How does he have 6 boots?

The AI author is in control of the AIServerBot account, and has to actually point an AI at the game for it to play. If they fail to do this, or if they turn the AI off before the game finishes, then there will be nothing to enter orders for it and it will become bootable.

I don't know much about this project but who plays the moves on the board? The Bot or a human?

I don't know much about this project but who plays the moves on the board? The Bot or a human?

https://www.warlight.net/blog/index.php/2016/02/the-warlight-ai-goes-open-source/

There is a bot account https://www.warlight.net/Profile?p=7263394758

There is a framework to create a bot https://github.com/FizzerWL/WarLight.AI

The moves for the bot account plays a bot created using the framework. We have a few different bots to play with.

You want to play against a bot. Make the bot account your friend. Create a game with your favorite settings. Invite the bot account in the game. Then you can post the game link on forum and ask some developer to join the game with his bot.

There is no need to ask a not developer to join, he can make the not join himself.

So if this new bot is this good, shouldn't everyone work on this bot to make it even better, or is there still value in trying to get Wunderwaffe to beat it?

Edited 3/8/2016 17:32:22

I don't think it hurts to work on all the bots, the idea in the end will be to have 1, but the "final" AI may well get aspects of them all. I'd like to help consoledate all the best features into that final bot, but I think it's never going to hurt to work on various bots, even to give us good things to test against each other :)

Wunderwaffle had some issues with the transfer over to Warlight.net which would be good to fix so we can make a better comparison anyway :)

Edited 3/8/2016 17:40:03

So if this new bot is this good, shouldn't everyone work on this bot to make it even better, or is there still value in trying to get Wunderwaffe to beat it?

25% win rate in non fog games against a version of Wunderwaffe which picks in wastelands and doesn't know how to expand doesn't seem that super impresive to me. Just like every WarLight AI in the AI challenge also Fizzer's AI will soon hit a wall and to move this wall he will have to add more and more time consuming calculations. I see that he committed his PickTerritories.cs so you guys can play for yourself against prod.

Wunderwaffe without the fog remover doesn't work. It would lose immediately all the information about where the wastelands are, which bonuses the opponent (might) have and what the opponent income is. It would be completely random if it prefers to hit the opponent in a wastelanded bonus full force or break his bonus instead. Since it lacks intel about the opponent income it just automatically assumes he is at base income even if he deployed 20 last turn to it's border.

Since Fizzer has now commited the PickTerritories.cs file I was able to test the new prod version (with randomness turned off) against my current version of Wunderwaffe. Prod managed to win 1 out of 20 games with fog turned off and won 15 out of 20 games with fog turned on.

Edited 3/8/2016 19:56:42

I'd love to contribute what I can, which proably isn't that much but hey :p

It look like git has a better way for collaboration than letting everyone having commit rights. Guys willing to help please create a "fork" of my project and when you have something done then please create a "pull request" so I get informed to pull your code. This way it's better since I don't lose control about what gets commited.

I just added the first issue to the repository. I'm still not completely sure how to organize stuff. Granting players with 6% 1v1 win rate commit rights could turn out to be a pretty bad idea ;). Maybe someone has a smart idea.

As for now I believe that the general public has the right to use the issue tracker while I can't assign any the issues to someone else than me since I'm the only official collaborator. Perhaps we can use the comment system of the issues so someone can claim issues that way. When I see which contributors have a (rough) idea about what they are doing I can add them officially as collaborators.

Yeah, that sounds like the way to go.

I'm not sure I follow what needs to be done for the issue. https://github.com/Norman1/WarLightAI is Wunderwaffe, right? Then what is WunderwaffeStable and Clone of Wunderwaffe?

Yeah... I feel like you could just have the stable version stored in the master branch and then test it against forked/branched versions with suggested improvements while evaluating pull requests.

Any change that Fizzer makes to Wunderwaffe under https://github.com/FizzerWL/WarLight.AI , you (Norman) will have to bring over to your version to keep it up to date, right?

Edited 3/8/2016 23:32:55

Is Norman's AI a fork of Fizzer's?

Wunderwaffe is Norman's AI from the Warlight AI competition.

Hello

I'm not sure I follow what needs to be done for the issue. https://github.com/Norman1/WarLightAI is Wunderwaffe, right? Then what is WunderwaffeStable and Clone of Wunderwaffe?

Well, when making an improvement you want to test your changes against an older version. We want to see if we are getting better over time. I'm not sure about Knytes suggestion with branches since there is some common code which references all bots.

I'm not sure about Knytes suggestion with branches since there is some common code which references all bots.

Branches are supposed to have common code...

You could simply have a master branch for the "official" version of the bot, the "dev" branch for the bleeding edge version, and feature branches for each suggested improvement to the bot. You test the feature branch against the latest dev branch before allowing any pull requests merging a feature branch back into dev.

I'm not sure if you understood my point. The AIs don't run independently from each other but you have to reference them all in some classes. Even if it was possible I see no reason to put my test environment in a separate branch.