Prediction time: opening night lineup

[Weave]

Looking at the pipeline of talent we got, some will get traded or become a buy out if they don't start performing.

Leino and stafford being the obvious choices. But Myers must be feeling the heat by now.

 

But lets say Ristolainen (NHL) and zadarov (AHL) perform well, i can see Regier thinking about putting Myers up on the trade block.

A team like edmonton would take Myers for sure.

Myers + stafford for Eberle and a 1st round pick :D

 

Shoot high and ask for Yakupov instead of eberle ?

 

Who was it in another thread that thought if he posted which players are not eligible for the AHL that it would stop the three-times-weekly need to correct posters here?

[Wraith]

A lot of people make the Corsi mistake. It's a problem of not understanding the difference between a signify value and a determine value, for the lack of better adjectives.

 

There are better adjectives: dependent and indepentent variables. This is important terminology is you're going to do a proper data analysis.

 

Steve Ott comes up ice, he knocks the puck into the corner making the goaltender handle it around the net, sometimes it results in a goal, as in last night. But Ott doesn't get a shot, so the Corsi's really don't like Ott.

 

Except that in this example, Ott gets credit for his icemates shot(s). So the Corsi's love him again.

 

Drew Stafford is just playing the role of the rich man 300 years ago who didn't want to pay tax. He is told that shots are important. So he takes shots. The Sabres love him, because he takes shots, but he's not effective, because it's not important. A usable data framework cannot allow Drew Stafford to control an action that signifies value, because the game is not about shots. You must look at data that determines value, honestly, it really clouds your judgment.

 

1) At a team level, Corsi is by far the most consistent predictor of future success. This is proven.

 

2) At an individual level, Corsi is also a great predictor but it needs context such as: individual offensive zone start %, individual quality of linemates, individual quality of competition, individual shot quality (distance and location). Amongst the Sabres forwards who played >10, there are two great examples: A) Steve Ott finished last in Corsi but he started in the offensive zone far less than his teammates (by far) and faced the highest quality of competition (again, by far); B) Marcus Foligno finished first in Corsi but he started very frequently in the offensive zone (only Grigorenko started there more) and faced a very low quality of competition (again, only Grigorenko faced worse competition).

 

3) Contrary to the naysayers, an individual taking a ton of low quality shots will not compile a good Corsi statistic. Drew Stafford is a great example of this. He led the team in shots but was very mediocre in Corsi (6th out of 12 forwards with >10 games played: Hecht, Porter, Grigorenko, Kaleta, Gerbe, and Ott were the players with a worse Corsi). Because of your choice of the example above, it's not clear to me that you understand that for the Corsi statistic, a player receives credit for his linemates shots and attempted shots. Good Corsi numbers are generated by a line that gets sustained pressure and multiple shots a shift while limiting their opponent's attempts to do the same. An individual firing low quality shots does not result in rebounds, does not result in possession, and does not result in sustained pressure, so the Corsi number suffers as a result.

 

4) The problem with Drew Stafford last season is best summed up here.

[LGR4GM]

Looking at the pipeline of talent we got, some will get traded or become a buy out if they don't start performing.

Leino and stafford being the obvious choices. But Myers must be feeling the heat by now.

 

But lets say Ristolainen (NHL) and zadarov (AHL) perform well, i can see Regier thinking about putting Myers up on the trade block.

A team like edmonton would take Myers for sure.

Myers + stafford for Eberle and a 1st round pick :D

 

Shoot high and ask for Yakupov instead of eberle ?

NOOOOOOOOOOOOOO!!!!

 

Grigorenko - NHL ONLY

Girgensons - NHL or AHL

 

Zadorov - NHL or JUNIORS NOOOOOOOO AHL!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Ristolainen - NHL or AHL

 

Who was it in another thread that thought if he posted which players are not eligible for the AHL that it would stop the three-times-weekly need to correct posters here?

