Advanced Stats- Expected Goal %

[JoeSchmoe]

I thought I'd start this thread for discussion. I'll provide regular updates on 5 on 5 team expected goal % to see where our guys are.

Essentially the stat quantifies the scoring chances for along with the quality of the chance vs the number of chances and quality of chance against you. It'll also do a "relative" measurement that compares how the team does when you're on the ice vs off.

Measuring shot quality is an imperfect science, but a lot of people try very hard to model this, and I think they're at least close. Different sites have different models as well. I use Moneypuck since it's free and what I'm used to.

The stat does not track shooting skill, so players like Thompson, Olofsson, Matthews, etc will be underrepresented by this stat given they do a better job of putting pucks away.

Players who consistently get defensive matchups will show worse since they inherently are used to stop goals instead of getting goals (i.e. Malenstyn). Also, if you consistently play with a good or bad player, it will affect you.

It also doesn't track special teams play which sometimes is important.

That said, we know our guys pretty good, so we can make some sense of the numbers... Here's what I see.

Peterka, Dahlin, and Thompson are our best players and that's clearly seen here.

Power and Byram are low. Byram sucked in this regard on COL. Are they actually getting the tough matchups? Last year Power was actually pretty good so it's surprising to see him that low.

Cozens isn't as bad as advertised, but still has to improve.

Lafferty, Malenstyn, and Krebs are low, but would their defense assignments justify this?

[JoeSchmoe]

Here's more food for thought. McLeod and Zucker are getting double to triple the defensive zone starts than most of the other forwards, yet are both mid-tier in expected goals. This far, they are looking like good pickups by Adams.

[Pimlach]

On-ice Expected Goals.

Off-ice Expected Goals?  

That's it!   We need guys that can score goals when they are off-ice.  

 

[mjd1001]

1 hour ago, JoeSchmoe said:

Here's more food for thought. McLeod and Zucker are getting double to triple the defensive zone starts than most of the other forwards, yet are both mid-tier in expected goals. This far, they are looking like good pickups by Adams.

On the other hand, Quinn, Benson, and Cozens are near the bottom of that list, making it look even worse for them.

[Slack_in_MA]

3 hours ago, JoeSchmoe said:

Essentially the stat quantifies the scoring chances for along with the quality of the chance vs the number of chances and quality of chance against you. It'll also do a "relative" measurement that compares how the team does when you're on the ice vs off.

Any chance this can be simplified even further for the layman/ casual fan? E.g., like you’re explaining it to an eight year-old. Thanks!

[JoeSchmoe]

6 minutes ago, Slack_in_MA said:

Any chance this can be simplified even further for the layman/ casual fan? E.g., like you’re explaining it to an eight year-old. Thanks!

Reading it again, it is a bit wordy.

The eight-year version is its kind of like a plus-minus for scoring chances.

The teenager version is the more chances you get, and the better chances you get, the more expected goals you should have. The more chances against along with the better chances against, the more expected goals against you get. The % is the ratio of the two.

[LGR4GM]

11 hours ago, Slack_in_MA said:

Any chance this can be simplified even further for the layman/ casual fan? E.g., like you’re explaining it to an eight year-old. Thanks!

Every shot taken on the ice, has a small chance to go in for a goal. Math breaks down how many shots become goals from that spot.

So let's use the faceoff dot as an example, and say for every 100 shots taken from the right faceoff dot, 10 go on. That gives us 10 out of 100 or 10% or for xgf, 0.1 because 10% as a decimal is 0.1

So, when you're on the ice you get .1 for every right faceoff dot shot your team takes and .1 against for every one the opposing team takes. So if you're good and are just allowing low quality point shoots but in turn get high quality shots yourself, you'll have a here xgf or expected goals for. xGF% is your xgf / xga

Note, shots have to be on net. 

[Slack_in_MA]

4 hours ago, LGR4GM said:

Every shot taken on the ice, has a small chance to go in for a goal. Math breaks down how many shots become goals from that spot.

So let's use the faceoff dot as an example, and say for every 100 shots taken from the right faceoff dot, 10 go on. That gives us 10 out of 100 or 10% or for xgf, 0.1 because 10% as a decimal is 0.1

So, when you're on the ice you get .1 for every right faceoff dot shot your team takes and .1 against for every one the opposing team takes. So if you're good and are just allowing low quality point shoots but in turn get high quality shots yourself, you'll have a here xgf or expected goals for. xGF% is your xgf / xga

Note, shots have to be on net. 

