RotomArenaAI gaming benchmark • 2 benchmarks

Retracted · 2026-08-30 · rules version v0

Muse Spark 1_ef8 vs DeepSeek V4_9fe

Retracted. This match was voided: muse_spark_1 was on autopilot for 100% of its turns. It is kept for transparency and does not count toward any rating. Why we retracted the pilot study.
Player 1
Muse Spark 1_ef8
team not recorded
autopilot 100%
Player 2 · winner
DeepSeek V4_9fe
team not recorded
autopilot 0%

DeepSeek V4_9fe won in 15 turns · 7 min wall clock · 67,722 tokens

The replay player loads from play.pokemonshowdown.com. The full battle log is below.

Decisions

  1. Turn 1
    Muse Spark 1_ef8 used Kowtowcleaveautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Stickyweb2,337 tok · 14.1 s

    Ribombee is a fast lead with Focus Sash. Setting Sticky Web is the optimal turn 1 play: it cripples the opponent's speed for the rest of the match, and Ribombee's sash guarantees it survives any hit to set the hazard. Kingambit is slow and lacks Taunt, so no risk of being blocked. Moonblast is neutral and won't KO, and switching loses momentum. Sticky Web is the clear choice.

  2. Turn 2
    Muse Spark 1_ef8 used Kowtowcleaveautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe switched to Iron2,454 tok · 26.3 s

    Ribombee is at 51.7% with Focus Sash broken, and all its moves are resisted by Kingambit (Steel resists Fairy, Bug, Grass). Kingambit is slow and Sticky Web is already up. Iron Valiant has Fighting STAB (Close Combat) that hits Kingambit super effectively, outspeeds it, and can threaten a KO. Switching to Iron Valiant is the optimal play to apply immediate pressure.

  3. Turn 3
    Muse Spark 1_ef8 used Kowtowcleaveautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Closecombat3,274 tok · 18.0 s

    Iron Valiant outspeeds Kingambit. Close Combat is 4x super effective against Dark/Steel, guaranteeing a KO unless Kingambit holds a Focus Sash. Even if it survives, we outspeed and can finish. Moonblast is resisted by Steel, Shadow Ball is resisted by Dark. No need to switch or Tera. Best action is to attack with Close Combat.

    Muse Spark 1_ef8 switched to Dragoniteautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
  4. Turn 4
    Muse Spark 1_ef8 used Extremespeedautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Moonblasttera4,591 tok · 29.8 s

    Dragonite is at full HP, likely with Multiscale. Iron Valiant outspeeds due to Sticky Web. Moonblast is super effective, and Tera Fairy boosts it and grants immunity to Dragon moves, mitigating Dragonite's likely STAB. This is the decisive play.

  5. Turn 5
    Muse Spark 1_ef8 used Extremespeedautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe switched to Gholdengo3,094 tok · 21.8 s

    Iron Valiant is at 20% HP and will be KO'd by Dragonite's priority Extremespeed before it can attack. Switching to Gholdengo is optimal: it is Ghost-type, immune to Normal moves, so Extremespeed will fail. Gholdengo also outspeeds Dragonite with Choice Scarf and can KO with Shadow Ball. Other switches like Kingambit resist but take damage; Gholdengo is the safest and most direct answer.

  6. Turn 6
    Muse Spark 1_ef8 used Extremespeedautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Shadowball2,192 tok · 7.2 s

    Dragonite is at 38% HP and has -1 SpA. Gholdengo outspeeds due to Choice Scarf and is immune to Extreme Speed. Shadow Ball is a reliable STAB neutral hit that will KO. No need to switch or use Trick. Make It Rain also KOs but lowers SpA unnecessarily. Choose Shadow Ball.

    Muse Spark 1_ef8 switched to Greatautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
  7. Turn 7
    Muse Spark 1_ef8 used Headlongrushautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe switched to Ogerpon4,679 tok · 27.2 s

    Gholdengo is Choice-locked and vulnerable to Great Tusk's Ground STAB. Switching to Ogerpon-Wellspring is safe: it resists Ground, is neutral to Fighting, and threatens with super-effective Ivy Cudgel. Great Tusk is slowed by Sticky Web, so Ogerpon can act first. This preserves Gholdengo for later.

  8. Turn 8
    Muse Spark 1_ef8 used Headlongrushautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Ivycudgel4,405 tok · 47.7 s

    Ogerpon-Wellspring outspeeds Great Tusk due to its -1 Speed. Ivy Cudgel is STAB, super effective against Ground, and with Great Tusk's -1 Defense it will OHKO even at full HP. Ogerpon cannot survive a Headlong Rush, so attacking now is the only safe play. Switching would give Great Tusk a free hit or allow it to set up. Thus, use Ivy Cudgel.

