Just Played a Few Games

AnoHito · 2014-11-26 09:22:41

Right, and that being the case, the better player can and should reset the stakes after building a lead, and so they will anyway. By doing it automatically it helps to encourage a back and forth game between two players, rather than having the first player to get a lead abandon the game instantly. Personally, I enjoy a good back and forth game more than I do having to leave matches constantly because not doing so would put me at a disadvantage. And if the other player is so much better than me that I have no hope of winning, I wouldn't stick around either way.

AnoHito · 2014-11-26 11:51:08

GGs to whoever I was playing last night/this morning. So I think that once I got a feel for how you were picking numbers the first round, and refined my betting strategy a bit, my strategy was starting to give me a statistical edge. It was slight, and the game essentially turned into a war of attrition, but after I reach the point of equilibrium I finally started to move out ahead. I noticed you were avoiding necessarily picking the highest value in your column with the lowest high value, and I was a bit surprised that it actually did seem to work in your favor. Not knowing what you column you picked was actually worse for me than you getting a low number in the first round. But it still wasn't quite enough. I would have liked to play more to fully test my approach, but I could tell it was going to take hours to come back at the rate we were going, and I was getting a little fatigued from playing for so long. One thing I'm sort of curious about now, is if I could write a bot that could play Cordial Minuet with an optimum strategy. Taking the guesswork out of what to pick for yourself in the first round greatly simplifies things. The biggest challenge would be figuring out optimal counter picking strategies for the first round, and implementing an optimal betting strategy. Still, it seems doable, and if actually playing the game is going to be such a slow process, writing a bot in the long run might be more fun than playing it myself...

jere · 2014-11-26 12:59:52

This will reset the stakes, and make it less likely the other player can gain back what they lost... It's a solution in that it allows you to continue playing against someone uninterrupted. It's extremely annoying to spend time playing against someone, trying to learn their tactics, only to have them leave the second they have a lead.

I briefly had this concern, but realized it makes no sense. Once you've lost 10 chips, you've lost 10 chips. Your opponent is also giving up the knowledge of your tactics and since they're the ones beating you, I'd say the loss is theirs. The whole concept of "winning back" what you lost smells like sunk costs and gambler's fallacy to me.

If the games are consistently short, that indeed seems annoying, but I haven't had that happen very often.

I have gained some serious insight into what an optimal strategy for Cordial Minuet should look like. It seems like in the first round, strategy should be dictated by the rows/columns with the lowest high number. In a typical first round, you should always give your opponent the column with the lowest high number, because it ensures a level or predictability (you know they cannot have a number above that number).

I'm not convinced. I don't think the word "always" should ever be mentioned while discussing column picking. If you're adopting the same column picking strategy every time, you're easily beaten. If your opponent was always giving you the row with the lowest high number, then they are easily beaten because you get like a 29 and you've probably got a 50% chance to give them sub 10.

My strategy yesterday was actually to pick one of the highest numbers to min/max my hand. I'd rather know out of the gate if I have the advantage or not.

If there ever is an "optimal strategy" I expect it to have more to do with betting than anything else.

I do think writing a bot would be interesting though.

Last edited by jere (2014-11-26 13:01:35)

jere · 2014-11-26 13:38:00

Like the TCD forum, I want a thread where I can say "IS THAT YOU?"

I just played a game and I'd bet a lot of coins it was Ano. At least they had Ano's optimal column picking strategy.

Yea, I think I'm playing Ano over and over again.... the "optimal strategy" is incredibly predictable.

Last edited by jere (2014-11-26 15:20:01)

donkeyspaceman · 2014-11-26 15:35:01

jere wrote:

This will reset the stakes, and make it less likely the other player can gain back what they lost... It's a solution in that it allows you to continue playing against someone uninterrupted. It's extremely annoying to spend time playing against someone, trying to learn their tactics, only to have them leave the second they have a lead.
I briefly had this concern, but realized it makes no sense. Once you've lost 10 chips, you've lost 10 chips. Your opponent is also giving up the knowledge of your tactics and since they're the ones beating you, I'd say the loss is theirs. The whole concept of "winning back" what you lost smells like sunk costs and gambler's fallacy to me.
If the games are consistently short, that indeed seems annoying, but I haven't had that happen very often.

