@ ghug: thank you for your explanation and I understand your meaning but you do not know what happened , so I think you are not able to judge but thank you for trying to help.

There are no strong nations or weak nations. That's part of the simply beauty of the game. The seven positions are for all purposes essentially balanced. Now some nations suit a particular play style more than others, but that does not make them inherently stronger or weaker. The perception that Italy and Austria are weak comes from players who most likely do not know how to play Italy or Austria. These nations properly played have every bit as much opportunity to win as any other nation. Each nation on the map face certain key challenges and risk events (English Channel, Black Sea, Venice/Trieste, Sweden, Tyrolia, etc.) at different points in the game. If you manage them properly, the very same perceived weakness can become a strength. Ultimately this game is about how you play, and how you interact with the other six players. The tactical superiority of one nation over another is a myth.

In the game you two had together, you mostly just overreact to a simple situation and refuse to let it deescalate. You were obnoxious in your public press, paranoid, and then you even held up a game because of your feelings about it. You also cast doubts on the viability of Eden as an ally for no reason. I think we are mostly able to judge, because we see almost all of your interaction with Eden.

RUFFHAUS, there are some positions that are tactically inferior. They have statistically always had worse results. Even the designer of the game says it is not a balanced game. I agree that Austria isn't as bad as it's made out to be, and Italy is strong if you get a good game going, but some powers don't have to deal with the same early challenges - English Channel is a contentious point, yes, but Black Sea can be bounced at almost no harm to Turkey, though a bit to Russia. Venice/Trieste is bad, which is why the game is unbalanced. Sweden is a negotiation tool, not a risk event. It is a reason for Russia to make friends with Germany. Tyrol is a problem for Italy/Germany/Austria, which is why Italy and Austria have weak positions. Only some powers can be attacked by significant force from three powers. Of these, Austria has no natural defensible border. Of course it's going to get murdered a lot. Italy has only three provinces to move to through land, of course it's going to end up in pieces or small. There are major weaknesses that are undeniable, and not every power shares them.

That's funny, as Austria I find that making an alliance with Germany with DMZs and a (never permanent) alliance between Austria and Italy, from there they have the best positioning to take Serbia and Greece, then move in on Russia too. Austria is my favorite country. Italy also works well. You have the advantage of CHOOSiNG who you want to attack

Turkey, just for it's amazing ability to bounce back from a terrible situation

Diplomacy is a game dominated by fashion and trends, and these fashions bring certain nations to the fore. When the fashion is for England and France to DMZ the Channel, England gets something of a bonus. In times when the Channel is far less certain England's fortunes have proved to be less impressive.

Tis a funny old game.

Ruffhaus - I agree with you wholeheartedly. And the data for draws shows that every country has an almost identical draw rate in the tournament, which supports what you're saying.

France is extremely clearly on a tier of its own, and then the other six countries are about the same.

The correction that one hopes might be made to the "site meta" is Italy attacking France more often. It usually works pretty well for all three of E/G/I so I'm surprised it doesn't happen more frequently.

Its because people are afraid of the stab coming from the East. For an Italian stab of France to work, he needs to go all in in 1901, which requires trusting Austria. Then it requires you to be vulnerable to an attack from the east for 1-2 years and hope Turkey can be contained. Its just too risky

I feel like Russia is the best if he can focus on either moving north or south. if he's threatened by England and turkey then it becomes very difficult to win.

I am surprised by England, but then again I haven't played England much within the past 20 or so gunboats I remember playing.

I do think that France tends to be stronger, but that also makes it a bit harder to play. Same with Russia. I've had success and struggles with both.

I personally like playing Austria (SC density makes it fun and easy if you aren't killed off by Italy) and Turkey because it is more likely I'll stay for a good game.

I believe my tendency as England was to convoy to Norway and go for stp... I should try attacking France first. Same with Italy. I usually go lepanto or stab of Austria. I should attack France more often :3

I think the rule here is that everyone should attack France more often haha.

I honestly can't remember the last time I opened to the channel. It just goes against my play style most of the time. And I think that's the larger issue. Every other nation can do an opening that's fairly noncommital, but if you are fighting France, you need to do a very blatant and obvious move towards him

I would be more interested in stats based on more games of all player skill levels, if we're trying to determine strength of various powers. I think the conversation here illustrates my objection- higher level (or at least more regular) players tend to be sensitive to trends and fashions. At the moment, without seeing any press, France can be almost 100% sure that Italy isn't opening to Piedmont, and that England is more likely to go for the northern opening than to open to the channel. That's prior knowledge that is both useful, and not likely to change much over the course of a single tournament.

If we're sticking to tournament games, I'd prefer analysis of multiple tournaments, but I think it'd be even better to stick to site-wide analysis- this has the advantage that it's easier to collect.

I think looking at the results of tournaments is certainly interesting- just that it is more likely to give an insight into current fashions.

