but with the current mirror system, it is symmetrical : we both start with then without the advantage. that way a player that wins both games is “better”.
2 Likes
wouldn’t change anything in my (and most people, i think) case, unless the distance is 0 or 1 hex ;).
If you play two chess game one with a knigth less and another with a knight most, if the oponnent is very weak you win the 2 games but often you win one. It is better to have twwo games with equality. And with platinum it is possible when there is symetric position in the beginning ( not in chess because white and black is a little different ).
1 Like
I wouldn’t say it’s “better”. The nature of chess demands for perfect symetry (although the first move advantage is huge). Platinum is perfectly valid with an imbalanced gameworld. And in my opinion, more interesting. As a go player I can tell you that symetry doesn’t assure balance.
But I wouldn’t be against having one symetric map among a umber of maps. Also, as suggested in the chat, if the score took into account the swiftness of the victory, mirror matches would be more telling. I’ll explain. In a mirror match, there is a side with an advantage, but sometimes, one IA will take that advantage and win in 35 rounds, then in the mirror match the other IA needs 150 rounds to win, with the same advantage. I think in that case, the first IA is stronger, and I don’t think it is taken into account for the scores.
honestly, with only a symetric map, every match would be the same after two days, with the top 3 having exactly the same strategy, Then us in the top 20 all almost identical (already the games are quite the same), then maybe a hundred players with roughly the same, and so on and so forth…
3 Likes
Not only a symetric map : I think that it should be a random position but symetric. there are a lot of possibility ( more that one billion surely ).
Try the position given at the end, in the top 60 red always win.
I found it after two or three tries.
distribution=HIGHISH
platinum=120
seed=467229491
noEmptyIslands=false
startPlatinum=0
map=0 1 0 2 0 3 0 4 0 9 0 10 0 11 1 0 1 1 1 2 1 3 1 4 1 9 1 10 1 11 2 0 2 1 2 2 2 3 2 5 2 9 2 10 2 11 3 0 3 1 3 2 3 5 3 7 3 8 3 10 3 11 4 0 4 1 4 2 4 6 4 7 4 9 4 11 5 0 5 1 5 2 5 4 5 6 5 9 5 11 6 0 6 1 6 3 6 4 6 5 6 7 6 8 6 10 7 0 7 3 7 4 7 5 7 7 7 9 8 0 8 1 8 3 8 4 8 5 8 6 8 8 8 9 9 1 9 3 9 4 9 8 10 2 10 4 10 5 10 9 11 2 11 6 11 7 11 9 12 3 12 7 12 8 12 10 13 2 13 3 13 5 13 6 13 7 13 8 13 10 13 11 14 2 14 4 14 6 14 7 14 8 14 11 15 1 15 3 15 4 15 6 15 7 15 8 15 10 15 11 16 0 16 2 16 5 16 7 16 9 16 10 16 11 17 0 17 2 17 4 17 5 17 9 17 10 17 11 18 0 18 1 18 3 18 4 18 6 18 9 18 10 18 11 19 0 19 1 19 2 19 6 19 8 19 9 19 10 19 11 20 0 20 1 20 2 20 7 20 8 20 9 20 10 20 11 21 0 21 1 21 2 21 7 21 8 21 9 21 10
startUnits=10
unitCost=20
1 Like
Example of algorithm to initialise a symetric position
On a half of board randomly place 60 platinum, 5 yelow pods, 5 red pods.
And then create the other half, with the same platinum, and same pods with color permutted.
I think that there is better but it is just to show that, the position is also random like.
2 Likes
Yes. That would be a lot better on ymetrical maps. No questions asked. But it forbids any non symetrical map. Case in point : my AI beats PDS (currently first) as red on the map you gave, in 79 turns. his destroys mine in 55. Clearly we can compute which one is better :D.
What I would propose is : the games are on X maps (say 2 next week, 3 the following week, etc). On symetrical maps, symetrical plat & spawns (as you"re proposing), on non symmetrical ones, well, let rand() decide. And we keep the mirror games, even on symmetrical maps. What about that? (I really liked the world maps from PR 1 !)
Anyway, tomorrow we’ll know the new rule and map 
Agree 100%! I started a separate discussion about this, under “Feature Requests”
I’ve noticed that many of the top bots have almost identical strategy for defence - just spawn in a location as soon as it is about to get taken. I believe that this is the optimal strategy with the current rules and it makes the games rather short and boring. So how about a new rule: spawning of new units happens deterministically and only in a few (pre-set) locations. In other words we do not control where it happens, it happens automatically as soon as we reach 20 or more platinum. This way we have to decide how to move new units to “hot” battle zones, instead of spawning them there. The games will be longer and perhaps a bit more interesting. A similar idea was used in Ants (Google AI Challenge) where spawning of new ants happened at ant hills.
4 Likes
This rules is like the first one (Moba Style), all the pods spawn at your base and you have to conquer the enemy base to win.
1 Like
good idea dkamenetsky but let the player choose if he wanna spawn in a pre set location (cost less maybe 10 or 15 platinum) or spawn whenever he wants in his terretory (20 plat cost).
2 Likes
You can change yourself your username in your settings.
Scroll to the top of this page, click on your username at top right -> settings/preferences -> change username.
To complete this algorithm, in the actuality map the symetric zone of id is (149-id). For each pair of symetric zones choose randomly one zone, then we have a half set and we can process the precedent algorithm.
Good idea!
I’d propose 20 platinum price for auto spawn, and 40 platinum for anywhere on own territory.
1 Like
Hi,
I like the idea of the rule that has been implemented yesterday, but I think it would be nice to push its concept a little bit further.
The motto of yesterday’s rule is “Your AI must adapt to every situation”. It is actually quite easy to detect which map the AI is playing on, and thus, writing several algorithms that will be chosen according to the selected map.
The ultimate challenge would be a randomly generated map in every match (well, every pair of matches, with reverse spawns, as it is right now).
I think it would require our AIs to be trully adaptive, instead of switching between N tactics as the map pool has a limited number of maps.
What do you think ?
11 Likes
I fully support the idea of randomly generated maps for each match. This will truly test our AIs. If this is not possible then there is another alternative. For the final testing you can reserve a set of maps (that we don’t see) and test our AIs on those maps.
1 Like
That is an extract from rules:
zoneCount = 154 (may be changed during the contest if balancing is required < 200)
linkCount = 306 (may be changed during the contest if balancing is required < 400)
However new map exceeds these limits with zoneCount = 368 and linkCount = 564.
It is not a big problem for me to correct my algorithm to stop crashing, but please fix either rules or map.
2 Likes
Yes you’re right.
Thanks for poiting this out.
Have you any preferences for generation algorithm? Binding of isaac, MiroorMoon, others?
I think that the only constraint should be the number of continents, like 1, 2 or 3 top. The less constraints, the more random the maps will be, and the more adaptive our AIs must be.