In episode 16 of the Global Guessing Weekly Podcast, we discussed the Bias, Information, and Noise Model of Forecasting with Ville Satopää and Marat Salikhov.
Every forecaster has wanted to know what the most important factor for improving forecasting accuracy is, but for a long time the answer was not clear. Thanks to a chance overlap of co-authors Ville Satopää and Marat Salikhov at INSEAD, however, a new paper was published alongside forecasting pioneers Philip Tetlock and Barbara Mellers that does a great job of providing a solution.
Their paper, “Bias, Information, Noise: The BIN Model of Forecasting,” deconstructs the forecasting process into its component parts of:
- Information: The inputs you use to move your forecast away from the base rate;
- Bias: Systematic error across a number of forecasts from a single forecaster; and
- Noise: Non-information that is registered as information in a forecast).
From there they test which of these parts is most critical to the accuracy of a forecast, and posit methods to improve in these areas.
In this episode we are lucky enough to sit down with Ville and Marat to discuss the origins of this paper and their follow up one on Crowd Aggregators, their findings from both papers, and the implications for the future of forecasting. We talk about possible avenues for further research based on the exciting results from Ville and Marat’s research, and even speculate on potential applications of the research in new and interesting environments.
We are both big fans of this paper and its authors, and we found the conversation with Ville and Marat extremely rich! Be sure to tune in to this episode, and then go read their paper and let us (and them) know what you think.
Listen On Other Platforms
Just look up Global Guessing Podcasts to find the episode! New episodes might take 6-24 hours to show up, so make sure to subscribe to the podcast so you'll know when the new episode goes live.
- Amazon Music
- Pocket Casts
- Podcast Addict