Partial account inputs: inspect the Breakdowns tab before relying on estimates.
Manual targeting throughout: full-clear, drink teals as available, burn stamina on visible resistant objects, then drink oranges and harvest. The first finisher keeps the selected fraction of starting stamina; later finishers stay untouched until needed. If no suitable dump appears, activate oranges at the reserve instead. After activation, manually descend when an exit is revealed and remaining rocks exceed the expected stamina budget. Auto full-clear stays enabled; this is not unattended Auto execution. Take an eligible rope before digging; delayed procs may still remove it. No rope is guaranteed.
Party
0Elixir belt
Built-in elixir presets
Each preset replaces only its equipped belt or memorized set. Party, expedition mode, entry and strategy stay unchanged. These pictured loadouts are not optimizer recommendations.
Selected elixir effects
Strategy
Runs continue until stamina is exhausted or no further progress is possible; phased Amber farming may instead take a visible rope. Cancel stops calculation. Auto speed changes decision timing only. Manual policies cannot see hidden exits. Fixed manual tools fall back to the basic tool when unaffordable.
Manual clear every 5 depths uses stamina-efficient targeting, fully clears depths 5, 10, 15 and so on, and otherwise descends at the first exit. With Hidden Elixirs unlocked and an equipped elixir, each fifth depth contains one refill. Clearing costs extra stamina and may not improve the final depth.
Use the current configuration and selected Normal/Manic mode to find the largest gain over each cave's previous maximum. Keep party order; exclude members beyond each cave's party limit. No party, strategy or elixir search is performed.
Keep the current party order, belt, memory, cave, mode, entry and tools. Use the Amber belt preset for three oranges and two teals if unlocked. Explore one/two finishers, 25/50/75/100% reserves, short/long harvest targets and early/late rope seeking. Refine promising depths and reserves twice, down to individual floors, including observed orange-activation depths. Shortcut runs concentrate on floors at or after entry instead of wasting most depth choices before entry. Distinct observed behaviors and extra parameter-space challengers survive screening; promising strategies get at least eight shared discovery trials. Fresh finalist trials start at the setting below and increase for close comparisons or uncertain rope rates, up to the cap. This stronger search takes longer. Targets are not guaranteed descents or full clears; zero seeks the next visible untouched rope immediately. Zero minimum allows any observed rope rate; higher values filter fresh finalists, not guarantee safety. Current phased settings remain a reference. Named configurations and configuration-search scope are not used.
Trial settings
Every legal combination in the selected scope receives screening trials. Cave and Normal/Manic mode stay fixed. Belts fill every possible slot from the checked elixirs, including unlocked duplicates. Empty slots remain only when that pool cannot fill them. Memories are optional; none checked means no memories. Current and checked named configurations are measured as references even when excluded, but excluded references cannot win. The strongest five screening setups and references receive fresh finalist trials. Up to four workers run independent trials in parallel.