webdiplomacy.net

RoganJosh wrote: ↑

Thu Nov 12, 2020 6:47 pm

... it always builds (A Tri, A Vie)

That should be (A Tri, A Bud).

Got a France win with no misorders to ruin the ending.
http://webdiplomacy.net/board.php?gameI ... ocache=385

http://webdiplomacy.net/board.php?gameID=332015

Very solid play throughout. It survived the first 50/50 I had for the win, but not the second. Lightyears better than Kestas Bot. Well done!

I'm still curious as to how repetitive it is.

RoganJosh wrote: ↑

Thu Nov 12, 2020 9:58 pm

http://webdiplomacy.net/board.php?gameID=332015

Very solid play throughout. It survived the first 50/50 I had for the win, but not the second. Lightyears better than Kestas Bot. Well done!

I'm still curious as to how repetitive it is.

I replayed Eden's game's start, and it bounced me in Pie instead of Mun in S02:
http://webdiplomacy.net/board.php?gameI ... #gamePanel

I promptly misordered my fleets. Please don't look at the rest of the game.

Oh ghug....your start had such promise.

I think the bot gets us because it doesn't go as hard at Germany out the gate as many of us are used to. It takes a little more time picking up SCs and overwhelms you.

Kinda reiterates my thoughts that going for the Med is never worth it as France.

Yeah, attacking Italy as France is something you only wanna do if Austria goes hard on Germany and Scandinavia, so obviously not against this bot.

ghug wrote: ↑

Thu Nov 12, 2020 10:24 pm

I replayed Eden's game's start, and it bounced me in Pie instead of Mun in S02:
http://webdiplomacy.net/board.php?gameI ... #gamePanel

I've seen it do two or three different order sets for S02, but they're all variants of the same Austrian rush strategy. For France it's like, send the fleet north, and make sure you get into Munich in 1902. It's still an open game, but you'll have a good shot at winning it.

ghug wrote: ↑

Thu Nov 12, 2020 10:24 pm

I promptly misordered my fleets. Please don't look at the rest of the game.

I felt this in my soul. At least mine was karmic punishment for bragging about the W before earning it.

https://webdiplomacy.net/board.php?game ... ocache=384

I used to be a good FvA player, but have become rather lousy. The bot's good, but not nearly unbeatable.

http://webdiplomacy.net/board.php?gameID=332148
Another one for the good guys. This one felt *much* easier than my last few. It going for that self bounce in Tyrolia in fall 1902 did not seem good. What else was I going to do with my army PIE here?

I think the nature of simultaneous turns is going to make an "unbeatable" AI impossible due to how much randomness there is. Better than any human should be doable, but not to such an extent we can never cheese off a win here or there.

It's still winning at an absolutely blistering 80% clip. To the average player it's probably pretty close to unstoppable.

I finally clutched out a win as France, and yes that means getting to 18.
http://webdiplomacy.net/board.php?gameID=332146

The bot's affinity for its One True Opening with either power - Romanian Anti-Alpine with Austria, Maginot with France - does make it a bit exploitable if you understand the RPS-esque mechanics of openings.

President Eden wrote: ↑

Fri Nov 13, 2020 4:54 am

It's still winning at an absolutely blistering 80% clip. To the average player it's probably pretty close to unstoppable.

I finally clutched out a win as France, and yes that means getting to 18.
http://webdiplomacy.net/board.php?gameID=332146

The bot's affinity for its One True Opening with either power - Romanian Anti-Alpine with Austria, Maginot with France - does make it a bit exploitable if you understand the RPS-esque mechanics of openings.

What's remarkable here is the bot creators were hoping for better performance than they're seeing and are making adjustments to make it even better. When for the rest of us, 80% win ratio is amazingly impressive. If they make it too much better we might have to ask for a worse version too and offer beginner vs expert difficulties.

jmo1121109 wrote: ↑

Fri Nov 13, 2020 5:45 am

President Eden wrote: ↑

Fri Nov 13, 2020 4:54 am

It's still winning at an absolutely blistering 80% clip. To the average player it's probably pretty close to unstoppable.

