By the Math of Hossa!

by mf37 on May 28, 2010 2:53 PM EDT

The Sporting News has posted a story that claims Hossa and his agent used an advanced statistical model to predict which teams were elite (likely to earn 100+ points in a season).

"When Marian hired us, one of his objectives consistently was to look for an elite organization to play for," [Agent Rich] Winter said. "We do a statistical analysis of performance to determine (elite teams)." Hossa and Winter weren't about to leave this huge decision to chance -- or feel. They, again, hand-picked Hossa's next team based on an intense study. Winter defines elite as a 100-point team, so he added Hossa's expected statistical performance to the projected performance of each team's current roster to find which teams would be contenders with him. "You can, using statistical models, determine with a high degree of probability, the 100-point teams," Winter said.

The Sporting News article is well worth the read, although I do wonder how much, if anything, can be mined from the limited details of Winter's so-called predictive model...

PensionPlanPuppets.com is a fan community that allows members to post their own thoughts and opinions on the Toronto Maple Leafs and hockey in general. These views and thoughts may not be shared by the editor of PensionPlanPuppets.com.

Tweet Comment 35 comments | 0 recs |

Do you like this story?

Comments

If the Hawks do win, I’ll be more impressed/weirded out that EA called it.

by A Lindros Jaw on May 28, 2010 3:05 PM EDT reply actions

haha i was thinking that the other day. Wierdest thing was, i didn my own NHL 10 simulation and the playoffs shook down almost exactly the same as they are now except chicago beat montreal in the final

There is no "I" in team, but there is an "M" and an "E"

by Matt_Roberts on May 28, 2010 3:28 PM EDT up reply actions

From that article

“Winter, with help from mathematical advisors, has determined exactly how many points a contending team needs from its top six forward group and top four defensemen, and the save percentage required from a goalie to become a 100-point team.

For example, if all thresholds are met from the defensemen and goalies, a team that gets at least 143 goals from its top six forwards will get 100 points. According to Winter, that number has stayed true every year since the lockout. He has calculations like that for every position.”

I really wonder what the requirement is for D-man and Goalies. Anyone good at reverse statistical analysis?

by MapleLeafMole on May 28, 2010 3:21 PM EDT reply actions

All I know is that if they ran the numbers with Toskala in goal their computer would have exploded.

The Guess Who sucked, the Jets were lousy anyway

by Plea From A Cat Named Felix on May 28, 2010 3:23 PM EDT up reply actions 2 recs

Impossible to determine based on the information in the article. Too many variables to back out the threasholds.

Leafs Nation: A drinking team with a hockey problem.

by nhlcheapshot on May 28, 2010 3:25 PM EDT up reply actions

Could you figure out the thresholds (or the vicinity) just by looking at what teams have earned 100 points? The toughest part would be figuring out how they predict things like save and shooting percentage going forward but I think figuring out the thresholds might be doable.

Pension Plan Puppets: A Toronto Maple Leafs blog and a group therapy session.
Like reading thoughts confined to 140 characters? I'm on Twitter too.

by PPP on May 28, 2010 3:28 PM EDT up reply actions

I think you could “guess” based on the actual data points of teams that managed over 100 points, but there’s no way to figure out what sv % or whatever stat was required to fit the requirements.

In a simple fashion, it’s kind of like saying X = A+B+C+D+E+F+G where X is # of predicted points, but they only gave us enough info to determine A and B. The combination of the rest of the variables could be any number of scenarios.

I’ve often thought about doing something similar just from a personal interest perspective for stat correlations, but turning something like that into a model for predictive success is a whole other beast. Especially once you have to incorporate things like player potential and cap management.

Leafs Nation: A drinking team with a hockey problem.

by nhlcheapshot on May 28, 2010 4:32 PM EDT up reply actions

More like X > A + B + C + D + E + F + G, I think

Then again, in their sample, they simply say C+D+E+F+G isn’t big enough to counter A+B. I’m sure there are counterexamples, but not ones they looked at.

Cидни Kросби: Александр Oвечкин, он твой папа теперь
Red Line Station: for Capitals fans who can bear reading something less intelligent than a story at Japers' Rink
(note name change)

by red army line on May 29, 2010 4:33 AM EDT up reply actions

As I said that was more of a simplistic example for the non-math inclined, the actual model i’d assume is similar to a basic linear optimization model:

MAXIMIZE a*X + b*Y + c*Z + d*W +e*U + f*V
subject to:
X > 143 Goals
Y > 50 Goals (defensemen)
Z > 0.910% (sv %)
W < Max Cap hit (W=sum of roster players contracts)
etc. for other unknown thresholds U,V and probably constants a-f, where the result is the # of expected points in the season.

