May 3, 2008

Philadelphia @ Montreal

Game 1:42.392.533.983.0
Game 2:
Game 3:22.981.631.693.2
Game 4:
Game 5:44.286.863.790.7
Series [1-4]1414.789.62017.090.7

Montreal gets another team they dominated in the regular season. Montreal should be prepared this time, however Philadelphia is a reasonable opponent and as you can see in the table below these teams are closely matched. Interestingly the two teams that went to game #7 will face off.

Even Strength
Power Play

The winner column displays the dominant team in that category. The more pictures of the team's logo the more dominant the team is in that category

All the non-percentage numbers are scoring rates. For example on the first row, the Montreal Canadiens have an even strength scoring rate of 2.48 goal for per hour. [GF = goals for, GA = goals against Exx = expected xx, SV% = shot quality neutral save percentage].

In the power play section in order to calculate the expected scoring rates I multiplied the goals for rate of one team and the goals against rate of the other and divide by the league average in order to get the expected rate for these two teams combined. So for example, the Montreal Canadiens have a power play goals for rate of 9.22 and the Philadelphia Flyers have a penalty killing goals against rate of 6.26. So 9.22*6.26/6.5=8.88 [league average is 6.5].

Outperforming expected goals for is a sign of a lucky team. Outperforming expected goals against is a sign of either a good goaltender or luck as well.

Each category listed has a different importance to winning, so be careful how you read these. Being able to score short-handed isn't going to win a lot of hockey games.


Topham said...

Did outperforming expected goals for at even strength and PP have anything to do with scoring short-handed?

JavaGeek said...

short answer: no

performing better than expected goals is a sign of luck and luck doesn't predict anything.

Jeff J said...

Is it possible that you've got the game 2 EGF reversed for MTL/PHI? I was surprised to see Philly with the advantage in g1 but chalked it up to my biased observations. But in g2 I was downright shocked.

Mtl had 27 EV+SH shots from an average of 38'. Flyers had 18 from an average of 40'. On the PP Mtl had 9 shots from an average of 25' to Philly's 5 from an average of 33'. The rebounds and shot types don't appear to be terribly one-sided either way. Price's SQN Sv% being higher than Biron's seems out of whack too.

If the numbers are right, what is it the Flyers are doing that is adding up to such high quality chances?

JavaGeek said...

In short two reason
I considers shots by the winning team (1 or 2 goals up) to have a better chance of going in. (I bet some people would disagree with this).
A better way to say the above is: it's harder to play goalie when your team is losing.
Or you can say:
A team down by 1 or 2 goals will give up better chances against in order to create offense. [Breakaways etc.]

Lupul had two wrist shots within 8-10' he didn't score, but the expected goals for those two shots were 1.2. (8:18 in 3rd)

Jeff J said...

"A team down by 1 or 2 goals will give up better chances against in order to create offense."

Yeah, I guess that makes sense.

But if you want to use the EGF numbers for predictive purposes, doesn't this somewhat make it a self-fulfilling prophecy? Forgive my poor phrasing - not sure what the proper stats terminology would be.

If a team has a 1 or 2 goal lead, wouldn't that usually imply that they are a better team? In cases where a worse team takes a lead might inflate their EGF number on subsequent chances. Put another way, wouldn't including the lead/behind game state compromise the independance of the EGF/EGA from the true GF/GA?

I suppose the same argument could be made about PP chances - that some PPs are better than others. But a good team will have way more leads than a poor team (thus inflating the "lead bonus" in calculating the EGF), while a good PP will still have roughly the same number of chances as a poor PP.

And one more thing: do the distance numbers you use take the angle into account? The pbp only list the distances while some pages (cbs) provide an approximate location on the ice.