I haven't played a ton in the last few days so I don't know if more of this is happening now, but as I was playing a lot this past weekend I very rarely (if ever) had anyone leave the match before somebody had been completely cleaned out. I had a lot more fun (and ultimately, profited a lot more) playing until I had an opportunity to get them to call my bluff and go all in. Even when I was the one getting cleaned out, I felt like I would have been able to make a comeback with a more cautious betting (I rarely fold, which is something I need to work on).

I actually feel pretty neutral on the idea of punishing a player for leaving in some way because I'm not convinced it happens very often (although it could be that people just like to play and are afraid they won't be able to get into another match immediately if they leave...this could be different when the game is public), but I'm inclined to say it's fine the way it is.

jere wrote:

I have gained some serious insight into what an optimal strategy for Cordial Minuet should look like. It seems like in the first round, strategy should be dictated by the rows/columns with the lowest high number. In a typical first round, you should always give your opponent the column with the lowest high number, because it ensures a level or predictability (you know they cannot have a number above that number).
I'm not convinced. I don't think the word "always" should ever be mentioned while discussing column picking. If you're adopting the same column picking strategy every time, you're easily beaten. If your opponent was always giving you the row with the lowest high number, then they are easily beaten because you get like a 29 and you've probably got a 50% chance to give them sub 10.
My strategy yesterday was actually to pick one of the highest numbers to min/max my hand. I'd rather know out of the gate if I have the advantage or not.
If there ever is an "optimal strategy" I expect it to have more to do with betting than anything else.
I do think writing a bot would be interesting though.

Are any of you familiar with the concept of "Donkeyspace"? As an overall goal, I often try to give my opponent "playable" values (20-30 range) while I aim for higher values on the first turn. If I give them something too low, they'll fold, but as long as I have a slight edge, I feel comfortable betting like I'm just trying to see how it plays out. When it gets down to the reveals, I try to trick them into thinking I have something low so they bet more (or I bet high when I actually have nothing and try to buy the pot).

After working the above strategies for a round, I assume the opponent knows them. So I adjust. I pick a column with a "playable" value for myself and select the column with the 36 in it for them, assuming they'll think I picked it.

The point is, I don't know if there's an "optimal" strategy for picking columns, because it doesn't always matter if you have the "optimal" score. I've played games with very low scores and still won big by bluffing the opponent into a fold. Playing "optimally" will win you a tiny payout, but navigating the donkeyspace is what gets you the big payouts.

AnoHito · 2014-11-26 15:54:14

donkeyspaceman wrote:

Are any of you familiar with the concept of "Donkeyspace"? As an overall goal, I often try to give my opponent "playable" values (20-30 range) while I aim for higher values on the first turn. If I give them something too low, they'll fold, but as long as I have a slight edge, I feel comfortable betting like I'm just trying to see how it plays out. When it gets down to the reveals, I try to trick them into thinking I have something low so they bet more (or I bet high when I actually have nothing and try to buy the pot).
After working the above strategies for a round, I assume the opponent knows them. So I adjust. I pick a column with a "playable" value for myself and select the column with the 36 in it for them, assuming they'll think I picked it.
The point is, I don't know if there's an "optimal" strategy for picking columns, because it doesn't always matter if you have the "optimal" score. I've played games with very low scores and still won big by bluffing the opponent into a fold. Playing "optimally" will win you a tiny payout, but navigating the donkeyspace is what gets you the big payouts.

Well, my goal right now is to figure out a strategy that "breaks" the game. In other words, a strategy where the best counter strategy the other player could use would result in a 50% or less chance of winning. In that sense, I'm not really looking for big payouts, I'm looking for ways to exploit the game that will guarantee profit (if only a little bit). That's what testers are supposed to do right? It's not like I'm going to get rich playing $0.01 matches.

But at the moment, I'm thinking that my first idea was a failure, because it seems that when the other player does not pick the highest number in the column with the lowest high number consistently, losing the advantage of knowing which number they picked takes away too much of your advantage for the strategy to by viable. Now I'm trying something new, and I defy anyone to figure out what strategy I am actually using.

Last edited by AnoHito (2014-11-26 15:54:49)

jere · 2014-11-26 16:13:15

Are any of you familiar with the concept of "Donkeyspace"?

