summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--strategy.py58
1 files changed, 56 insertions, 2 deletions
diff --git a/strategy.py b/strategy.py
index 0cba570..1f1249f 100644
--- a/strategy.py
+++ b/strategy.py
@@ -65,6 +65,19 @@ class Strategy:
# if however there's no enemy to avoid, chase food or jizz randomly around
else:
def edible(cell): return (cell.is_food) or (cell.mass <= sorted(c.player.own_cells, key = lambda x: x.mass)[0].mass * 0.75) and not (cell.is_virus)
+ def rival(cell, food):
+ if cell.is_virus or cell.is_food: return False
+ if cell.cid in c.player.own_ids: return False
+
+ if cell.mass < 1.25*my_smallest:
+ return food.is_food or cell.size > 1.25*food.size
+ else:
+ return False
+ def splitkiller(cell):
+ return not cell.is_virus and not cell.is_food and cell.mass > 1.25*2*my_smallest
+
+ def nonsplitkiller(cell):
+ return not cell.is_virus and not cell.is_food and 1.20*my_smallest < cell.mass and cell.mass < 1.25*2*my_smallest
if self.target_cell != None:
self.target = tuple(self.target_cell.pos)
@@ -79,8 +92,49 @@ class Strategy:
if not self.has_target:
food = list(filter(edible, c.world.cells.values()))
- def dist(cell): return math.sqrt((cell.pos[0]-c.player.center[0])**2 + (cell.pos[1]-c.player.center[1])**2)
- food = sorted(food, key = dist)
+ def quality(cell):
+ dd_sq = max((cell.pos[0]-c.player.center[0])**2 + (cell.pos[1]-c.player.center[1])**2,0.001)
+ sigma = 500
+ dist_score = -math.exp(-dd_sq/(2*sigma**2))
+
+ rivals = filter(lambda r : rival(r,cell), c.world.cells.values())
+ splitkillers = filter(splitkiller, c.world.cells.values())
+ nonsplitkillers = filter(nonsplitkiller, c.world.cells.values())
+
+ rival_score = 0
+ for r in rivals:
+ dd_sq = max(0.001, (r.pos[0]-cell.pos[0])**2 + (r.pos[1]-cell.pos[1])**2)
+ sigma = r.size + 100
+ rival_score += math.exp(-dd_sq/(2*sigma**2))
+
+ splitkill_score = 0
+ for s in splitkillers:
+ dd_sq = max(0.001, (s.pos[0]-cell.pos[0])**2 + (s.pos[1]-cell.pos[1])**2)
+ sigma = (500+2*s.size)
+ splitkill_score += math.exp(-dd_sq/(2*sigma**2))
+
+ nonsplitkill_score = 0
+ for s in nonsplitkillers:
+ dd_sq = max(0.001, (s.pos[0]-cell.pos[0])**2 + (s.pos[1]-cell.pos[1])**2)
+ sigma = (300+s.size)
+ nonsplitkill_score += math.exp(-dd_sq/(2*sigma**2))
+
+ density_score = 0
+ sigma = 300
+ for f in filter(lambda c : c.is_food and c!=cell, c.world.cells.values()):
+ dd_sq = (f.pos[0]-cell.pos[0])**2 + (f.pos[1]-cell.pos[1])**2
+ density_score -= math.exp(-dd_sq/(2*sigma**2))
+
+ wall_score = 0
+ wall_dist = min( cell.pos[0]-c.world.top_left[1], c.world.bottom_right[1]-cell.pos[0], cell.pos[1]-c.world.top_left[0], c.world.bottom_right[0]-cell.pos[1] )
+ sigma = 100
+ wall_score = math.exp(-wall_dist**2/(2*sigma**2))
+
+ return dist_score + 0.2*rival_score + nonsplitkill_score + 5*splitkill_score + 0.1*density_score + 5*wall_score
+ ##print (density_score)
+ #return density_score
+
+ food = sorted(food, key = quality)
if len(food) > 0:
self.target = (food[0].pos[0], food[0].pos[1])