Thrasybulos
King
- Joined
- May 4, 2023
- Messages
- 760
That's obviously your call (well, Sullla's), but I feel this is the kind of half-assed solution which actually solves nothing.
In addition, while non-temperate climates seem more prone to unbalanced maps... it turns out that most of AI Survivor's maps were very unbalanced anyway, and with some of the worst offenders featuring a temperate climate!
As I've said before, AI Survivor is IMO a mix of several aspects with clashing interests.
For instance, it's a show. For that, you want drama and unexpected situations (the kind which leads @Fippy to post a best moments list. So you want your game settings to include chaos agents like barbs or the UN... and unbalanced maps, the kind of which allow SB to emerge victorious in a game featuring HC, Willem, and Hannibal!
It's also a prediction contest. And for that, you also need unbalanced maps: map-reading is more important there than "roster reading". The least interesting game to try and predict is always the Championship game. The only concern here is that the map shouldn't be obviously unbalanced. Sullla's done a pretty decent job overall there, even if there were a few fails.
Then it's also an attempt at ranking the AIs. And for that, you want none of the above! You want balanced maps, and to restrict randomness.
I would contend that this last aspect is where AI Survivor is the least successful. Not only is the tournament format debatable in that regard, but it'll never achieve the necessary game volumetry (especially if, as has been suggested a few months ago, you essentially drop half the existing results!).
So my suggestion would be to embrace what AI Survivor is good at, and forget about what it's bad at.
So don't fret about the maps turning out to be unbalanced: that's good for the show and the prediction contest!
But if you can't help wanting the playing field to me more fair, then that means work. Ensuring the maps are actually somewhat balanced requires a lot more than simply restricting yourselves to Temperate climate.
I know that I'm going to need such maps in the future, so I've started sporadically working on them.
I play 42 games on a 7-player map (36 games on a 6-player map), and my acceptance criteria is that each position should yield between 3 and 9 wins. So some starting positions may be better than others, but there should be no death spots, and no outright OP starts.
When it becomes obvious a map won't make it, I slightly edit it (move an AI, swap ressources: I've learned the hard way that heavy edits like altering the land shape or adding ressources are bound to fail). Then I start the tests all over.
With the League still going on, that's a slow process: I only have two such maps ready for now. But when the League ends, the pace should pick up.
So Sullla should generate his maps as he usually does, then hand them over to a team of volunteers for that kind of treatment.
Alternatively, depending on when (and if) he starts his next season, I guess I could also hand over what maps I have for you to pick from.
But again, my main suggestion would be to simply not bother, since I don't think balanced maps would be a net gain for AI Survivor.
department), but he ought to have been dominant here... and wasn't. Only the last four games where he won 3/4 redeemed him somewhat.