A lot of the determination for the thresholds and constants was likely built on a regression and significance-testing, with probably a touch of trial-and-error until their model achieved back-dated success.

Leafs Nation: A drinking team with a hockey problem.

by nhlcheapshot on May 29, 2010 4:48 AM EDT up reply actions

I suppose you’re right, but hey, whatever works I guess for Hossa. With the variation each year in goals, going by goals isn’t the best way to go (I’m sure Gabe Desjardins would advocate something with Fenwick or Corsi). If Hossa though could find a team with stability in the top forwards and D going forward, it could work. Chicago, if they weren’t so utterly cap screwed, might have had that. Even at the very top, they still do.

Gosh, 50 goals from D is pretty hefty.

by red army line on May 29, 2010 6:03 AM EDT up reply actions

Ya other than the 143, i just made those numbers up for example purposes. The actual figures im sure were determined by regression, trial-and-error or otherwise model-driven.

143 goals (from your top 6 I believe it said in the article), that’s an average of just less than 24 goals/player. Damn. And they say that stat has stayed consistent since the lockout.

Leafs have a lot of work to do on our top 6 if that model is an accurate predictor of reality.

Leafs Nation: A drinking team with a hockey problem.

by nhlcheapshot on May 29, 2010 9:16 AM EDT up reply actions

going with goals

is actually the ideal. Goals for and against are the most closely correlated to win percentage.

Going by shot totals or puck possession would just end up being a correlate to goals for and against… but those two would still be the highest correlates to win percentages.

A win is based on scoring more goals and allowing fewer goals, not having more shots and allowing fewer shots.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 30, 2010 11:05 AM EDT up reply actions

Ya I just made up the 50 goals for example, they don’t give the actual number in the article.

I would think total Goals allowed or some type of GAA * 82 would be used. But in the article it states sv %… so who knows.

Leafs Nation: A drinking team with a hockey problem.

by nhlcheapshot on May 30, 2010 12:04 PM EDT up reply actions

Yeah because SV%

is how goalies are typically compared in terms of ability…. even strength SV% is even better.

If you can figure out how many shots the team will give up roughly (and most NHL teams give up between 28 and 32 shots a game on average), then you’re talking somewhere around 2460 shots on goal (30*82). If the pythagorean ratio for goals for and against works out to a win percentage of around .560, and you assume they only allow 2.60 goals against per game (these are 100 point teams remember), then you know they allowed 213 goals on the season (2.6*82=213.2). That means the goalies would allow 213 goals on 2460 shots as a team, which works out to a team save percentage of .913 (unfortunately I haven’t eliminated empty net goals here but this is all just spit balling anyway).

That’s the ideal situation, and I don’t expect the Leafs goalies to limit the opposition to a goals against average of 2.60 goals against for an entire year. Assume the Leafs are closer to 2.85 goals against per game and you end up seeing them let in 234 goals on the season, which gives them a team save percentage of around 0.905 (again without excluding empty netters which would bump the requisite save percentage higher).

This is why I’m arguing the save percentage needs to be ABOVE .910 assuming a roughly typical number of shots faced.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 30, 2010 2:45 PM EDT up reply actions

why would an empty netter increase the required save percent? At the time an empty netter is scored you were about to lose that game anyway, so all you’re doing is messing up the pythagorean stats by including them. It makes the team look more porous than it really was.

Glory glory Man United, AND the other MU, AAAAnd the Leafs. I think I need a drink now.

by Wan Ihite on May 31, 2010 2:16 AM EDT up reply actions

No you aren't screwing up the pythagorean stats

Since the correlation to win percentage is completely separate from the calculation of save percentage.

I’m saying the relationship between Save Percentage and goals against per game for the team is skewed by empty netters. Empty netters will necessitate an increased save percentage because they aren’t included in that calculation.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 31, 2010 8:20 PM EDT up reply actions

50 goals

from your top 6 D is pretty damn ridiculous. Most teams get around 30 from their top 4, expecting another 20 from the bottom 2 seems absurd.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 30, 2010 11:06 AM EDT up reply actions

NNNNNNEERRRRRRRRRRRD!

Pension Plan Puppets -A Toronto Maple Leafs Blog: Pineapple Free Thanks to Mikhail Grabovski

by Sergei Puckizin on May 29, 2010 12:55 PM EDT up reply actions

That’s what I figured.

by MapleLeafMole on May 28, 2010 3:45 PM EDT up reply actions

I e-mailed Hawerchuk from Behind The Net to see if he wanted to take a crack at it.