I finally clutched out a win as France, and yes that means getting to 18.
http://webdiplomacy.net/board.php?gameID=332146

The bot's affinity for its One True Opening with either power - Romanian Anti-Alpine with Austria, Maginot with France - does make it a bit exploitable if you understand the RPS-esque mechanics of openings.

What's remarkable here is the bot creators were hoping for better performance than they're seeing and are making adjustments to make it even better. When for the rest of us, 80% win ratio is amazingly impressive. If they make it too much better we might have to ask for a worse version too and offer beginner vs expert difficulties.

What's the win % with this ai vs the old one we had in FvA? Have you guys been able to test that out?

Rename it to UnfairBot and make the beginner version FairBot.

Ezio wrote: ↑

Fri Nov 13, 2020 5:49 am

jmo1121109 wrote: ↑

Fri Nov 13, 2020 5:45 am

President Eden wrote: ↑

Fri Nov 13, 2020 4:54 am

It's still winning at an absolutely blistering 80% clip. To the average player it's probably pretty close to unstoppable.

I finally clutched out a win as France, and yes that means getting to 18.
http://webdiplomacy.net/board.php?gameID=332146

The bot's affinity for its One True Opening with either power - Romanian Anti-Alpine with Austria, Maginot with France - does make it a bit exploitable if you understand the RPS-esque mechanics of openings.

What's remarkable here is the bot creators were hoping for better performance than they're seeing and are making adjustments to make it even better. When for the rest of us, 80% win ratio is amazingly impressive. If they make it too much better we might have to ask for a worse version too and offer beginner vs expert difficulties.

What's the win % with this ai vs the old one we had in FvA? Have you guys been able to test that out?

I don't have that easily available because Kestas Bot also played classic games. I'll hop into the db in the next day or two and pull those numbers though.

jmo1121109 wrote: ↑

Fri Nov 13, 2020 5:52 am

Ezio wrote: ↑

Fri Nov 13, 2020 5:49 am

jmo1121109 wrote: ↑

Fri Nov 13, 2020 5:45 am

What's remarkable here is the bot creators were hoping for better performance than they're seeing and are making adjustments to make it even better. When for the rest of us, 80% win ratio is amazingly impressive. If they make it too much better we might have to ask for a worse version too and offer beginner vs expert difficulties.

What's the win % with this ai vs the old one we had in FvA? Have you guys been able to test that out?

I don't have that easily available because Kestas Bot also played classic games. I'll hop into the db in the next day or two and pull those numbers though.

ooooh awesome!

Trying to play a game but the bot doesn't finalise - any idea why?

President Eden wrote: ↑

Fri Nov 13, 2020 4:54 am

It's still winning at an absolutely blistering 80% clip. To the average player it's probably pretty close to unstoppable.

The End is near, machines are beating up the humans. I hope there is John Connor somwhere out there getting ready.

Djantani wrote: ↑

Fri Nov 13, 2020 8:59 am

Trying to play a game but the bot doesn't finalise - any idea why?

I have this question also. Trying it out, but spending a lot of time waiting on the bot. Am I missing something?

The game setup screen says "anytime you ready up your orders, the game will immediately move to the next phase" - but that is not happening in the game I created. I've readed my Spring 1901 orders and the Bot is just sitting there with a big fat next to it...

Djantani wrote: ↑

Fri Nov 13, 2020 8:59 am

Trying to play a game but the bot doesn't finalise - any idea why?

Sorry about that. There's a couple rare bugs that cause it to crash. We're fixing them as they happen, but this one happened at 3AM. The bot should be back up soon.

Ezio wrote: ↑

Fri Nov 13, 2020 4:14 am

I think the nature of simultaneous turns is going to make an "unbeatable" AI impossible due to how much randomness there is. Better than any human should be doable, but not to such an extent we can never cheese off a win here or there.

Yeah that's right. By "unbeatable" we mean "unbeatable on average". So if you play enough games, the bot will always win more than it loses. Much like how a rock-paper-scissors player that always randomly chooses an action with 1/3rd probability is unbeatable on average. We're clearly not there yet, but that's the goal.