How apropos, since that video is supposedly part of the inspiration for Cordial Minuet. I think I'll watch it now.

donkeyspaceman · 2014-11-26 16:13:47

That's fair. I do think it would be interesting (and valuable for Jason) to know if the game can be "broken" mechanically, I just don't think anyone would want to use such a strategy if it did exist.

Let us know how that bot works out.

donkeyspaceman · 2014-11-26 16:15:48

jere wrote:

Are any of you familiar with the concept of "Donkeyspace"?
How apropos, since that video is supposedly part of the inspiration for Cordial Minuet. I think I'll watch it now.

I wasn't aware of that! Now I understand why I was so attracted to the concept of this game when I first heard about it.

jere · 2014-11-26 17:12:24

Now I'm trying something new, and I defy anyone to figure out what strategy I am actually using. wink

Well, the player I've been playing against has been going on all in on a large number of hands. I think they are simply going all in when they have a high first tile. Playing against this style is nerve-wracking, though I've won out against it at least as much as I've lost.You either get bullied into folding, or at best, you shove knowing you've got a 50-50 shot. On the other hand, if them shoving tips you off to which row they picked, that gives you quite a bit of information.

AnoHito · 2014-11-26 18:51:30

Yes, going all in so much was something of an experiment, but I've revised my strategy a bit because the way I was doing things gave too much information to the other player (they always knew if I had a good opening hand, because it's almost always a bad idea to go all in on a complete bluff). I think the way I'm doing things now is working better though, as my stake has been increasing for the first time since I started experimenting with fixed strategies. But it's still too early to tell if my new strategy is viable. At the moment, I'm mainly concerned with what betting strategy best complements my picking strategy.

jasonrohrer · 2014-11-26 22:41:57

Long, great thread. Some quick answers as I read through.

I don't want to give the player too many strategy crutches. The graph on the side is just doing the math for you, especially in the end game, where any skilled player would be taking out a calculator (because knowing when you have a score that is higher than any of your opponent's scores is important, for example).

I made the game in C++ because I want it to run on old, slow computers. It currently runs at 60FPS on my 14-year-old laptop (my main dev machine). Once a bunch of brush strokes are added to the board (by the end of the game), that slows down to 30FPS or something like that. In v9 (not released yet), my 14-y-o laptop is able to recompute the possible score graph in realtime as you drag, exploring 14,000 possibilities without a framerate stutter.

These kinds of things would barely be possible on a modern beast machine using a non-native language (and most people actually don't have beast machines). And forget about my 14-y-o laptop. I did play around with Flash a bit at the start, and found that I could barely draw 10 circles on the screen at a reasonable framerate.

Porting the game to iPhone (assuming that apple would approve it) would require changing one file, because C++ is an iPhone target language. And then it would be a real app, not running in a browser.

C++ code is far more portable than people might think. I had the game running on Raspberry Pi on my TV set yesterday---it just compiled and ran there.

As far as dev time... well, I have so much networking code and stuff written already from past games (this is C++ game 11 for me) that it's not very much work. I'm coding very little except the game logic itself. The protocol is all HTTP, and I already have very easy functions written to do that. Loading a sprite from TGA and then drawing it on the screen is two function calls. And I have a built-in game recorder (that I coded up three games ago), translation engine, settings engine, etc.

It also took time to hand-draw and scan the graphics and just generally get the whole thing done. There are more moving parts that you might think. On the server side, there was lots of API stuff to deal with for processing credit cards and sending checks out.

But yes, please do port or rewrite your own client for browsers and such!

jasonrohrer · 2014-11-26 23:27:35

More responses:

Yes, this game is 100% skill in that there are no random elements. It is not "subject to chance" according to legal interpretations.

You're right that very few people bet money on "hard skill" games like chess. You can rank players with such fine granularity that the pool of evenly matched players at a given level is very small. And only a fool would play for money (or play at all, really) against someone above their own skill level.

But this game is an experiment in what I'm calling "soft skill." What are you doing in CM to win? You're not searching the 500,000-leaf game tree (though doing that would help you). You're not crunching loads of math and probability (though doing that would help you). You're learning to read and predict the behavior of your opponent and learning to be unpredictable yourself.

So, whereas hard-skill games involve complex (in both senses) algorithms to play well, and thus require years of study to get good at, with an amateur absolutely standing no chance against a more skilled opponent.... games of soft skill are different.

