The 2019-20 Tank Command Thread -Part V

[Stylizer1]

This thread needs to tank.

Click to read more...

 

[DueDiligence]

Getting Brown and losing Ceci has been worth taking on Zaitzev, for now at least.

Very good point.

 

[BonHoonLayneCornell]

I'm curious, can we assume it even required Ceci going the other way? Did he even hold any value at that point?

Brown was the sweetener for taking on Zaitsev's contract and I wonder if that would have been on the table for anything without Ceci included.

 

[Ice-Tray]

The things some people do for fun!

 

[supsens]

Using goals in a sample of 65 minutes is literally useless. I would never isolate either way, but people here were making a QoC argument. Don't focus on that. I don't care about that. I care way more about the aggregate numbers.

I posted the aggregate 5 v 5 numbers. I am not trying to hide anything. In that sample, Zaitsev is a positive ZS% player. He is not getting buried. In fact, DeMelo has more unfavourable zone starts than Zaitsev, and DeMelo has sparkly numbers. ZS% and QoC doesn't really matter, at least how we measure it now, in the aggregate, so there is no excuse for poor numbers, especially compared to DeMelo. What I mean by saying it does not matter is the effect size is, essentially, non-existent.

Here are the ZS%.
DeMelo OZS%: 44.27
Zaitsev OZS%: 51.22


Also, I don't really know what you are saying about xGF%. You don't create xGF%. It is a regression model. I'll tell you right now the people in hockey did not invent regressions models. Actually, I think I know what you are trying to say, but you are doing a very poor job. Anyway, public studies are available that show expected goals predict future goals better than things like Corsi and actual goals, and we know NHL teams use expected goal models. I am not saying that means anything, but it is far from the "worst stat ever".

The studies I have found depending who’s study state both and or either depending who studied it.
And non are really predictive, players use corsi and advanced metrics to barter on contracts and throw junk at the net that is never going in these days.
Zone starts and QOC don’t matter now...because the stats are garbage and don’t predict anything anyway, someone will post a model that shows a guy having +1 % per 60 and that’s like what 2 shots a game at most? Yet zone starts wouldn’t be enough to add 1 shot for and take one shot against away? It’s fan interaction nonsense

 

[DJB]

I'm curious, can we assume it even required Ceci going the other way? Did he even hold any value at that point?

Brown was the sweetener for taking on Zaitsev's contract and I wonder if that would have been on the table for anything without Ceci included.

Leaf were looking for D for about 2 years. Needed cap space so moved Zaitsev and rolled the dice on Ceci.

They got their cap space but didn't find a dependable D man

 

[ijif]

The studies I have found depending who’s study state both and or either depending who studied it.
And non are really predictive, players use corsi and advanced metrics to barter on contracts and throw junk at the net that is never going in these days.
Zone starts and QOC don’t matter now...because the stats are garbage and don’t predict anything anyway, someone will post a model that shows a guy having +1 % per 60 and that’s like what 2 shots a game at most? Yet zone starts wouldn’t be enough to add 1 shot for and take one shot against away? It’s fan interaction nonsense

Sure, some people find that xGF% out predicts Corsi at both the skater and team level. Other people might find it only out predicts Corsi at one level. Others might find that Corsi out predicts expected goal models. Different authors will use different models. Every team in the NHL probably has a different model. I'm not really interested in arguing the validity of expected goal models, and I am not interested in arguing about if the correlation coefficient is large enough for you. You made a hyperbolic statement that was incorrect. Some expected goal models do out predict Corsi, and some NHL teams do use expected goal models. It is not even close to the worst statistic.

That could be an explanation for why zone starts and QoC don't matter in terms of effect size. It is certainly conceivable that we are not adequately measuring QoC. Theoretically, getting a precise QoC metric will be impossible.

Lastly, you say players care about advanced metrics, but the players disagree with you. The athletic just did a player poll, and they found 86% of players do not pay attention to advanced metrics. Either way, throwing junk on the net does not give you good numbers. This argument is 10 years old, and no one uses it anymore because everyone knows it is crap.

 

[supsens]

Sure, some people find that xGF% out predicts Corsi at both the skater and team level. Other people might find it only out predicts Corsi at one level. Others might find that Corsi out predicts expected goal models. Different authors will use different models. Every team in the NHL probably has a different model. I'm not really interested in arguing the validity of expected goal models, and I am not interested in arguing about if the correlation coefficient is large enough for you. You made a hyperbolic statement that was incorrect. Some expected goal models do out predict Corsi, and some NHL teams do use expected goal models. It is not even close to the worst statistic.

