Debriefing - Autumn AI 2020 (Code of Kutulu)

Autumn AI 2020 is over! The main motive was to involve existing members in Code of Kutulu for a bot with higher rank as well as bring new members [especially from my university] to the community.

This thread is open for you to share your strategy, and/or feedback!

You can also view the strategies from the official contest here.

You can check the rankings here: Autumn AI 2020

4 Likes

I find the game quite random, not just because of how the leaderboard works. I always considered it very random.

For example in 2v2 scenarios, if 2 bots disagree on best moves they go different ways and the other “team” might stick together.

EDIT: forgot to mention most importantly
It’s great these kind of contests take place. They aren’t real contests, but it’s like a “let’s pick a game and play it” event
I think it’s a fun game when you try it for the first time.

1 Like

Congrats and kudos to @Crypticsy for organizing this! I agree with AntiSquid, but regardless the inherent randomness in 4-player games, bringing in two dozen new coders from his school and 100+ new players in general is no small feat!

Regarding my strategy: it was unconventional yet simple: I did not submit…:slight_smile:
I had a 1-y old mid-Bronze level bot in the arena with some very basic heuristics. As Kutulu is one of the most complex game in terms of number of different game entities & rules), in the end I decided I will not have enough time to write a bug-free simulator that is necessary for proper search algos.

This “strategy” led me to a respectable “minus 19” result in the ranking progression leaderboard overtaking dozens of other players who lost more places… :wink:

4 Likes

But to avoid the trolling-only mode, let me share my very basic algo:
In each turn I calculated the distance from my explorer to all other entities using floodfill.
Then for each of the possible moves I calculated the score as a function of the distance of the closest wanderer and the number of players within Plan range. I chose the best of such moves, but override it with light, yell or plan if the action was applicable and some conditions were true. Conditions included: is a wanderer too close, are other players close enough, is my health low, etc.

1 Like

I wish the contest would have been promoted a bit more on the page,
I usually never click on the bell next to the profile picture i have been coding on codingame the last 5 days and never realized the contest was on. apart from that keep em coming and i hope everyone enjoyed the contest :slight_smile:

2 Likes

Those unofficial contests are really a good idea. But 7 days is too short, it should last, at least, 10 days with 2 week-ends (like PR2). I think 2 weeks with 3 week-ends would be even better (since it’s unofficial it’s not like we’re in hurry).

4 Likes

Yes, having a longer contest would be better but on my end there was a rush. My university was starting a new semester on 20th September and given it started relatively late, I was left with 7 days at most. But for future events (if any) I will try to increase the timeframe as long as possible.

And congrats to you on winning again.