Thanks. That's very clear.

[R_Dudley]

5 hours ago, LGR4GM said:

Every shot taken on the ice, has a small chance to go in for a goal. Math breaks down how many shots become goals from that spot.

So let's use the faceoff dot as an example, and say for every 100 shots taken from the right faceoff dot, 10 go on. That gives us 10 out of 100 or 10% or for xgf, 0.1 because 10% as a decimal is 0.1

So, when you're on the ice you get .1 for every right faceoff dot shot your team takes and .1 against for every one the opposing team takes. So if you're good and are just allowing low quality point shoots but in turn get high quality shots yourself, you'll have a here xgf or expected goals for. xGF% is your xgf / xga

Note, shots have to be on net. 

Very well explained in layman's terms and appreciated for us not master class tactician fans. 

[mjd1001]

6 hours ago, LGR4GM said:

Every shot taken on the ice, has a small chance to go in for a goal. Math breaks down how many shots become goals from that spot.

So let's use the faceoff dot as an example, and say for every 100 shots taken from the right faceoff dot, 10 go on. That gives us 10 out of 100 or 10% or for xgf, 0.1 because 10% as a decimal is 0.1

So, when you're on the ice you get .1 for every right faceoff dot shot your team takes and .1 against for every one the opposing team takes. So if you're good and are just allowing low quality point shoots but in turn get high quality shots yourself, you'll have a here xgf or expected goals for. xGF% is your xgf / xga

Note, shots have to be on net. 

So, it helps, in the offensive end, explain who gets better chances or who helps generate better chances. However, for someone like Cozens, it doesn't. We know he is still getting good chances, just over the long run, he is under a 10% shooter. 

So you kinda have to combine this with how good a shooter is (or how good of shooters their linemates are) to get an accurate picture, at least in the offensive zone how good they are, or will be.

[SwampD]

I love this stat. It would be perfect if it also factored in the quality of the shooter and goalie (it is a league average, so it doesn't differentiate between Samuelson shooting on Bobrovsky from Crosby shooting on Houser from the same spot), but it's close enough.

[GASabresIUFAN]

9 hours ago, LGR4GM said:

Note, shots have to be on net.

That eliminated most of the Sabres forwards.

[JoeSchmoe]

2 hours ago, SwampD said:

I love this stat. It would be perfect if it also factored in the quality of the shooter and goalie (it is a league average, so it doesn't differentiate between Samuelson shooting on Bobrovsky from Crosby shooting on Houser from the same spot), but it's close enough.

I'd really be curious to know what kind of data actual NHL teams have. It wouldn't surprise me if there were teams that had data similar to this. It's all about the budget and how much you want to buy in to the math / analytics vs the artform of the eyetest.

From the looks of it, it seems this offseason the Sabres were looking for value on under represented point / expected goal players that might have had tougher assignments. McLeod, Zucker, Aube-Kubel, Lafferty, and Malenstyn all fell under this umbrella. With our cap space, I'm not sure we needed to penny pinch like we did, unless a deal on a top 6 fell through and we don't know about it.

Edited October 21 by JoeSchmoe

[SwampD]

9 minutes ago, JoeSchmoe said:

I'd really be curious to know what kind of data actual NHL teams have. It wouldn't surprise me if there were teams that had data similar to this. It's all about the budget and how much you want to buy in to the math / analytics vs the artform of the eyetest.

From the looks of it, it seems this offseason the Sabres were looking for value on under represented point / expected goal players that might have had tougher assignments. McLeod, Zucker, Aube-Kubel, Lafferty, and Malenstyn all fell under this umbrella. With our cap space, I'm not sure we needed to penny pinch like we did, unless a deal on a top 6 fell through and we don't know about it.

It didn't. They just pinch their pennies.

[LGR4GM]

5 hours ago, SwampD said:

I love this stat. It would be perfect if it also factored in the quality of the shooter and goalie (it is a league average, so it doesn't differentiate between Samuelson shooting on Bobrovsky from Crosby shooting on Houser from the same spot), but it's close enough.

