Every forecast is cryptographically sealed before bid day — hashed, timestamped to the Bitcoin blockchain, and mirrored to a public ledger. When the official results publish, we score ourselves. No cherry-picking possible, no edits after the fact.
Three steps, none of which require trusting us:
The full prediction file is hashed (SHA-256) and the hash is published before the letting. Changing a single digit afterward would change the hash.
Every hash is anchored to the Bitcoin blockchain via OpenTimestamps and pushed to GitHub — two clocks nobody here controls.
When TxDOT publishes the official bid tabulation, the prediction is revealed and scored against it — wins and misses, permanently on the record.
12 sealed commitments to date, all timestamped. Verification instructions come with every prediction.
Real lettings, scored against the published TxDOT bid tabulation. Swipe or use the arrows →
Average prediction miss, first model generation vs the current one. Every scored letting becomes training data; the model retrains monthly and only ships changes that prove themselves against held-out results.
Harvested continuously from published bid tabulations across Texas and beyond — the raw material behind every forecast.
Every contractor who pulls plans has a bidding history. We price their likely bid from it, line by line — that's where the 77% head-to-head win rate comes from.
When the evidence isn't there, we say so and commit no number. Misses are published with the wins. That's why the wins mean something.
This edge only works if your competitors don't have it. Seats are exclusive: one contractor per work type in each territory, first come, first served. When your seat is taken, it's taken.
Send your details and someone will get in contact with you — usually within one business day.