TSN's Updated Playoff Predictions for Canadian Teams

[SID.IS.SID.ME.IS.ME]

On 2016-12-03 at 10:54 AM, WiDeN said:

How in any way could your prediction be more wrong?  

If you understand what PDO predicts, then you understand that their results should improve from here.  

Unless they stay this decimated by injury their results should improve.

 

They are only out of a playoff spot by 3 points thanks to a weak division.

 

I don't see what you are arguing over.  Oldnews, although combative, has the statistical interpretation bang on.

 

The bolded statement has already proven to be wrong, so that doesn't set the rest of your predictions up as something I'd bet on.

In defense of @baumerman77, I think all he's saying is that he predicted that the Canucks would play their best hockey in the first 1/3 of the season, and then a depleted roster (due to injuries and potential TDL selloffs) would lead to them falling later in the standings.

 

The current PDO issue suggests the Canucks might be a bit better than their results currently indicate. Regression, in a vacuum, should sort that out over a larger sample of games.

 

So maybe the Canucks are a bit worse off right now than he predicted. And maybe they'll improve (due to regression) over the next stretch. But he believes they'll still crash by season's end.

 

Hard to say whether or not that will happen. But it's not an altogether unreasonable prediction. There are depth issues on this team. The young kids currently playing well may not be able to sustain that over an entire season. So if more injuries hit and we end up selling assets, we might be headed for a cliff.

 

So if he's proven correct and the roster does become seriously depleted, then his projection actually carries weight. PDO can regress to the mean, but if the roster is severely weakened, we could see shots differentials plummet, even with shooting percentage and save percentage rising.

 

The net effect of that could be an even worse win percentage in the later season (than what we've seen to-date).

 

And it's also possible that regression in PDO won't actually create enough improvement in actual winning percentage for the team to climb all that many spots in the standings. We could just keep pace with the crucial teams ahead of us, but fail to catch them.

 

I don't necessarily agree with the prediction he's making (mostly because I'm loath to make predictions--too many variables and unknowns). But I can understand where he's coming from. And I don't think it's from a place of not understanding PDO or regression.

 

It's just that @baumerman77 (at least as I've read what he's posted here) doesn't believe those factors will be sufficient to undo the problems he's predicting will come later in the season (due to roster depth issues and how this team will cope with injuries and possibly selling key components at the TDL).

[baumerman77]

5 hours ago, SID.IS.SID.ME.IS.ME said:

In defense of @baumerman77, I think all he's saying is that he predicted that the Canucks would play their best hockey in the first 1/3 of the season, and then a depleted roster (due to injuries and potential TDL selloffs) would lead to them falling later in the standings.

 

The current PDO issue suggests the Canucks might be a bit better than their results currently indicate. Regression, in a vacuum, should sort that out over a larger sample of games.

 

So maybe the Canucks are a bit worse off right now than he predicted. And maybe they'll improve (due to regression) over the next stretch. But he believes they'll still crash by season's end.

 

Hard to say whether or not that will happen. But it's not an altogether unreasonable prediction. There are depth issues on this team. The young kids currently playing well may not be able to sustain that over an entire season. So if more injuries hit and we end up selling assets, we might be headed for a cliff.

 

So if he's proven correct and the roster does become seriously depleted, then his projection actually carries weight. PDO can regress to the mean, but if the roster is severely weakened, we could see shots differentials plummet, even with shooting percentage and save percentage rising.

 

The net effect of that could be an even worse win percentage in the later season (than what we've seen to-date).

 

And it's also possible that regression in PDO won't actually create enough improvement in actual winning percentage for the team to climb all that many spots in the standings. We could just keep pace with the crucial teams ahead of us, but fail to catch them.

 

I don't necessarily agree with the prediction he's making (mostly because I'm loath to make predictions--too many variables and unknowns). But I can understand where he's coming from. And I don't think it's from a place of not understanding PDO or regression.

 

It's just that @baumerman77 (at least as I've read what he's posted here) doesn't believe those factors will be sufficient to undo the problems he's predicting will come later in the season (due to roster depth issues and how this team will cope with injuries and possibly selling key components at the TDL).

That's pretty much what I am saying. 