Pension Plan Puppets: A Toronto Maple Leafs blog and a group therapy session.
Like reading thoughts confined to 140 characters? I'm on Twitter too.

by PPP on May 28, 2010 3:27 PM EDT up reply actions

It would be pretty interesting to see.

by MapleLeafMole on May 28, 2010 3:45 PM EDT up reply actions

In fact it might just be better to see with advanced stats.

by A Lindros Jaw on May 28, 2010 4:37 PM EDT up reply actions

Was it really hard to predict that Detroit, Pittsburgh, and Chicago were going to be 100pt teams?

by A Lindros Jaw on May 28, 2010 3:39 PM EDT up reply actions

I’d be fun to use and see where the Leafs need to get, vis-a-vis this mathematical model, to be a 100 point team. i.e . our D needs to score x amount of goals. Our goalies need to put up x sv% and so on.
Then we can move onto fantasy land, where Kulemin is a 30 goal, 70 point player

by MapleLeafMole on May 28, 2010 3:47 PM EDT up reply actions

That’s the other thing. It’s interesting to wonder how many teams he looked at. It’s not like there are tonnes of Cup contenders.

Pension Plan Puppets: A Toronto Maple Leafs blog and a group therapy session.
Like reading thoughts confined to 140 characters? I'm on Twitter too.

by PPP on May 30, 2010 11:42 PM EDT up reply actions

It's actually not impossible

The way to determine points would rely on a determination of points resulting from a certain winning percentage, which can be determined from a pythagorean model using goals for and goals against as a determinant.

If you assume a 100 point team since the lockout usually ends up with around 46 wins (as an average of 9 different seasons that garnered within 2 points of 100 it worked as a win total) that means the team needs a winning percentage of around 0.561. The average 100 point team this past season allowed roughly 2.60 goals against per game. Thus to achieve a winning percentage using the pythagenpuck method (described by Alan Ryder) they would need to score 2.865 goals per game. That translates to 235 goals for.

If the top 6 forwards have to score 143 goals, that’s almost exactly 60% of the team’s offense. The top 4 D probably need to account for another 30 goals or so (12.7% of the offense). That leaves 65 goals to be produced by the bottom 6 forwards and the bottom 2 D men (plus any call ups).

The goalies are likely to required to maintain approximately a .910 save percentage to end up with a .560 winning percentage based on the numbers I’m using above.

Obviously if the goaltending is weaker, then the goals for has to go up, which means the team needs to score more to earn 100 points, but frankly I don’t think a team makes it to the 100 point plateau if their goaltending is playing with a combined save percentage that is far below .910.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 28, 2010 5:28 PM EDT up reply actions

So if you look at the Leafs

that means Gustavsson and Giguere need to play with a .910 save percentage over the course of the year (not impossible), and the Leafs top 4 D need to produce 30 goals. Last year Phaneuf, Beauchemin, Kaberle, and Schenn would have combined for 29 goals, so again, not far off.

The issue is the forwards. The Leafs bottom 6 forward group produced roughly 57 goals. Gunnarsson, Finger, and Exelby contributed 6 goals. So the lower end contributed 63 goals (the target was 65 remember).

Up front the Leafs got only 119 goals from Kessel, Bozak, Kulemin, Ponikarovsky, Stajan, Hagman, and Grabovski (the players I consider to have been top 6 for the Leafs for most of the year). They fell 24 goals short. Since Kessel produced 30, that means you either need to spread out those 24 amongst the other 6 forwards, or else his production needs to go up a lot.

Given Kessel’s talent, I wouldn’t be shocked if he scores 40 next year. Bozak only scored 8, Grabovski only had 10, and Kulemin only potted 16. If you put Bozak up around 15, and Grabovski around 15, and Kulemin around 20, that’s an increase of 16 goals… If Kessel scores 40 that’s another 10 which is an increase of 26 goals overall. Losing Ponikarovsky, Stajan, and Hagman means the Leafs need to replace the 55 goals they scored though.

The incumbent top 6 forwards remaining from last year would thus be Kessel, Bozak, Kulemin, and Grabovski, and if things go reasonably, they can produce 90 of the required 143 goals. That means the guys moving up into the top 6 (whoever they are) need to produce another 53 goals for the team to approach 100 point status. Something tells me the Leafs aren’t going to accomplish that anytime soon.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 28, 2010 5:39 PM EDT up reply actions

For the Leafs to reach the 40 win plateau (approximately playoff range)

That’s a .488 win percentage. If the Leafs can allow 2.85 GA per game (234 goals against), then they only need to score 2.8 goals for per game (230 goals for). That means we really only need 130-135 goals from the top 6 forwards. Which still means another 40-45 goals coming from somewhere.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 28, 2010 5:44 PM EDT up reply actions

So basically every other game someone has to score one extra goal. Interesting.

