News Article: Who tracks the game for us like Dubas does now for Leafs?

[Kellogs]

This is a big issue with adv stats currently imo. The NHL's real time stats data is notoriously bad, and even some of the simpler stuff on the various advanced stats sites isn't consistant.

For example, Toews GA at 5v5 in 2014 on the three maj sites:

Extraskater: 2.4 GA/60, 3.5 GF/60
BTN:2.44 GA/60, 3.69 GF/60
Stats.Hockeyanalysis: 2.45 GA/60, 3.52 GF/60

Part of the problem is they can't even agree on TOI;

BTN has Toews at 15.21 TOI/60
HockeyAnalysis only lists total TOI at 1124:11
ExtraSkater has Toews at 14.60/60 (which is close to 1124:11 divided by the number of games played, 76 but that would have given 14.80/60)

Which is probably why teams have their own people track their stuff internally, and don't use any of the data from websites like BTN or ExtraSkater, that way they remove the rink effect from the data collected.

 

[operasen]

Part of the problem is they can't even agree on TOI;

BTN has Toews at 15.21 TOI/60
HockeyAnalysis only lists total TOI at 1124:11
ExtraSkater has Toews at 14.60/60 (which is close to 1124:11 divided by the number of games played, 76 but that would have given 14.80/60)

Maybe we'll see the idea of adding a chip into the skate somewhere that links to on around the bench area that tracks the TOI - simple to do.

 

[Neiler]

Can these advanced stats be forecasted with any accuracy? If not, I'm not sure I understand the use for them. Anyone can tell you what happened last year in any business.

 

[BonkTastic]

Can these advanced stats be forecasted with any accuracy?

Can "regular" stats be forecasted with any more accuracy than the advanced ones?

 

[Micklebot]

Maybe we'll see the idea of adding a chip into the skate somewhere that links to on around the bench area that tracks the TOI - simple to do.

I think Sportsvu would make that redundant. They are able to track all that using an array of HD cameras.

 

[StefanW]

I think Sportsvu would make that redundant. They are able to track all that using an array of HD cameras.

The biggest issue they are currently having with adapting Sportsvu is tracking the puck, which can easily be hidden under players.

 

[StefanW]

Can these advanced stats be forecasted with any accuracy? If not, I'm not sure I understand the use for them. Anyone can tell you what happened last year in any business.

There is some work done that correlates some of the stats with success. The fancy stats people tend to get pretty good predictive results. In terms of validation there is some disagreement. The best source for info re this type of issue can be found at http://hockeyanalysis.com/ David is pretty meticulous and asks the hard questions.

 

[Micklebot]

The biggest issue they are currently having with adapting Sportsvu is tracking the puck, which can easily be hidden under players.

Foxtrax's glow puck to the rescue!

 

[PierresGabriels]

#HireYost2014



...

 

[Laoghaire]

Can these advanced stats be forecasted with any accuracy? If not, I'm not sure I understand the use for them. Anyone can tell you what happened last year in any business.

They aren't predictable, they are predictors.

 

[Mr Invidious]

They aren't predictable, they are predictors.

Precisely.

There's more to these teams than these numbers, but:

CF% 5v5, score close:
1) LAK - 57.3%
2) CHI - 55.7%
3) BOS - 55.0%
4) NJD - 54.6%
5) SJS - 53.8%
...
26) MTL - 47.2%
27) CGY - 45.8%
28) EDM - 43.4%
29) TOR - 42.1%
30) BUF - 41.1%

FF% 5v5, score close:
1) LAK - 56.7%
2) CHI - 55.2%
3) SJS - 54.6%
4) BOS - 54.1%
5) NJD - 53.9%
...
26) CGY - 47.3%
27) COL - 46.8%
28) EDM - 43.6%
29) TOR - 41.5%
30) BUF - 41.0%

Compare those numbers with how the regular season/playoffs shaped out and there's a clear correlation to be found.