 

I could be right or I could be completely wrong or somewhere in the middle. I made the prediction in the "time capsule thread" of which the entire point was to look back after the season had ended to see who made good predictions and who made bad ones. 

[baumerman77]

On 12/5/2016 at 2:25 PM, SID.IS.SID.ME.IS.ME said:

In defense of @baumerman77, I think all he's saying is that he predicted that the Canucks would play their best hockey in the first 1/3 of the season, and then a depleted roster (due to injuries and potential TDL selloffs) would lead to them falling later in the standings.

 

The current PDO issue suggests the Canucks might be a bit better than their results currently indicate. Regression, in a vacuum, should sort that out over a larger sample of games.

 

So maybe the Canucks are a bit worse off right now than he predicted. And maybe they'll improve (due to regression) over the next stretch. But he believes they'll still crash by season's end.

 

Hard to say whether or not that will happen. But it's not an altogether unreasonable prediction. There are depth issues on this team. The young kids currently playing well may not be able to sustain that over an entire season. So if more injuries hit and we end up selling assets, we might be headed for a cliff.

 

So if he's proven correct and the roster does become seriously depleted, then his projection actually carries weight. PDO can regress to the mean, but if the roster is severely weakened, we could see shots differentials plummet, even with shooting percentage and save percentage rising.

 

The net effect of that could be an even worse win percentage in the later season (than what we've seen to-date).

 

And it's also possible that regression in PDO won't actually create enough improvement in actual winning percentage for the team to climb all that many spots in the standings. We could just keep pace with the crucial teams ahead of us, but fail to catch them.

 

I don't necessarily agree with the prediction he's making (mostly because I'm loath to make predictions--too many variables and unknowns). But I can understand where he's coming from. And I don't think it's from a place of not understanding PDO or regression.

 

It's just that @baumerman77 (at least as I've read what he's posted here) doesn't believe those factors will be sufficient to undo the problems he's predicting will come later in the season (due to roster depth issues and how this team will cope with injuries and possibly selling key components at the TDL).

 

On 12/5/2016 at 8:20 PM, baumerman77 said:

That's pretty much what I am saying. 

 

I could be right or I could be completely wrong or somewhere in the middle. I made the prediction in the "time capsule thread" of which the entire point was to look back after the season had ended to see who made good predictions and who made bad ones. 

I feel vindicated. 

[SID.IS.SID.ME.IS.ME]

30 minutes ago, baumerman77 said:

 

I feel vindicated. 

Was wondering why I was getting a notification from this thread.

 

But yeah, funny how that all played out.

 

I can't be bothered to read through all the earlier posts, but seems like your prediction ended up pretty accurate.

 

PDO regressed (increased) to 99.4 (from around 96 or 97 at the time of your post IIRC). You predicted we'd be in the 99-99.5 range for the season. Pretty damn close.

 

But even with that regression, the bottom fell out in terms of our differentials. By season's end, we'd plummeted to 27th in CF% and GF%.

 

You predicted we'd need an extremely high PDO to have any hope to offset these issues.

 

Again, pretty much correct.

 

Injuries took a huge toll. We moved Hansen and Burrows at the TDL. 

 

And we had a terrible record in the final 1/3 of the season.

 

All pretty much as you predicted.

 

Yeah, I can see how you'd feel vindicated.

 

Especially after all the heat you took in this thread.

[baumerman77]

19 hours ago, SID.IS.SID.ME.IS.ME said:

Was wondering why I was getting a notification from this thread.

 

But yeah, funny how that all played out.

 

I can't be bothered to read through all the earlier posts, but seems like your prediction ended up pretty accurate.

 

PDO regressed (increased) to 99.4 (from around 96 or 97 at the time of your post IIRC). You predicted we'd be in the 99-99.5 range for the season. Pretty damn close.

 

But even with that regression, the bottom fell out in terms of our differentials. By season's end, we'd plummeted to 27th in CF% and GF%.

 

You predicted we'd need an extremely high PDO to have any hope to offset these issues.

 

Again, pretty much correct.

 

Injuries took a huge toll. We moved Hansen and Burrows at the TDL. 

 

And we had a terrible record in the final 1/3 of the season.

 

All pretty much as you predicted.

 

Yeah, I can see how you'd feel vindicated.

