Data Source
Seed distribution and mechanics are based on ForgottenArbiter's Harvest Mechanics Research.
The default settings err on the side of caution - actual profits should be slightly higher than calculated.
How the Optimizer Works
The optimizer uses dynamic programming to find the best harvest order for any grove state.
At each step, you choose which plot to harvest. Harvesting gives you lifeforce, upgrades seeds of other colors, and may cause the paired plot to wilt. The DP evaluates all possible choices and picks the one with highest expected value.
Mean-field approximation: Instead of simulating random upgrades, we assume all seeds upgrade by their expected amount. For example, if T1→T2 has 25% chance, we treat it as each T1 seed becoming 0.25 of a T2 seed. This gives exact expected values without Monte Carlo noise.
Color weighting: Your atlas reduction settings affect how often each color appears. We enumerate all possible grove color patterns, weight each by its probability, and combine the results.
Mono-color optimization: When the DP reaches a state where all remaining plots are the same color, we short-circuit with an exact calculation. Since harvesting same-color plots doesn't trigger upgrades, we compute: for single plots, E[LF]; for pairs, E[max]×wiltProb + 2×E[LF]×(1-wiltProb). This captures the "pick the better one first" advantage using Monte Carlo-derived E[max(LF₁,LF₂)] at each upgrade level.
Advanced Settings
Upgrade probabilities, drop rates, and lifeforce-per-drop values can all be adjusted via the "Advanced" toggle. The defaults are conservative estimates from community research.
45% Reduction Requirement
At least one color must have 45% reduction to ensure frequent color switches for effective upgrading.