We want Notice Price to be very good at predicting the next trade price, and we consider this to be the best way to measure Notice Price accuracy.
To measure predictive ability, we calculate the difference between the Notice Price immediately prior to a trade and the price of the trade. The goal is to have this difference be as small as possible. We then look at this difference over many trades and companies to calculate the +/- % range in Notice Price that will include 80% of the next trades. We call this the 80% confidence predictive range.
If we look across all trades in our database, 80% fall within a range of +20% and -20% of the current Notice Price. This means that if we are printing a price of $100 today, there’s an 80% chance that the next trade tomorrow will fall between $120 and $80. While this may seem like a large 80% confidence predictive range, remember that Notice Price prints prices for 1,100+ companies, many of which are thinly traded.
Predictive range benchmarking
80% confidence predictive range for companies with high private market consensus (indicated by a dark blue Notice Price Drivers icon):
High end of 80% confidence range = +9% (9.7% of next trades will be above the upper end of range)
Low end of 80% confidence range = -9% (10.3% of next trades will be below the lower end of the range)
80% confidence predictive range for companies with low private market consensus (indicated by a light blue Notice Price Drivers icon):
High end of 80% confidence range = +20% (8% of next trades will be above the upper end of range)
Low end of 80% confidence range = -20% (12% of next trades will be below the lower end of the range)
We re-run our predictive range benchmarking analysis periodically and tweak the pricing algorithm if it improves our ability to predict future trades. We updated the Notice Price algorithm to v1.3 in September 2023.
Notice is the only private market data platform that benchmarks our pricing algorithm and publishes the results.