This week, we look at the stupidity of computers, the limitations of current hockey data as well as some in-depth looks at a couple of young centermen.
Matt Fenwick shows us that simulations and models mirror the assumptions of their human creators and therefore can be fooled by randomness just like any innumerate dunce.
Matt notes that the best simulation of the New Jersey Devils season by the simulator over at Hockey Reference had them at 70 points back on December 24th. A couple months later and the simulator has been shown to be completely wrong. Spectacularly wrong.
That projection/simulation is at Hockey-Reference.com, and with due congratulations to the man behind it for winning the TrueHoop Stat Geek Smackdown in both 2007 and 2008, it's useless. Ultra-useless. Somehow, a mathematical(?) model of an NHL team produced a result saying that the best of 1000 scenarios for the upcoming 48 games is that they would be 4 games over .500.
This shouldn't even pass the smell test for this year's post-apocalyptic version of the Oilers, let alone mine or anyone's smell test for the team that had the lowest SA/gm in the league.
The Devils had okay shooting numbers to start the season, but their percentages cratered across the board. The HF model obviously didn't anticipate regression to the mean.
At the start of his post, Matt also reminds us that we still have a lot to learn when it comes to quantitative analysis in hockey. Tyler Dellow carries on in the same vein with this post on Dion Phaneuf and his declining performances on the power-play. He notes:
We don't really measure whatever skills might lead to a good PP in hockey. We know that more shots are, generally speaking, better than fewer, and that's been a problem in Toronto (although Phaneuf is producing around his historical level). We know that shooting percentage is important as well, although it's subject to vicious fluctuations. If we really wanted to get an idea of the skills that matter, we'd be charting every touch of the puck on the PP and getting an understanding of where a good PP team looks different from a bad one. Is it an inability to gain the line? A tendency to shoot from bad spots? A tendency to shoot without making the goalie move? With better data, all of this could be answered. As it stands, it can't.
Those limitations in mind, Robert Cleave does good work examining the progression of Flames rookie Mikael Backlund and compares it to fellow Swede and former first round pick Patrik Berglund. Flames coach Brent Sutter has heavily sheltered Backlund this year, resulting in limiting responsibilities and opportunities, much to many fans chagrin. Robert plots the use of Berglund in St .Louis during his first three seasons in the league and determines that Sutter's use of Backlund isn't really outside of the norm.
Finally, John Fischer has an exhaustive audit of Travis Zajac's season over at In Lou We Trust, ranging from counting numbers to underlying stats. The results are somewhat mixed, although overall the news is good for Devils fans.