Are there better metrics for how a team is playing than SCF% and HDCF%? What do you use?

[Sensators]

I'm trying to get a view into how teams are playing looking ahead to the playoffs, and I would like to find a metric to help me do this - but tampa is not that good at SCF/HDCF and we all know tampa is the best team in the league. On top of this washington was AWFUL at it in the second half of last year and went on to win the stanley cup. In fact washingtons 3rd line was just awful all last year and then amazing in the playoffs.

Are there other measures I should use to judge how a team is playing? It's pretty clear to me that these two are flawed. Although Vasilevsky probably does make Tampa look stronger than they are - whenever I watch them they are just everywhere in the offensive zone.

Click to read more...

 

[Michael Farkas]

This is not intended to be:
1) a hijack
2) dumb

Though, it may be #2.

As a tag a long, has there been any correlation between "extreme" PDO values and playoff, namely - first round - success? i.e. do teams with a regular season PDO <990 have a winning record in the first round?

 

[SniperHF]

Are there other measures I should use to judge how a team is playing? It's pretty clear to me that these two are flawed. Although Vasilevsky probably does make Tampa look stronger than they are - whenever I watch them they are just everywhere in the offensive zone.

These kinds of stats aren't really my bag, but might it be that these ones you are looking at aren't particularly amazing for the Lightning because the Lightning happen to have one of the best combinations of special teams(1st PP, 3rd PK) in the league and most of those numbers are for 5 on 5? So if you're looking at a 5 on 5 number you're basically not looking at a key advantage for the Lightning.


Also perhaps another thought, but naturalstattrick also tracks "medium" danger scoring and the Lightning are killing it there. They have 22 more medium danger goals than the 2nd place team at all strengths. So this is just a total butt pull theory, but the idea being that a "medium" danger chance for a team like the lightning is more likely to go in due to the skill of the players and or scheme the team runs.

 

[Sensators]

These kinds of stats aren't really my bag, but might it be that these ones you are looking at aren't particularly amazing for the Lightning because the Lightning happen to have one of the best combinations of special teams(1st PP, 3rd PK) in the league and most of those numbers are for 5 on 5? So if you're looking at a 5 on 5 number you're basically not looking at a key advantage for the Lightning.


Also perhaps another thought, but naturalstattrick also tracks "medium" danger scoring and the Lightning are killing it there. They have 22 more medium danger goals than the 2nd place team at all strengths. So this is just a total butt pull theory, but the idea being that a "medium" danger chance for a team like the lightning is more likely to go in due to the skill of the players and or scheme the team runs.

Even when I used all strengths the lightning were still not amazing, I'll take a look into these medium chances, thanks. I'm not sure what makes something a medium or high scoring chance besides where the shot it taken, would be nice to know that.

 

[SniperHF]

Even when I used all strengths the lightning were still not amazing, I'll take a look into these medium chances, thanks. I'm not sure what makes something a medium or high scoring chance besides where the shot it taken, would be nice to know that.

Demystifying Hockey Analytics, Part 2: Stats Sites
Shot Quality And Expected Goals: Part 1.5 | Corsica

To tl;dr it to the best my limited ability, I would say what they are saying is medium danger shots generally go in 9% of time as an average.

The lightning are currently shooting a 13.42% of medium danger shots which is top of the league.

I dunno if NSS counts it the same way Corisca does or not though.

This could all be meaningless or noise or both, hey good teams score goals right
But if you wanted to know a stat the Lightning were at the top of this is one of them. It's not just scoring but also raw numbers of medium danger shots for. So they both generate and score on a lot of medium danger chances.

I think the for and against % are interesting but sometimes a high raw number is worth noting.

 

[Ralph Spoilsport]

So, what determines shot quality? The biggest factor is shot location.

Is this true? I don't think so. Not by a long shot (hehehe).

Seems to me the biggest factor is the position of the goalie, something these analyses of "chances" don't consider. The odds of scoring on a shot from 20 feet out in the slot will be different against a goalie playing deep in his crease compared to a goalie playing at the top of his crease. What matters is how much net there is to shoot at.

Regardless of where the shot comes from, if the goalie is out of position its a high danger chance. If he's square to the puck it's a low danger chance.

How do you measure this? Beats me. Doesn't make "shot location" any more meaningful.

Just my opinion. I will admit my skepticism keeps me from looking too deeply into these heat map studies. Seems a waste of time, but maybe others see value in it, so good for them.

 

[Pookie]

Not sure this is the place for it but coming with an open mind. Help me with...

“Score Effects”

If we are proposing that Possession wins games. And we try to measure it, we find that slightly more games tend to be won while being outshot.

As we know, teams with a lead will play more defensive to protect the single point or the W. in short, they aren’t playing a “possession game”.

So their shot proxy totals drop.

Fine. Let’s just measure that. Maybe a combination of high danger shots and solid defensive play wins games. Let’s go where the raw data takes us.

Adding score effects seems intellectually dishonest in evaluating the theory.

——

1. Theory: possession wins games

2. When winning, teams change tactics and don’t play a possession game.

3. We weight any “Corsi event” arising from this tactic change as more important.

4. We adjust the final totals.

5. And conclude possession wins games even when they played the Trap or Collapsed the Net.


How is this even remotely logical?

 

[Filthy Dangles]

Not sure this is the place for it but coming with an open mind. Help me with...

“Score Effects”

If we are proposing that Possession wins games. And we try to measure it, we find that slightly more games tend to be won while being outshot.