 

[GimmeMyJetpack]

Precisely.

There's more to these teams than these numbers, but:

CF% 5v5, score close:
1) LAK - 57.3%
2) CHI - 55.7%
3) BOS - 55.0%
4) NJD - 54.6%
5) SJS - 53.8%
...
26) MTL - 47.2%
27) CGY - 45.8%
28) EDM - 43.4%
29) TOR - 42.1%
30) BUF - 41.1%

FF% 5v5, score close:
1) LAK - 56.7%
2) CHI - 55.2%
3) SJS - 54.6%
4) BOS - 54.1%
5) NJD - 53.9%
...
26) CGY - 47.3%
27) COL - 46.8%
28) EDM - 43.6%
29) TOR - 41.5%
30) BUF - 41.0%

Compare those numbers with how the regular season/playoffs shaped out and there's a clear correlation to be found.

except NJ and COL

 

[Mr Invidious]

except NJ and COL

There will always be outliers. These stats are not an exact science, and that's the problem, people believe that you can only be in one of two groups:

1) "Advanced stats are nothing but nonsense".

2) "Advanced stats are the holy grail".

Possession numbers predicted COL's downfall going into the playoffs, as did their PDO.

 

[starling]

It's in team's best interest not to admit they track advanced stats even if they do.
Otherwise you'll see lots of low percentage plays where players just walk in and shoot the puck just to rack up some corsi.

 

[Laoghaire]

There will always be outliers. These stats are not an exact science, and that's the problem, people believe that you can only be in one of two groups:

1) "Advanced stats are nothing but nonsense".

2) "Advanced stats are the holy grail".

Possession numbers predicted COL's downfall going into the playoffs, as did their PDO.

Plus prediction and deduction are different. There should be outliers, that's because it isn't so clear cut as if A then B. The logic is: If A you are more likely to have B occur.

 

[Knave]

They're pretty awful predictors looking at those two lists. Hot garbage. And I also don't see a use for team statistics. "Oh look good teams are good and bad teams are bad". Wonderful.

 

[SlapJack]

Possession numbers predicted COL's downfall going into the playoffs, as did their PDO.

It's tough to call a blown 3-2 series lead or a loss in Game 7 OT a solid predictor. Could've gone either way.

 

[StefanW]

They're pretty awful predictors looking at those two lists. Hot garbage. And I also don't see a use for team statistics. "Oh look good teams are good and bad teams are bad". Wonderful.

There is some truth to what you say. However, the value is in establishing teams that are underrated or overrated. The analytics crowd was saying that Toronto would fall hard when they were still firmly in a playoff spot, and they felt that LA was set for a long playoff run. Those points were not obvious to everyone at the time they were made. Ditto that Pittsburg, Colorado, and the Ducks were set for a fall in the playoffs.

 

[Mr Invidious]

They're pretty awful predictors looking at those two lists. Hot garbage. And I also don't see a use for team statistics. "Oh look good teams are good and bad teams are bad". Wonderful.

You do your own research and come to your own conclusion. It's a fact that a positive correlation exists when you know what you're looking for. It's about probability.

But it already sounds like you have your mind made up.

It's tough to call a blown 3-2 series lead or a loss in Game 7 OT a solid predictor. Could've gone either way.

A #1 seed came into the playoffs playing relatively poorly following a big hot streak throughout the season and saw them get bounced in the 1st round. Again, these statistics aren't an exact science but I don't think it's a coincidence.

 

[DrEasy]

Can "regular" stats be forecasted with any more accuracy than the advanced ones?

That's exactly it: one point of looking for advanced stats is to find those that are more predictable and yet are good indicators of future performance.

In baseball we already know a few things: batting average varies a lot from year to year, whereas on base percentage is more stable (and walk percentage and K percentage even more so). While it would be hard to predict future batting average, it's actually easier to predict future OBP, and in fact it's more useful too, as it correlates better with wins anyway.