SI.com Advanced Analytic Model Failing in Spectacular Fashion

[polak]

Saw this on HF and since we’re the skyscraper sized thorn in the Analytics crowds ass, I figure it was worth a share on here.

http://www.si.com/nhl/2015/04/14/201...key-fancystats

Sports Illustrated claimed their advanced analytics model would predict 75% of the series outcomes correctly (11 out of 15). Well so far they’re 2/6 and can’t get any more incorrect if they are going to reach their goal. They picked the Capitals and the Lightning too so that could drop to 2/8 quite easily. Also they were wrong in the # of Games in the series they picked correctly but I guess we can let that slide... Even though they predicted they would both go to 7 and they only went to 5

The best part? They got one of the analysts 4 year old daughter to make some predictions and she is beating them! (okay only by 1, but still)

Hopefully this trend continues and people stop underestimating the powers of intangibles, shot blocking and shot quality.

Either way, their ineptitude is quite satisfying for this Flames fan.

[Geeoff]

Sounds like SI made up their own metric combining a bunch of different things.

[trackercowe]

Our model says that it will be almost as close as the Blues–Wild series, and that the Canucks will squeak through to round 2, where they’ll either get smashed by the Jets or have a competitive series against Anaheim.

Haha, they sure nailed that one!

[cannon7]

Even if the Flames make the Stanley Cup Finals, the moment they lose you'll have some announcer saying, "well, you knew it couldn't last...blah, blah, blah."

[RyZ]

- Only 5 of the top 10 "corsi" teams made the playoffs this year.
- 3 of the bottom 10 made it.
- The Presidents Trophy winner NYR finished 19th.

This being the analytics movements big break out year into the mainstream was pretty much a catastrophic fail.

Its a useful stat, but it's just another stat to add to the line. There is nothing special or predictive about it over and above stats we have already had for years and people that use it exclusively as a measure of how good a team is are foolish.

[Five-hole]

Originally Posted by polak

Saw this on HF and since we’re the skyscraper sized throne in the Analytics crowds ass, I figure it was worth a share on here.

[polak]

Originally Posted by Five-hole

Damn it.

[DJones]

I guess they had Winnipeg at 70% chance to beat the Ducks.

Winning one goal games apparently means you are lucky not good.

Actual Numbers; http://www.depthockeyanalytics.com/u...i-predictions/