Luck Plays No Part in Diplomacy

[Your Humble Narrator]

One last point:

> I'd argue that's probably why your solo rate in classic press is 1/3rd mine and a number of other players who make a habit of winning those situations or avoiding them altogether.

Quis custodiet ipsos custodes when the custodes are hurting my feelings and I have no authority to cry to?

[RoganJosh]

@Swordsman, and anyone that cares.

There is no such thing as "optimal" moves. There will be a bunch of reasonable moves. You will have to compare the strengths of these moves somehow. You should assign weights to them, to the best of your ability. (Is that trivial? Not at all. But that's a different question.)

But here's the thing: What do you do once you have assigned weights? Say you got it down to three choices A, B, and C, with weights .6, .2, .2. Which option to you choose? Clearly, you think that A is the strongest move. So, do you simply pick A?

Always picking A is a sub-optimal strategy. What these weights tell you is that 6 times out of 10 you should play A, and 2 times out of 10 you should play B, etc. In a specific game, how do you choose? Optimal strategy is to choose at random according to the weights.

(No, I don't think that you always pick A. Most players know that always playing the same moves is a bad strategy, and they will use gut feeling or intuition or whatever to convince themselves to sometimes play B or C.)

For RPS you are aware of that optimal strategy is to make a random choice. It is not clear to me, from what you have written above, whether you are aware of that for diplomacy the optimal strategy is to make a weighted random choice.

The argument for that RPS is not random is that humans can't make random choices. That argument clearly applies to F2F diplomacy as well. But if we were playing on webRPS.net making choices in a dropdown menu with 24h phases, then nothing would stop you from actually using a random number generator.

These mechanics are the most clear in 1v1 diplomacy, where they are taken to the extreme. But they also apply to gunboat and to full press. But, let me emphasize this: there is a bunch of players from the top 10 of the 1v1 ELO ranking that regularly use random number generators. There is just no denying that, even in practice, random is part of this game.

[Kremmen]

First off, has anyone actually flipped a coin or used RNG to determine their moveset beyond maybe the first set of moves in a gunboat game?

Interesting question. In gunboat, no, but I almost never play gunboat.

Having played many hundreds of different board games, I've had discussions of this type before. Whatever the game, there are times when you have to make a decision and the decision of another player is crucial in how that will pan out. Let's say the other player is really good at reading the intent and strategy of others. Let's say you are an excellent technical player of the game, but not great at reading others or manipulating (by press in Diplomacy or by whatever other methods might be available) their actions. If you have a 50/50 (or even 60/40) decision to make and that opponent correctly guesses the actions of others 75% of the time, what is your best method to bring the odds back to 50/50 (or 60/40 or whatever)?

Could you, in the available minute or two (or even hours/days in online Diplomacy), learn all the psychology and other skills that your opponent has learned over years of play? Don't think so. You can, however, pull out your favourite randomiser (coin/die/phone app) and go with its choice. Advantage negated.

Most players won't do it, because we intrinsically prefer to be in control of our own destiny, even if we know that another course of action is safer. (The majority of people are happier to drive a car than to fly, despite cars being 2000 times more dangerous.)

[Squigs44]

Follow up:

If I were playing chess, and I got to a state where I personally believed that I had two possible moves that were equal in my success rate (Let's say that both of these moves *could* be countered by my opponent, but only by one specific move the next turn). So I have a firm belief that my success is a 50/50. Then I flip a coin to determine my move. In the heads outcome, I move my pawn, my opponent missed the counter, and I win. In the tails outcome, I move my bishop, my opponent sees the counter, and I lose. Thus, my outcome depends on the coin. Does this mean chess has some luck in it because I as a player can introduce luck into my personal decisions?
I understand that this situation is not the same as a diplomacy situation, but the principle remains. Someone deciding to flip a coin on a belief that a decision is 50/50 does not mean that the game has luck in it.

[EvW]

Your chess analogy does not really work. Certainly, there are situations where you have several (let's say 2) winning attempts, that will all result in a draw if your opponent find the right continuation. However, there is no guessing at all involved - or at the very least, it will be an educated guess with a reason. In these cases, once the first player has spotted his 2 winning attempts, he will evaluate which continuation is the most difficult for his opponent to spot or correctly calculate. Now, it's not a given that there is a clear choice here, and his move selection (or winning attempt) could indeed be random (although, more likely, based on their instinct). Whichever option is chosen, though, his opponent won't have to guess at all. Regardless of the first player's choice, the second player needs only find the good response to the move the first player chose. So the outcome never depends on the coin, as whatever the first player does, the 2nd players can respond with the counter move, as he already has seen the move the first player picked.
Sorry for that long explanation, but I thought I'd explain, as I'm crap at diplomacy but at least mediocre at chess.