That could be an explanation for why zone starts and QoC don't matter in terms of effect size. It is certainly conceivable that we are not adequately measuring QoC. Theoretically, getting a precise QoC metric will be impossible.

Lastly, you say players care about advanced metrics, but the players disagree with you. The athletic just did a player poll, and they found 86% of players do not pay attention to advanced metrics. Either way, throwing junk on the net does not give you good numbers. This argument is 10 years old, and no one uses it anymore because everyone knows it is crap.

Lol ya 86% of players have no idea a few more shots on net is going to get them more money, I highly doubt that, I guess every city they play in they love to.
What’s crap is Expected goals for, it’s like saying all of Canada has averaged 365 inches of rain a year for the last 5 years then expecting 1 inch of rain in Calgary every day never mind expecting exactly 365 inches of rain for the year...
it does not account for who the goalie is/was it does not account for who was shooting or what events actually led to the goal going in. It’s just fun with averages...
The chances of Everyone being average is nil, averages go up and down people are at the far end on each side of the average.

 

[ijif]

How do these numbers account for D partners, FO% when on ice and home vs away splits, zone starts. I’m just asking because DJ deploys so differently at home than road.

Good question!

Unfortunately, I don't think I can see who Zaitsev is paired with in each individual sample. All I can see is a "with or without you" for Zaitsev and all the other defenders over the entire year. The pattern is very clear. He drags everyone down. That could be because of QoC, but the individual matchups suggest that is not the case.

FO% probably has little effect. The individual samples are super small, but I guess it is possible it could be impactful, but you'd have to say that Zaitsev is generally getting terrible face-off numbers all the time, which seems really unlikely.

Home and away is an interesting variable. Some statistical models do take that into account when evaluating players, but those models are not fans of Zaitsev.

I think I can check all the zone starts, but overall, Zaitsev is a positive zone start player at 5v5. DeMelo has a lower offensive zone start percentage than Zaitsev. DeMelo still puts up good numbers because he actually is a top-four defender.

This is not really directed at you, but more to any Zaitsev supporters (anyone that considers him a top-four defender) Compare any top-four defender on any team to Zaitsev, and you will see massive differences in the on-ice metrics. That is not because Zaitsev is getting the hardest usage in the league. It is because he is not a top-four defender. It really is that simple.

 

[ijif]

Lol ya 86% of players have no idea a few more shots on net is going to get them more money, I highly doubt that, I guess every city they play in they love to.
What’s crap is Expected goals for, it’s like saying all of Canada has averaged 365 inches of rain a year for the last 5 years then expecting 1 inch of rain in Calgary every day never mind expecting exactly 365 inches of rain for the year...
it does not account for who the goalie is/was it does not account for who was shooting or what events actually led to the goal going in. It’s just fun with averages...
The chances of Everyone being average is nil, averages go up and down people are at the far end on each side of the average.

You have no idea what you are talking about. We are done here.

 

[supsens]

Good question!

Unfortunately, I don't think I can see who Zaitsev is paired with in each individual sample. All I can see is a "with or without you" for Zaitsev and all the other defenders over the entire year. The pattern is very clear. He drags everyone down. That could be because of QoC, but the individual matchups suggest that is not the case.

FO% probably has little effect. The individual samples are super small, but I guess it is possible it could be impactful, but you'd have to say that Zaitsev is generally getting terrible face-off numbers all the time, which seems really unlikely.

Home and away is an interesting variable. Some statistical models do take that into account when evaluating players, but those models are not fans of Zaitsev.

I think I can check all the zone starts, but overall, Zaitsev is a positive zone start player at 5v5. DeMelo has a lower offensive zone start percentage than Zaitsev. DeMelo still puts up good numbers because he actually is a top-four defender.

This is not really directed at you, but more to any Zaitsev supporters (anyone that considers him a top-four defender) Compare any top-four defender on any team to Zaitsev, and you will see massive differences in the on-ice metrics. That is not because Zaitsev is getting the hardest usage in the league. It is because he is not a top-four defender. It really is that simple.

Chabot gets lit up goals for/against worse than z, what’s that say? Is he a number 7 dman? Or is there some random stat that means more than actual results?

 

[supsens]

You have no idea what you are talking about. We are done here.

