With all the improvements of this website the biggest downfall I personally think is the AI.

It is quite stupid and often downgrades to a spoiler/kamikaze than an actual opponent.

It not only makes games with AIs frustrating but it makes single player games outside of the challenges boring.

My two cents.

Perhaps with the Google AI challenge thing some AI's will be made that could then be incorporated into the game as options. IE you'd have options for which AI(s) you wanted to play against. They could be rated based on general difficulty to help with selection as well. That's some ways off though I'd guess. Additionally I don't think there's a whole lot of demand for better AI's in relation to other features and it would probably be hard to do.

You should consider putting something to this effect on uservoice and see how many people vote for it. That will help gauge the demand for such a thing and help Randy to evaluate if he should re-prioritize. As it is now I'm sure he has a ton of things he wants in and most of them probably have a much higher priority and are probably not so difficult and time consuming as an AI. Although if it was a higher priority and with the AI google challenge more effort might be made to help people in that aspect (like making the visualizer).

One improvement I could see adding would be to maybe give a little "personality" to each of the AIs by randomly assigning an aggressiveness factor to each of them at the start of each game. It would make the interactions less predictable from game to game. I assume there is an algorithm that determines when they attack which could have a modifier added or something. Says the man who has no idea how these things work.

I want the Montezuma AI on my team.

The AI never using abandons is one thing I think would improve it greatly.

It should use the abandons in obvious situations, better than giving a player free reign to mow the AI down knowing he doesn't have to worry bout a huge abandon.

Max, the AI never uses any card but R.. and those as soon as it gets them *presuming no players on same team*
means you can count on 'alot' of things..

Not using cards is a hard one to overcome. That is pretty sophisticated decision making. Placing armies can follow a reasonably limited rule and work. But using cards is a level of detail that will be hard to achieve. It involves not only decisions about the AIs behavior, but the ability to interpret the human player's intentions and respond accordingly. Very tough to do.

Actually, discussing this raises an interesting idea. What about an AI design challenge? And there can be a tournament style "robot wars" if enough people design some.

Whoops, already been discussed: http://blog.warlight.net/index.php/2011/04/warlight-as-a-candidate-to-the-google-ai-challenge/

Good idea! I encourage those with the skills to pitch in.

It would be nice if there was some amount of human intervention when it comes to AIs. Maybe, when creating a game, there could be human intervened AI and standard AI.

In a game where there are no cards, or at least no gift cards, it is impossible to get the AI to take the last piece for a bonus if that piece is already controlled by a teammate. Wow, run on sentence, I apologize.

The human intervened AIs can be controlled completely or in partial by the teammate with most games played, or least amount of boots by percentage. Or maybe, if the game is 3v3, then controlled by most games won by that specific game type.

I almost feel that this would be easier to engineer than it would to engineer different AI levels, ie beginner, easy, intermediate, hard, veteran, etc...

These are just a few thoughts, probably stupid to boot!

In team games, if a team mate gets booted or surrenders and the AI takes over, there should be an option for the remaining team members to take control of the substitute AI and if necessary revise its orders.

^Agreed. A team could still fight to win if they control the AIs territory.

Alternatively, a randomly chosen player on the booted/surrendered player's team could take control of the territories controlled by the booted player. So the team would lose the 5 (or however many) army base reinforcement but a human player would control the rest of the reinforcements and have strategic position.

If that didn't make sense, essentially what I mean is that the colour of the territories controlled by the booted player would switch to the colour of one of their teammates'.

The problem with one person controlling all of that players territories is that idea is it allows for being booted to be a legitimate strategy that could help the team win. I don't like that being an option.

For example, if there is a landlocked player making 30 armies per turn, it might be advantageous for him to get booted so another player can then make 30+ armies per turn and use that to bust through the enemy team.


I do think letting a teammate control the booted player's turn would be a good idea, I just don't think that giving those territories to the player would be right.

Re:Impaller

I would think it'd be advantageous if the player who overtook the guy stockpiling 30 would now be able to stockpile 50 or more with their combined income.

But teammates do that all the time, transferring the troops they put on board.

While I do see the advantage, the player would still need to be booted, and he's there to play he would try his best not to.

OR, how about this, in games where the option for teamates to overtake teammates AI, none of the team's players are allowed to boot their own teammate.

Then you might have some arses that go away to get booted by the other team so their team might get an advantage but let's be realistic and understand that would be a glaring minority.

Also, so as not to turn this into too big a mess how about setting limitations of how many teamates one teammate could control, as so:

3 Player Team=Only 1 AI can be overtaken, so no 1 guy controlling 3 players
4 Player Team=Same as above, only 2 live players can continue all their teammates
5 Player Team=Must have at least 3 live players
6 Player Team=Must have at least 3 live players

I agree with Impaller.

Team mates can transfer the troops they put on the board, but in the situation Impaller is describing, they are not near the front line and will take several turns to be useful.

Getting booted is easy. For one, in some games there is the auto boot. If you don't have that option, oftentimes you just have to annoy the other team with consistent turn overages.

Being able to modify orders for the AI is a much fairer solution, IMO. There's no reason to combine territories if you can put some human intelligence into the AI.

the Imp,
I don't think anyone is suggesting 'combining' players.. the landlocked players armies won't suddenly be playable in the Regents country.. only in the landlocked players country.. so the armies would still take just as long to get to the front as if said player didn't get himself booted.. anything besides that in that situation would be highly illogical, imho..

Actually, Perrin. That is exactly what was suggested. Not that the armies would "combine" but that one player would get the income of another player added to their own income. It doesn't matter if the armies combine or not, it could still be a huge income boost to one player who is currently balls-deep in combat with an enemy player. Those extra armies to deploy each turn would certainly be on the front lines immediately.

Exactly .. especially on huge maps where airlift cards are scarce merging a booted player's territories with the rest of the team would be too big an advantage. Plus there might be meta bonuses (think big new york or east asia map) that get unlocked once a single player holds all the territories formerly controlled by his booted team mate.
Revising AI orders would be much fairer. The right to change AI orders should be rotated among the remaining players of a team, so each turn another player gets to control the AI.

I see the suggestion differently... I read it more similar to a player running two accounts.. the two players would be more cohesive then two actual players, but a landlocked player sends his armies to the non-landlocked ally anyways.. as I said in my previous post.. the idea of giving the income of two players to one merely because that one is acting regent of one player is highly illogical.

honestly.. what's the difference between player A placing 10 armies in Mexico to send to South america every turn to reinforce Player B, and Player B controlling player A to place 10 armies in mexico to send to south america every turn..

no one mentioned a complete collusion of the player and would be AI's empires.. merely a sister nation that is the would be AI that is controlled by the player..