 

Especially after all the heat you took in this thread.

Ya. Well you were the only one to actually understand my point and not jump to a conclusion.

[oldnews]

20 hours ago, SID.IS.SID.ME.IS.ME said:

 

PDO regressed (increased) to 99.4 (from around 96 or 97 at the time of your post IIRC). You predicted we'd be in the 99-99.5 range for the season. Pretty damn close.

 

where do you get this?

 

.986 according to sportingcharts.

https://www.sportingcharts.com/nhl/stats/team-pdo-numbers-save-plus-shooting-percentage/2016/

 

didn't really regress beyond midseason.

started at a league worst .976 - regressed to the mid point of the season to near the mark it ended at.

[DADDYROCK]

I FEEL GOOD NOW , I JUST HEARD THE EDMONTON OILERS MADE THE PLAYOFFS FIRST TIME SINCE

 

GET THIS "SINCE 2006" WOW .THE LEAFS HAVE MADE IT TWICE.

 

I GUESS THEY "HAVE" HAD A BIT LONGER TO GET A WINNER TOGETHER THAN OUR CANUCKS.

 

BE PATIENT OUR TIME WILL COME SOON.

 

 

[SID.IS.SID.ME.IS.ME]

1 hour ago, oldnews said:

where do you get this?

 

.986 according to sportingcharts.

https://www.sportingcharts.com/nhl/stats/team-pdo-numbers-save-plus-shooting-percentage/2016/

 

didn't really regress beyond midseason.

started at a league worst .976 - regressed to the mid point of the season to near the mark it ended at.

http://stats.hockeyanalysis.com/teamstats.php?disp=1&db=201617&sit=5v5&sort=PDO&sortdir=DESC

 

99.4 for the season (according to the above source). 

 

EDIT: just checked on Corsica and they also have 99.44 

 

http://www.corsica.hockey/teams/

 

 

 

[oldnews]

37 minutes ago, SID.IS.SID.ME.IS.ME said:

http://stats.hockeyanalysis.com/teamstats.php?disp=1&db=201617&sit=5v5&sort=PDO&sortdir=DESC

 

99.4 for the season (according to the above source). 

 

EDIT: just checked on Corsica and they also have 99.44 

 

http://www.corsica.hockey/teams/

 

 

 

That's odd.

statshockey has the Canucks' sv% at 92.47.

 

NHL has MIler at .914 and Markstrom at .910.

 

Curious where they come up with the sv% numbers.  Seem inflated tbh.

And a 6.9% shooting percentage.  PDO also seems inflated.

 

Another thing that doesn't quite make sense to me is  why - with a depleted roster - someone would expect PDO to continue to regress (as is suggested above).

To me, that is counterintuitive.

 

For example, in this context:

As the Canucks lose Granlund, Eriksson, Baertschi, Hansen, Burrows....

And they are replaced by Shore, Cramarossa, Biega, Boucher, Megna, Chaput...

Why would we expect regression? - when half that statistic is shooting percentage - and the relative quality of talent - shooters - the ability to put the puck in the net - has been significantly downgraded in the process?  Those replacement players simply can't be expected to have comparable shooting percentages over any kind of extended sample size.

It seems to me that we could expect the end of regression with the downgrades to the lineup.  And that's what actually appears to have been the case.

Regression would be relative - the quality of lineup initially - vs at an eventual point, a team with 4 of it's initial 12 forwards in the lineup.

 

For example - we saw guys like Sutter, Eriksson etc get off to horrible starts - very low shooting percentages - which regressed as we approached mid-season - because we've seen them over careers put up far better shooting% - the small samples were originally misrepresentative.

And we saw the team in fact climb into a playoff spot after the mid point - and their PDO regress closer to what we'd expect (relative to their more comparable roster the previous season, or season before that) - and I believe their record was in the range of 23-21-6 or thereabouts at the end of January -  in fact in a playoff spot - in front of Calgary actually, for a short period.

What happened from that point forward - wadr, PDO, or even corsi - don't really tell you a whole lot about their 'expected' team, or what it will be 'expected' to look like next year - and can most certainly not be simply reduced to the effect of a couple trade deadline deals.  Nor was it merely 'expected' injuries.  It obviously diverged well beyond the 'normal' or 'expected'.

