I Can Take The Numb Out Of Numbers- I Promise (3/19/2012)

(This article was originally published by me as a guest author at the site called “Surly and Scribe”)

The Kings are on track to be the Champions of the Pacific Division, finish 3rd with home ice, and vindicate the much-maligned El Deano. That’s right, I said it: if the teams involved simply continue to win as often as they have all season at home and on the road, the Kings outpace Dallas, San Jose and Phoenix to take the Pacific Division Title.

This may sound bold, or probably biased, and although I like to think I am both of those things, this projection is based on more than dart throwing, my Ouija Board and Madame Moldavia’s Tarot Card Reading. I am using (gasp) “Math.” According to my pipe dream of a standings predictor, the Scoracle, the Kings go in hot and it’s the other teams that are bothered.

Previously, the Scoracle projected the final standings of 6 teams based solely on their point-getting percentages over the entire season. The result was for the Stars to be 3rd, the Coyotes to be 7th and the Sharks to be 8th, followed by the Flames, Kings and Avalanche. That sucked, and just seemed wrong somehow. If I projected just home and road percentages accumulated over the full season, I reached the “wrong” result, even though that way seemed completely logical. But, I am told, there is more than one way to skin a cat.

First, a few quick thoughts about skinning a cat. It would appear that: (1) Some people really hate cats. (2) Some people probably eat cats. (3) If people know how to skin a cat in more than just one way, that cat has to be really pissed off by now. (4) Cats’ skin apparently grows back, and almost certainly very quickly. (5) Furballs are either completely wasted or are used as stock for transplant back to the cat.

Since we’ve accepted the cat skinning rule, let’s apply it to the projecting of standings. Maybe just using season-total point-getting percentages creates an incomplete picture. (If you were to ask Bobby and Surly, they would tell you that “spreadsheets” is a term that should only be used by hotel maids while cleaning a room, but I am using the “no filters and no censorship” rule against them)  Surely, Shirley, there must be other important factors, than just season-long points percentages.  I have decided to look at one, and trust me, it will be short-bus simple. What if we try to include a team’s recent play, their recent performance? For example, Dallas.

In their last ten games, they are 7-2-1. They lost their last game, so before that they were 8-1-1, even hotter. And remember, all the original Scoracle projections dated from before games were played on March 13, so for projection purposes we adjusted Dallas’ record over their last 10 games to what it was on March 13. Between March 13 and now, Dallas lost two games, meaning that as of March 13, Dallas “Last 10 Games” was 9-0-1, a points-percentage of .950.  Holy shit, that’s hot. I hate Dallas.

9-0-1 is record-setting hot if kept up over a full season, by a long shot. Crazy hot, almost like Detroit’s home win streak of 23. But we knew that number was unsustainable. Still, it was a real number, and one cannot deny that Dallas was on a tear. If Dallas was on a tear, they were playing above their season’s average, and to project standings based solely on their average over the full season would not give the full value to how well they are playing now.

If our only criterion was a team’s previous 10 games, we could base the final standings on how hot, or not, each team is playing now, during crunch time. So here’s that “math,” the projections according to just the last ten games played by each playoff contender as of March 13:

As you can see, as of March 13 and based on recent play alone the Kings would have still missed the playoffs, finishing in this order: the Stars win the division, the Avalanche finish 7th, the Coyotes 8th, the Flames 9th, Kings 10th and the Sharks 11th. These results are very different than the original projection made using season-long percentages. A side by side comparison of the last 5 teams as projected by the two methods is here:

The Avalanche go from 11th to 7th, the Coyotes drop from 7th to 9th and the Sharks drop from 8th to 11th. If the two results are so different, it seems that neither is correct if taken alone. We can’t just use recent overall play to project, as much as I would like to since it drowns the Sharks. Sorry, no can do; just as it makes no sense to project Dallas staying at .950 for the rest of the season, it makes no sense to apply just the “Last 10 Games Average” to the other teams. This alone is not good enough. And also, the season point-getting average is not good enough alone, either. We have to find a way to use both numbers together; this is inevitably subjective, and is called “weighting.”

If a team was great at home all year, like Detroit winning 23 straight, but then they lost 3 of 4 and were badly outplayed in the 5th game (against the Kings), then we have to decide how to describe that team’s current level of play. Are they the 23-straight winners, or are they 3 of 4, or 4 of 5 losers? How do we incorporate both, into one number?

My belief is that the two numbers are almost equally important. You can’t ignore either. To describe how good a team is, or how hard it would be to beat them, we have to include what they have done all year, but also how well they are playing in the most recent past. And again, home and road must be accounted for in the recent past, just as it is in the season long number. So, I am going to use the season-total percentage of points won, and the team’s percentage of points won over their last ten games, and I am going to add them together along with the team’s last 5 games at home or on the road, depending on which venue they will play at. Let me make that more clear.

