... win by Sudoku
-
Is there no way to win by scissors beating paper?
-
Is there no way to win by scissors beating paper?
That's correct -- scissors just scores paper.
-
I haven't really done any work on parsing the logs. I've been playing around with writing a live extension to make the UI a bit friendlier.
It might be a decent idea to implement the parser/stats/game state manipulation stuff in JS/coffeescript so that it can be combined with browser extension for an AI.
-
One question: Any case of 'Rock Beats Scissors'?
-
Proof that Sudoku is impossible
-
Proof that Sudoku is impossible
It might be possible but simply not applicable since you already win by many other ways when you are tucking massive echo cards and splaying all the colors in-order to make way for Sudoku. Special achievements will give you the win beforehand :-\
-
Proof that Sudoku is impossible
It might be possible but simply not applicable since you already win by many other ways when you are tucking massive echo cards and splaying all the colors in-order to make way for Sudoku. Special achievements will give you the win beforehand :-\
yep,not to mention that when you have sudoku it's easy to call 10,pull off human genome/puzzle cube/social network and win through one of those
-
I wonder there are so many people won by social networking. I think using puzzle cube and human genome were easier.
My next aim: win by sudoku
yep,not to mention that when you have sudoku it's easy to call 10,pull off human genome/puzzle cube/social network and win through one of those
Not really. The process of sudoku was actually "meld a card -> win if you have 9 different bonus -> execute the dogma of the melded card".
-
yeah i mean,it's easier to usualy hunt for one of those card than to use sudoku effect to hunt for the remaining bonus value's
-
Won by Bio and Self Service today. :)
-
I'm surprised a win by score after drawing >10 is so much rarer than wins by achievements.
-
this is great stuff! would it be possible to do personalized statistics, (like just for me haha?).
-
I'm surprised a win by score after drawing >10 is so much rarer than wins by achievements.
I'm not. I'll venture to say most of my games end before an 8 is melded, and that the distribution of the wins feels roughly along the lines of what I've experienced.
-
I'm surprised a win by score after drawing >10 is so much rarer than wins by achievements.
i didn't even know this was a game ending condition until past my 40th game.
-
I'm surprised a win by score after drawing >10 is so much rarer than wins by achievements.
i didn't even know this was a game ending condition until past my 40th game.
This probably contributes to its infrequency. If people don't know it's a game end condition, they don't really go for it.
-
Winning on score feels like 3-piling to me.
-
I'm looking into doing stuff with the Innovation logs. Many interesting statistics are going to require parsing the game logs and reconstructing the game state, such as how likely you are to win based on your number of achievements, current tech difference, particular cards melded, and so on. This is of course daunting. Has anyone made a start on it? Is your code available?
I wrote this code when the logs didn't include the "long header", so it has to be modified to take care of the new log format I think.
https://github.com/anntzer/innovation-stats
To parse a game log, run
python innovation_parser.py <logfile>.html
Requires Python3 but probably easy to make it work with Python 2.
This does not reconstruct the game state (because the log info is actually not sufficient for that right now, I believe) but outputs a list of "events" (basically by matching against a huge regexp), from which you can e.g. get the running scores or the running achievement count.
I would be happy to work more on that... some day.
-
Winning on score feels like 3-piling to me.
Holy crap, this is so correct.
-
I'd be interested in seeing which cards' dogmas are applied most.
-
Your wish is my command. Probably would be more interesting broken out by age, may do that later.
(http://i.imgur.com/r6x0Odw.jpg)
-
Awesome.
Can you correlate that with whether or not the applying player went on to win or lose?
e: Also, how are you generating these stats? Looks like R output to me. Is there a database you downloaded, or did you brew your own parser et al.?
-
Would it make more sense to compare each dogma's activations with the average for that Age?
Alternatively, have a series of graphs, one for each Age, with a line showing the average for that age.
EDIT: Never mind, you already mentioned that.
-
Your wish is my command. Probably would be more interesting broken out by age, may do that later.
(http://)
I can't believe Writing is so high. Why would anyone activate it more than once or twice?
-
Because they are bad.
Because they find it humorous that it is equivalent to their standard draw action at the moment, so they activate it
-
I lost an echoes game by Saxophone. When it happened I just shrugged (was ahead in achievements at the time) since its generally not worth worrying about your opponent winning in this manner. Also have a Social Networking win.
-
I had a social networking win in 3 player. Pretty epic. I was ahead on achievements, but had lost icon control so everyone was catching up.
-
Because they are bad.
Because they find it humorous that it is equivalent to their standard draw action at the moment, so they activate it
I totally do it for the second reason.
...
...
And maybe the first :/
-
Your wish is my command. Probably would be more interesting broken out by age, may do that later.
(http://)
I can't believe Writing is so high. Why would anyone activate it more than once or twice?
Sharing is caring!
-
Wow, Writing? Philosophy is the other surprise. People like teching and splaying far too much!
-
I often find that Philosophy's scoring thing is actually handy for getting a card that I don't want to play out of my hand (perhaps I've got Domestication and a couple nice 3s and a scraggly 1 and we're on the 3s or 4s pile), with the splay being almost a bonus.
OK, maybe not often.
Sometimes?
Occasionally?
Yeah, we'll go with "occasionally".
-
I mean any card that scores from hand can't be that bad. It scales nicely as the game goes on.
-
Because they are bad.
Because they find it humorous that it is equivalent to their standard draw action at the moment, so they activate it
Writing is not equivalent to your standard draw action if you are sharing it. I will often activate writing to get two cards and toss my opponent one, especially at an age break! (Similar to Oars)