Flames & Team Metrics (Analytics)

[Resolute 14]

This is just an unnecessarily overcomplicated way of saying the team has had some good games and some bad games. Beyond that, I can't really add much since "bubble team" is undefined. How is that calculated? The 16th place team today? Or the 16th place team at each date in the schedule? Or what?

Also, correct me if I am wrong, but is your second chart inverted? If it is measuring CA, SCA, HDSCA, etc, then I would expect our worst games to be the highest numbers above the line, not below.

[Oil Stain]

Natural stat trick further breaks down corsi into leading, trailing, and tied.

The Flames rank there:

8th best when trailing.
12th when tied.
20th when leading.

3 teams have a corsi share greater than 50% when leading.
27 teams are over 50% when trailing.

The Flames lead the league in 5 on 5 minutes spent trailing.

Because the Flames are good at playing from behind, it has really supercharged their overall numbers.

[Bingo]

Originally Posted by Resolute 14

This is just an unnecessarily overcomplicated way of saying the team has had some good games and some bad games. Beyond that, I can't really add much since "bubble team" is undefined. How is that calculated? The 16th place team today? Or the 16th place team at each date in the schedule? Or what?

Also, correct me if I am wrong, but is your second chart inverted? If it is measuring CA, SCA, HDSCA, etc, then I would expect our worst games to be the highest numbers above the line, not below.

You can be so sweet ... thanks

Since 16 teams make the playoffs, each stat is compared to the 16th best team in each stat category

The numbers invert, the chart doesn't. Always a + to be above the line.

[delayedreflex]

I think someone in another thread suggested that maybe it might be more meaningful if the data was normalized, and perhaps add horizontal lines for the averages for the best and worst teams in the league as well. Is +10 Corsi events/60 min pretty good, or really good? How about +10 SC/60? As is, about the only thing you can take away from the graph is that the Flames have generally above-average shot metrics for most of their games - but otherwise gives no information on relative performance compared to the best and worst teams.

Like right now, looking at the bars, it looks like the Flames should be well below average for total shot generation (ie. the gray bars are mostly below the line) - yet the Flames are supposedly 4th in the league for CF/60, according to the number in the parenthesis? One of those must be wrong, or I am misunderstanding something on the graph.

[Resolute 14]

Originally Posted by Bingo

You can be so sweet ... thanks

Since 16 teams make the playoffs, each stat is compared to the 16th best team in each stat category

The numbers invert, the chart doesn't. Always a + to be above the line.

Ok, then what does that tell us? It doesn't measure our quality in any absolute terms, as if we played like the 29th best team in the league, but our opponent played like the 31st, then relatively, we could well still end up looking better than than the 16th place team for that sample size of one game. Also, the median team changes from day to day. So what represents 16th place today could be considerably higher or lower than what represented 16th place last week or last month. That will skew the data and make any given game of ours look better or worse than it was. Your baseline just doesn't make sense to me in the context you are using it.

I dunno. I think that if you want to display the data in this format, with an eye toward showing how good or bad the Flames are playing relative to a middle of the pack team, you'd be better off using rolling averages for both what represents the median for those stats and the Flames' performance.

[Bingo]

All these stats add to zero as a shot attempt of any degree of danger is both a shot attempt for for one team and against for the other.

So the median team in each case is zero.

This is just looking at the 16th ranked team which in a 31 team league is easier than team 15.5 and also a good pick as 16 teams make the playoffs.


Sent from my iPhone using Tapatalk

[Resolute 14]

Yeah, I get what you are trying to do. I just don't think this format works. There are a multitude of reasons why a single game may appear to be better or worse than the 16-game average of your median team: score effects (i.e.: the Pens debacle), quality of opponent, etc. And 16th place is a moving target.

I'm also trying to figure out how we can appear to be so bad at CF according to that table while being one of the top teams in the league in CF, CF/60 and CF%. That table has us below the median team in 12 of 16 games and above in only three.

I'm not saying the idea is bad. I am suggesting that the idea needs to be refined.

[Bingo]

Originally Posted by Resolute 14

Yeah, I get what you are trying to do. I just don't think this format works. There are a multitude of reasons why a single game may appear to be better or worse than the 16-game average of your median team: score effects (i.e.: the Pens debacle), quality of opponent, etc. And 16th place is a moving target.

I'm also trying to figure out how we can appear to be so bad at CF according to that table while being one of the top teams in the league in CF, CF/60 and CF%. That table has us below the median team in 12 of 16 games and above in only three.

I'm not saying the idea is bad. I am suggesting that the idea needs to be refined.

Are you mixing up CF and CF%?

The look of the two graphs suggests the Flames are super elite in CA (or prevention of shot events), not CF the generation of shot events. So the CF vs the 16th place team is hit and miss, while the CA line in the second graph is always above the bubble team and dominant.

If your CA is always above the bubble team your CF% is going to be pretty damn good, but that doesn't mean your CF necessarily is.

[Kovaz]

Originally Posted by Bingo

Are you mixing up CF and CF%?

The look of the two graphs suggests the Flames are super elite in CA (or prevention of shot events), not CF the generation of shot events. So the CF vs the 16th place team is hit and miss, while the CA line in the second graph is always above the bubble team and dominant.

If your CA is always above the bubble team your CF% is going to be pretty damn good, but that doesn't mean your CF necessarily is.

We're 4th in the league at CF/60, but looking at the graphs I'd guess we're well below average. For example, the recent Ducks game we were at 70.88 CF/60, and that shows up as just a hair below the 16th best team in the graph. So that suggests the median team is at ~71 CF/60. However San Jose is 2nd in the entire league at 70.5 CF/60, so something's definitely off with those numbers.