It was me and I am starting a mother ###### thread about it because I am so sick of hearing about how player X is going to the AHL when he can't play there for another 2years!!!!!!!!!!!!!

[drnkirishone]

 

NOOOOOOOOOOOOOO!!!!

 

Grigorenko - NHL ONLY

Girgensons - NHL or AHL

 

Zadorov - NHL or JUNIORS NOOOOOOOO AHL!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Ristolainen - NHL or AHL

 

 

It was me and I am starting a mother ###### thread about it because I am so sick of hearing about how player X is going to the AHL when he can't play there for another 2years!!!!!!!!!!!!!

so zadorov is playing for rochester this year? Lol

[SwampD]

There are better adjectives: dependent and indepentent variables. This is important terminology is you're going to do a proper data analysis.

 

 

 

Except that in this example, Ott gets credit for his icemates shot(s). So the Corsi's love him again.

 

 

 

1) At a team level, Corsi is by far the most consistent predictor of future success. This is proven.

 

2) At an individual level, Corsi is also a great predictor but it needs context such as: individual offensive zone start %, individual quality of linemates, individual quality of competition, individual shot quality (distance and location). Amongst the Sabres forwards who played >10, there are two great examples: A) Steve Ott finished last in Corsi but he started in the offensive zone far less than his teammates (by far) and faced the highest quality of competition (again, by far); B) Marcus Foligno finished first in Corsi but he started very frequently in the offensive zone (only Grigorenko started there more) and faced a very low quality of competition (again, only Grigorenko faced worse competition).

 

3) Contrary to the naysayers, an individual taking a ton of low quality shots will not compile a good Corsi statistic. Drew Stafford is a great example of this. He led the team in shots but was very mediocre in Corsi (6th out of 12 forwards with >10 games played: Hecht, Porter, Grigorenko, Kaleta, Gerbe, and Ott were the players with a worse Corsi). Because of your choice of the example above, it's not clear to me that you understand that for the Corsi statistic, a player receives credit for his linemates shots and attempted shots. Good Corsi numbers are generated by a line that gets sustained pressure and multiple shots a shift while limiting their opponent's attempts to do the same. An individual firing low quality shots does not result in rebounds, does not result in possession, and does not result in sustained pressure, so the Corsi number suffers as a result.

 

4) The problem with Drew Stafford last season is best summed up here.

The only thing I've ever seen a Corsi number do is try to show me that a crappy player actually isn't as crappy as I thought.

 

It's funny that you don't need a number to know when a good player is good.

[nfreeman]

Couple of items from WGR yesterday:

 

1. Duff really likes Larsson.

 

2. Hamilton thinks Ristolainen will make the Sabres out of camp and in fact thinks he's already made it.

 

So maybe:

 

Myers-Ehrhoff

Pysyk-Tallinder

Weber-Ristolainen

[darksabre]

Couple of items from WGR yesterday:

 

1. Duff really likes Larsson.

 

2. Hamilton thinks Ristolainen will make the Sabres out of camp and in fact thinks he's already made it.

 

So maybe:

 

Myers-Ehrhoff

Pysyk-Tallinder

Weber-Ristolainen

 

I really really wish someone would bump Hank out of the lineup.

[Peppy22]

Could be a tallinder and sulzer swap now and then

[nfreeman]

I really really wish someone would bump Hank out of the lineup.

 

I understand the sentiment -- Hank did plenty to earn it during his time here -- but I also think a D group needs experience. The Sabres' D group is pretty green.

[darksabre]

 

 

I understand the sentiment -- Hank did plenty to earn it during his time here -- but I also think a D group needs experience. The Sabres' D group is pretty green.

 

His experience won't make up for his declining ability.

[TrueBlueGED]

A lot of people make the Corsi mistake. It's a problem of not understanding the difference between a signify value and a determine value, for the lack of better adjectives. Let me start with a bit of history that will serve as an analogy

 

If you've taken some tax law, you've come across the story where the king decided to tax his people by counting windows, because big houses were built by rich people, and the king could get more money out of the rich people. The taxman had it easy, walked up to the house, counted the windows, collected the tax.

