Troll vs Castles Brainstorm

AI Battles
#1

20 players in the leaderboard and not a single comment on the contribution, but we all knew nobody reads the statements anyway, right? Especially when they explicitly tell you you’re on your own.

So I’ll be making my plea a bit more persistent here: please express your thoughts on the ruleset, as trivial as they may appear. Or comment on the “Evolution” section of the statement. Spark some discussion!

1 Like
The JBM Streams
#2

As I wrote on the chat : it is highly probable that the game does not have a pure optimal strategy.

Getting an edge on the opponent will therefore require detecting that he doesn’t play an optimal mixed strategy on average, and countering that bias.

This can not be done in the current format where a single game is played. It would probably be better to have many games against the same opponent so that we can actually try to adapt to him.

1 Like
#3

If you stick to a limited number of maps with reasonnable size, it may be possible to precalculate a mixed strategy offline for each map (haven’t checked the memory requirement for this…)

This becomes a combination of : do we actually precalculate a correct mixed strategy offline ? Can we pack it efficiently to play on CG ? Can we try and detect/adapt to the opponent behavior on the fly ?

As an example : I think there is a rock/paper/scissors contest online somewhere. Make the underlying game solvable and you still have a somewhat interesting concept.

2 Likes
#4

Add elemental attacks, ice fire lightning.

  • Ice slows troll.
  • Fire make the troll angry and run faster!
  • Lightning stuns the troll but makes him charged, once charged it will discharge on first occasion it lands next to one of the players. The charges keep stacking up until it discharges!
1 Like
#5

Add beer. Slows the troll down. :stuck_out_tongue:

#6

I thought that only worked on dwarves.

#7

Fair point. I quite like @AntiSquid’s ideas though

#8

Here is a live chart for score progression
https://docs.google.com/spreadsheets/u/1/d/e/2PACX-1vRdR4SpYZ-t8o11tqqMpMQruRPaMjRChAhyJAcWd38W41uExbPFK_I-TKpXjKll3-f6JkUod14jffDY/pubchart?oid=1837678358&format=interactive

First time I saw your game I thought of Age Of War, which is basically a game with 2 castles and you train units that attack opponent, you just choose which unit and when.
https://www.crazygames.fr/jeu/age-of-war

It would be amazing to see some of this game mechanics implemented in the most epic crossover!

(btw AoW 2 is even better but graphics are clearly too advanced for JBM)

1 Like
#9

That could possibly even be another multi

1 Like
#10

Just beat impossible level.

Evolved to last level.
Got 4 of the most powerful turrest.
Just afk a bit.
Spam super soldiers (100k ones)
Win :o

I think the issue with the game you suggest is that it follows a somewhat different idea also games might be too long for CG. Reminds me of the old civilization map in warcraft3 custom mode, long gone good times.

2 Likes
#11

We already have a BotG :wink:

1 Like
#12

No idea what you mean.

#13

Evolution Plan

Following much discussion with many of you, the upcoming changes are going to be:

  1. [Possibly breaking] the game’s random seed is going to be available as an input parameter.

    • it may break your code if you were reading figures as a line-independent stream scanner. (if you kept the CG stub, you’ll be fine)
    • I needed it so I can let the boss give you reproducible plays, while still having a stochastic algorithm.
    • if you use random numbers, you’re actually encouraged to use it too so your games are reproducible to other players. Just remember to salt the seed with a secret constant of your own if you don’t want it predictable in arena play. (generate a 64-bit locally once and XOR it to the input seed before you feed it to your RNG)

  2. In a match, the game will be replayed on the same map multiple times. This aims to favor:

    • ranking stability
    • opponent pattern prediction

  3. The map range will be extended.

    • Initially to random values in the current maps’ range
    • If this brainstorm manages to come up with some bounds and relations that determine what makes a map “interesting”, we’ll use that.

FWIW, I also intend the leagues to match that.

Per usual, comments and feedback welcome!

4 Likes
#14

So, any more opinions?

The demo is still playable here.

#15

Add some rule modifiers:

  • Troll moves 2 steps if difference between stones is odd/even
  • Troll does NOT move if difference between stones is odd/even
  • Troll become enraged and moves 2 steps after getting hit by X stones
#16

Additional criterion for interestingness : rocks > 3x roadlength

No other criteria considering the multiround that you already have in your roadmap.

#17

To celebrate TVC’s 99th upvote, here’s a public apology: yes, I know I’m behind schedule, I’m sorry, I haven’t forgotten and I will get back to it. (there has been progress, but not enough to get an updated publishable state)

The plan hasn’t changed much, I’m just going to be more lax on delivery dates. It’s still available in the game’s statement. Comments are still open here and on the contribution page.

Happy stoning!

1 Like
#18

When this game will be available for production?

#19

We’re out of budget. Do you have new sponsorship available?

#20

Ask the CEO of TrollerPact. I’m sure he’ll find the neccessary funds.