A team has a number of road points. That number is a percentage of points possible. The road percentage is usually less than the home percentage. A team’s road percentage over a full season can be different than the team’s road percentage over their last 5 road games; same with their home percentage. Also, a team’s combined road and home percentage over a season can be different from a team’s combined road and home percentage over their last ten games.

It is my belief that when judging a team’s chance to win their next game, their season-long percentage is equally important as their most recent play. Their most recent play can be defined by their last ten games overall, and by their last 5 games either at home or on the road, depending on where the game in question will be played. If a team is going to play at home, I would look at: (1) their season-long percentage at home, (2) their last ten games overall percentage and (3) their last 5 home games percentage.

As an example, for Detroit at home, I would take their total season points percentage at home (.853) and add it to their last ten games points percentage (3-6-1 = .350) and add those to their last 5 home games points percentage (2-2-1 = .500), and then I would average those numbers. So, (.853 + .350)+(.853 + .500) = 2.556, and 2.556 divided by 4 = .639.

Now, I think the Red Wings should be considered a .639 team at home, based on their full season at home as well as their recent play at home and their recent play overall. To do this for the Red Wings on the road, I would simply use their road season average, their last 5 road games, and their last 10 overall. From these I would have a new number to use when calculating according to the Scoracle format.

Here then, with all the adjustments for recent play factored in, are the projected standings for the last 5 teams as of right now, according to the old (season only) rules and the new (season plus recent play) rules:

As you can see, with the adjustments for each team’s recent play on the road and at home, balanced against their full-season play on the road and at home, the Kings beat Dallas and everybody else. The margin is very slight, but the Kings win. If either team varies just slightly from their average, to the good or to the bad, the Division Title hangs in the balance. And, Dallas is almost certainly going to win a tiebreaker, because as of today they have 3 more regulation/overtime wins than the Kings. If the Kings are going to win the division, they have to beat Dallas by one point or more. If we look at it objectively, the numbers are so close between Dallas and the Kings that it could be a tie, which would give Dallas the Division. Since I am a Kings fan, it’s like any other game: the score is tied which means the Kings are winning.

From doing this work, I learned that the toughest remaining home opponents go to the Flames at an average of .522, then Sharks at .521, Stars at .520, Avalanche at .513, Kings at .490 and Coyotes at .470.

The team facing the hardest average remaining road opponents is the Stars at .619, then the Sharks at .609, Coyotes at .577, Avalanche at .575, Flames at .553 and then the Kings with the easiest road opponents averaging .524. Considering that the Kings are averaging a weighted number of .603 (combining recent play with season long play) and facing an average of .524, the Kings have by far the best prospects for picking up points away from Staples Center.

The Kings have done the hard work, which is to draw even with teams that were above them. And when I say hard work, I mean extremely hard work. Take Pittsburgh, for example, fighting with the Rangers for the Atlantic Division title. As of February 20, Pittsburgh was 10 points back.  Since then, Pittsburgh has won eleven straight games and finally lost last night in OT, to go 11-0-1. In their last ten, the Rangers are 4-5-1.  But even with all that, as of today Pittsburgh is still 1 point back. So, even if you win every game while the other team falters, that other team is not going to suddenly lose every single game while you go undefeated. The Kings need to keep going they way they have been, and hope the other teams don’t suddenly turn into Pittsburgh.

The standings, and the projections, are so close that factors we can’t measure using statistics are likely to be the difference. If you have to pick a winner, the numbers say one thing, and even though those numbers include the results of intangibles over the full season, the intangibles are hardest to predict over a short span of just a dozen games or so.

The good news for the Kings is that all this data cannot measure which team is gonna be the one that just refuses to lose, the one that simply plays their guts out, gets a few bounces and storms into the playoffs having owned the last month of the season to get there. If I look at the Kings, they seem to be a team that is fully capable of doing just that. I think it is an unbiased view, if that is possible, to look at the improvement of this team since Carter got here and feel that it gives the Kings a significant edge, and also imagine them sustaining that level of play.

Even though the numbers are great in projection, those numbers come out very, very close. To make it into the playoffs, the Kings have to not just equal the other teams in points, but due to tiebreakers the Kings have to finish above them.  Right now, the Kings have the edge. And, we know how hard it is for our competitors to jump teams, much less jump 3 or 4 teams. My head says Kings in 3rd but more likely 7th, according to numbers; my heart says Kings in 3rd by a point over the Stars and Sharks, and two points over the Avalanche and Flames, but the Kings may need to do it by beating San Jose in regulation twice, to end their season. I think they can do that, and I want to believe they can at least get in, but now we know just how brutally difficult it is going to be.


This entry was posted in Math Predicts Future? Ya Right., Really Long Stuff. Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>