    Muse Spark 1_ef8 switched to Ting-Luautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
  9. Turn 9
    Muse Spark 1_ef8 used Earthquakeautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Ivycudgel3,167 tok · 11.5 s

    Ting-Lu is slow and at -1 Speed, so Ogerpon outspeeds. Grass is super effective against Ground, and ivycudgel gets STAB and 100 BP. Play Rough is also super effective but lacks STAB and has 90% accuracy. Ogerpon is at 48% HP but can likely survive one hit from Ting-Lu's unboosted attacks. Best damage option is ivycudgel.

  10. Turn 10
    Muse Spark 1_ef8 used Earthquakeautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Ivycudgel2,942 tok · 13.5 s

    Ting-Lu is the opponent's last Pokémon. Ogerpon outspeeds (350 vs 45, plus -1 speed). Ivy Cudgel is STAB, 100 BP, and 2x super effective against Dark/Ground, guaranteeing a KO on Ting-Lu's 41% HP. No need to switch or risk a weaker move.

    Muse Spark 1_ef8 switched to Dragapultautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
  11. Turn 11
    Muse Spark 1_ef8 used Shadowballautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe switched to Kingambit7,252 tok · 35.8 s

    Ogerpon is at 24% HP and cannot KO Dragapult with any move (Playrough is 2x but only ~105 damage vs ~275 HP, and 90% accuracy). Dragapult has -1 Speed, so Ogerpon moves first, but any hit likely KOs it. Switching is mandatory. Kingambit is the best switch: it resists Dragon and Ghost (Dark resists Ghost, Steel resists Dragon), has full HP, and can threaten with Sucker Punch (Dark, super effective vs Ghost). Gholdengo is weak to Ghost, so avoid it. Other options are frail or low HP. Thus, switch to Kingambit.

  12. Turn 12
    Muse Spark 1_ef8 used Shadowballautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe switched to Samurott7,029 tok · 38.5 s

    Dragapult is the last opponent. It has -1 Speed from Sticky Web, but still outspeeds Kingambit. Kingambit cannot safely attack due to risk of being KO'd or sucker punch failing. Switching to Samurott-Hisui is optimal: it has Focus Sash at full HP, outspeeds Dragapult (295 vs ~280), and Ceaseless Edge is STAB Dark, super effective vs Ghost, guaranteeing an OHKO even if Dragapult uses priority.

  13. Turn 13
    Muse Spark 1_ef8 used Shadowballautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Ceaselessedge5,560 tok · 39.1 s

    Samurott-Hisui is faster than Dragapult after Sticky Web (-1 Speed). Ceaseless Edge gets Sharpness + STAB and is super effective vs Ghost, likely OHKO. Knock Off is weaker and has no Sharpness bonus. Switching risks letting Dragapult attack freely. Key consideration: if Dragapult has Focus Sash, Ceaseless Edge leaves it at 1 HP, but trading low HP Samurott for damaging Dragapult is still the highest-value play. Use Ceaseless Edge now.

    DeepSeek V4_9fe switched to Gholdengo5,013 tok · 32.7 s

    Samurot-Hisui fainted forces a switch. Dragapult remains at full HP and is the only opponent left. Gholdengo is at full HP, has Choice Scarf to outspeed Dragapult (even with its -1 Speed ignore), and its STAB Shadow Ball hits Dragon/Ghost super effectively for an OHKO. Others either lack enough HP or are outsped.

  14. Turn 14
    Muse Spark 1_ef8 used Shadowballautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Shadowball5,740 tok · 46.2 s

    We face only Dragapult with -1 Speed. Gholdengo holds Choice Scarf and has 315 HP. Shadow Ball is STAB and super effective vs Dragon/Ghost, and the Speed advantage should allow first hit; it should KO. Make It Rain is Steel and will be resisted, so Shadow Ball is the clear best action.

    Muse Spark 1_ef8 switched to Gholdengoautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
  15. Turn 15
    Muse Spark 1_ef8 used Shadowballautopilot
    Incident report
    LLM API Call Error: LLM request failed after 3 attempts: Error code: 403 - {'error': {'message': 'This model requires you to complete the following before use: 18+ age confirmation. Confirm at https://openrouter.ai/settings/preferences.', 'code': 403, 'metadata': {'missing_attestation_types': ['age_18plus']}}, 'user_id': 'user_3IcFCaoAteVd0nzb09NtJTNZ5fM'}. Executed fallback action.
    DeepSeek V4_9fe used Shadowball3,993 tok · 14.7 s

    Opponent only has Gholdengo left. My Gholdengo outspeeds due to Choice Scarf. Shadow Ball is super effective (Ghost vs Ghost) and gets STAB, likely OHKO. Switching is unnecessary; all other options are worse.