Evbayiro-IRR Million-Case Benchmark Dataset for Capital Budgeting Algorithm Comparison

Version 3.0.0 of the Evbayiro-IRR benchmark dataset archives a 1,262,557-case SEC/Damodaran public-company quarterly cash-flow proxy benchmark for comparing capital-budgeting IRR engines under decision-mismatch evaluation. The archive supports research on IRR-based capital-budgeting decision mismatch. It compares Evbayiro-IRR 0.2.0 with an Excel-style default-guess proxy, Newton-Raphson, Secant, numpy_financial.irr, PyXIRR, and known-bracket bisection and Brent controls. Bisection and Brent are included only as known-bracket controls because they are supplied with the Evbayiro decision-relevant sign-changing bracket. The benchmark contains 1,262,557 source-backed public-company quarterly cash-flow proxy cases, 1,459 companies with usable cases, 55 Damodaran industries, 11 sectors, 145,620 conventional one-sign-change cases, 1,116,937 non-conventional multi-sign-change cases, and 14,801,872 method-result rows. Required-rate-of-return inputs are Damodaran industry WACC proxies. Cash-flow cases are SEC Companyfacts rolling-window free-cash-flow proxy sequences, not confidential internal capital-budgeting project files. The archive includes the compact benchmark input, quality profiles, benchmark protocol, output summaries, no-boundary audit outputs, diagnostic samples, paper-ready figures and tables, scripts, logs, and SHA-256 file hashes. The 12.7 GB raw expanded CSV is excluded because the compact input is the citable benchmark dataset and can regenerate the benchmark evidence with the included scripts.