These comparisons cover the committed benchmark scenarios, not whatever you just typed into the planner. A tool agreeing here does not certify your own study plan.
How to read a difference
external agreement Same sample size under matched assumptions, or a one-observation difference from integer rounding.
different question The tools define the effect, test variant, or predictor differently. Both can be right; they are powering different quantities.
needs review The numbers differ and no declared assumption accounts for it. A prompt to investigate, never an automatic finding that a tool is wrong.
not externally checked No external tool implements this design, so nothing was verified against. This is not a pass.
Why there is no “best tool” ranking. A leaderboard would misrepresent tools that are correct for a different estimand. Where two tools disagree under matched assumptions, the independent simulation is the tiebreaker, because it shares no code with any analytic formula. Where assumptions differ, there is no most accurate answer — there are two different questions.
External tool results are precomputed and committed to the repository, so every comparison is reviewable in version history. Package links and full citations are on the credits page.