People don't just come back to poker because they might get lucky next time. They come back because "reading their opponent" and "bluffing better next time" seem like reasonably attainable goals. The soft skill is based on intuition, and it feels more like an "art" than a "science."

But yes, Poker also has a strong random element. This game tries to build a system of soft skill, akin to the soft skill present in poker, from the ground up with no random element.

As far as I'm aware, there is no precedent here, so I really have no idea where the player ecosystem will end up in the long run.

Regarding your strategy of always giving the opponent the row with the lowest high number:

Let's say you do that consistently. You know your opponent will never get more than a 25 or whatever on turn one. You can box your opponent in that way, and over time, train them to always pick the row with that 25 (or whatever) in it. Over time, you can predict exactly what row they will pick for themselves.

But once you have them trained that way... would you really give them the 25 every time? There's certainly some lower number in the row that they picked. If you knew what they would pick, you'd probably switch at some point and stick them with a lower number.

A pure Nash equilibrium only happens when both players know what their opponent will pick and still don't want to switch. In your example, this only operates in one direction. Your opponent knows what column you will pick for them and still doesn't want to switch (they want to take the 25). BUT, it's not a Nash equilibrium because you, knowing that your opponent will do that, would want to switch columns.

jasonrohrer · 2014-11-26 23:38:18

And yeah, "Life and Death and Middle Pair" is one of the greatest talks in the history of the video game industry.

Frank is the one who got me thinking about Poker. He gave another talk the next year about how game theory saved the world from a nuclear holocaust.

donkeyspaceman · 2014-11-27 02:23:02

jasonrohrer wrote:

And yeah, "Life and Death and Middle Pair" is one of the greatest talks in the history of the video game industry.
Frank is the one who got me thinking about Poker. He gave another talk the next year about how game theory saved the world from a nuclear holocaust.

Hear, hear! I loved that talk (obviously, considering my username).

I'm not sure I've seen the other one you're referencing. Link?

.. · 2014-11-27 12:18:48

GGs to whoever I was playing last night/this morning.

It was me again. Partway through I actually had an insight not into your strategy but into the game itself, which is why you went from winning to losing. In fact I switched from trying to determine your strategy to just assuming that you hadn't had the same realisation. Obviously I can't tell you what it is However you played so conservatively that I gave up on extracting much from you and just tried to get you to concede one coin at a time. I definitely could have done far better there.

BTW, we played 57 rounds in that second game!

I noticed you were avoiding necessarily picking the highest value in your column with the lowest high value, and I was a bit surprised that it actually did seem to work in your favor.

I couldn't parse this sentence, so I can't confirm that I was doing whatever you mean. Later I switched to something definitely completely different.

(Haven't read the rest of this thread; bedtime)

Last edited by .. (2014-11-27 12:22:46)

jasonrohrer · 2014-11-27 22:13:44

donkeyspaceman wrote:

I'm not sure I've seen the other one you're referencing. Link?

I don't know if the GDC version is online for free anywhere, but he re-gave the talk at NYU later:

http://vimeo.com/88998276

CORDIAL MINUET ENSEMBLE

#26 2014-11-26 09:22:41

Re: Just Played a Few Games

#27 2014-11-26 11:51:08

Re: Just Played a Few Games

#28 2014-11-26 12:59:52

Re: Just Played a Few Games

#29 2014-11-26 13:38:00

Re: Just Played a Few Games

#30 2014-11-26 15:35:01

Re: Just Played a Few Games

#31 2014-11-26 15:54:14

Re: Just Played a Few Games

#32 2014-11-26 16:13:15

Re: Just Played a Few Games

#33 2014-11-26 16:13:47

Re: Just Played a Few Games

#34 2014-11-26 16:15:48

Re: Just Played a Few Games

#35 2014-11-26 17:12:24

Re: Just Played a Few Games

#36 2014-11-26 18:51:30

Re: Just Played a Few Games

#37 2014-11-26 22:41:57

Re: Just Played a Few Games

#38 2014-11-26 23:27:35

Re: Just Played a Few Games

#39 2014-11-26 23:38:18

Re: Just Played a Few Games

#40 2014-11-27 02:23:02

Re: Just Played a Few Games

#41 2014-11-27 12:18:48

Re: Just Played a Few Games

#42 2014-11-27 22:13:44

Re: Just Played a Few Games

Board footer