I'm about ready to launch NFL next weekend. I have back tested data and a pretty solid look at the previous years. Yes, I know its back tested. I have heard the argument before. My question for the board is that as I have looked at the current season as it happens...I have made some minor tweaks. Obviously, I continue to look at how these tweaks effect previous years performance. 2009, 2010, 2011 and 2012 are solid at 60%+... but 2008 falls off a cliff and hits 48.5%.... I know the kick off rule etc... but has the game changed that dramatically. I know it is more of a passing league now than 2008... but I really can't pinpoint anything that is that dramatic.... Thoughts...
I can't exactly tell if that is the challenger.. but I have to assume...
it is. Very helpful... I put on my NFL picks and my college picks. I won't bet NFL until next week. But college I am betting now... you can laugh later. Enjoy Dave
Hey Dave, good luck w/ the upcoming season; it's exciting to launch a model live!
I have a question; for your model parameters, how did you optimize those? I use some built-in optimizers in Matlab, but was wondering what exactly is the objective function; do you use your backtester for that?
You are going to laugh but I optimize using a brute force method. All of my statistics are obviously downloaded in excel and as a whole I find excel to be bordering on worthless. I attempted several times to use the Solve Function, but since I am using multiple tabs and pages Excel won't let me. I have all my outputs centralized on one page, so as I change the weighting of the parameters, I can see how it affects the last 3-5 years of data and outputs. To answer your question, I need a better way to optimize and if you have any suggestions I am all for it. Now the positive of using the brute force method is that it allows you to dive into the numbers more because you are not solely relying on the computer and a formula.
Right, so you're essentially using your backtesting results for the past 3-5 years as your objective function.
I'd recommend porting over that data to Matlab, write a backtesting script, then use that as the objective function for something like the built-in simulated annealing optimizer. It'll automate exactly what you're doing manually (let it spin around for a day or so)!
....If I attempt the import...I will probably do it in the off season. Btw, I got rejected from Cornell's engineering program back in 88...LOL. R you up for a conference call. I could use your input of importing the data over. Ty in advance.
SBR Top-Rated Sportsbooks
Modeling question about the NFL...
Reply With Quote