[RoganJosh]

If I were playing chess, and I got to a state where I personally believed that I had two possible moves that were equal in my success rate (Let's say that both of these moves *could* be countered by my opponent, but only by one specific move the next turn). So I have a firm belief that my success is a 50/50. Then I flip a coin to determine my move. In the heads outcome, I move my pawn, my opponent missed the counter, and I win. In the tails outcome, I move my bishop, my opponent sees the counter, and I lose. Thus, my outcome depends on the coin. Does this mean chess has some luck in it because I as a player can introduce luck into my personal decisions?
I understand that this situation is not the same as a diplomacy situation, but the principle remains. Someone deciding to flip a coin on a belief that a decision is 50/50 does not mean that the game has luck in it.

Optimal strategy in chess does not have randomness. Which, loosely speaking, means that if you introduce randomness into you strategy then you will play worse.

Optimal strategy in Diplomacy has randomness. Which, loosely speaking, means that if you (accurately) introduce randomness into your strategy you will play better.

Your example is weird because you are comparing a situation which you allegedly believe is 50/50 (though, in fact it is not, and you should know better) with a situation that, strategically, is 50/50, and where (if you have no further information) your best choice is to randomize.

[Squigs44]

Optimal strategy in Diplomacy has randomness. Which, loosely speaking, means that if you (accurately) introduce randomness into your strategy you will play better.

I think this is the main point of disagreement. I don't think randomness is the optimal strategy, I think outguessing your opponent can yield better results. Jmo's 75% success rate at alledged 50/50s is a case study of this point.

[Restitution]

I think this is the main point of disagreement. I don't think randomness is the optimal strategy, I think outguessing your opponent can yield better results. Jmo's 75% success rate at alledged 50/50s is a case study of this point.

It is Game Theory Optimal, 100%, to behave randomly. In any given particular game, against non-optimal opponents, you can outperform optimal strategy.

"Optimal strategy" is the strategy against which the optimal strategy is the best counter. So, fex, in RPS, the optimal strategy is to play each one at 33%, because it can't be exploited by any other strategy. But even though playing each at 33% is optimal, against an opponent only playing paper, obviously you ought to play only rock if you want to maximize performance.

RoganJosh is talking about the game theory conception of optimal strategy. Perfect Diplomacy players will behave randomly. This isn't the case for Chess or Go or whatever, which are games which have discrete solutions.

[RoganJosh]

I think this is the main point of disagreement. I don't think randomness is the optimal strategy, I think outguessing your opponent can yield better results. Jmo's 75% success rate at alledged 50/50s is a case study of this point.

If Jmo's opponents would have been smart enough to use a random number generator, then Jmo's success rate at 50/50s would have been 50%.

[Restitution]

If Jmo's opponents would have been smart enough to use a random number generator, then Jmo's success rate at 50/50s would have been 50%.

That is a great way to put it.

[swordsman3003]

I think this is the main point of disagreement. I don't think randomness is the optimal strategy, I think outguessing your opponent can yield better results. Jmo's 75% success rate at alledged 50/50s is a case study of this point.

If Jmo's opponents would have been smart enough to use a random number generator, then Jmo's success rate at 50/50s would have been 50%.

And if Jmo had been foolish enough to use a random number generator, then Jmo's success rate at 50/50s would have been 50%.

The idea of literally using a random number generator to make your decisions is something, in all honesty, I have not considered and perhaps never even really heard of? Like, the idea is so outrageous to me that I might have seen it somewhere before and just forgotten or interpreted it metaphorically.

In my opinion, this is the most interesting concept to come out of the conversation. It raises a practical question: "When should you use a random-number generator to make your decisions in Diplomacy?"

There's nothing confusing about this question, but I don't think the answer is as obvious as some commentators are making it out to be.

I'll tell you right now, I became a top player on this site using the answer "never." Maybe using a random-number generator would make me better...or maybe it would make me worse.

I think I will start researching and write a companion piece for my blog.

