HowaquantitativeresearchersharpenspredictionmarketforecastswithBigdata.com
A quantitative investor and researcher prices Kalshi contracts with Bigdata.com as the research layer behind every forecast. On Boeing's Q3 deliveries the forecast missed by 0.2 aircraft. On Domino's same-store sales, by 0.1 of a point.
Assaf Amitay, a quantitative investor and researcher, prices contracts on Kalshi. Each market settles on a single reported number, so the whole job is forecasting that number more precisely than the market has priced it.
Using Bigdata.com, he verifies the reported history, rebuilds the Street’s own track record, backtests his estimator against real errors, and scores the ladder before it closes. On Boeing’s Q3 deliveries the forecast missed by 0.2 aircraft. On Domino’s same-store sales, by 0.1 of a point.
The challenge: a market that settles on one number
Pricing a Kalshi contract well demands three things: verified history, time before close, and calibrated error.
A prediction market resolves on a specific reported figure. Direction is not enough: the ladder prices strikes in narrow increments, half a point apart on a restaurant comp, so width matters as much as centre.
Public data gets you close and no further. A run-rate put Boeing’s quarter near 160 with a standard deviation of 8, and at that width every strike near 160 looked fairly priced.
The workflow needed data that was:
Verified at source: a history entered by hand had already produced three wrong values.
Long enough to backtest, so the error itself could be measured, not assumed.
Measurable against consensus, to know where the Street goes wrong, not only what it expects.
Searchable within minutes of a release, so the result can be scored and the model updated.
The solution: Bigdata.com as the research layer behind every forecast
Bigdata.com supplies the evidence behind one reusable, six-stage workflow that runs on any KPI market. It reads the settlement rules, researches the company, logs every data point with its source and timestamp, backtests, scores the ladder against fair value after fees, and re-weights each input once the actual prints.
Bigdata.com delivered:
Verified reported history. Every comp since 2022 taken from transcripts, 8-Ks and 10-Qs, correcting three hand-entered values.
The Street’s track record. Consensus reconstructed for 13 consecutive prints, making the Street’s own error measurable.
Third-party trackers. An independent monthly delivery tracker, verified against Boeing’s releases and backtested over 17 months.
Searchable within minutes. Domino’s 8-K and Boeing’s release searchable within minutes, so each outcome is scored and the model re-weighted the same day.
Impact: a narrower distribution, and the edge that comes with it
On 7 October the workflow put Boeing’s Q3 commercial deliveries at 161.2 with a standard deviation of 2.5, against a public run-rate of about 160 with a standard deviation of 8. Boeing reported 161 on 13 October. The miss was 0.2 aircraft, or 0.08 standard deviations, and the Kalshi ladder settled at 161 on every strike.
Narrowing the distribution is the whole point. At a standard deviation of 8 every strike near 160 looked fairly priced; at 2.5 two mispriced strikes stood out, and both resolved the way the narrower distribution said they would.
Boeing Q3 deliveries, forecast and 17-month backtest | Public data | With Bigdata.com |
|---|---|---|
Forecast centre | ~160 run-rate | 161.2 |
Standard deviation | 8.0 | 2.5 |
Reported | 161 | 161 |
Avg monthly miss | 6.10 | 1.35 |
Quarter-end months | 9.80 | 2.20 |
Both paths used the same raw September input; the difference in the centre is the backtest’s bias correction.
“Two of the three months were already official before the market was priced. The forecast was really a forecast of September, and the tracker found through Bigdata.com is what made that month measurable.”
Assaf Amitay, quantitative investor and researcher, linkedin.com/in/assaf-amitay
The second KPI: Domino’s Q3 same-store sales
The same six stages ran on a restaurant comp rather than an industrial volume. The Street expected +1.1%, card-spend data alone pointed to a decline of 1.3%, and the Kalshi ladder was centred just below zero. A first pass built on five hand-entered quarters also pointed the wrong side of zero. The full Bigdata.com pass, verified back to 2022 and backtested across twelve quarters, forecast +0.20%. Domino’s reported +0.3%.
Result | Domino’s Q3 same-store sales |
|---|---|
Miss on Domino’s Q3 comp | 0.1 point |
Average miss, vs 0.99 for Street consensus | 0.70 |
More accurate than consensus over 12 quarters | 29% |
Consecutive prints of Street consensus rebuilt | 13 |
One reusable workflow, six stages
Read the rules
Research
Log every data point
Backtest and calibrate
Score the market
Learn after the print
The research problem is the same at any size
Prediction markets are no longer only a retail venue: some of the world’s largest hedge funds and financial institutions are already building the rails to access them. The research problem is the same at any size. Whoever forecasts the settling number most precisely prices the contract best, and that is a question of verified history, measurable error and evidence that arrives before the market closes.