Team or private models could get that specific. I don't think public models do but some may have a little of that. You'd have to track each player and where they score from and also where goalies get scored on. The issue is goals are rare events so there might be too much noise but that's above my understanding of stats or large data models. 

[SwampD]

19 minutes ago, LGR4GM said:

Team or private models could get that specific. I don't think public models do but some may have a little of that. You'd have to track each player and where they score from and also where goalies get scored on. The issue is goals are rare events so there might be too much noise but that's above my understanding of stats or large data models. 

It doesn't really matter anyway. There is only one stat that matters, and the Sabres have have been substandard at that for a long f'n time.

[JoeSchmoe]

Here's the update for 5 on 5 expected goal % this season after last night's game. 

Not sure what I did for the first one but I qualified this by min 30min played (it's a slider and hard to get 30min exactly so it might actually be 29min).

 

[sabresparaavida]

On 10/21/2024 at 1:42 PM, mjd1001 said:

So, it helps, in the offensive end, explain who gets better chances or who helps generate better chances. However, for someone like Cozens, it doesn't. We know he is still getting good chances, just over the long run, he is under a 10% shooter. 

So you kinda have to combine this with how good a shooter is (or how good of shooters their linemates are) to get an accurate picture, at least in the offensive zone how good they are, or will be.

To be fair, it probably still works well for Cozens considering shots have to be on net to be counted for the xGF. 

[JoeSchmoe]

Update. A couple new takeaways...

I think I'd sit Lafferty once Aube-Kubel comes back.

Nice to see Clifton go from the near the cellar last year to where he is now. I didn't think he was great last year, but shedding Eric Johnson and his so called veteran "knows how to win" presence has helped him greatly.

[JoeSchmoe]

I haven't updated this in a while and thought I'd post.

I threw the last xG% chart in the Byram thread so I won't repost that.

Here's the team 5 on 5 xG%. We're still clipping along at a bubble playoff pace. If Cozens and Quinn ever get it together, and the goaltending does their part... Maybe, just maybe...

[mjd1001]

On 10/20/2024 at 4:39 PM, JoeSchmoe said:

 

That said, we know our guys pretty good, so we can make some sense of the numbers... Here's what I see.

Peterka, Dahlin, and Thompson are our best players and that's clearly seen here.

Power and Byram are low. Byram sucked in this regard on COL. Are they actually getting the tough matchups? Last year Power was actually pretty good so it's surprising to see him that low.

Cozens isn't as bad as advertised, but still has to improve.

Lafferty, Malenstyn, and Krebs are low, but would their defense assignments justify this?

Ok question on this stat...does it take into account shooting percentage?

From what I understand its about the chances you have and how good those chances are. But some players, you can give them great chance after great chance and if they aren't good shooters, it doesn't matter.

If I'm understanding this correctly, someone like Cozens who is a very bad/lower percentage shooter might be over-rated by this stat. He is getting chances, but if he (or his teamates) don't put them in, it doesn't matter much?  For the Forwards at least, I would think this state might be more valuable if there was a version of it that incorporated shooting percentage into it.

Like maybe take this state, and multiply it by the expected shooting percentage of a player or line (use their lifetime numbers) and that may be a more an indication of how they are playing?  Benson can be near the top of the list and Cozens in the middle, but if they are naturally guys who shoot 50% lower than someone like Tage or Peterka, than wouldn't it be better to evaluate the overall play with Tage and Peterka being even 50% higher than they are (relative to Benson and Cozens)?

I'm asking because, I'm not sure exactly how this stat is generated..but if my guess is correct I would think it over-values guys like Krebs and Cozens offensive performance.

[LGR4GM]

6 minutes ago, mjd1001 said:

Ok question on this stat...does it take into account shooting percentage?

From what I understand its about the chances you have and how good those chances are. But some players, you can give them great chance after great chance and if they aren't good shooters, it doesn't matter.

If I'm understanding this correctly, someone like Cozens who is a very bad/lower percentage shooter might be over-rated by this stat. He is getting chances, but if he (or his teamates) don't put them in, it doesn't matter much?  For the Forwards at least, I would think this state might be more valuable if there was a version of it that incorporated shooting percentage into it.

Like maybe take this state, and multiply it by the expected shooting percentage of a player or line (use their lifetime numbers) and that may be a more an indication of how they are playing?  Benson can be near the top of the list and Cozens in the middle, but if they are naturally guys who shoot 50% lower than someone like Tage or Peterka, than wouldn't it be better to evaluate the overall play with Tage and Peterka being even 50% higher than they are (relative to Benson and Cozens)?

