[Data Analytics] Predicting the cricket scores using DLS
The data in the spreadsheet contains data on ODI matches from 1999 to 2011.
- Using the first innings data alone in the above spreadsheet, find the best fit 'run production functions' in terms of wickets-in-hand w and overs-to-go u. Assume the model Z(u,w) = Z0(w)[1 - exp{-b(w)u}]. Use the sum of squared errors loss function, summed across overs and wickets. You will find one Z0(w) and b(w) for each w. In your report, you should provide a plot of the ten functions, report the (20) parameters associated with the (10) production functions, and the error per point. Your function should be named:
Z0, b = DuckworthLewis20Params()
- Now assume the model Z(u,w) = Z0(w)[1 - exp{-Lu/Z0(w)}] and use the sum of squared errors loss function, summed over overs and wickets. Note that this new regression forces all slopes to be equal at u = 0. In your report, you should provide a plot of the ten functions, report the (11) parameters associated with the (10) production functions, and the error per point. This function should be named:
Z0, u = DuckworthLewis11Params()