CRYPTONITEpresents

REAPERS OFACHERON

A hooded reaper surrounded by charcoal smoke on a pale sage background

CONGRATULATIONS REAPERS

Hades has selected you to collect souls for eternity

Top 3 Finalists Standings
Position Name Username Best Bot Name Approach
1 Winner emblem
Aseem Sahoo @aseemplaysbass ArgonV8
2 1st Runner-up emblem
Santhosh S @hallucinate2746 Regen90
3 2nd Runner-up emblem
Artham Das @primeTuffboi123 megatron

FINAL LEADERBOARD

Final Leaderboard Submissions
Rank Username Bot Name TrueSkill Score
#1 @aseemplaysbass ArgonV8 34.1100
#2 @aseemplaysbass ArgonV6 34.0914
#3 @hallucinate2746 Regen90 33.6527
#4 @hallucinate2746 Regen92 33.2026
#5 @primeTuffboi123 megatron 33.1290
#6 @hallucinate2746 Regen91 32.8932
#7 @Altster Physcon 32.5784
#8 @primeTuffboi123 Prime1 32.5556
#9 @Altster WolfPack 32.4919
#10 @hecker77 nerdycat 32.3584
#11 @Aaryan127 AA-v8 32.3144
#12 @Altster DalMakni 32.2953
#13 @Aaryan127 AA-v10 32.2677
#14 @cheeese01 dawgwtf 32.2508
#15 @hecker77 boombot 31.9227
#16 @primeTuffboi123 Optimus 31.7433
#17 @Altster SkyShots 31.7312
#18 @Altster BuketSan 31.7209
#19 @Aaryan127 AA-v6 31.6861
#20 @molecularity Atom-6 31.6721
#21 @naman0414 soullead 31.5824
#22 @Aaryan127 AA-v7 31.5649
#23 @molecularity Atom-3 31.4645
#24 @agento07 ca3s4_4 31.4576
#25 @aarav0202 CHARONIC 31.3888
#26 @aarav0202 KLEOS 31.3435
#27 @naman0414 new1 31.2202
#28 @agento07 br0t0s_1 31.1322
#29 @naman0414 dracaris 31.1056
#30 @Aaryan127 AA-v1 31.0925
#31 @molecularity Atom-5 31.0674
#32 @hecker77 oiia_cat 30.7768
#33 @aseemplaysbass ArgonV5 30.7360
#34 @Aaryan127 AA-v2 30.7246
#35 @molecularity Atom-4 30.6970
#36 @molecularity Atom-1 30.6676
#37 @cheeese01 tscoc 30.5325
#38 @agento07 br0t0s_2 30.5065
#39 @hecker77 nc_01 30.4403
#40 @naman0414 leadplus 30.4070
#41 @Aral-549 b4dB4rn 30.3968
#42 @aarav0202 ASPHODEL 30.3233
#43 @hallucinate2746 DefPB04 30.3077
#44 @hallucinate2746 Regen70 30.3027
#45 @naman0414 soulaim 30.2467
#46 @hecker77 doom_cat 29.9516
#47 @Aral-549 badB5rn 29.9484
#48 @aarav0202 CERBERUS 29.9405
#49 @primeTuffboi123 primefut 29.7463
#50 @primeTuffboi123 tuf2prim 29.7458
#51 @Deftsalt Acheron3 29.6363
#52 @molecularity Atom-2 29.6114
#53 @naman0414 Jon_snow 29.5932
#54 @aarav0202 ERINYS 29.5161
#55 @hecker77 scarycat 29.4533
#56 @Deftsalt Acheron4 29.4226
#57 @aseemplaysbass ArgonV4 29.3998
#58 @Indubrata ApexV6 29.2387
#59 @Anon0608 ANON2 29.0681
#60 @Anon0608 ANON4 28.9919
#61 @hallucinate2746 Regen60 28.9695
#62 @mishface123 FableV3 28.9507
#63 @mishface123 FableV2 28.7668
#64 @Deftsalt Acheron2 28.7554
#65 @Indubrata Champion 28.7056
#66 @Anon0608 ANON3 28.6599
#67 @spacenets Reap04 28.1360
#68 @Eternal666 deathv6 28.0650
#69 @spacenets Reap03 28.0338
#70 @Indubrata CombatV4 27.7503
#71 @Deftsalt Acheron1 27.3306
#72 @aseemplaysbass ArgonV3 27.1452
#73 @Altster GenZ 27.0597
#74 @primeTuffboi123 primetuf 26.8937
#75 @shr3ne soulbowl 26.7009
#76 @mishface123 Fable 26.6737
#77 @shr3ne reaper67 26.6033
#78 @aseemplaysbass ARGONV2 26.0863
#79 @Aral-549 b4dB2rn 26.0014
#80 @Aral-549 b4dB3rn 25.9086
#81 @Baseline Midas the Hoarder 25.7858
#82 @drnkwait Styx09 25.0142
#83 @Aral-549 badB0rn 24.9807
#84 @drnkwait Styx08 24.9234
#85 @Aral-549 b4dB1rn 24.8748
#86 @mishface123 Syntax 24.7400
#87 @aarav0202 OBOLUS 24.7234
#88 @cheeese01 tscsoc 24.4970
#89 @Indubrata V3 24.1900
#90 @drnkwait Styx07 23.8814
#91 @tushamagarwal Tusham_2 23.3325
#92 @Baseline Charon the Guide 23.2560
#93 @cheeese01 tcsoc 22.0288
#94 @drnkwait Styx04 21.7186
#95 @agento07 ca3s4_2 21.6367
#96 @ZinderScone Soulnap 21.6307
#97 @ZinderScone LilGal 21.4706
#98 @Mohmed-Faiz Bot 21.3136
#99 @Vidhan-152 Santa_3 21.2387
#100 @agento07 ca3s4_3 21.1211
#101 @agento07 ca3s4_1 20.7922
#102 @tushamagarwal Tusham_1 20.6334
#103 @ZinderScone Stranger 20.5935
#104 @Vidhan-152 Santa_02 20.2938
#105 @Anon0608 ANON1 20.2310
#106 @shr3ne Reaper01 20.1045
#107 @ReigningRaptor vittal5 20.0934
#108 @spacenets Reap02 19.9638
#109 @cheeese01 csoc2 19.8836
#110 @KAP-9 Ch1r0nV3 19.8796
#111 @tushamagarwal Tusham_6 19.8230
#112 @ReigningRaptor vittal4 19.6972
#113 @Eternal666 deathv3 19.5359
#114 @Eternal666 deathv4 19.2980
#115 @spacenets Reap01 19.2547
#116 @Eternal666 deathv5 19.0338
#117 @yashwandhr Model_04 18.7353
#118 @yashwandhr Model_03 18.6983
#119 @Arhaancyber Yazuka01 18.4883
#120 @tushamagarwal Tusham_4 18.3634
#121 @yashwandhr Model_01 18.2726
#122 @ReigningRaptor vittal3 18.0053
#123 @broskiilol monsterX 17.6115
#124 @Eternal666 deathv2 17.5515
#125 @cheeese01 tcsoc5 17.3836
#126 @RutujN model2 17.0218
#127 @yashwandhr Model_05 16.8210
#128 @broskiilol Version 16.4376
#129 @yashwandhr Model_02 16.3439
#130 @Vidhan-152 Santa_01 16.2319
#131 @tushamagarwal Tusham_3 15.8025
#132 @KAP-9 Heracles 15.7765
#133 @RutujN model1 15.5260
#134 @Arjun-08K MODEL_02 15.4071
#135 @Arhaancyber Caraxes0 15.0022
#136 @Indubrata DensV2 14.8572
#137 @KAP-9 FatUsagi 14.6827
#138 @drnkwait Styx01 14.2093
#139 @KAP-9 Ch1r0n 12.9301
#140 @Eternal666 deathv1 12.4593
#141 @tushamagarwal Tusham_5 11.6816
#142 @RutujN model3 11.3333
#143 @ReigningRaptor vittal2 11.0225
#144 @drnkwait Rule01 11.0042
#145 @ReigningRaptor vittal1 10.9510
#146 @ZinderScone Boat01 10.9395
#147 @AgrimGreat trial001 10.2291
#148 @broskiilol model2 9.6083
#149 @KAP-9 Ch1r0nV2 7.5970
#150 @Baseline Cerberus the Warden 5.5695
#151 @Baseline The Lethean Shade 4.0596
#152 @Arjun-08K Model_01 -3.5041
Solution Writeup • 1st Place • ArgonV8 (@aseemplaysbass)

