Guide · by PickleArtGeek

Understanding ProbCalc

The original guidebook behind DDTSR's BaseCalc & ProbCalc — how PickleArtGeek builds a polling average and a Monte Carlo win-probability model from a plain Wikipedia polling table, in Google Sheets.

Synopsis

ProbCalc started as a one-off Google Sheets election predictor in November 2025 and has since grown to cover sports, First-Past-the-Post constituencies, and a more structured winner-identification system. The guide walks through it in three parts.

Part I introduces the inverse gamma distribution ProbCalc is built on — its shape (α), scale (β), and position (p) parameters — and how a bigger α tightens the distribution around a result while β simply rescales the axis.
Part II covers "Simple ProbCalc": BaseCalc first, where polls are copied from Wikipedia, cleaned up, and combined into a single α per candidate via (poll result × sample size) ÷ (days until election) — so bigger, more recent polls count for more — then normalized into a percentage; then ProbCalc itself, which feeds those same α values into GAMMA.INV(RAND(), α, β) to simulate an election outcome, repeated hundreds of times and tallied with INDEX/MATCH/COUNTIF to get a win probability per candidate.
Part III gets more advanced: classifying a simulated win as a plurality, majority, or supermajority; extrapolating results for a subgroup (like a demographic or constituency) that lacks its own local polling by adjusting the topline α by that group's historical lean; and converting simulated percentages back into a feasible raw vote count.

This is the same core method DDTSR's BaseCalc & ProbCalc tool runs on this site — α is computed directly from polling data this way, and β is fixed at 1.

Understanding ProbCalc.pdf · 11 pages · by PickleArtGeek
Download PDF

Read it here

If the viewer above doesn't load (some mobile browsers don't embed PDFs), use the download button instead.