For informational purposes only. Not investment advice. Forecasts and market prices shown are historical. Data as of 13 October 2026.
Assaf Amitay, a quantitative investor and researcher, prices contracts on Kalshi. Each market settles on a single reported number, so the whole job is forecasting that number more precisely than the market has priced it.
Using Bigdata.com, he verifies the reported history, rebuilds the Street’s own track record, backtests his estimator against real errors, and scores the ladder before it closes. On Boeing’s Q3 deliveries the forecast missed by 0.2 aircraft. On Domino’s same-store sales, by 0.1 of a point.
The challenge: a market that settles on one number
Pricing a Kalshi contract well demands three things: verified history, time before close, and calibrated error.
A prediction market resolves on a specific reported figure. Direction is not enough: the ladder prices strikes in narrow increments, half a point apart on a restaurant comp, so width matters as much as centre.
Public data gets you close and no further. A run-rate put Boeing’s quarter near 160 with a standard deviation of 8, and at that width every strike near 160 looked fairly priced.
The workflow needed data that was:
Verified at source: a history entered by hand had already produced three wrong values.
Long enough to backtest, so the error itself could be measured, not assumed.
Measurable against consensus, to know where the Street goes wrong, not only what it expects.
Searchable within minutes of a release, so the result can be scored and the model updated.
The solution: Bigdata.com as the research layer behind every forecast
Bigdata.com supplies the evidence behind one reusable, six-stage workflow that runs on any KPI market. It reads the settlement rules, researches the company, logs every data point with its source and timestamp, backtests, scores the ladder against fair value after fees, and re-weights each input once the actual prints.
Bigdata.com delivered:
Verified reported history. Every comp since 2022 taken from transcripts, 8-Ks and 10-Qs, correcting three hand-entered values.
The Street’s track record. Consensus reconstructed for 13 consecutive prints, making the Street’s own error measurable.
Third-party trackers. An independent monthly delivery tracker, verified against Boeing’s releases and backtested over 17 months.
Searchable within minutes. Domino’s 8-K and Boeing’s release searchable within minutes, so each outcome is scored and the model re-weighted the same day.
Impact: a narrower distribution, and the edge that comes with it
On 7 October the workflow put Boeing’s Q3 commercial deliveries at 161.2 with a standard deviation of 2.5, against a public run-rate of about 160 with a standard deviation of 8. Boeing reported 161 on 13 October. The miss was 0.2 aircraft, or 0.08 standard deviations, and the Kalshi ladder settled at 161 on every strike.
Narrowing the distribution is the whole point. At a standard deviation of 8 every strike near 160 looked fairly priced; at 2.5 two mispriced strikes stood out, and both resolved the way the narrower distribution said they would.
Boeing Q3 deliveries, forecast and 17-month backtest | Public data | With Bigdata.com |
|---|---|---|
Forecast centre | ~160 run-rate | 161.2 |
Standard deviation | 8.0 | 2.5 |
Reported | 161 | 161 |
Avg monthly miss | 6.10 | 1.35 |
Quarter-end months | 9.80 | 2.20 |
Both paths used the same raw September input; the difference in the centre is the backtest’s bias correction.
“Two of the three months were already official before the market was priced. The forecast was really a forecast of September, and the tracker found through Bigdata.com is what made that month measurable.”
Assaf Amitay, quantitative investor and researcher, linkedin.com/in/assaf-amitay
The second KPI: Domino’s Q3 same-store sales
The same six stages ran on a restaurant comp rather than an industrial volume. The Street expected +1.1%, card-spend data alone pointed to a decline of 1.3%, and the Kalshi ladder was centred just below zero. A first pass built on five hand-entered quarters also pointed the wrong side of zero. The full Bigdata.com pass, verified back to 2022 and backtested across twelve quarters, forecast +0.20%. Domino’s reported +0.3%.
Result | Domino’s Q3 same-store sales |
|---|---|
Miss on Domino’s Q3 comp | 0.1 point |
Average miss, vs 0.99 for Street consensus | 0.70 |
More accurate than consensus over 12 quarters | 29% |
Consecutive prints of Street consensus rebuilt | 13 |
One reusable workflow, six stages
Read the rules
Research
Log every data point
Backtest and calibrate
Score the market
Learn after the print
The research problem is the same at any size
Prediction markets are no longer only a retail venue: some of the world’s largest hedge funds and financial institutions are already building the rails to access them. The research problem is the same at any size. Whoever forecasts the settling number most precisely prices the contract best, and that is a question of verified history, measurable error and evidence that arrives before the market closes.
For informational purposes only. Not investment advice. Forecasts and market prices shown are historical. Data as of 13 October 2026.
Developers
© 2026 Bigdata.com by RavenPack. All rights reserved.
Developers
© 2026 Bigdata.com by RavenPack. All rights reserved.