forum.webdiplomacy.net
http://forum.webdiplomacy.net/

Movement statistics?
http://forum.webdiplomacy.net/viewtopic.php?f=12&t=1176
Page 1 of 1

Author:  Tim [ Mon Aug 19, 2013 6:10 am ]
Post subject:  Movement statistics?

Apologies if this has already been discussed - I did a quick search and couldn't find it.

There are a lot of game histories saved on webdiplomacy. Has there been any work on generating statistics for various openings and/or wins given certain moves? I'm imagining an automatically generated page containing statistics like those found in some diplomacy strategy articles, (for example, this one):

Quote:
Essentially there are only four English openings, which account for 92.12% of all recorded UK postal games (some 2,132 in all). They are:

F(Lon)-NTH, F(Edi)-NWG, A(Lpl)-Yor 39.07%

F(Lon)-NTH, F(Edi)-NWG, A(Lpl)-Edi 22.61%

F(Lon)-ENG, F(Edi)-NTH, A(Lpl)-Yor 23.45%

F(Lon)-ENG, F(Edi)-NTH, A(Lpl)-Wal 6.99%


If this hasn't been considered before, would work towards automatically generating a page of data like this be welcomed? It would be interesting to see historical trends (eg, opening X is more popular this year than last year), and I'm sure there are many other interesting facets in the data.

Author:  kestasjk [ Mon Aug 19, 2013 11:29 pm ]
Post subject:  Re: Movement statistics?

I vaguely remember that Draugnar set up some kind of stats page based on our publically available statistics, but can't remember exactly. Definitely we've got a huge amount of data, and it is available to anyone who wants to research strategies, so if you've got something in mind let me know and I'll refresh the data. (It will come in SQL format though, so please don't ask for the data unless you know how to load and manipulate SQL data)

Author:  Tim [ Tue Aug 20, 2013 2:02 am ]
Post subject:  Re: Movement statistics?

I would love to have a play with the data! I have a few ideas, but here are a some off the top of my head:

General stats wise:
  • Common openings, and their win/lose percentages
  • Chances that a particular SC is held at the time of a win/draw
  • Most invaded/least invaded territories
  • Balance - are there countries that win more/less often? Do they stay the same from (real) year to year?
Interesting game lookups:
  • Situational lookup - eg, show me games that had say Austria in this particular position
  • Turnarounds - eg, what sort of moves were made by people who successfully bounced back from 1 SC?

SQL is no problem. Is this database layout documentation here still current? How would I go about access to the data?

Author:  kestasjk [ Wed Aug 21, 2013 12:01 am ]
Post subject:  Re: Movement statistics?

No worries Tim, I'll do a backup and create a new snapshot of the data hopefully this weekend, and post up a link here.

Author:  Tim [ Wed Aug 21, 2013 12:30 am ]
Post subject:  Re: Movement statistics?

Great! I have some time Sunday morning (GMT +10) to do an initial investigation.

Looking forward to it!

Author:  kestasjk [ Sat Aug 24, 2013 6:23 am ]
Post subject:  Re: Movement statistics?

Here's the data: http://webdiplomacy.net/webDiplomacy.net-SQLdata.zip

Let us know if you get anything out of it.

Author:  Tim [ Sun Aug 25, 2013 6:40 am ]
Post subject:  Re: Movement statistics?

Thanks, kestasjk. Here's an initial investigation, looking at classic, full-press games which have entries in the moves table for turn 0 (the early games didn't have move history recorded).

There are 23,825 such games, of which 13,978 have been won, and 9,847 have been drawn. I'm not sure yet whether I should do something special about games with absent players.

Here's a quick summary of the games described above:

Code:
                     % of all wins           % Games defeated in       % of draws participated in
England                12.63%                   33.99%                      59.76%
France                 16.05%                   31.75%                      61.37%
Italy                  9.08%                    43.61%                      50.31%
Germany                13.44%                   44.71%                      51.87%
Austria                11.38%                   55.07%                      44.49%
Turkey                 17.92%                   32.85%                      60.74%
Russia                 19.50%                   45.39%                      47.20%


The defeats and draws columns don't add up to 100%, because multiple players can be defeated or draw in each game.

It's interesting, as it seems to show a fair amount of imbalance. Also interesting, is that these numbers produce a different ranking to those at eg http://www.diplom.org/Zine/F2006M/Burto ... powers.htm. Clearly there is a lot of noise. I might look at number of unique players who win with each power, to see if we can spot "easy" vs "hard" powers.

In this data, it's interesting is that Russia wins the most games, but isn't a participant in as many draws as five other nations. Also interesting is that Italy is least likely to win, but not the most likely to be eliminated.

Author:  kestasjk [ Mon Aug 26, 2013 11:47 pm ]
Post subject:  Re: Movement statistics?

Interesting, poor Italy :(

Author:  Flame [ Fri Sep 27, 2013 12:10 pm ]
Post subject:  Re: Movement statistics?

It will be the coolest thing if we got the stats for game opennings.

Example: (R.Sharp debut classification)

FRANCE:
1 FRANCE F BRE-MAO A PAR-BUR A MAR S A PAR-BUR debut "Maginot"
2 FRANCE F BRE-MAO A PAR-PIC A MAR-SPA "Picardy"-opening
3 FRANCE F BRE-MAO A PAR-BUR A MAR-SPA "Burgundy"-opening
4 FRANCE F BRE-MAO A PAR-GAS A MAR-SPA "Gascony"-opening
5 FRANCE F BRE-ENG A PAR-PIC A MAR-SPA "English attack"-opening
so on...
.....
.....

1) number of games the opening was played
2) per-centage of eliminations playing this opening
2) per-centage of survivings playing this opening (including WINS)
3) per-centage of WINS playing this opening

Author:  Baskineli [ Wed Dec 23, 2015 4:38 pm ]
Post subject:  Re: Movement statistics?

The reason your ratio is different from the link is because you didn't normalize it (one sums to 100%, the other to 45.31%. By normalizing it (simply multiplying your values by 0.4531), you can see that the difference is not that big:

Code:
country   % wins webdip   %wins link   %wins webdip adj   ratio
England   12.63   6.46   5.72   89%
France   16.05   7.48   7.27   97%
Italy   9.08   4.56   4.11   90%
Germany   13.44   6.39   6.09   95%
Austria   11.38   5.92   5.16   87%
Turkey   17.92   6.32   8.12   128%
Russia   19.5   8.17   8.83   108%


We can see that Turkey is much more successful on webdiplomacy, at the expense of most other countries (except Russia).

Page 1 of 1 All times are UTC
Powered by phpBB® Forum Software © phpBB Group
https://www.phpbb.com/