Ruffhaus is right. The game has been played enough times that there is perfect information. When educated players are involved, there is no inherently weaker or stronger nations. There can be, however, weaker and stronger players.

Speaking of which, anyone interested in joining me for a series of 7 gunboats? I will be making a thread for it.

I strongly disagree with you Lando- perfect information is not the same thing as balance. In fact, perfect information is almost always assumed in game theoretic problems, and that certainly doesn't trivialize the whole field.

Statistically, France has a significantly larger collection of openings, the highest win and draw percentages than other countries, among te widest ranges of soloing possibilities, a nice range for army and fleet combinations and the win/draw percentages hold for both veteran games and games among varied skill levels.

By and far, I also agree that France has a setup that. Lends toward success.

Some arguments agains this are that Russia and Turkey solo much quicker, Italy and Russia have a higher spread of methods of soloing, if We discount games where Russia is eliminated, we see a much higher success rate compared to games where France is eliminated, turkey has far fewer opening move choices available allowing newer players fewer ways of making game-altering mistakes in 1901, and Germany along with France, has the easiest chance of obtaining 3 builds in 1901.

While each country has it's own strengths and weaknesses, France has more benefits available to it than nearly all other countries.

To be fair, each country does have a "fair" chance in this game. If we look at the strongest (statistically strongest) opening of each country, we see very comparable results.

I was going to stay out of this, but I couldn't help myself.

To really address the question of what country is best and whether it matters for good players versus poor players, we would need to do more descriptive statistics. You would need to do regression, and in this case, either multinomial logistic regression (win/draw/defeat) or do logistic regression for win and draw separately controlling for players' skills/ghostrating for when they participated in the game and the interactions of players' skills/GR and country-assignment. Then you analyze the odds ratios of how likely a country would win controlling for players' skills, how much players' skills matter, and whether players' skills interacted with country-assignment significantly.

So the dataset would need to contain country assignment, each player's skill at the time the game was played, and the outcomes for each game. Then you would run three main models, one with just the 7 countries, one with the 7 countries and the players' skills, and one with the 7 countries, the players' skills, and the interaction between country and skills.
Note: these are simple models not accounting for things like phase lengths, CDs, tournies, and so forth. Those factors may or may not matter, but the models become significantly more difficult if they do matter and the parameter estimates would then be biased.

If you don't have those covariates in the models, then all you are doing is looking at average statistics and you can't take players' skills into account.

If I have a ton of time one summer, I'll work with Alderian and Kestas to create the dataset and play around with it. But really, just from personal observation, I would hazard a guess that the outcomes for countries would be different for good players versus poor players as players' skills would equalize starting conditions ie countries.

I took the post to mean what you personally liked

It was meant to be a more statistical discussion, like Tru and zultar took it.

Zultar, I don't think CD's would matter if we assume they are randomly distributed, correct? And I'm not sure if a logit model would be most appropriate. We could do it by SC count, which would show the overall success of nations just as well as win/draw statistics.

CDs would matter because you can't assume that they are randomly distributed. My guess is that Italy and Austria would be more likely to CD than other countries. You can't do SC count because it doesn't indicate win/draw, which are categorical outcomes and the only ones we care about. It would make the model much easier to work with if we do SC outcomes but it wouldn't represent what we think of as win/draw. So a logit model is really the only way to go. And upon further thinking, I'm not sure how we would deal with draws because it is shared among multiple countries unlike win where only a single country wins.
Point is, notwithstanding descriptive statistics, to really analyze this sucker would take a lot of effort and econometrics.

I would look at it as a WTA game so SC count is useless. You can have 16 SC but if you lost Munich to the other guy you still lose.

@Zultar: could you give each country a score for each game, with a solo counting as 1, a draw counting as 1/n, with n being the size of the draw, and then 0 for anything else? Problem with that of course, is that you lose the distinction between draws and solos. Could you make each size of draw a different outcome?

Just throwing out ideas based on my one experimental design course 4 years ago!

Yeah, you're probably right zultar. I just don't know how to do multinomial logit models, so I was trying to think of ways to simplify the regression.

SC count is numerical data, not categorical. I do a lot of statistics with SC count, including one-tailed confidence intervals. SC count can indicate the capabilities of a country, but outcomes aren't as specific as one would like.

Nm, misread the post.

I was reading through the statistics you were talking about...good luck with that. Not sure what you would use as explanatory variables. You'd have to select those in such a way that when a person wants to determine which country is best for them, they would have to be able to identify these variables for themselves, and that becomes a recognizably harder to do.

@peterwiggin, the statistics he's describing are essentially this:

You obtain information about a group of people (for example years playing diplomacy, age, or other factors you believe contribute to a person's ability to win). You then run these through complex tests to determine the outcome that occurs given the above parameters. This would have to be done for each country. The assumption is that people that share specific characteristics will have similar outcomes with each country.

The problem I'm stating is that defining these characteristics becomes difficult for the tester to obtain and even more difficult for others to identify in themselves.