Sunday, March 07, 2010

Sunday sports page 

[] Awful predictions of the past year get their annual recap from Gregg Easterbrook.
After the first weekend of the 2009 NCAA men's basketball tournament, all 5 million ESPN bracket entries were wrong.
If you must make predictions, keep 'em simple...
My off-price, ultra-generic prediction -- Home Team Wins -- went 156-111 this season ... On ESPN's "Sunday NFL Countdown" Mike Ditka finished with 157 correct, and likely wasted some time by thinking.
Also: awful predictions about the stock market, economy, how the universe will end (not disproven yet), a million dead by global warming, global cooling, hurricanes, nuclear power, glaciers, sunspots ... and more sports...
Pro Football Weekly publishes two or three "best bets," to entice readers to sign up for a Handicapping Inner Circle product that costs $109.95 annually. In 2006, the PFW Best Bets went 31-34-2; in 2007, 32-36; in 2008, 35-32-1. For this season, Best Bets went 31-37. That's a four-year total of 129-139-1, meaning when Pro Football Weekly pundits are certain they are right, they are usually wrong.

PFW promotes its Handicapping Inner Circle by claiming 67 percent accuracy -- yet over a four-year period, its actual published picks were 48 percent accurate.

Peter King makes so many predictions, it's hard to know what to take seriously...

In late September, King said, "Minutes ago I spoke to people in Washington who told me there is absolutely no chance Jim Zorn is in trouble with the Redskins." ... Two weeks later, King said Zorn would be fired no later than the following week, to be replaced by Jerry Gray. Zorn wasn't fired until the season ended, and Gray was shown the door too. King said there was "no possibility" Jay Cutler would be traded by Denver. For the season, King forecast a Super Bowl of New England over Chicago -- the Bears did not make the playoffs -- and predicted the Saints would finish 7-9...
Remember folks, the news pages of the newspapers operate just like the sports pages, and the news channels on TV operate just like ESPN.

[] Dave Berri discusses books about sports, thinking, and thinking about sports.

[] As baseball moves from steroids scandals to to human growth hormone scandals, J.C. Bradbury observes: Banning HGH only signals to players that it works. To keep players from using it, make it legal and let them see it doesn't work.

[] The expiration of the salary cap in the NFL during the just-started free agent signing season gets a look from Brad Humphries at The Sports Economist.

[] Steady return generally has more value than inconsistent return at the same average rate. Phil Birnbaum discusses this regarding predicting game outcomes from team average for-against scores.

As a simple example of why this is so, say a baseball player hits six home runs. If he hits them all in one game he may feel great about setting a record -- but most of the home runs probably will be wasted running up the score, and none of them would help his team win any other game. In contrast, if he hits one home run each in six games he could help his team win two, three, or four games.

As it happens, this week I got in a discussion on this topic in the comments to a post at Pro Football that presented a counter-intuitive finding that higher pass completion percentage for NFL quarterbacks is associated with scoring fewer points scored and fewer games won -- while higher numbers in other metrics, such as average yards per pass attempt (AYA) are associated with more scoring and winning.

Now, I much prefer AYA as a measure of QB performance over pass completion percentage, since the objective of passing is to advance the ball downfield, not to successfully complete a very high percentage of passes that don't advance the ball.

(And for the record, the NFL's official passer rating is probably the worst metric of all. It is so biased towards completion percentage that a QB can increase his rating by completing passes that lose yards: hit 10 of 10 for minus 10 yards each and he'll get a rating of 79 for losing 100 yards.)

But even with AYA being my preferred measure, logic says completion percentage should have some additional positive value. Analogous to the home run example, while the goal is to advance the ball, advancing it any given total amount at a steady rate (with a high completion percentage) should be preferable to doing so hit-or-miss (with a low completion percentage).

So I got out my spreadsheet software, plugged in the NFL 2009 season passing numbers, and sure enough multiple regression produced a formula that predicts points scored from passer data weighting both AYA and completion percentage positively, while giving the former twice the weight of the latter.

And from that I can produce my very own passer rating formula! Looking at the numbers, they actually correlate with scoring better than do AYA, the NFL's official rating, or any other rating method I've found in a couple days. In principle, from that I can rate passing defenses (offense in reverse) then rate whole teams ... then pick winners, and best bets against the spread...

Come the start of the 2010 NFL season I may have my own sport web site, be rating teams and predicting game outcomes ... and for $150, be selling my Premium Best Bets.

A year from now, you may be reading about me in Easterbrook's column!