[Bingo]

Originally Posted by Kovaz

We're 4th in the league at CF/60, but looking at the graphs I'd guess we're well below average. For example, the recent Ducks game we were at 70.88 CF/60, and that shows up as just a hair below the 16th best team in the graph. So that suggests the median team is at ~71 CF/60. However San Jose is 2nd in the entire league at 70.5 CF/60, so something's definitely off with those numbers.

Fair enough ... I'll take a look

Sorry guys

And you're right 70+ games are huge, they have to be way up

[Bingo]

Fixed and updated

[tkflames]

Hey Bingo- I really appreciate the curves. I believe over time they do put a quantitative value against the results we are seeing on the ice.

I hate to request more complexity to this graph, but would you be able to put a similar band as the 16th team for either Tampa or Nashville onto the second graph? It is clear the the Flames are playing above a 16th place team, but context for dropoff against the top teams would be really interesting for scaling.

[csnarpy]

Originally Posted by Bingo

Fixed and updated

[FanIn80]

If above is good and below is bad... then these charts indicate we destroyed Vancouver in GM1 (yet we lost 7-5), and dominated Pittsburgh in GM10 (a 9-1 blowout against!).

It also says we got our asses kicked by Nashville and NYR, yet we won both of those games 3-0 and 4-1, respectively.

This is why I have such a huge problem with "advanced stats." We lost our best game of the season (according to the bars and lines) 7-5, and one of our best games 9-1, yet we won our two absolute worst games of the season (again, bars and lines) by a 7-1 combined goal differential.

[Snuffleupagus]

It's said it takes 20-25 games for a team to learn a new system from a new coach.

Excited to see the final product

[Bingo]

Originally Posted by FanIn80

If above is good and below is bad... then these charts indicate we destroyed Vancouver in GM1 (yet we lost 7-5), and dominated Pittsburgh in GM10 (a 9-1 blowout against!).

It also says we got our asses kicked by Nashville and NYR, yet we won both of those games 3-0 and 4-1, respectively.

This is why I have such a huge problem with "advanced stats." We lost our best game of the season (according to the bars and lines) 7-5, and one of our best games 9-1, yet we won our two absolute worst games of the season (again, bars and lines) by a 7-1 combined goal differential.

You shouldn't.

Calgary trailed the whole Vancouver game so they ran up the metrics chasing a game.

The Penguins led by a huge margin and pretty much mailed it in after the first few minutes in the second.

The Ranger game was exactly as the stats read above ... the Flames were dominated but saved by Rittich.

The Nashville win had a lot of shots put towards Smith, remember the Forsberg quote after two periods about "sure lots of shots, but I think I could have saved them all"

[DeluxeMoustache]

This leads to a question about what the whole value of the statistical analysis is, though

The high level argument (loosely framed, please don’t nitpick) is that shots are a proxy for possession / control, etc. People throw around save percentages like they apply more or less uniformly, and implicitly in the long run, proportionally share of possession translates to share of goals and share of wins.

Many of us have studied statistics to some extent and know that a good fit will have enough data points so that results can be expressed as accurate, say, within a certain margin or percentage, 19 times out of 20.

There are a lot of shots, sure, but few game results. Look at the Flames season. Apply analytics and compare to game results. Translating shot counts to wins has so many other contributing factors, and ultimately outliers in terms of results, that getting to a level of statistic significance is impossible.

If you have, say, 18 games and 4 outliers, the model is likely not good enough.

[Bingo]

Originally Posted by DeluxeMoustache

This leads to a question about what the whole value of the statistical analysis is, though

The high level argument (loosely framed, please don’t nitpick) is that shots are a proxy for possession / control, etc. People throw around save percentages like they apply more or less uniformly, and implicitly in the long run, proportionally share of possession translates to share of goals and share of wins.

Many of us have studied statistics to some extent and know that a good fit will have enough data points so that results can be expressed as accurate, say, within a certain margin or percentage, 19 times out of 20.

There are a lot of shots, sure, but few game results. Look at the Flames season. Apply analytics and compare to game results. Translating shot counts to wins has so many other contributing factors, and ultimately outliers in terms of results, that getting to a level of statistic significance is impossible.

If you have, say, 18 games and 4 outliers, the model is likely not good enough.

I really think this is another example of over thinking things.

First off nobody is calling anything a model, you won't see that coming out of my mouth or through my keyboard.

It's counting. Counting isn't a model. Summarizing a bunch of counting stats isn't a model, its a summary.

Score effects explains a lot of the 4/22 examples that don't fit. The Flames got out numbered in Nashville and Manhattan because they led the whole way. They out numbered teams that beat them in Vancouver and the Penguins because they trailed the whole way.

But nothing can be 100% predictive about any counting stats, including simple ones like shots on goal that we used to use. You can lose a game while out shooting the opponent, and you can win a game when getting out shot. But I'm willing to guess over a large enough sample size it's better to get more shots than your opponent if you want to win.

Similarly it's a pretty sound assumption that you probably win more games than you lose if you get more scoring chances than the opponent.

I'm just summarizing that ... it's not a model.

[DeluxeMoustache]

I agree, I know it’s not a model. Let’s call it a concept.

I actually like the concept to a degree. But the closer what you are analyzing is to the data, the more useful.

Look at all of the Bennett conversation. It is very clear that this year he is having an impact on the ice. He is generating scoring chances, has been snakebit, and not been finishing.

Generally the concept works for individuals in helping analyze their play relative to peers

However translating shots and things you can count to something that can be predictive of actual team results just has too many levels in between, and factors which can’t be isolated and factored in appropriately. So it isn’t as you say a model.

I have read enough of your commentary (and thanks for it) know that you know this, and know that you are clear about what you do and don’t try to represent using the data.

I think the stats are interesting. I also know they are simple and not predictive. I think we see things the same way.

[awildermode]

D’Arcy McGraph