[Squigs44]

I think this is the main point of disagreement. I don't think randomness is the optimal strategy, I think outguessing your opponent can yield better results. Jmo's 75% success rate at alledged 50/50s is a case study of this point.

If Jmo's opponents would have been smart enough to use a random number generator, then Jmo's success rate at 50/50s would have been 50%.

If jmo used a random number generator, his success rate would have been lower, and therefore his strategy not optimal (against this opponent). If his opponent had been using a random number generator, then his strategy would have netted him a 50% success rate, the same as using a random number generator. However, jmo doesn't know what strategy his opponent will apply, so using the outguess strategy has a higher success rate overall.

Optimal strategies against optimal opponents are not optimal strategies against irrational opponents. Take this 538 Riddler for example: https://fivethirtyeight.com/features/ar ... t-warlord/
In the express Riddler, if everyone were playing optimally, then the optimal strategy would be to choose a 1. Yet the answer in this game is never 1, because there are people who don't play optimally. So the optimal strategy against nonoptimal players is to choose some number higher than 1. Did the winner of this challenge win because he was lucky? No, he won because he outguessed other people.

[RoganJosh]

@Swordsman, Squigs, & Restitution

Optimal strategy in Diplomacy in the game theory sense includes randomness. If your opponents don't play by optimal strategy, then you should try to exploit that. I think we're all in agreement here.

Here's a (maybe the only..?) situation when randomizing your moves make perfect sense:

Maybe you don't play by Optimal Strategy. Maybe you have a preference for Berlin. And you are playing jmo. And he is so smart that he is gonna try to exploit this.

That means that jmo is not playing by Optimal Strategy. Which you can exploit, by playing Munich against him.

But that means that you are not playing by Optimal Strategy. Which jmo can exploit by predicting that you are gonna try to exploit him.

But that means that jmo is not playing by Optimal Strategy. Which you can exploit by predicting that he is gonna try to exploit that you are gonna try to exploit him.

The chain doesn't end. And if both you and your opponent are competent enough to see the whole chain, well... Then there really isn't any way of exploiting your opponent. And the only way you can make sure that you are not being exploited is to randomize your choice.

[mhsmith0]

I think the trivial answer to "when should you use a RNG to make a decision" is something like "when there are 2+ options of approximately equal attractiveness and when I have no real way to sort between then", bonus points for actively wanting to avoid being predictable.

As a 1v1 player who's at least sometimes been top 10 (even if i did just get bumped down ), I can say that if I'm playing against a top player, I'll use RNG if I don't feel particularly confident at outguessing them. Of course, sometimes "RNG" is more of a gut-type "what do I feel like playing this game" but it's not super hard to grab a coin and flip it either.

If I'm playing gunboat on an anon board, my openings will usually be whatever I think is the optimal opening (a more random process would be when I wasn't really sure what is best - for instance, as Russia I'd open to STP instead of Ukr a non-zero percent of the time; whereas as England I'm probably opening NTH/ENG/Yor basically always because I at least *THINK* that's best for England [among other things, the other two openings kind of SUCK if there's realistic odds of France opening to the Channel; the Wales army is useless if you bounce, and a vacated channel is brutal if France gets in] ). Maybe if I was on a board with people who knew I was a player and who had some insight into my tendencies I might vary that a bit (ditto if I was on a board w/ people whose tendencies I'd had insight into), but that's kind of where I'm at.

I'd probably also RNG some decisions around coinflip scenarios if I lacked any real insight into what my opponents might be doing, but I'd definitely prefer to just have that insight.

PS A more TLDR: flipping a coin prevents your opponents from getting the better of you, but it also prevents you from getting the better of your opponent. As a strategy it becomes more attractive the more likely it is that you're outmatched, and less attractive the more likely it is that you think you have your opponent(s) outmatched. Or to put it another way: I'd be likelier to flip a coin against RoganJosh in a 1v1 than I would against some newer player who I felt confident about outguessing

[Squigs44]

I will concur that if every single person in the world were to use a "coin flip optimal strategy" everytime they get into a Diplomacy "50/50" situation that jmos strategy is no more or less optimal than using a coin flip.

But that's not the case, and never will be the case, and never has been the case. The game of Diplomacy was designed for humans, who do not always make optimal moves. The game was not designed to have coin flip optimal decisions and will never have coin flip optimal decisions. If you think diplomacy has luck in it because of coin flip optimal decisions, you misunderstand the game, or humans in general.

