Improved VHL Regression - Goals, Assists, Points

[eaglesfan036 4,603]

If you haven't already read my regression articles before, I suggest you check out my first article to understand what I'm talking about in regards to interpreting the tables, and also feel free to check out my 2nd article as well. I've also linked my data sheet here in case any of you nerds want to check out the excel spreadsheet I used for this data. If you don't care to read through all the detailed math and stuff, just scroll down to the summary of my findings. To quickly summarize how to interpret the tables, look at the probability column, and any attribute under 0.1 (10%) is statistically significant, and the coefficients with the highest number have the strongest correlation (positive or negative).

 

How the Data Was Gathered

 

I looked at every players' season-end attributes and their total stats for each individual regular season, spanning seven seasons from S59 to S65. Because I can't keep track of how players upgrade their attributes throughout the entire season this isn't a perfect system, but it should still be relatively accurate. I deleted all entries from any computer players like CGY LW2 because every bot has the same stats, and I also deleted entries for players who only played like 40 games in a season versus 72. as the lack of games played could skew the results as well. The last time I ran my regression, I was only looking at S63 which was a small sample size of only 76 observations, versus the 570 observations I gathered from all seven seasons, so this regression should be much more accurate. 

 

Goals

 

 

Every statistically significant attribute is significant at the 1% level, with these attributes being defense, experience, faceoffs, passing, puck handling, penalty shot, and scoring.

 

As Expected

 

• Scoring unsurprisingly has the strongest positive correlation to goals scored
• Defense has the 2nd strongest positive correlation, which makes sense as winning puck battles helps your team have more puck possession, resulting in more goals
• Passing has a negative correlation to goals scored, because you miss 100% of the shots you don't take - Wayne Gretzky - Michael Scott
• Puck handling has a positive correlation to goals scored because better dekes help you keep possession of the puck and get into high danger scoring areas
• Faceoffs has a positive correlation to goals scored because winning the faceoff helps your team possess the puck more, and players with high faceoffs most likely have high tpe in general

 

Surprises

 

• Experience has a decently strong positive correlation to goals scored, but I think this is because players with high experience have high tpe totals in general
• Penalty shots has a decently strong negative correlation to goals scored, and I have no idea why that would be
• To all my skating lovers out there, seven seasons of data shows it has minimal impact on goals scored

 

Assists

 

 

Defense, discipline, and puck handling are significant at the 1% level. Experience, passing, and faceoffs are significant at the 5% level. Skating is significant at the 10% level.

 

As Expected

 

• Faceoffs has a positive correlation to assists scored because winning the faceoff helps your team possess the puck more, and players with high faceoffs most likely have high tpe in general
• Skating has a positive correlation to assists because higher skating helps your player not get knocked off the puck, resulting in higher time on attack

 

Surprises

 

• Defense, not passing, has the strongest positive correlation to assists, which actually might make sense because your team overall having the puck more results in more assists (primary or secondary)
• Discipline has a decently strong negative correlation to assists, and I have no clue how discipline would impact anything other than taking less penalties, other than it just being a tpe waste?
• Passing only has a weak positive correlation to assists, which I would not expect due to it's impact in your player's decision making process (pass, shoot, or skate)
• Puck handling has the second strongest positive correlation to assists, proving once again it's one of the most important attributes
• Experience has about the same positive correlation to assists as passing, but that's likely just because players with the highest tpe have the most experience on average.

 

Points

 

Defense, discipline, experience, faceoffs, puck handling, and scoring are significant are the 1% level. Penalty shots is significant at the 10% level.