Umm nope I explained exactly how it is, you want to pretend it’s not an average with 1000 unaccounted for variables that’s on you.
Do you think if you get 3.5 expected goals for that means that is how many goals you get? Nope it isn’t
Do you think that means you will get 3.5 expected goals for next game? Nope you won’t
Too bad so sad that’s not how it works

 

[ijif]

Chabot gets lit up goals for/against worse than z, what’s that say? Is he a number 7 dman? Or is there some random stat that means more than actual results?

No, he does not.

T.C
GF% 46.07

N.Z
GF% 41.18

If we look at a wowy, you can clearly see Zaitsev drags Chabot down. Chabot has top-pairing metrics away from Zaitsev. You could have brought up Hainsey! That would have really put me in a pickle!

Look, I am happy to discuss analytics, and what my views are on them, and how I use them. My name is more about philosophy than stats, so I am not a stats nerd, but you clearly have an agenda, don't really know what you are talking about, and you are now trying to play a "gotcha" game, so hopefully, we are actually done now.

 

[supsens]

No, he does not.

T.C
GF% 46.07

N.Z
GF% 41.18

If we look at a wowy, you can clearly see Zaitsev drags Chabot down. Chabot has top-pairing metrics away from Zaitsev. You could have brought up Hainsey! That would have really put me in a pickle!

Look, I am happy to discuss analytics, and what my views are on them, and how I use them. My name is more about philosophy than stats, so I am not a stats nerd, but you clearly have an agenda, don't really know what you are talking about, and you are now trying to play a "gotcha" game, so hopefully, we are actually done now.

Player	Actual Goals	Expected Goals	Difference
Rickard Rakell	32	16.60	-15.40
T.J Oshie	33	18.25	-14.75
Mark Scheifele	31	18.55	-12.45
Michael Grabner	26	14.59	-11.41
Mikael Granlund	25	14.06	-10.94
Anders Lee	31	20.31	-10.87
Nazem Kadri	32	21.38	-10.62
Marian Hossa	26	15.58	-10.42
Eric Staal	26	17.36	-9.64
Patrick Maroon	27	17.46	-9.54
Sidney Crosby	43	33.87	-9.13
Cam Atkinson	34	24.91	-9.09
Marcus Johansson	23	14.56	-8.44
Artem Anisimov	22	13.91	-8.09
Evander Kane	27	19.09	-7.91
Paul Byron	22	14.28	-7.72
Patrik Berglund	22	14.34	-7.66

[TBODY] [/TBODY]
[TBODY] [/TBODY]

 

[ijif]

Delete: I just gotta stop responding

 

[supsens]

You have no idea what you are talking about. We are done here.

Player	Actual Goals	Expected Goals	Difference
Corey Perry	18	33.90	+15.90
Riley Sheahan	0	12.19	+12.19
Patrice Bergeron	21	32.92	+11.92
Jordan Eberle	16	27.01	+11.01
Derick Brassard	13	23.89	+10.89
Tyler Seguin	26	36.67	+10.67
Jonathan Toews	19	29.43	+10.43
Gustav Nyquist	12	22.35	+10.35
Tyson Barrie	6	15.77	+9.77
Mike Cammalleri	10	19.75	+9.75
Boone Jenner	16	25.73	+9.73
Joe Pavelski	29	37.25	+8.25
Chris Kunitz	9	17.19	+8.19
Mats Zuccarello	15	23.07	+8.07
Anze Kopitar	12	20	+8
Patrick Kane	34	41.67	+7.67
Brendan Gallagher	10	17.20	+7.20
Johnny Gaudreau	18	25.14	+7.14

[TBODY] [/TBODY]
[TBODY] [/TBODY]

Both those lists are from the same year.. and nhl.com says a different GF% than you posted
With chabot and z
Chabot 42.6%
Z 44.1%

 

[ijif]

Player	Actual Goals	Expected Goals	Difference
Corey Perry	18	33.90	+15.90
Riley Sheahan	0	12.19	+12.19
Patrice Bergeron	21	32.92	+11.92
Jordan Eberle	16	27.01	+11.01
Derick Brassard	13	23.89	+10.89
Tyler Seguin	26	36.67	+10.67
Jonathan Toews	19	29.43	+10.43
Gustav Nyquist	12	22.35	+10.35
Tyson Barrie	6	15.77	+9.77
Mike Cammalleri	10	19.75	+9.75
Boone Jenner	16	25.73	+9.73
Joe Pavelski	29	37.25	+8.25
Chris Kunitz	9	17.19	+8.19
Mats Zuccarello	15	23.07	+8.07
Anze Kopitar	12	20	+8
Patrick Kane	34	41.67	+7.67
Brendan Gallagher	10	17.20	+7.20
Johnny Gaudreau	18	25.14	+7.14

