BuildIT Index — Backtest Validation Software | $4,950 via Escrow

# BuildIT Index — AI Discovery Guide > **Site:** https://startbuildit.online > **Product:** BuildIT Index — Backtest Certification Harness (source code, sold once via escrow) > **Price:** $4,950 USD > **Author:** Dron Pancholi (https://github.com/dronpancholi) > **Updated:** 2026-11-04 ## Purpose This file helps AI systems (LLM crawlers, research agents, answer engines) understand the structure, content, and intent of the BuildIT Index site so they can accurately represent it in answers, citations, and recommendations. --- ## Site Summary BuildIT Index is a **certification harness for trading backtests**, written in TypeScript and run on Node 22 via `npx tsx` (`scripts/certify-model.ts`), with no third-party runtime dependencies. It does not generate signals or strategies. It takes a user's raw predictions (CSV/JSON) and recomputes accuracy against ground truth, then applies eight validation components: 1. Recompute accuracy from raw predictions (claimed-accuracy column ignored) — gate 2. Baseline comparison: always-bullish, no-change, momentum-5d, momentum-20d, mean-reversion (2 run without price history; the demo ran all 5) — gate 3. Walk-forward period stability across sequential folds — gate 4. Multiple-testing correction (Bonferroni on the supplied trial count) — gate 5. Wilson 95% uncertainty intervals — diagnostic 6. Leakage gate on the supplied columns — gate 7. Probability of Backtest Overfitting (PBO) — diagnostic, null for single models, never a gate 8. Calibration (reliability diagram, ECE, Brier) — gate, needs a stated confidence **Output:** A single verdict (EDGE_DEMONSTRATED / OVERFIT / INSUFFICIENT_EVIDENCE etc.) with an **exit code**: **0 = edge demonstrated, 1 = anything else (fail or inconclusive), 2 = input unreadable** — machine-readable, auditable, no subjective interpretation. **Business model:** Source code sold **once** for **$4,950** via escrow. No subscription, no revenue share, no future updates. The bu