Round III setup

[annadhm]

As a basis for Round III, I implemented some improvements to our bot based on the behaviour in Round II #14.

• Dont run headless to the next pot of gold
  The gold value is now weighed against distance, movement costs, and enemy proximity. If the pot is not worth the sprint or an enemy is closer, the bot backs off and explores instead.

• Include exploration of the map
  I integrated the frontier exploration logic of the Explorer bot from @the-other-thanos Explorer dora #4 . The bot now actively seeks out unknown map areas instead of idling when no gold is reachable.

• Refine the amount of moves per round
  An affordability check ensures we never spend more than we can gain. The number of moves per round is now staggered based on pot value — higher value pots justify more aggressive sprinting.

I made some test runs on different maps. We are performing quite ok, I would say. But our winning rate is quite dependent on the type of map we are playing in. So I think it would make sense to adapt the baviour based on the map we are playing in, but also including a proper default behaviour if we are playing a new map.

Regarding enemy avoidance: I reduced the path-blocking from 3 tiles to 2, since blocking 3 tiles slowed our bot down too much. This sometimes causes collisions now. A possible solution for Round III could be to switch from avoidance to collision-exploiting — actively using enemy positions to our advantage rather than routing around them.

[juwei95]

Remarks

Really nice job, you absolutely killed it with this update! 🚀
Choosing movement speed depending on pot value is smart, and I like that you integrated the exploration logic.

Evaluation

Here are the results across 500 games with 100 rounds each on random maps, compared to the last statistic:

• Your new bot implementation performs significantly better than the old version (283% more wins)
• The difference in performance stays roughly constant for longer games over more rounds.
• Beats Adlhart in about 46% of games (used to be 16%).

Conclusions

• Thanks to the new strategy the bot now manages its gold much more conservatively and efficiently.
• Overall, this seems like a huge improvement to what we had before! I'd definitely pick the new movement algorithm as our submission for round III.

[the-other-thanos]

wow, very nice!