Tuesday, January 20, 2015

NFL Pro Pick-em Strategy - Wisdom of Crowds

Yahoo! Pro-Pickem Strategy - Fail?

Click for larger image. 
It always pays to evaluate the success (or failure) of your strategies, and base assumptions... but I would love for Yahoo! to some data to support the evaluation.

Are Crowds Smarter than Individuals 

My final results for the 2014 Pro-Pickem season were a marked improvement over my standing from last 2013. Overall improvement was from ~2 games missed per week in 2013 (28 games out of first place), to 12 games missed over 17 weeks. That is, Niam Knows! won this year with 186 correct picks, and I only had 174.

Strategy Changes from 2013 to 2014

Previously, I would watch the Pick Distribution and do a gut-check on my perception of the game. For 2014 I also made a couple of simple rules: Always turn in a pick list (in 2012 I missed a week, which doomed my score), and always review the pick list just prior to game time. I didn't make many changes through the year, but a simple trick - watching for injuries to the starting QB - probably saved me 4-5 games. (RGIII and the Redskins hurt that strategy as well).

Something Happened

Business Insider has published an article: Las Vegas Screwed Up - by setting a bad line for the Super Bowl.

... "In the case of the Super Bowl, that didn't happen. About 80% of the bets on the Big Game in the first 24 hours were on the Patriots, according to PreGame.com.

What is even more intriguing about this big swing is that the early betting action is typically done by the more seasoned and often professional gamblers making big bets (AKA "sharps") while Joe Public typically waits until a day or two before the game or even game day.
This means that the usually smarter gamblers felt the opening line with the Seahawks favored was a terrible line and immediately put a ton of money on the Patriots..."

Below is a screen clip from Yahoo Sports Pro-Pickem showing the pick distribution for the Super Bowl as of 1/20. It clearly shows the Patriots as the favorite. So how did Vegas miss? And, was the early money correct, or not? How do the "sharps" adjust for any injuries over the period between Divisional Championships and the Super Bowl? Doesn't uncertainty decrease... as you get closer to game time? 

Bottom Line

The Business Insider article indicates that smart money bets early. In Pro-Pickem, the early bets tend to be homers - people picking their home team. The distribution gets more accurate over the season as the wheat is separated from the chaff. As the season wears on the ability to make up ground on the Pick-em leaders gets more difficult as picks become more obvious and upset games are widely missed. In effect, the leader board becomes self-governing as everyone reflects the tendency towards average.

Is the final Distribution (right before game time) accurate? Is the Crowd more accurate than the individual? 

Wisdom of Crowds - More Data Please

Yahoo Pro-Pickem could really provide a cool data element called "The Crowd". Simply create a player called "The Crowd" and assign all of the picks from the weekly Distribution to this player. For the tie breakers: take the team picked most often for high and low scores, and the average of the numeric scores. (Or, just put Oakland as Lowest Scorer).

Step 2 -- Create a player for each affinity "Fans of San Francisco", "Fans of Dallas" etc. and rank the average "affinity" score. This would provide proof that specific fans know their football, and other fans are just junior varsity. (We all know who you are - rookies!)

Step 3 -- Find the "tail" you know, the guy that makes the outrageous picks... and is correct! Identify the guy that has the most correct picks against the average.

Somewhere at Yahoo! is a math geek that can work this out.


No comments:

Post a Comment

Strictly moderated for language.
Moderately moderated for content.