 

People started building enormous houses without windows. What the king learned was that windows signified wealth, they didn't determine wealth, and if you allow those who are being valuated to control those signifiers, you will get bad valuation.

 

300 years later Steve Ott comes up ice, he knocks the puck into the corner making the goaltender handle it around the net, sometimes it results in a goal, as in last night. But Ott doesn't get a shot, so the Corsi's really don't like Ott.

 

Two minutes after that, Drew Stafford will cross the red line and shoot at the goaltender. The goaltender will pass to a defenseman, and the opponent will start coming back. A terrible play, the goaltender coach will yell out "Way to go Drew," and the GM will wonder despite winning the last 20 trades, why his team loses.

 

Drew Stafford is just playing the role of the rich man 300 years ago who didn't want to pay tax. He is told that shots are important. So he takes shots. The Sabres love him, because he takes shots, but he's not effective, because it's not important.

 

A usable data framework cannot allow Drew Stafford to control an action that signifies value, because the game is not about shots. You must look at data that determines value, honestly, it really clouds your judgment.

 

Aside from vehemently disagreeing with this analysis of the Corsi number, it does nothing to justify the use of +/- as a measure of quality play. It's hilarious you argue against Corsi in that it doesn't measure the impact of every single play, yet use +/- as if it does. If a team starts a rush and scores, but one of the defensemen made a line change during the rush, the Dman going off gets no reward from +/- even if he helped clear the puck to start the rush while the Dman stepping on gets a +1 even though he had nothing to do with the play and goal. Extremely stylized examples are easy to create to discredit any statistic. Here's the thing: there's research upon research upon research showing +/- is largely random and determined by team shooting and save percentages, and thus is a terrible measure of individual performance. At the same time, there's a ton of research showing Corsi is positively related to scoring and winning at the individual and team levels. It's not a perfect stat, and as Wraith said context is important...but it's infinitely better than +/-.

 

There are better adjectives: dependent and indepentent variables. This is important terminology is you're going to do a proper data analysis.

 

 

 

Except that in this example, Ott gets credit for his icemates shot(s). So the Corsi's love him again.

 

 

 

1) At a team level, Corsi is by far the most consistent predictor of future success. This is proven.

 

2) At an individual level, Corsi is also a great predictor but it needs context such as: individual offensive zone start %, individual quality of linemates, individual quality of competition, individual shot quality (distance and location). Amongst the Sabres forwards who played >10, there are two great examples: A) Steve Ott finished last in Corsi but he started in the offensive zone far less than his teammates (by far) and faced the highest quality of competition (again, by far); B) Marcus Foligno finished first in Corsi but he started very frequently in the offensive zone (only Grigorenko started there more) and faced a very low quality of competition (again, only Grigorenko faced worse competition).

 

3) Contrary to the naysayers, an individual taking a ton of low quality shots will not compile a good Corsi statistic. Drew Stafford is a great example of this. He led the team in shots but was very mediocre in Corsi (6th out of 12 forwards with >10 games played: Hecht, Porter, Grigorenko, Kaleta, Gerbe, and Ott were the players with a worse Corsi). Because of your choice of the example above, it's not clear to me that you understand that for the Corsi statistic, a player receives credit for his linemates shots and attempted shots. Good Corsi numbers are generated by a line that gets sustained pressure and multiple shots a shift while limiting their opponent's attempts to do the same. An individual firing low quality shots does not result in rebounds, does not result in possession, and does not result in sustained pressure, so the Corsi number suffers as a result.

 

4) The problem with Drew Stafford last season is best summed up here.

 

Terrific post.

[shrader]

so zadorov is playing for rochester this year? Lol

 

They're actually sending him to RIT. d4rksabre will be thrilled.

