|
on Forecasting |
| By: | Steffen Q. Mueller (Chair for Economic Policy, University of Hamburg) |
| Abstract: | This study explores the forecasting of Major League Baseball game ticket sales and identifies important attendance predictors by means of random forests that are grown from classification and regression trees (CART) and conditional inference trees. Unlike previous studies that predict sport demand, I consider different forecasting horizons and only use information that is publicly accessible in advance of a game or season. Models are trained using data from 2013 to 2014 to make predictions for the 2015 regular season. The static within-season approach is complemented by a dynamic month-ahead forecasting strategy. Out-of-sample performance is evaluated for individual teams and tested against least-squares regression and a naive lagged attendance forecast. My empirical results show high variation in team-specific prediction accuracy with respect to both models and forecasting horizons. Linear and tree-ensemble models, on average, do not vary substantially in predictive accuracy; however, OLS regression fails to account for various team-specific peculiarities. |
| Keywords: | Attendance, Major League Baseball, Random forest, Conditional forest, Sport demand, Sports forecasting, Ticket sales, Variable importance |
| JEL: | C44 C53 |
| Date: | 2018–06–27 |
| URL: | http://d.repec.org/n?u=RePEc:hce:wpaper:065&r=for |
| By: | Fortin, Alain-Philippe (HEC Montreal, Canada Research Chair in Risk Management); Simonato, Jean-Guy (HEC Montreal, Department of Finance); Dionne, Georges (HEC Montreal, Canada Research Chair in Risk Management) |
| Abstract: | When forecasting the market risk of stock portfolios, is a univariate or a multivariate modeling approach more effective? This question is examined in the context of forecasting the one-week-ahead Expected Shortfall for a portfolio equally invested in the Fama-French and momentum factors. Applying extensive tests and comparisons, we find that in most cases there are no statistically significant differences between the forecasting accuracy of the two approaches. This suggests that univariate models, which are more parsimonious and simpler to implement than multivariate models, can be used to forecast the downsize risk of equity portfolios without losses in precision. |
| Keywords: | Value-at-Risk; Expected Shortfall; Conditional Value-at-Risk; Elicitability; model comparison; backtesting; Fama-French and momentum factors |
| JEL: | C22 C32 C52 C53 G17 |
| Date: | 2018–06–18 |
| URL: | http://d.repec.org/n?u=RePEc:ris:crcrmw:2018_004&r=for |
| By: | Yen H. Lok |
| Abstract: | The contribution of this paper is two-fold. The first contribution is the development of a filter-combine scheme for trading strategies to diversify model risk. Multiple statistical machine learning models are used to predict the price direction of multiple assets. We demonstrate the effectiveness of model-averaging after under-performing models are removed via a filtering algorithm. The second contribution is the identification of appropriate measures of performance for selecting models. In the literature, different measures are usually designed for different applications and purposes, and it is not always clear as to whether certain measures are relevant to a particular trading strategy. By identifying relevant measures, one can identify the key drivers underlying well-performing models, and allocate more resources in optimising and improving the appropriate models. |
| JEL: | C51 C52 |
| Date: | 2018–06–22 |
| URL: | http://d.repec.org/n?u=RePEc:jmp:jm2018:plo493&r=for |