In a nutshell, if you're looking at the objective outcomes as of December, or January - what actually transpired thereafter - from mumps, flu, food poisoning, etc - to suggest that one 'expected' that, or the extreme departure in results - hard to take seriously wadr.   Feeling vindicated - is the equivalent of 'predicting a recession', but getting a drought, and crediting one's 'predictive models' of the market..  But, hey, if it floats one's boat.

[SID.IS.SID.ME.IS.ME]

@oldnews: wondering if the stats differences are from some sites using 5v5, others "even strength" (which in some cases can include everything from 5v5 to 3v3 to empty net situations), and others looking at all situations?

 

As for the regression issue, I suppose one could believe that Sv% or Sh% were unsustainably low due to misfortune and would rebound regardless of talent. Like if a healthy team of legitimate NHL players was putting up "sub replacement level" percentages teamwide. That would just be bad luck. Even if that team gets banged up, and has a bunch of tweeners come in on injury replacement, there still could be an improvement in overall PDO.

 

I dunno. 

 

As far as "vindication" goes, I think Baumer felt a little ganged up on. I was late to the conversation and honestly haven't read through the entire thread, so could have missed out on something that justified the backlash. But it did seem like he was taking a lot of heat for something that didn't read as all the crazy to me at the time. And what he was predicting didn't really make me feel like he didn't understand PDO. 

 

I could see a hypothetical situation where everything he was predicting could actually happen (and could all fit with the way he was describing the stats). That was the reason I defended him (even though I didn't really have a dog in the fight and was a bit of a Johnny-come-lately to the overall discussion).

 

As it turned out, it seems (to me anyway) that he was more right than wrong on the prediction. I don't think it was based on anything more than a hunch (and I don't recall him claiming anything more than that). But it ended up being correct (or at least within the ballpark) on several points.

 

Whenever people bump old threads like that, they tend to be carrying a little "hurt" over something. Why else would they bother? I can hardly remember half the stuff I post (or people post in their replies) beyond the last week or so.

 

And this goes back months.

 

Seems like Baumer just wanted to point out that he wasn't so crazy after all.

 

So "vindication."

 

I see no harm in that.

[oldnews]

1 hour ago, SID.IS.SID.ME.IS.ME said:

@oldnews: wondering if the stats differences are from some sites using 5v5, others "even strength" (which in some cases can include everything from 5v5 to 3v3 to empty net situations), and others looking at all situations?

 

As for the regression issue, I suppose one could believe that Sv% or Sh% were unsustainably low due to misfortune and would rebound regardless of talent. Like if a healthy team of legitimate NHL players was putting up "sub replacement level" percentages teamwide. That would just be bad luck. Even if that team gets banged up, and has a bunch of tweeners come in on injury replacement, there still could be an improvement in overall PDO.

 

I dunno. 

 

As far as "vindication" goes, I think Baumer felt a little ganged up on. I was late to the conversation and honestly haven't read through the entire thread, so could have missed out on something that justified the backlash

 

Whenever people bump old threads like that, they tend to be carrying a little "hurt" over something. Why else would they bother? I can hardly remember half the stuff I post (or people post in their replies) beyond the last week or so.

 

And this goes back months.

 

I guess some conversations we jump into we don't necessarily know where the sticking points are.

I think the reactions may have something to do with not being particularly drawn to baumerman's repeated, arrogant attacks on Benning - which ironically depart from 'thinking'/'analytics'/pedantic posturing, into....

 

On 2015-06-27 at 8:21 AM, baumerman77 said:

Dim Jim.

 

Quote

Dim Jim

 

On 2015-08-10 at 8:01 PM, baumerman77 said:

a potato would not be able to make bad trades and bad signing (no trades and signings in fact) which might make it better than Benning.

 

On 2015-10-06 at 6:08 PM, baumerman77 said:

Benning's approach to hockey is from the stone age.

 

Must be frustrating to see a cave man still running the team when baumerman has hero chartz, delta corsiz, and superior hockey intelligence.

 

Anyhow, the arrogance, and the appetite and desire for this team's failure is tedious.

 

The 'analtyics community' tends to like to over-state things, ironically, and overestimate their analytics.

