Here’s an interesting article about how a computer simulation run 10,000 times has determined that while Peyton Manning and the Colts will win Superbowl XLI, they will not cover the spread. For those scoring (betting) at home, Accuscore predicts that the game will come down to the wire with Indianopolis prevailing by about a field goal.
What struck me as interesting was the methodology: they don’t just base their simulations on player statistics, but consider many different factors. From the article…
”We don’t input any stats into our engine,” [COO Gibby] McCaleb said. ”I can’t go into too much detail because of our intellectual property, but we actually play the game one play at a time.
”Stats are completely misleading. Weather, field position, score, time on the clock, coaching tendencies — you name it, it’s in there. This is why our simulation engine is so complex and takes days to run a full NFL season.”
It’s a far cry from how the engine originally was conceived.
”In studying evolution, I worked on … simulations where a population’s segment of DNA was modeled one generation at a time,” Oh said. ”Each generation [featured] things like mortality rates, birth rates, mutation rates and population movement rates, which were used to simulate how each generation ‘lived.’
”To determine what these rates are, we reviewed extensive research by geneticists, mathematicians, archeologists and anthropologists. The evolutionary program would use the average rates and variability to simulate a single generation. It would then simulate the same sample for a predetermined number of generations and output how this population looks ‘genetically.”’
Comments RSS
| TrackBack URI
It was inevitable, I suppose. Look at how things like the Iowa “Political Stock Market” and the Hollywood “Fame Market” have done in predicting outcomes. Even UBC’s own “election prediction site” - which uses nothing more than volunteered comments written by political junkies of one stripe or another - has a solid prediction record, both federally and provincially.
This somewhat makes me wonder why I would bother (if not now, then someday very soon) pay anyone a management fee for managing a mutual fund. Not only don’t computers need flash cars, fancy houses, $6 lattés and expensive trophies of all sorts, they are far less likely to allow sentiment to cloud their judgement when it’s time to unload a position that isn’t doing well. Really, only a very good contrarian - someone who sees what the “simulation” can’t, not having had it programmed in yet - is worth a fee these days.
Comment by Bruce Stewart - January 27, 2007 3:20 pm
Hey - I work for the Colts and I’m blogging about Wikinomics here…http://www.patcoyle.net/2007/01/30/things-fans-will-do-with-animals-youtube-and-myspace/
Comment by Pat Coyle - January 30, 2007 11:11 am
Hey Mike,
As a fellow sports-geek I’ve been a huge fan of http://whatifsports.com.
They also come at it from a simulaton point of view, and have done similar exercises in the past.
Cheers.
Comment by Ian Ketcheson - January 31, 2007 3:41 pm
Hi Pat:
Thanks for the mention in your blog. I checked it out…its now in my favourites (note the Canadian spelling — goes with our preference for 3-down football).
Mike
Comment by Mike Dover - February 1, 2007 9:28 am