Glory glory Man United, AND the other MU, AAAAnd the Leafs. I think I need a drink now.

by Wan Ihite on May 30, 2010 1:53 PM EDT up reply actions

The way I look at it is

We need another 2 guys scoring 20-25 goals. If Stålberg wants to be one of them, I’m fine with that. I’m guessing Burke will either try to add another one via free-agency or trade… so here’s looking at Versteeg/Sharp/Bolland/Brouwer in the Chicago situation, or perhaps Jamie Benn from Dallas… or David Backes from St. Louis, or SOMEBODY… but yeah… he’s going to need to add scoring somehow if he thinks we’re going to compete for the playoffs next year.

The idea that Kadri will jump in and score 20-25 is dreaming in technicolour for those of you who are thinking that. Eric Staal had 11 goals and 31 points as a 19 year old rookie, Marc Savard didn’t hit the 20 goal plateau until he was 22 years old and was in his 3rd NHL season. Even our current offensive leader Kessel produced only 11 goals and 29 points as a 19 year old. The most we should expect from Kadri is something around 10 goals, and that would be solid production. If he tops that, then that’s awesome, but it’s not likely to happen.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 30, 2010 2:37 PM EDT up reply actions

The NHL equivalency figure for the OHL is 0.29. Kadri had 1.66 ppg last year, so that’s the equivalent of 0.48 ppg in the NHL. 36% of Kadri’s points come from goals, so over 82 games a typical OHLer with Kadri’s stats could be expected to post 14-25-39 over 82 games.

You have to adjust that a little given that Kadri was 19, not 18 or 17, and given that if he makes the team he’ll get plenty of PP time and 2nd line minutes (while many guys do time on the 3rd line to ease in).

In short, if healthy and playing 2nd line minutes, I think it’s unduly pessimistic to say 10 goals is the “most we should expect”. Unless there’s something about his game that will translate poorly to the NHL, 15-25-40 is possible.

On that last point I know people who have watched him say “he’ll get killed it he tries that junk in the NHL”, but isn’t that true of a great many players who ultimately adjust and become effective NHLers?

by The '67 Sound on May 31, 2010 1:54 PM EDT up reply actions

I think

it’s a huge over-estimate to assume he’s going to play a full 82 games as a rookie in the NHL.

That’s largely why the numbers tend to be lower. I’d assume something closer to 65 or 70.

Kessel played 70 as a rookie, Legwand played 71, Savard only played 28, Nathan Horton played 55. I don’t think Kadri will be particularly durable as a rookie, and I think he’s likely to go back and forth between the NHL and AHL as he learns the defensive aspects of the game.

In a 65 game season your numbers would have him at 31 points with 11 goals. Not quite as far off as we were before now are we.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 31, 2010 8:26 PM EDT up reply actions

I see what you’re saying. Yes, I was just projecting /82 rates, which obviously isn’t realistic.

I guess my only disagreement is I’d say 11-20-31 is a realistic projection rather than a best case scenario.

by The '67 Sound on Jun 1, 2010 11:51 AM EDT up reply actions

I think the best case scenario

is so rare, that I’m not really prepared to consider it… if we get the season Matt Duchene had, then great… but I’m certainly not anticipating that.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on Jun 2, 2010 9:23 PM EDT up reply actions

Either way

as many on here have already said, the model is an approximation that just combines goals for and goals against, which means with the multiple variables you’re doing a fairly complicated regression, and I’m not sure how detailed you can get unless you run some pretty involved predictions on player performance going into the future.

I think the 143 goal thing is what guarantees results. The fact is, if you want to be SURE you’re going to a 100 point team, then you’d pick a team that’s a virtual lock to far exceed the minimums. When you look at Chicago’s output from their top 6, they scored 146… I’m not 100% certain that Hossa expected that output. In fact, the large part of what got Chicago to where they ended up would be their team defense.

Vancouver’s top 6 scored 173 goals this year. Nashville’s top 6 only scored 115 goals. Obviously the 143 number isn’t set in stone either way.

"Success is the ability to go from one failure to another with no loss of enthusiasm."

- Sir Winston Churchill

I'm pretty sure he's talking about the Leafs.

by Steve Burtch on May 28, 2010 5:54 PM EDT reply actions