As we know, teams with a lead will play more defensive to protect the single point or the W. in short, they aren’t playing a “possession game”.

So their shot proxy totals drop.

Fine. Let’s just measure that. Maybe a combination of high danger shots and solid defensive play wins games. Let’s go where the raw data takes us.

Adding score effects seems intellectually dishonest in evaluating the theory.

——

1. Theory: possession wins games

2. When winning, teams change tactics and don’t play a possession game.

3. We weight any “Corsi event” arising from this tactic change as more important.

4. We adjust the final totals.

5. And conclude possession wins games even when they played the Trap or Collapsed the Net.


How is this even remotely logical?

This might be anecdotal but it sounds like you’re oversimplifying and overstating it.

Teams don’t stop the playing when they take a lead, especially a one goal lead and especially when it’s not late in the game.

When teams are up multiple goals later in a game they might just get the puck deep more than they otherwise would and not risk a turnover. Maybe the trailing team gets a few more perimeter shot attempts they otherwise wouldn’t take.

Also sometimes a team has a big lead and the ice is tilted completely in their favor and they pour it on.

I think you think this effect is a lot bigger than it really is

———

That’s not even to mention the whole debate about the value of Possession vs Shots which are two different things even if the latter is supposed to measure the former.

 

[supsens]

This might be anecdotal but it sounds like you’re oversimplifying and overstating it.

Teams don’t stop the playing when they take a lead, especially a one goal lead and especially when it’s not late in the game.

When teams are up multiple goals later in a game they might just get the puck deep more than they otherwise would and not risk a turnover. Maybe the trailing team gets a few more perimeter shot attempts they otherwise wouldn’t take.

Also sometimes a team has a big lead and the ice is tilted completely in their favor and they pour it on.

I think you think this effect is a lot bigger than it really is

———

That’s not even to mention the whole debate about the value of Possession vs Shots which are two different things even if the latter is supposed to measure the former.

It’s not that teams stop playing it’s that the losing team starts to push harder for a goal and forgets about being defensively sound or in position. Everyone starts taking chances and rushing in more often then not they get a few meh chances and give up a 2 on 1 or a breakaway. But unless it’s a big lead or late in the game most teams keep it cool and play the right way

 

[SotasicA]

I'm trying to get a view into how teams are playing looking ahead to the playoffs, and I would like to find a metric to help me do this - but tampa is not that good at SCF/HDCF and we all know tampa is the best team in the league

I'd say SCF/HDCF did a good job predicting the playoffs...

 

[supsens]

Is this true? I don't think so. Not by a long shot (hehehe).

Seems to me the biggest factor is the position of the goalie, something these analyses of "chances" don't consider. The odds of scoring on a shot from 20 feet out in the slot will be different against a goalie playing deep in his crease compared to a goalie playing at the top of his crease. What matters is how much net there is to shoot at.

Regardless of where the shot comes from, if the goalie is out of position its a high danger chance. If he's square to the puck it's a low danger chance.

How do you measure this? Beats me. Doesn't make "shot location" any more meaningful.

Just my opinion. I will admit my skepticism keeps me from looking too deeply into these heat map studies. Seems a waste of time, but maybe others see value in it, so good for them.

, when most goalies save% are within .03 of each other using R2 as validation seems somewhat iffy at best
and when a low scoring chance is what causes the goalie to be down and out and a rebound in the high danger scoring area goes in how does all the credit go to “high danger” same thing with a tip in.
It’s non predictable non repeatable score keeping after the game and a ton of random noise
A chart saying If you bang in rebounds and make nice tip ins you will score more than shots from way out. Ooooo that’s ground breaking.
The chart says the better teams do this.....lol yep so does the win/loss record.

 

[mynamejeff420]

Is this true? I don't think so. Not by a long shot (hehehe).

Seems to me the biggest factor is the position of the goalie, something these analyses of "chances" don't consider. The odds of scoring on a shot from 20 feet out in the slot will be different against a goalie playing deep in his crease compared to a goalie playing at the top of his crease. What matters is how much net there is to shoot at.

Regardless of where the shot comes from, if the goalie is out of position its a high danger chance. If he's square to the puck it's a low danger chance.

How do you measure this? Beats me. Doesn't make "shot location" any more meaningful.

Just my opinion. I will admit my skepticism keeps me from looking too deeply into these heat map studies. Seems a waste of time, but maybe others see value in it, so good for them.

Alex Novet had a great series of posts on Hockey-Graphs that added pre-shot movement and screens and stuff to xG models and found that shot distance and angle were still by far the most important thing when it comes to the likelihood of a shot becoming a goal.

Expected Goals Model with Pre-Shot Movement, Part 1: The Model

I know that doesn't address your issue about goalie positioning, but in terms of evaluating a goalies positioning, that would be reflected in their GSAx numbers (Goals Saved Above Expected), as a goalie with good positioning would likely save more goals than expected, and a goalie with poor positioning would save fewer goals than expected. If we're talking pure predictiveness then I guess goalie positioning would be important, but if we're talking player evaluation it shouldn't really be included in an xG model.

 

[JaegerDice]

There is no single stat that gives you a complete picture. The beauty of advanced stats is how easily they can be parsed and sliced by context. There's no one magic number to look at and say 'yes, that's the one, that will predict the winner'.

Personally, I look at CF% and xGF% for team stats, along with PDO.

For players, the first thing I look at is their sh%. After that, CF% and rel CF%, then xGF% and rel xGF%. Finally, ZSR and QoC numbers.