Approach Overview

ArgonV8 combines reinforcement learning with calculations of collection routes, cargo delivery, combat exposure, and resource regeneration. Its neural policy adjusts action scores supplied by a mechanics prior strategy. A separate navigation guard directs full-cargo returns when the surrounding area is clear.

Fixed preprocessing identifies resources, docks, walls, opponents, beams, and cargo, yielding 701 inputs (605 spatial planes, 32 mechanics features, and 64 farming features). The learned actor (909,896 parameters across 701 → 768 → 384 → 192 → 8) produces bounded score adjustments using tanh activations. Training started from ArgonV6 using PPO and GAE in a league against rule-based bots and exploiters, choosing the 137.5M-decision checkpoint (963 ONNX nodes).
Solution Writeup • 2nd Place • Regen90 (@hallucinate2746)

Approach Overview

Regen90 executes a hand-written strategy compiled into a fixed ONNX graph without learned neural-network weights. It checks firing and aiming opportunities before collection or delivery, prioritizing combat over delivery even with a full hold. Breadth-first search estimates arrival distances, value iteration guides movement, and map matching supplies return headings when docks are out of view.

The rules execute strictly in order: fire at hittable opponents, turn to aim, deliver full holds, select resources with opponent-arrival discounts, choose movement via value iteration, and bank cargo or patrol. Parameters were tuned through simulated competition using CMA-ES. The exported model contains 671 ONNX nodes with verified 100% agreement on all 8,000 checked states.
Solution Writeup • 3rd Place • megatron (@primeTuffboi123)

Approach Overview

Megatron learns to imitate Prime1, the previous rule-based bot, reproducing its tactical decisions with substantially lower latency per move. Prime1 executed complex localization, routing, docking, firing, and opponent beam-exposure risk modeling through a large ONNX graph.

It converts RGB observations into six binary spatial planes and five cargo features, feeding four 3×3 conv layers (32 channels), a 256-unit fully connected layer, and an eight-unit action output (1,024,328 parameters, 51 ONNX nodes). Training used behavior cloning followed by DAgger self-play on ~653k positions, achieving 99.33% agreement with Prime1 at ~0.38 ms mean inference time.