The Battle for Wesnoth Forums

Hi, I'm not experienced but as player I can tell that any reward may be reduced to gold (aside from leader kill ofc). Collecting and holding villages has a numerical value as much as killing an unit which have costed some amount on the enemy side. On Isar's for example, detecting an easy kill and r...

Do you think it is necessary for Reinforcement Learning algorithms? From what I gathered, one would need to save the current state and action, continue for the next state, and then update the return value for the previous state. Would that be possible? What you are referring to is a play-through. R...

Thanks mattsc for all your help (on this thread and offline) in understanding how a new AI can be written in Wesnoth and how the AI functions in general. So far I have come to the conclusion that simulating a sequence of moves (in a single turn or otherwise) is not possible with the current AI frame...

Hi all, I have been trying to develop an AI based on deep reinforcement learning and have had some interesting discussions regarding the same on this thread. Right now I am trying to get some basic debugging stuff working - display game state information and deciding on what input features would be ...

This could be a basic count of useful attributes like number of attacked and defended hexes, expected unit attack damage on both sides etc. If someone can tell me how this information can be extracted it'll be very helpful to visualize and examine first, using real replays. I do not know how useful...

Since a lot of replies have been made and most of them pertain to what is or what is not possible within the realm of AI algorithms today, I will try to answer them the best I can. But first, I would like to talk about some technicalities regarding the possible design of input features. The biggest ...

Im just a player but, isn't wesnoth too non deterministic for that? Yes Wesnoth is very non deterministic, but that doesn't mean that an AI can't be developed for it. For example, poker playing AIs exist and are pretty good at the game. Basically a Wesnoth AI would play the game pretty much like a ...

Sorry for replying more than a year later. I think this question is still relevant today and in the past year or so after gaining my PhD in Computer Science, I think I am slightly better equipped at providing some ideas and brainstorm with all of you better. I previously mentioned using the old game...

I don't know if this is the proper forum to discuss this, but lately, after noticing all the great advancements in the field of machine learning, especially deep learning, I was wondering whether anyone has thought about the 'Writing a Wesnoth AI is hard' statement in a different light. Notwithstand...

Just discovered this faction (and the EoM in general) and think it is awesome!! I was wondering if there could be more campaigns with the Windsong on 1.8? (or 1.9) The faction looks good, and feels good to play. I hope to play some 1v1s with it sometime soon. Unfortunately, I see that this thread is...

@ Quetz : Please re-run your implementation with the modified games history file that I sent you with the original parameters. We need to see if the implementations are well understood and consistent first, and then go ahead with parameter changes. Some of the ranks do look good I must say, but this...

I think in cases of all win or nearly all win (less than 10% games lost) scenarios it is very hard to set an upper limit to the rating. Since either a 1700 or a 2800 could have won 30 games against 1500s, we'll never know the players' real strength until s/he lost. Trueskill still works well as the ...

Hi

I was busy with work so couldn't post this earlier. Here are the graphs for the players mentioned :

http://xntrick.comuf.com/wesnoth/trueskill_graphs.htm

Cheers!

PS: Nice to see the site up and running again.

I'm with the people who'd want to keep the replays. Reasons being, we might need the information about factions and units and maps later on to build statistics. Also old replays could be played on 1.8 or 1.9 as well if we just changed some WML (?) Not too sure about this though. My suggestion will b...

Upon the suggestion of tiboloid I looked up trueskill again and figured that the 'RD' in trueskill indeed takes into account the outcome of a game and doesn't always decrease. I was quite happy about this and wrote the implementation. Here are the results : http://xntrick.comuf.com/wesnoth/trueskill...