A reality check on the "odds" provided by Sportsclubstats

I'm pretty sure I was the first in the Kings blogosphere to hawk the greatness of sportsclubstats. If I wasn't, I apologize to whomever I'm slighting; I've been linking to them for years, in any case...since I was a mere commenter even. I mention that because I've noticed a crucial misperception in the way people are using the SCS numbers lately.

First, what SCS does:

  • They generate one hundred million simulated seasons...
  • ...starting from the current standings...
  • ...using a virtual coin flip to determine the outcome of each future game
  • (and they use a second, "weighted", model, in which the coin-flip is skewed slightly to take position in standings and home ice into consideration).
  • and then they present the results of those randomized simulations in various charts
  • giving percentages on how frequently team x finished with y points, and how often y points landed them in zth place.

What SCS does not do:

  • They do NOT make any serious attempt at qualitative analysis. The only straying from "coin flip" is a mathematical weighting (in the weighted model) of certain generic advantages, like home-ice, and the assumption that having a better record might give a slight edge to a team.
  • They do NOT, for example, make any adjustments for injuries, or slumps, or winning streaks, or fragile team psychologies, or curses.
I only look at the uninflected coin-flip model. Because, as a Kings fan, I don't believe that having a better record than, say, St Louis, means we're more likely to beat St. Louis. So, let's take a moment to look at what the current SCS numbers say,
  • The coin flip model says that the Kings make the playoffs in 99% of their 100,000,000 simulations.
  • It also says, in the 541,000 simulations in which the Kings lose their last three games, they miss the playoffs 13% of the time. Feel better? I don't.
  • Why? Because I know that if Dallas wins their next their last three and the Kings lose their last three, the chances of the Kings making the playoffs are zero, unless Chicago loses two of its last three in regulation.
  • If the Kings lose all three, they will lose any tie-breakers with anyone they might be tied with because they will be stuck at 36 wins.
  • So, lose all three, and the Kings need either (1) Dallas to lose once in regulation, or (2) Chicago to lose twice in regulation.
  • I don't think Chicago is going to lose twice in regulation.
  • So...is Dallas going to beat the Avs twice and the Wild once? Now maybe you'll see why the randomized coin-flip model is somewhat different than reality. SCS treats each of those games as singular random events, each with a 50% likelihood of resulting in a Dallas win. Over three games, the odds of the coin-flip coming up "win" three times is 12.5%.
  • But Dallas is on a roll. And its opponents suck. And with each win, the likelihood of momentum tipping in Dallas's favor increases exponentially.
  • And don't forget that Dallas isn't playing in a vacuum. In order for the worst-case scenario to play out, the Kings will be losing while Dallas is winning. And that, too, will have the effect of increasing Dallas's momentum while simultaneously feeding the Kings' descent.
  • In fact, if I were to assign (completely unscientific, uninformed) odds to Dallas's and LA's remaining games, I would say that Dallas has about a 90% chance of winning each of its games (by which I mean each of its games, taken individually), while the Kings have about a 75% of losing each of its remaining games. You might disagree. That's fine. It's completely subjective. I base my analysis only on the fact that Dallas is winning and LA is depleted, psychologically fragile, and they have to rely on two things (goaltending and defense) which just got crushed in its last game.
  • If you multiply out those odds, the individual odds of those six outcomes, you arrive at a 30% likelihood of the Stars winning out while the Kings lose out. And, to me, that sounds a lot closer to reality than the 1% of that happening assigned by the SCS randomizer.
  • Since I made up those numbers, it's actually the odds of the worse-case scenario happening according to my damaged Kings-fan psyche, which you may or may not relate to.
  • Yes, everything I've said is entirely cockamamie and unscientific. But that's the point. The whole process is irrational. The playing of these several games is also driven by forces that are not entirely rational.
  • And, as Kings fans, I think we all know that if the Kings lose all three remaining games, Dallas will win out. They just will.