You've seen it with Sbisa and dug into goal based vs 'possession' metrics.

baumerman on the other hand maintains he's the worst defenseman in the NHL, tries to prop it up with a single, cherry-picked possession metirc = weak 'analytics' - accompanied with that distinct air of intellectual superiority...

 

I have to wonder if 'analytics' and discussion boards are ruining the sport.  It used to be enjoyable - just watching the game - now we literally suck the life out of the game.

 

 

 

[SID.IS.SID.ME.IS.ME]

2 hours ago, oldnews said:

I have to wonder if 'analytics' and discussion boards are ruining the sport.  It used to be enjoyable - just watching the game - now we literally suck the life out of the game.

I hear ya. Gets to me at times too. And I'm more pro-analytics than most. And also someone who devotes much of my hockey fandom to online places like discussion boards.

 

I definitely take issue with some of the hubris we see from certain corners of the analytics community. I don't really have issues with posters here (or the other boards--even that "cesspool" one most CDCers hate). Bring your opinions--I love reading all the different perspectives. So long as things stay relatively civil. But I can't stand some of the media types and bloggers who serve as Pied Pipers for some really bad analytics.

 

Eg: Sbisa=worst Dman in the NHL, Sutter=replacement level defensive forward, etc.

 

That kind of stuff wears on me. Just because it does such a disservice to the responsible people doing good work with numbers. I hate the hyperbolic claims. Usually backed up with pretty shoddy evidence (dressed up as something else). And packaged into easily digestible nuggets that are ready-made for Twitter.

 

Doesn't bug me too much when I see these things repeated here. That's inevitable. It's the sources where these things originate that draw my ire.

 

That all said, analytics hasn't really changed the game for me. Not the actual viewing experience anyway. At the rink, I still turn back into a little kid gazing in wonder at my heros. Watching the team play is still magical for me. The only numbers I think about on the scoreboard and the jerseys. The goals, the clock, and the players. Couldn't care less about any of the other numbers. 

 

Between games is when I dig into the other stuff. It used to be radio shows and the sports section. Now it's boards, blogs, podcasts, and playing with data from the various sites. A hundred ways to dissect each game. And I still want more.

 

But even when we get to the point (likely through the new computer tracking methods and algorithms) where there are microstats available for literally everything done during the games, it's not gonna ruin hockey. 

 

The game will always be magical. Analytics can't change that. I shut all that stuff off in my mind when the puck drops. And just get immersed in the action. At least when I'm actually there (have to admit I'll look at certain stats occasionally if I'm watching a televised game alone).

 

Maybe it's ruining how we talk about hockey during the time between the actual games. But even then, I'm not so sure. It's still pretty easy to have conversations about the game that don't ever involve statistics. And when I've had the opportunity to talk to real "hockey people" (other than some actual numbers guys), analytics are about the last thing I've wanted to bring into the conversation. Would have seemed gauche.

 

Same goes for when I talk hockey with my dad and the other old timers in his circle of friends. We talk about the players in terms of their skating, shot, puck skills, etc. Not their team relative metrics or heat maps. If we talk coaching, it's about preparation and strategy, not shift charts and deployment/usage. And when we discuss the front office, it's about personnel and team building, not the cap numbers and "asset management."

 

But sometimes I like digging into the numbers. Testing my assumptions against the data. Seeing if my eye test jibes with what the metrics say. 

 

And analytics so allow you to make somewhat intelligent observations about teams and players that you might not have been able to watch enough to otherwise gauge. So long as you recognize the limitations and avoid over-reliance on single metrics and simplistic applications.

 

Analytics will never have a Matrix moment where the ice surface becomes revealed as just a sheet of weird green code. 

 

Players are more complicated than hero charts. And none of the metrics are really all the effective, on their own, in measuring value. Never mind forecasting future performance (just check the r^2 for any predictive analytics used in hockey).

 

Goals and wins are the only stats that really matter. The other stats allow us to

make somewhat educated guesses as to future outcomes. But most metrics remain fairly blunt instruments in terms of their ability to actually predict results.

 

That's why they play the games.

 

Otherwise, we could just hand out the Cup and league awards at the start of the season, based on how all the teams and players shake out in terms of their "expected goals" and "wins above replacement."