[Restitution]

If there is a minute fraction of a chance that your opponent will flip a coin, then there is some amount of luck in Diplomacy.

[Carl Tuckerson]

I think this is the main point of disagreement. I don't think randomness is the optimal strategy, I think outguessing your opponent can yield better results. Jmo's 75% success rate at alledged 50/50s is a case study of this point.

If Jmo's opponents would have been smart enough to use a random number generator, then Jmo's success rate at 50/50s would have been 50%.

Optimal strategies against optimal opponents are not optimal strategies against irrational opponents. Take this 538 Riddler for example: https://fivethirtyeight.com/features/ar ... t-warlord/
In the express Riddler, if everyone were playing optimally, then the optimal strategy would be to choose a 1. Yet the answer in this game is never 1, because there are people who don't play optimally. So the optimal strategy against nonoptimal players is to choose some number higher than 1. Did the winner of this challenge win because he was lucky? No, he won because he outguessed other people.

Off-topic but I am curious about how this problem works.
I think I get why the optimal answer is one: whatever your starting presumption for others' guesses is, the correct solution is always lower than that, and so as you work through sequential levels of understanding of the problem, the correct guess always gets lower and lower, until it finally hits the floor of one. If we presume at least three players, no one person going against the grain can swing the average in his favor, so there's no incentive not to match the progressively downward trend of the levels.
And this guy got it right by guessing a higher number than one because many participants didn't bother or didn't know to keep moving downward.
Is that explanation on the right track?

[Restitution]

Off-topic but I am curious about how this problem works.
I think I get why the optimal answer is one: whatever your starting presumption for others' guesses is, the correct solution is always lower than that, and so as you work through sequential levels of understanding of the problem, the correct guess always gets lower and lower, until it finally hits the floor of one. If we presume at least three players, no one person going against the grain can swing the average in his favor, so there's no incentive not to match the progressively downward trend of the levels.
And this guy got it right by guessing a higher number than one because many participants didn't bother or didn't know to keep moving downward.
Is that explanation on the right track?

Yes. The optimal strategy is the one that cannot be exploited, not the one which does the best against sub-optimal strategies.

An optimal strategy is fundamentally *not* the same as the "best strategy".

[RoganJosh]

The game was not designed to have coin flip optimal decisions and will never have coin flip optimal decisions. If you think diplomacy has luck in it because of coin flip optimal decisions, you misunderstand the game, or humans in general.

I think you are the only one using the term "luck" here.

The game of diplomacy has randomness in its optimal strategy. In certain positions, your best choice is to flip a coin. That's not an opinion, that's simply how the game is designed.

[Squigs44]

I think you are the only one using the term "luck" here.

Im confused. The entire discussion is built around the viewpoint that "Luck Plays No Part in Diplomacy".

3 Reasons Why "Luck Plays No Part in Diplomacy"
Solo Win Tip #2: There's No Such Thing As Luck

There may be no "luck" involved but there sure is a good amount of "variance" in these games concerning effects restricted to an ongoing game which you cannot influence.

I believe luck comes into play when all players are behaving rationally

Of course there's luck. Diplomacy has Nash equilibriums.

Even in a CHAOS gunboat, which should in theory be the most luck driven game possible, I was able to outmaneuver people with strategies like my convoys in Autumn 06 and 09. http://webdiplomacy.net/board.php?gameI ... #gamePanel

I think the claim that "luck" plays no part in Diplomacy isn't true for most colloquial uses of the term, which really means that there are situations where you can make optimal decisions every time and still lose.

Basically, I think it's a valid definition of "luck" to say that a game entirely devoid of luck has no such variance (e.g. chess or go).

No, optimal strategy in Diplomacy will always be of the form: Submit order set A 75% of the time, and submit order set B 25% of the time. In each specific game, you should pick at random. This is what introduces an element of luck to the game. (I'm not making this shit up: it happens regularly in high level 1v1 games that people even use pseudo-random number generators.)

I believe that I was unironically using the RNG abbreviation to describe whether you happen to rand an opponent who will use a friendly or unfriendly strategy against you, particularly in spring 1901 (and if they choose opening moves off of some kind of rand metric, then there's obviously luck involved in which such move they choose in this particular game).

That's 9 people (including yourself) that have used the term luck besides me. I am going to give you the benefit of the doubt and believe that you just left out some context to your statement here and that you didn't actually believe that I was the only person using the term luck.