I'm asking because, I'm not sure exactly how this stat is generated..but if my guess is correct I would think it over-values guys like Krebs and Cozens offensive performance.

Xgf would be calculated by the average chance of a goal from that spot on the ice.  So no, it wouldn't take into account individual sh skill but instead uses league averages. Auston Matthews will out perform his xgf% and Cozens underperform. The percentage comes from taking shots given v. Shots taken. So Buffalo is giving up slightly better chances than they get as a team... considering Cozens line is terrible, that's about right. 

[JoeSchmoe]

2 hours ago, mjd1001 said:

Ok question on this stat...does it take into account shooting percentage?

From what I understand its about the chances you have and how good those chances are. But some players, you can give them great chance after great chance and if they aren't good shooters, it doesn't matter.

If I'm understanding this correctly, someone like Cozens who is a very bad/lower percentage shooter might be over-rated by this stat. He is getting chances, but if he (or his teamates) don't put them in, it doesn't matter much?  For the Forwards at least, I would think this state might be more valuable if there was a version of it that incorporated shooting percentage into it.

Like maybe take this state, and multiply it by the expected shooting percentage of a player or line (use their lifetime numbers) and that may be a more an indication of how they are playing?  Benson can be near the top of the list and Cozens in the middle, but if they are naturally guys who shoot 50% lower than someone like Tage or Peterka, than wouldn't it be better to evaluate the overall play with Tage and Peterka being even 50% higher than they are (relative to Benson and Cozens)?

I'm asking because, I'm not sure exactly how this stat is generated..but if my guess is correct I would think it over-values guys like Krebs and Cozens offensive performance.

Obviously my initial post was TLDR. I cover that in the post.

"The stat does not track shooting skill, so players like Thompson, Olofsson, Matthews, etc will be underrepresented by this stat given they do a better job of putting pucks away"

[SwampD]

On 11/7/2024 at 4:29 PM, mjd1001 said:

Ok question on this stat...does it take into account shooting percentage?

From what I understand its about the chances you have and how good those chances are. But some players, you can give them great chance after great chance and if they aren't good shooters, it doesn't matter.

If I'm understanding this correctly, someone like Cozens who is a very bad/lower percentage shooter might be over-rated by this stat. He is getting chances, but if he (or his teamates) don't put them in, it doesn't matter much?  For the Forwards at least, I would think this state might be more valuable if there was a version of it that incorporated shooting percentage into it.

Like maybe take this state, and multiply it by the expected shooting percentage of a player or line (use their lifetime numbers) and that may be a more an indication of how they are playing?  Benson can be near the top of the list and Cozens in the middle, but if they are naturally guys who shoot 50% lower than someone like Tage or Peterka, than wouldn't it be better to evaluate the overall play with Tage and Peterka being even 50% higher than they are (relative to Benson and Cozens)?

I'm asking because, I'm not sure exactly how this stat is generated..but if my guess is correct I would think it over-values guys like Krebs and Cozens offensive performance.

On 11/7/2024 at 7:08 PM, JoeSchmoe said:

Obviously my initial post was TLDR. I cover that in the post.

"The stat does not track shooting skill, so players like Thompson, Olofsson, Matthews, etc will be underrepresented by this stat given they do a better job of putting pucks away"

And this.

On 10/21/2024 at 4:28 PM, SwampD said:

I love this stat. It would be perfect if it also factored in the quality of the shooter and goalie (it is a league average, so it doesn't differentiate between Samuelson shooting on Bobrovsky from Crosby shooting on Houser from the same spot), but it's close enough.

 

[JoeSchmoe]

Update and some thoughts...

Tage is far and away the class of the team. Keep in mind a player with his shooting skill is going to far outperform his expected goals, which is the best on the team by a good amount.

After some good games Cozens, Byram, and Quinn are improving greatly. They are still at the bottom, but their relative numbers have improved closer to the mean. Byram was -15% when I started this.

I kept the minimum minutes at 50min so he's not on here, but Gilbert is -24.9% relative to the team. I know people like that he plays tough but we can do better with other guys on the roster.

Edited Saturday at 04:02 AM by JoeSchmoe