Calibration, Forecasting and Dynamics Event as iCalendar


16 March 2018

11:30am - 12:30pm

Venue: The University of Auckland Business School, Level 2, Room 205, 12 Grafton Road, Auckland, 1010

Presenter: Professor Sergiu Hart, The Hebrew University of Jerusalem

How good is a forecaster? Assume for concreteness that every day the forecaster issues a forecast of the type "the chance of rain tomorrow is 30%." A simple test one may conduct is to calculate the proportion of rainy days out of those days that the forecast was 30%, and compare it to 30%; and do the same for all other forecasts. A forecaster is said to be calibrated if, in the long run, the differences between the actual proportions of rainy days and the forecasts are small - no matter what the weather really was.  The classical result of Foster and Vohra (1998) is: calibration can always be guaranteed by randomised forecasting procedures (a short proof will be provided).

We propose to smooth out the calibration score, which measures how good a forecaster is, by combining nearby forecasts. While regular calibration can be guaranteed only by randomised forecasting procedures, we show that smooth calibration can be guaranteed by deterministic procedures. As a consequence, it does not matter if the forecasts are leaked, ie, made known in advance: smooth calibration can nevertheless be guaranteed (while regular calibration cannot). Moreover, our procedure has finite recall, is stationary, and all forecasts lie on a finite grid. We then consider “smooth calibrated learning” in n-person games, and show that in the long run it is close to Nash equilibria most of the time.

The talk is self-contained; no prior knowledge will be assumed.

Sergiu is the Professor of Economics and Professor of Mathematics, at the University of Jerusalem.  He received his PhD at Tel Aviv University in 1976.  His previous academic appointments were at Stanford University, Tel Aviv University, and Harvard University(visiting). The main area of research of Sergiu Hart is game theory and economic theory, with additional contributions in mathematics, computer science, probability and statistics. Among his major contributions are studies of strategic foundations of cooperation; strategic use of information in long-term interactions ("repeated games"); adaptive and evolutionary dynamics, particularly with boundedly rational agents; perfect economic competition and its relations to models of fair distribution; and riskiness.

View the paper "Smooth Calibration, Leaky Forecasts, Finite Recall, and Nash Dynamics" by Dean P Foster and Sergiu Hart.

For more information contact:
Haiping Zhang
Ext: 81152