Hannah bot improvement

[HannahSophie08]

Hey, I added a new bot called r1 (round3_v1) based on the round 2 bot. Here is what I changed:

• choose exploration move function looked for tiles not visited yet , so I changed it to unknown tiles. Through that the bot doesn't need to actually visit each tile
• if no path to gold can be found with danger active, retries without danger
• in maps and caves the bot tended to oscillate and get stuck, I included a function "is_oscillating" that checks for that and if the bot is stuck sprints away from that location (10 steps and ignoring danger)
• I replaced the old exploration function with a new one that counts unknown tiles within the visibility range around each possible next tile, so the bot only moves during exploration when it would actually reveal new information
• I lowered the sprint threshold in the early rounds from 30 to 10
• I removed am_i_closest to make the bot still try when someone else is closer as we could still be faster when we sprint
• whether the gold can be reached is now calculated including sprints and not only single steps

I testes this with 10 games per map with 1000 rounds each (the bot I pushed is r1):

Benchmark results (QUICK_TEST=False)
N_GAMES_PER_MAP = 10, ROUNDS = 1000

--- random ---
bot wins win% mean median stdev min max

Test 0 0.0% 238.6 174.5 167.3 99 549
Beatme 0 0.0% 862.4 727.5 295.0 590 1376
Round2 bot 0 0.0% 1251.0 1360.5 354.1 692 1873
Scout 5 0 0.0% 1365.7 1232.0 549.5 542 2117
r1 10 100.0% 7437.2 7375.5 1395.0 5493 9393
r2 0 0.0% 685.5 463.0 655.7 1 2119

--- maze_map ---
bot wins win% mean median stdev min max

Test 0 0.0% 324.8 371.0 157.2 99 536
Beatme 6 60.0% 1798.0 1784.5 385.0 1317 2519
Round2 bot 2 20.0% 1246.4 1145.0 454.6 463 1931
Scout 5 2 20.0% 1271.2 1010.5 613.3 744 2546
r1 0 0.0% 368.2 31.0 546.1 2 1369
r2 0 0.0% 219.8 179.5 262.2 2 897

--- floodfill_map ---
bot wins win% mean median stdev min max

Test 0 0.0% 177.2 99.0 102.1 99 324
Beatme 0 0.0% 383.3 342.5 126.4 249 659
Round2 bot 0 0.0% 1315.1 1325.5 373.2 773 2160
Scout 5 0 0.0% 892.4 1048.5 828.2 27 2322
r1 10 100.0% 11529.2 11091.5 1134.6 10236 13632
r2 0 0.0% 483.7 176.0 542.3 4 1495

--- inverse_floodfill_map ---
bot wins win% mean median stdev min max

Test 0 0.0% 385.1 383.0 213.4 99 825
Beatme 1 10.0% 2096.9 2070.5 479.9 1224 2775
Round2 bot 3 30.0% 2388.1 2253.0 514.3 1777 3220
Scout 5 4 40.0% 2268.6 2190.5 389.5 1684 2749
r1 2 20.0% 872.3 181.5 1413.3 0 3864
r2 0 0.0% 154.2 69.0 213.3 1 642

--- random_coverage_map ---
bot wins win% mean median stdev min max

Test 0 0.0% 196.0 172.0 108.0 99 379
Beatme 0 0.0% 561.2 643.5 174.3 239 739
Round2 bot 0 0.0% 1367.3 1208.0 594.8 776 2861
Scout 5 0 0.0% 513.2 445.0 493.8 27 1174
r1 10 100.0% 10698.8 10804.0 1311.9 8542 12450
r2 0 0.0% 649.9 530.5 591.7 0 1692

--- mazes_and_caves ---
bot wins win% mean median stdev min max

Test 0 0.0% 237.4 236.0 108.9 99 369
Beatme 9 90.0% 1568.5 1514.0 395.7 1063 2335
Round2 bot 0 0.0% 498.8 528.5 215.4 99 783
Scout 5 0 0.0% 532.9 481.0 200.3 291 896
r1 1 10.0% 382.4 4.5 895.9 0 2895
r2 0 0.0% 11.9 4.5 26.1 0 86

=== OVERALL (60 games) ===
bot wins win% mean median stdev min max

Test 0 0.0% 259.9 255.0 159.8 99 825
Beatme 16 26.7% 1211.7 1153.5 723.4 239 2775
Round2 bot 5 8.3% 1344.5 1285.5 694.6 99 3220
Scout 5 6 10.0% 1140.7 954.0 800.9 27 2749
r1 33 55.0% 5214.7 4678.5 5005.6 0 13632
r2 0 0.0% 367.5 141.0 497.2 0 2119

Total runtime: 1432.7s

The bot performs poorly in the maze map and the maze and caves map. A next step would be to further look into that.

[amkilar]

Hi guys, is that the final version of your bot for today?

[HannahSophie08]

Hi, no sorry that is a pull request I meant to direct to our fork and accidentally chose the main repository. Naja opened the pull request with the final bot: round 3 bot (#101)

[amkilar]

Okay 👍🏻