[TBODY] [/TBODY]
[TBODY] [/TBODY]

Both those lists are from the same year.. and nhl.com says a different GF% than you posted
With chabot and z
Chabot 42.6%
Z 44.1%

I am only responding to clear up data differences to anyone reading this conversation. I know what a regression model is. Your graphs are doing nothing but wasting space in the thread. Please do not respond to this post as I do not want to derail this thread anymore. Again, I am only responding to you to clear up the data differences.

NHL.com is using EV GF%, and I am using 5 v 5 GF%.

 

[supsens]

I am only responding to clear up data differences to anyone reading this conversation. I know what a regression model is. Your graphs are doing nothing but wasting space in the thread. Please do not respond to this post as I do not want to derail this thread anymore. Again, I am only responding to you to clear up the data differences.

NHL.com is using EV GF%, and I am using 5 v 5 GF%.

Ohh regression model, again no one is average every year, no team is average every year and when most the goalies are within 3% of each other (starters even closer) it’s not really that hard to get a good R2 value as ‘proof’ is it?
And with your GF% if Z’s line scores what one goal, two tops it’s a tie? Oooo I guess one goal is the difference from a number 6 dman and a top talent.
It’s also a team sport 5 guys vrs 5 guys expected goals against has a lot of blame or credit to go around.

 

[JD1]

Sure, some people find that xGF% out predicts Corsi at both the skater and team level. Other people might find it only out predicts Corsi at one level. Others might find that Corsi out predicts expected goal models. Different authors will use different models. Every team in the NHL probably has a different model. I'm not really interested in arguing the validity of expected goal models, and I am not interested in arguing about if the correlation coefficient is large enough for you. You made a hyperbolic statement that was incorrect. Some expected goal models do out predict Corsi, and some NHL teams do use expected goal models. It is not even close to the worst statistic.

That could be an explanation for why zone starts and QoC don't matter in terms of effect size. It is certainly conceivable that we are not adequately measuring QoC. Theoretically, getting a precise QoC metric will be impossible.

Lastly, you say players care about advanced metrics, but the players disagree with you. The athletic just did a player poll, and they found 86% of players do not pay attention to advanced metrics. Either way, throwing junk on the net does not give you good numbers. This argument is 10 years old, and no one uses it anymore because everyone knows it is crap.

Interesting little discussion you guys are having.

The Athletic poll comment was funny. 86% of players don't pay attention. You know what they care deeply about? +/-. Go figure. The stats guys versus the players. I wonder who understands the game better?

 

[JD1]

Good question!

Unfortunately, I don't think I can see who Zaitsev is paired with in each individual sample. All I can see is a "with or without you" for Zaitsev and all the other defenders over the entire year. The pattern is very clear. He drags everyone down. That could be because of QoC, but the individual matchups suggest that is not the case.

FO% probably has little effect. The individual samples are super small, but I guess it is possible it could be impactful, but you'd have to say that Zaitsev is generally getting terrible face-off numbers all the time, which seems really unlikely.

Home and away is an interesting variable. Some statistical models do take that into account when evaluating players, but those models are not fans of Zaitsev.

I think I can check all the zone starts, but overall, Zaitsev is a positive zone start player at 5v5. DeMelo has a lower offensive zone start percentage than Zaitsev. DeMelo still puts up good numbers because he actually is a top-four defender.

This is not really directed at you, but more to any Zaitsev supporters (anyone that considers him a top-four defender) Compare any top-four defender on any team to Zaitsev, and you will see massive differences in the on-ice metrics. That is not because Zaitsev is getting the hardest usage in the league. It is because he is not a top-four defender. It really is that simple.

