Lies, Damn Lies and Diplomacy
Last issue I tried to distill whatever useful information there might be in the judges' registration databases. This issue we'll take a look at the Diplomacy Hall of Fame, which is faithfully maintained by Nick Fitzpatrick. What could be more appropriate for a Winter Adjustment issue than a summary of the latest Diplomacy results?
The Hall of Fame maintains a record of the outcome of every game played on any judge since 1991 (as well as some earlier non-judge games). I used version 18, which is dated September 1, 1995.
Of the 635 games which ran to completion (and were eligible for Hall of Fame points), almost exactly half resulted in a single power conquering all of Europe. The other half ended in some kind of draws. There were even two cases of a seven-way draw! Here's the complete breakdown:
It is interesting to note that two-way draws are much less common than three-way draws. This is no doubt because of the difficulty of arriving at an even 17-17 split without somebody getting greedy and grabbing that 18th center!
Of course, the question burning in everyone's mind is "which country wins the most?"
The answer is France, but it's a tight race. Poor Italy really is noticeably less likely to win than the other countries, but I'll bet you already knew that. Don't take these percentages too seriously, however. Even with about 45 wins per country, the margin of error is still plus or minus 1.4 points. We'll have to play a lot more Diplomacy to know the rankings for sure.
Solo wins only account for half of the games, so let's look at which countries are most likely to be included in a draw:
This chart shows that the three corner countries are almost 50% more likely to be included in a draw than the central countries (including Russia). Presumably this is because they are harder to eliminate. It is also interesting that Russia is the second most likely to win alone, but the second least likely to be included in a draw. Perhaps this confirms the common feeling that Russia is a "make or break" country -- you sometimes roll right over everyone, but if things go sour it can get pretty ugly. Italy still brings up the rear, but the gap is much smaller this time. The margin of error here is about 2.3 points.
So how can we reconcile these two charts? Which country is really best to play? If your goal is to maximize the number of Hall of Fame points you will receive, then we should calculate the average number of Hall of Fame points awarded to each country. This should appropriately balance the wins and the draws:
France is still on top, but England and Turkey now edge out Russia and Austria to share second place. For ease of copy and paste, here is the optimized Hall of Fame preference list:
The average game awards 4.9 Hall of Fame points and the average player receives 0.7 points per game completed. At that rate you can expect to finish 14 games to just make it onto the bottom of the Top 100 Players list. To catch Dan Shoham at the top of the list you can expect to play about 133 games!
Note that Austria and Italy do not occur in any of these, and these top seven represent 25% of all draws! Since there are too many possible combinations of allies to make an easy picture, let's just concentrate on the pairwise combinations. The small charts below show the number of games in which each country shared in a draw with each other country. The chart is symmetrical, since a two-way EF draw shows up in both England and France's charts. A three-way EFT draw is counted six times, twice in each of England, France and Turkey's charts.
The three combinations EF, ET and FT stand way up above the rest. Note also that France is the most likely ally of every country except for Turkey, where France is a close second to England. To simplify this even further I have created an "alliance map", in which I have drawn an arrow from each country to its two most likely allies. The thickness of the arrow (crudely) indicates the likelihood of the alliance.
Note that these are not necessarily the best or most likely alliances to form at the beginning or middle of a game. They are simply the ones mostly likely to finish the game.
I recently saw a reprint of an article by Mark Nelson (originally published in Electronic Protocol) where he analyzed the games played in EP to find the average length of time it takes each country to win. Here are the updated results using the latest Hall of Fame data.
The top chart shows the number of games which were won in each year. The earliest wins are in 1905, and the latest in 1938 (just in time to start WWII!). The lower charts break out the data for each winning country. The red arrows indicate the average winning time for each country.
It is immediately apparent that Russia wins sooner than the other countries. Fully half of all wins in 1905, 1906 and 1907 are Russian.
The average winning times are all virtually identical for all countries (1911.7), except Austria which tends to win one year earlier (1910.7), and Russia which tends to win two years earlier (1909.5). In computing these averages, I excluded games ending after 1925 (2% of the data) since these outliers perturb the results strongly, and there are too few of them to draw useful conclusions about long-running games.
Well, that's all for this issue. (Here's the raw data from which most of the charts were derived.) I'm pretty much fresh out of databases to analyze, so if you have some interesting information lying around but haven't had time to scrutinize it, send it my way....
If you wish to e-mail feedback on this article to the author, and clicking on the mail address above does not work for you, feel free to use the "Dear DP..." mail interface.