Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Reading Two-Spirit Synergy Data

mindwanderer and the community-run SI Digital Data Analysis series surface per-spirit win rates well, but they don’t directly answer the question 2-player groups actually ask: “given that I’m playing Spirit X, which other spirit should my partner pick?”

Heidi Kalbe’s Team Spirit Island Tableau dashboard is the closest public source. This chapter is a reader’s guide.

Source

Built on Spirit Island Digital stats data (same pipeline as mindwanderer’s broader dataset). Outputs a pairwise win-rate heatmap: rows = Spirit A, columns = Spirit B, cells = win rate of the pair across games meeting the dashboard’s filters.

What the cells actually measure

Each cell in the heatmap is:

Win rate for all digital games where {Spirit A} + {Spirit B} were the two spirits, filtered by (adversary, level, scenario, player-count).

The dashboard lets you toggle those filters. The result cell is a raw win rate, not a synergy-corrected metric. That matters:

  • A cell reading “Lightning + Shadows 70% vs. England L3” is not strictly measuring synergy. It’s measuring the pair’s combined performance — Lightning would probably be at 68% with many partners at that level, so the “synergy premium” for Shadows specifically is ~2%, not 20%.
  • To estimate actual synergy: cell_win_rate − (expected_win_rate_from_individual_stats). Expected rate is a rough average of each spirit’s solo performance at that adversary/level.

How to read it well

  1. Filter hard. Default view blends all adversaries and levels. Always filter to the adversary-level you’re about to play. Unfiltered numbers are close to noise.
  2. Check n per cell. Tableau should surface a sample size. Two-spirit cells are always smaller than solo cells — the dataset partitions by pair, which sparsifies fast. Cells with n < 30 are suggestive at best.
  3. Compare to same-row baseline. Pick your spirit’s row. The median cell is your spirit’s rough “average partner” win rate. Cells above the median are plausible synergies; cells far above suggest either genuine synergy or small-n variance.
  4. Cross-check with BGG SI Digital Data Analysis Series. The series (entries 5–13) has pair-performance prose that adds context the raw heatmap hides — e.g., “Lightning + Shadows plays well at L3 because Shadows’s fear econ feeds Lightning’s fear-contingent Powers, not because of element synergy.”

Known limitations

  • Digital-only: the dashboard is built on Spirit Island Digital data, so the Digital vs. Tabletop caveats (see that chapter) all apply. Shuffle quality, rules-strictness, and player-skill distribution skew tabletop numbers away from what the dashboard shows.
  • Skill-heterogeneous: the dashboard doesn’t weight by player skill. Pairs chosen mostly by new players will look weaker than they “should” and vice versa.
  • Methodology changes: same BGG thread 3555502 methodology caveat that applies to mindwanderer — post-Handelabra-migration data is not identical to pre-migration data. Where the dashboard aggregates across that boundary, treat older comparisons with extra skepticism.
  • Aspect variants may or may not be disambiguated in the dashboard depending on version. Check the filter panel; if aspects aren’t separated, treat each base-spirit cell as an aspect-blend.

Using the dashboard for game prep

Before a 2-player game:

  1. Filter to your adversary + level + scenario.
  2. Find your spirit’s row (or your partner’s column).
  3. Sort by win rate descending. The top 3 partner candidates are your leading pair picks.
  4. Sanity-check with n. If n < 20 on your top pick, it’s a tie with the spirit below it.
  5. Cross-reference the archetype recommendations in Two-Player Dynamics — the dashboard surfaces quantitative signal, but archetype-coverage reasoning is the qualitative backstop. They should agree; when they don’t, the disagreement usually means small-n or a specific matchup quirk you should dig into.

Cross-references