[LGR4GM]

Couple of items from WGR yesterday:

 

1. Duff really likes Larsson.

 

2. Hamilton thinks Ristolainen will make the Sabres out of camp and in fact thinks he's already made it.

 

So maybe:

 

Myers-Ehrhoff

Pysyk-Tallinder

Weber-Ristolainen

I am starting to really like Larsson.

[darksabre]

 

 

They're actually sending him to RIT. d4rksabre will be thrilled.

 

Oddly enough, wouldn't he technically be able to do that?

[chileanseabass]

Oddly enough, wouldn't he technically be able to do that?

 

Nope, he lost NCAA eligibility once he played in the OHL.

[darksabre]

 

 

Nope, he lost NCAA eligibility once he played in the OHL.

 

Ah, I wasn't sure what juniors would do. Would it vary by league?

[LGR4GM]

Ah, I wasn't sure what juniors would do. Would it vary by league?

Any league that pays players screws your NCAA eligibility. So the 3 main CHL (WHL, OHL, QMJHL) all do that.

 

 

I am convinced that Johan Larsson will be on this team to start the season. We have our 3rd line center IMPO. This kid is a competitor. Very excited he is a Sabre.

[Marvelo]

Here's my take.

 

Trade Vanek for ? - Hodgson-Foligno

Leino-Ennis-Trade Stafford for ?

Ott- Girgensons-Larsson

Tropp-Grigorenko-Armia

Extras- Porter, Flynn, Kaleta

 

Ehrhoff, Myers, Pysyk, Weber, McBain, McNabb

Extras- Ruhedal, Tallinder

 

Enroth - Hackett

 

I say give Armia a shot on the 4th line, see how he does. I like Girgensons w Larsson. Not sure Grigorenko is ready but the shorter depth of the net this year could work to the advantage of quick guys with long reach like him. I want Stafford and Vanek traded for a couple skill guys with grit. If I'm Regier, I get something for Vanek before it's too late. We'll need a miracle if we get something for Stafford. If I think you send Zadorov and Ristolainen to get major minutes with Roch and develop. I'd like to also trade Miller.

[LGR4GM]

Here's my take.

 

Trade Vanek for ? - Hodgson-Foligno

Leino-Ennis-Trade Stafford for ?

Ott- Girgensons-Larsson

Tropp-Grigorenko-Armia

Extras- Porter, Flynn, Kaleta

 

Ehrhoff, Myers, Pysyk, Weber, McBain, McNabb

Extras- Ruhedal, Tallinder

 

Enroth - Hackett

 

I say give Armia a shot on the 4th line, see how he does. I like Girgensons w Larsson. Not sure Grigorenko is ready but the shorter depth of the net this year could work to the advantage of quick guys with long reach like him. I want Stafford and Vanek traded for a couple skill guys with grit. If I'm Regier, I get something for Vanek before it's too late. We'll need a miracle if we get something for Stafford. If I think you send Zadorov and Ristolainen to get major minutes with Roch and develop. I'd like to also trade Miller.

Why did I strike your entire statement? You think we can send Nikita Zadorov to Rochester. Your entire post in now invalid.

[inkman]

 

Why did I strike your entire statement? You think we can send Nikita Zadorov to Rochester. Your entire post in now invalid.

Liger says growl

[Marvelo]

Why did I strike your entire statement? You think we can send Nikita Zadorov to Rochester. Your entire post in now invalid.

 

oh I forgot, God was in the room.

[darksabre]

 

 

oh I forgot, God was in the room.

 

But seriously, there's a whole thread dedicated to helping you not make these easy mistakes.

[Marvelo]

who cares. details details.

[darksabre]

who cares. details details.

 

The people who have to read your posts care. Try harder.

[rakish]

There are better adjectives: dependent and indepentent variables. This is important terminology is you're going to do a proper data analysis.

 

 

 

Except that in this example, Ott gets credit for his icemates shot(s). So the Corsi's love him again.

 

 

 

1) At a team level, Corsi is by far the most consistent predictor of future success. This is proven.