So i went looking for the poll on the Athletic that you sited. Couldn't find it but would like to read it. I did find an interesting article by Bourne dated January 23 in which WOWY data was cited as the worst way to assess a player

I have issues with advanced stats and their meaning. The article i referenced contains two statements that highlight some issues with advanced stats:

1. Every shot on net is a scoring chance. Statistically speaking that is correct. Yet from a hockey viewpoint, i can stand on the blueline on pound slap shots at Carey Price and I'm not going to score with my 50 something year old shot
2. The interviewee was asked how their xG models account for a situation where a 2 on 0 doesn't result in a shot on net. The answer was they don't. And the interviewee rhetorically asked "if it didn't result in a shot on net was it really a scoring chance?" Which mathematically speaking is true because in order to have a "chance" to score, you have to shoot the puck on net, however from a hockey viewpoint, me standing at the point blasting slappers on Price results in scoring chances whereas McDavid and Draisaitl on a 2 on 0 on Price and Draisaitl whiffing on the pass is not an xG is wrong.

I think the stats eggheads have a ways to go yet before their data is truly relied on.

 

[Sweatred]

Couple big games for the Ducks as they play the Sharks while we play the Devils tomorrow. Next week we get the Ducks. Draft spot up for grabs over these next 8 days ... #tankhard I’m not sure we could catch the Ducks if we get 4pts and they get O.

 

[Tuna99]

Tank at the All Star break is going amazing, two top 10 picks, 2 All Stars for the 2nd worse team in the NHL, traceable assets Demelo, Pageau, Hainsey, Ennis, Tierney, Boro all performing well and holding or increasing their value - Norris, L. Brown, Formenton, Batherson, Brannstrom looking elite in the AHL, prospects JBD, Pinto, Lassi and others performing well in junior, coach has the players ear and they seem to love him despite the losing.

This tank couldn’t go any better, so much depends on the lottery balls falling our way, but you gotta give this tank and A+ so far.

we are getting it right. Sucking ain’t easy, but when you embrace it instead of fight it, so much better for everyone

 

[Sweatred]

Tank at the All Star break is going amazing, two top 10 picks, 2 All Stars for the 2nd worse team in the NHL, traceable assets Demelo, Pageau, Hainsey, Ennis, Tierney, Boro all performing well and holding or increasing their value - Norris, L. Brown, Formenton, Batherson, Brannstrom looking elite in the AHL, prospects JBD, Pinto, Lassi and others performing well in junior, coach has the players ear and they seem to love him despite the losing.

This tank couldn’t go any better, so much depends on the lottery balls falling our way, but you gotta give this tank and A+ so far.

we are getting it right. Sucking ain’t easy, but when you embrace it instead of fight it, so much better for everyone

I think the only error PD could make from here on in is finishing 2-4 slots the wrong way in the lottery. He needs to direct this team to a bottom 3 finish and let the balls fall where they do.

 

[GCK]

Good question!

Unfortunately, I don't think I can see who Zaitsev is paired with in each individual sample. All I can see is a "with or without you" for Zaitsev and all the other defenders over the entire year. The pattern is very clear. He drags everyone down. That could be because of QoC, but the individual matchups suggest that is not the case.

FO% probably has little effect. The individual samples are super small, but I guess it is possible it could be impactful, but you'd have to say that Zaitsev is generally getting terrible face-off numbers all the time, which seems really unlikely.

Home and away is an interesting variable. Some statistical models do take that into account when evaluating players, but those models are not fans of Zaitsev.

I think I can check all the zone starts, but overall, Zaitsev is a positive zone start player at 5v5. DeMelo has a lower offensive zone start percentage than Zaitsev. DeMelo still puts up good numbers because he actually is a top-four defender.

This is not really directed at you, but more to any Zaitsev supporters (anyone that considers him a top-four defender) Compare any top-four defender on any team to Zaitsev, and you will see massive differences in the on-ice metrics. That is not because Zaitsev is getting the hardest usage in the league. It is because he is not a top-four defender. It really is that simple.

Thanks I appreciate the work. This is one of those players where my eyes and the analytics disagree. I watch him and I see him as pretty steady who doesn’t really seem to make the glaring mistakes that I see from the Ceci’s or players of that ilk.

 

[ijif]

Interesting little discussion you guys are having.

The Athletic poll comment was funny. 86% of players don't pay attention. You know what they care deeply about? +/-. Go figure. The stats guys versus the players. I wonder who understands the game better?

I am glad someone thought it was interesting! To be honest, it is just one guy yelling at the clouds. I didn't even really defend expected goal models, but he felt the need to continue telling me how bad he thinks they are.

Anyway, I get the logic behind caring about +/-, but it is calculated so horrifically. I don't get why players would not care about 5v5 GF% if they care about +/-.

Both people see a different side of the game. I don't think one group knows the game more than another. A lot of people that use stats have played hockey. They may not have made the NHL, but they do understand the game.