Sunday, November 28, 2010

What Should "Most Valuable" Encompass?

Even though 82 games seems like a long time and a large sample size, it really isn't. There's a heavy amount of luck in the standings, as there is in individual performances. We see players like Jonathan Cheechoo, Lee Stempniak, and so on go on hot runs for parts of seasons or for a season or two, then fall back into invisibility.

The same applies to players whose normal levels of play may not be in invisibility: Henrik Sedin last season, for example. Sedin is a point-per-game player who suddenly saw his competition level drop, his own shooting percentage skyrocket, and his teammates' shooting percentages skyrocket (Alex Burrows, Mikael Samuelsson). Sidney Crosby shot 17% last season, well above normal and higher than he can sustain, almost certainly. Even our own Alex Ovechkin benefited from high on-ice shooting percentage; how else would his 1.3 points/game-ness jump to 1.5, and in fact 1.6+ prior to his second suspension?

Corsi ratings are less influenced by luck since the sample size is around 10 times bigger than goals, if not more. But should "value" encompass random performance variability ("luck"), or be based on repeatable performance? Should players receive a bump in value for benefiting from factors they essentially cannot control? I'm of two minds on this.

I can see the argument for not considering "luck." It's impossible to tell how much is a player's own luck (like wearing socks that doesn't rub his flesh raw, taking a new brand of antihistamines, using Alex Ovechkin's stick, or playing particularly inspired hockey against a rival) versus others' luck (the other team was playing street hockey in front of the White House the morning of the game, the starting goalie's wife gave birth to their first child that day and the goalie decided to pitch a shutout behind you, your coach always puts you out on the first power play unit with four superstars). While the luck is beneficial to a team, it makes the team bank on it at times, which can be harmful if the luck decides to suddenly disappear (see '09-'10 Capitals). Moreover, are you really helping your team much if you're giving away lots of scoring chances in your own end and not generating in the offensive end? Only people you're helping then are the other team and whoever makes the brand of aspirin your coaches buy.

On the other hand, the bounces that go your way and other uncontrollable factors do help your team. We don't have much idea how events would go without those bounces; fact of the matter is that they did happen. Should Jeff Schultz be severely penalized for his 106.9% PDO last season? It wasn't his fault, after all, that it was so high. Who knows, maybe his teammates had more confidence to go with quantity over quality with Schultz on ice. Maybe Schultz is a great locker room leader. Discounting luck means discounting whatever slight control a player may have on his luck, like leadership, clutch play, and who knows, maybe even stealing pucks.

Whatever effects luck has, at the end of the day, still have plenty of value. For example, the Capitals' shooting luck may have accounted for roughly forty extra goals for. Putting 273 goals into the Capitals' pythagorean (1.12* [GF^2]/[GF^2 + GA^2] yields about 109 points, great but well below the 121 points the team actually posted. The Capitals' seeding wouldn't have changed, but what about San Jose, whose shooting and stopping pucks luck edged them one point ahead of Chicago (113-112) for first place in the West? San Jose, instead of drawing underrated tough Nashville, drew the Colorado Avalanche, a team that jumped from bottom-5 to 12th overall in the NHL purely on the basis of luck (though don't tell them that). Whatever luck the Avs had was most certainly valuable (though if I were them I'd have taken that top-5 draft pick to pick up Cam Fowler, Vladimir Tarasenko, or who knows, Taylor Hall if they'd have won the lottery, over the playoff appearance).

The line on whether or not to consider luck becomes even more blurred with goalies, for whom random performance variation gives us Cristobal Huet putting up both best and worst save percentage numbers within a five year span and the luck of a single goalie can change a team's fortunes drastically--another reason why the Sharks finished ahead of Chicago was because of Evgeni Nabokov's overperformance, and another reason the Capitals won the President's Trophy was because Jose Theodore performed as well at even strength as Martin Brodeur, .924 at even strength.

My thoughts? Don't vote goalie for the Hart unless his performance vastly exceeds anyone else's (I'm talking along these lines). Discount luck if without it the guy would be a net negative relative to everyone else on his team. Otherwise, discount some of it if the guy would already be a huge positive (like Alex Ovechkin).

If the luck was negative, on the other hand (like the Hawks' 27th overall PDO), then discount it entirely, unless the performance was actually a huge underperformance (like Semin's 44 shots without a goal, which is less than 1% probable).

With all of this, make sure to look at whether points come from goals or assists, and which types of assists as well, plus zone starts (looking at you, Patrick Kane). Also make sure to look how much damage was done on the power play versus at even strength or shorthanded. My personal preference is to favor even strength goals and to discount luck save personal shooting luck.

My second question: should games played matter? It depends on whether we're discussing most valuable in a relative sense or most valuable in an absolute sense. Player A may play 80 games and have a GVT of 24, while Player B played 70 with a GVT of 22; Player A has more total value in the season with 24 extra goals, but Player B when in the lineup provides more value to his team on a per-game basis. I think when within 10 games played rates are better to use, since missing a few games here and there is uncontrollable luck for the most part, and players should not be docked all that much for bad luck.

My third question: should salaries matter? When we're talking about Nicklas Backstrom versus Alex Ovechkin in 09-10, for example, while we probably have a good idea of the value of each as individuals, Backstrom allowed the addition of another $7 million in cap hit over Ovechkin. While the Capitals didn't use that space all that much, playing on a below-market-AAV contract has value in allowing the team to use the difference between market value and the cap hit to sign another player. Sidney Crosby, for example, I'm sure took a pay cut which indirectly allowed the Penguins to hang onto Chris Kunitz. Backstrom's pay cut may allow the Capitals to keep Alexander Semin.

I think that this question becomes especially relevant when the player is in fact paid at market value, but then gets better and outperforms his contract. Look at Craig Anderson last year--his performance might have been worth around $6 million, if not more, yet he was paid only $1.5 million. That's means a team spending a little under the cap can pay market value for each player, and instead of being only average, can be a "contender" with some players outperforming.

I think the first case of players taking a discount should not be counted, but the second case should be, of players simply playing better than the contract (except in the case of entry-level contracts, which have salary ceiling restrictions).

In short: sorry Steven Stamkos, Sidney Crosby has been the MVP so far this season.

No comments:

Post a Comment