 

2) At an individual level, Corsi is also a great predictor but it needs context such as: individual offensive zone start %, individual quality of linemates, individual quality of competition, individual shot quality (distance and location). Amongst the Sabres forwards who played >10, there are two great examples: A) Steve Ott finished last in Corsi but he started in the offensive zone far less than his teammates (by far) and faced the highest quality of competition (again, by far); B) Marcus Foligno finished first in Corsi but he started very frequently in the offensive zone (only Grigorenko started there more) and faced a very low quality of competition (again, only Grigorenko faced worse competition).

 

3) Contrary to the naysayers, an individual taking a ton of low quality shots will not compile a good Corsi statistic. Drew Stafford is a great example of this. He led the team in shots but was very mediocre in Corsi (6th out of 12 forwards with >10 games played: Hecht, Porter, Grigorenko, Kaleta, Gerbe, and Ott were the players with a worse Corsi). Because of your choice of the example above, it's not clear to me that you understand that for the Corsi statistic, a player receives credit for his linemates shots and attempted shots. Good Corsi numbers are generated by a line that gets sustained pressure and multiple shots a shift while limiting their opponent's attempts to do the same. An individual firing low quality shots does not result in rebounds, does not result in possession, and does not result in sustained pressure, so the Corsi number suffers as a result.

 

4) The problem with Drew Stafford last season is best summed up here.

 

1- People say this, but that's all I see. I read something like this

 

The writer concludes with this: What is interesting is I can use the Weka function CfsSubsetEval which tells me which features are contributing the most to the accuracy of the classifier. I am surprised to see it is: Home/Away location, Goals Against and Goal differential. These are not advanced statistics, they are the traditional statistics that are making the biggest difference on predicting of winning a single game. It should be reiterated that this is NOT me disproving the use of advanced statistics such as Fenwick Close but rather saying in predicting in the short term of a single game there is still value in these traditional statistics.

 

 

What he does is first proves that traditional statistics are making the biggest difference on prediction, then turns around and denies it, because it isn't what he believes, really crappy analysis.

 

2- I just want to note that using my framework Pominville saw the most difficult competition against opposing forwards, and against opposing defensemen. That Corsi people combine the two numbers is a weakness in the analysis because the two numbers mean totally different things.

 

3- So it seems that the worst forwards on the Sabres using your framework are

Stafford

Hecht

Porter

Grigo

Kaleta

Gerbe

Ott

 

So really, you are asking me to study up on a framework that has Steve Ott as the worst Sabre forward? So I can understand it better? Really? So those are the bottom 6 or so of the Corsi chart, if you go plus/minus, the valuation, off the top of my head, will look something like

 

Ott

Hecht

Flynn

Vanek

Kaleta

Gerbe

Grigo

Foligno

Hodgson

Ennis

Stafford

 

Now we could use a different framework, goals, and it would look like...

 

Vanek

Hodgson

Ennis

...

 

You get the point, the important thing is whether the framework captures the value of the player. Now I don't post a lot, I really doubt I have something to say to the people that see a lot more hockey than I do, I live in Virginia, last year I got the Caps and the NBC games. I listen to RJ a lot. I try to keep people posted on things they might not see, like moving Ennis to wing will be bad.

 

So the question is, which framework best captures the value of the players. I will ask the users who read this to tell me, just post, rakish (that's me), you are nuts, Steve Ott is the worst player on the Sabres, I watch them and I know. I might be wrong, I would like those who watch more to tell me.

 

Now if you chart team goal differential vs team points, you get a nice line. If you chart team shooting percentage vs team points, you get a slightly worse line, though the better teams do have better shooting. If you chart shot differential vs team points, you get a very sloppy chart. Sloppy enough to not base my analysis on shots.

 

4- I wrote the same thing in a post a couple months back. When Stafford plays well, he has a lot of 10' shots, when he plays poorly, he has a lot of 15' shots. It has nothing to do with quantity.