Wonderful Vaft. Really, really awesome.

I'd like to make a visual representation of the most commons sets for each country. I think I have editable diplomacy board like that kicking around somewhere on my computer.

Pretty cool dataset Vaft, I certainly appreciate all of your effort in this regard, its a lot of fun to peruse.

Out of curiosity, and with Kestas's permission of course, could we just not setup a server that hosts these stats and let anyone run queries? It'd be fun for the SQL junkies amongst us. I'm sure we could even program a GUI to let anyone look up at least basic stats.

ive already compiled a color coded SC control by a soloing country using photoshop. I have dubbed it the "solo radar" and as i told vaft, theres a pretty clear-cut trend for most of the countries about the stalemate line. France and Russia are quite a bit different, R more than F, but the rest often take 17 from one side and one or more from the other. I will eventually post the maps to a website ive been working on once my computer gets fixed.

@Tru, would they happen to look like the maps in this article? http://www.diplom.org/Zine/F2007R/Burton/statistician3.htm

Damn. That really one or two ups what I was going for.

Of course, those maps let me to make this meme: http://imgur.com/TpDoF

Theyre pretty similar but i think mine are easier to read. they dont contain numbers but the zones are color coded with blues and deep reds for 80% and higher and lighter reds, oranges and yellows for those below that and non colored for 10% and less.

That and i have the stalemate line located for an easy view.

Tru Ninja, yes, it was inspired by your solo radar, I should have attributed it to you, I'm sorry I didn't. I'm currently looking into auto-producing svgs from the data in a similar way to the page dubmdell linked to. Looking into the svg code, it doesn't seem that hard, and would avoid the time-consuming process of photoshopping.

Poozer, maybe, you could certainly ask him. Although considering server load, I don't know how much he would make public. The table containing moves has 21'186'273 rows times 11 columns. The table with territory status has 43'978'066 rows times 5 columns. It's a lot of data, and with some slight miscalculations, it takes a really long time to process it. Trust me, I know :)

Dont stress over it vaft. Ive been doing various stats for years for this game. I just havent published most of them yet. If you beat me to the punch, then so be it.

I don't see it as beating you to the punch, and I hope you don't. We have different methods, I approach it as a programmer (can I automate this to spew out data?), you as an actual statistician (but what does it really say?). I think a combination of the two is good. I value everything you write, and I try to learn from it. If you ever need data, I can produce it for you, or anyone else who asks. I hope you can share some of your T tests soon, it's fascinating and while I think I grasp the theory, the practical application is beyond me :)

T-tests are among the easiest to calculate. i can tell you how to do it in Excel, on a graphing calculator or by hand. If by hand, you just need to look up the critical values on t-tables located online.

You guys really want to one up that old map of centres in a solo?
Mapping country growth during a solo would make an awesome short clip. I wonder if there'd be a discernible pattern of how a country grows.

@Yonni: good idea. Maybe it can be done as an animation of maps showing the likelihood of a power owning each supply center, progressing year by year, where the probabilities are generated from games where that particular power soloed.

You know what? I think I can actually do that. I'll be back.

Haha. Awesome vaft. You're our very own diplocentric ask Jeeves.

Good luck. Id love to see your results, although im sure youll find some rather well structured things in the first few years but i think that there are so many possibilities for the remaining half dozen centers or so that the structure and predictability will begin to break down.

Lemme know what you find. It might be that each country has several routes that are common.

Hey Vaft - I've just seen 'Noi the albino'. Now in my mind you are a skinny hairless guy.

Lol very nice

fiedler, I have hair to my shoulders and a full beard. Sorry about that. I'm pretty skinny though. Or just slightly flabby. Tomato, tomato.

just his kind of guy ; )

UPDATE - I wrote a script that produces images. http://www.draugnar.com/VaftStats/ now has SC statistics in gloriously ugly html and illustrations. Make sure you press F5 when you go to the index page to get the new version!

Example image:
http://www.draugnar.com/VaftStats/scwin/regular/england.png

This has nothing to do with statistics, but I played around with the script mentioned above to produce this:

http://draugnar.com/VaftStats/finished.gif

It's an animation of the classic gameID=79015 won by Barn3tt. It's an early version without any units, just territories. It's pretty cool though. Do you have any ideas what I could use this code for? Also, make sure to read my previous post.

http://draugnar.com/VaftStats/test2.gif

I added units to the gif. No movement arrows or such yet, and it only shows the end result of a turn (for autumn, move, retreat and build phases are just one phase). Please help me spot any errors!

Here, have some more random stats: http://draugnar.com/VaftStats/autumn1901.txt

I checked how often a country owns a certain centre by autumn 1901, and then how often owning that centre in 1901 leads to a win. It only covers regular games, and as with all my stats, from version 0.82 of webDiplomacy. Before I do anything else with this data, I'd like your feedback on whether or not it's really useful, if it shows anything except "Early expansion is good".

What it DID give me was some really fucked up games that I thought were errors at first. I dug up the gameIDs, I present to you, gentle diplomats, the only game where Austria owns Ankara in 1901 - gameID=16259 - and the only game where England owns Portugal in 1901 - gameID=10735

It saddens me those never lead to a win.

That is not a very good way of doing those stats.

First percentage = own X/all games
2nd percentrage = own X/all wins
Wouldn't the percentage own X/all wins with X owned a better indicator of how important it is to own a certain centre?

From the looks of the maps though, it seems gunboat games are quite narrowly defined in the way that countries win. In press games the choice of winning SC's is much wider; in gunboat it's nearly always Tun - Mar - Mun - StP for France for example.

Gobbledy, I assume you mean the opposite, (wins with X)/(owned X). But what would that really say? It's comparing different orders of magnitude. Taking Norway as England has 1089 wins, 12037 occurrences, so 9.05% of taking Norway leads to a win. Taking Brest has 52 wins, 441 occurrences, so 11.79% of taking Brest leads to a win. Does that mean that taking Brest is more important than taking Norway? The difference in sample size is enormous. I think it's far more interesting to look at the number of wins and how many times it had a certain centre. The number of occurrences is mostly there as a point of reference.

Although I just realised that the biggest flaw is that it only looks at centres in an isolated way. My "England holds Holland" includes situations where England took only Holland, where Norway was also taken, where Brest was also taken, where London was lost, etc. I need to rewrite it to look at complete situations, it's pretty meaningless as it is now.