Posts Tagged ‘fancy stats’

The month of January has been a very interesting one the Minnesota Wild, to say the least. They lost their two best skaters to injury in Mikko Koivu and Zach Parise, plus one of the best feel-good stories in the league in Josh Harding, who was really holding the team together through their offensive struggles earlier in the season. And yet, the club is 8-3 during this month, and as of today they sit in the 8th spot of the Western Conference playoffs–with a four point lead over Phoenix, who has two games in hand. The computers give the Wild a 51% chance to make the postseason, and the Yotes a 38% chance, so it’s really anybody’s game at this point. Or is it? Stats guys get a bad rep for being the bearers-of-bad-news, of throwing ice water on everything…and unfortunately today I’m going to step right into that stereotype. Read on…

 

Today I take a minute to reflect on why people doing advanced stats are like zen monks and how the blogosphere zeitgeist makes this a very exciting time for sports analytics. Also why I named my NHL 14 skater Carlos Danger. Article referenced: bit.ly/1cPkWiB

Hockey is a game, and fantasy hockey is a game based around a game (though not a game within a game, that would be Inception!)

In game theory, an important distinction is whether the players have perfect information or imperfect information.

In games with perfect information (chess and checkers are good examples,) all the factors are known by all players. Each player can see all of his opponent’s pieces, and there is no ‘hidden’ information except for the plans and strategies inside the other guy’s head.

In games with imperfect information (most card games, blackjack and poker for example,) some of the factors are unknown by the players, which is where the complexity, the skill, and the intricacies of the game come into play. If everyone knew the dealer’s face down card in blackjack, or what everyone else at the poker table had as their pocket cards, those games would be a lot less fun.

So, do we have perfect information or imperfect information in fantasy hockey? Some would foolishly say it’s the former–they look at the stats for last night’s games or a guy’s production last season, and think it’s perfect information. After all, there’s the numbers right there in black and white, there’s everything that happened in the game, right in the box score. But when I think about how advanced stats like Corsi, PDO, zone starts, and qual comp contribute to fantasy, I think about how they make those boxcar stats look imperfect in an awful hurry. Sure, this guy got that many points, but he did it against really soft competition, or he got some really good puck luck, and those points may not really be an accurate reflection of his skill or a predictor of how many points he can put up going forward.

But on the other side of the coin, fantasy doesn’t care about a player’s QoC or his PDO…goals are goals. Let me clarify: if you drafted Patrick Marleau or Thomas Vanek this season, you got yourself a whole heap of goals in the first few games of the season, and you almost certainly won your first couple matchups. Whether those goals came on ridiculously high shooting percentages (they did) or came easier because of favorable zone starts (they did,) they still counted. Last April, I got myself on board the Pascal Dupuis express during my fantasy playoffs, and I enjoyed a nice little hot streak, to the tune of around a point per game over a couple weeks. Did I know that production was unsustainable? Certainly. Did I keep sending him out there night after night? You’re damn right I did! But I digress…

The point I’m trying to make is that the role of more sophisticated stats (or advanced stats, or underlying stats, or fancy stats, or whatever you want to call them, ultimately it doesn’t really matter) is to provide more *context* to a hockey player’s production. And in fantasy, that context can be supremely helpful. It can give us strong signals on whether to buy low or sell high on a team or player, and those signals (if we choose to heed them) can give us a leg up on our competition who is ignorant to even basic stuff like individual Sh%.

These stats don’t give us perfect information, by any means. No self-respecting stat guy (or gal) would tell you that. And we as a community are constantly trying to improve our methods, to develop new numbers and metrics that are meaningful and useful and not downright crazy (if you’re plugged in to the #fancystats community on Twitter, you may have heard about the paper that got accepted to the Sloan Sports Analytics Conference that makes that case that Alexander Steen is one of the most important players in the league. But I digress again…)

Bottom line: in fantasy hockey, we are dealing with imperfect information. But the value of Corsi and Fenwick and the usage charts and all that is that they give us slightly *less imperfect* information. One step closer to knowing when the dealer is about to flip over that suicide king, or that the guy on the button is working with deuce-seven offsuit.

^^The podcast is embedded in this page. Click the play button above to begin streaming, or click here to download an .mp3 file.

Today I interview one of my favorite hockey researchers, Rob Vollman of Hockey Prospectus and Hockey Abstract. Rob has developed a number of great tools and stats that have become a major part of the hockey analytics community.

“A good hockey stat first of all has to be useful.”

“The real achievement is to take something that’s complex and make it simple.”

LINKS:

Hockey Abstract

Rob’s personal page where he puts all his work. This site is a wealth of information–everything from Quality Starts to “Oz Coke charts” (AKA player usage charts) to historical comparisons. Your one-stop-shop for Vollman’s writing.

Hockey Prospectus

Another great site with a ton of great hockey writing from a bunch of today’s smartest minds.

>>E-mail me your fantasy hockey questions at hashtaghockey [at] gmail [dot] com, and follow me on Twitter, @Hashtag_Hockey