 

As Expected

 

• Defense has the strongest positive correlation to points, which actually might make sense because your team overall having the puck more results in more assists (primary or secondary) and more goals scored.
• Faceoffs has a decently strong positive correlation to points, because winning the faceoff helps your team possess the puck more, and players with high faceoffs most likely have high tpe in general
• Scoring has a strong positive correlation to points

 

Surprises

 

• Discipline has a decently strong negative correlation to points, and I have no clue how discipline would impact anything other than taking less penalties, other than it just being a tpe waste?
• Experience has a decently strong positive correlation to points, but I think this is because players with high experience have high tpe totals in general
• Puck Handling has the 2nd highest positive correlation to points after defense, once again showing how important of an attribute it is. 
• Penalty shots has a slight negative correlation to points for some unknown reason, I can't imagine why it would matter other than being a tpe waste
• All my skating lovers, it is not statistically significant in relation to points

 

Summary

 

As a quik recap, in order of importance the following attributes have the greatest impact on these statistics. + means positive correlation, - means negative correlation, ? means I am unsure if totally correct.

 

Goals - Scoring (+), Penalty Shot (-), Passing (-), Faceoffs (+), Defense (+), Experience (+), Puck Handling (+)

Assists - Defense (+), Puck Handling (+), Discipline (-), Skating (+), Passing (+), Experience (+), Faceoffs (+)

Points - Defense (+), Puck Handling (+), Scoring (+), Discipline (-), Faceoffs (+), Experience (+), Penalty Shots (-)

 

Notice I didn't mention strength the entire time? I would think that it contributes towards points by helping your player not get bumped off the puck and by bumping other players off the puck with your checks, but it was not statistically significant in any of these categories. Also, it still looks like skating is not a very valuable attribute, even though a lot of hall of fame players who had tons of tpe in every attribute had lots of skating as well  

 

[eaglesfan036 4,603]

Over 1,000 words - for the weeks ending 4/14 and 4/21

[Renomitsu 934]

Awesome. I would have liked this post just for the 'defense matters' tag but I'm always glad to see statistics.

 

Was this data normalized for ice time? And also, would it be possible to do a similar analysis for hits and shots blocked?

[eaglesfan036 4,603]

5 hours ago, Renomitsu said:

Awesome. I would have liked this post just for the 'defense matters' tag but I'm always glad to see statistics.

 

Was this data normalized for ice time? And also, would it be possible to do a similar analysis for hits and shots blocked?

 

As the VHL currently stands / was, everyone was either in the 1st or 2nd line so ice time is pretty close for everyone being just a few minutes off, but not enough that it would effect the data. I also ran this for points per 20 minutes to normalize for ice time, and the results were essentially the same for as points. Also, I deleted all bots and anyone who didn't play 70 or more games from this regression. 

 

And yes, my next article will cover hits, shots, PIMs, and other stats. Unfortunately the adjusted R squared for shots blocked is really low, meaning that the attributes don't really predict shots blocked, as that is more based on being on a sucky team with lots of shots against.

[Renomitsu 934]

1 minute ago, eaglesfan036 said:

 

As the VHL currently stands / was, everyone was either in the 1st or 2nd line so ice time is pretty close for everyone being just a few minutes off, but not enough that it would effect the data. I also ran this for points per 20 minutes to normalize for ice time, and the results were essentially the same for as points. Also, I deleted all bots and anyone who didn't play 70 or more games from this regression. 

 

And yes, my next article will cover hits, shots, PIMs, and other stats. Unfortunately the adjusted R squared for shots blocked is really low, meaning that the attributes don't really predict shots blocked, as that is more based on being on a sucky team with lots of shots against.

Thanks for following up. Glad to hear you're hitting all the stats, even if the R2 for SB is poor.

 

Great, great work - and super helpful for strategizing GMs!

[eaglesfan036 4,603]

19 minutes ago, Renomitsu said:

Thanks for following up. Glad to hear you're hitting all the stats, even if the R2 for SB is poor.

 

Great, great work - and super helpful for strategizing GMs!

Thank you, and just a note for everyone while I do think people should use this to change how they prioritize some attributes, I don't think you should do something as extreme as completely ignoring upgrading skating or strength, as I'm sure they still have uses that just aren't showing up for some reason

[Fire Tortorella 2,653]

Theory - faceoffs is a "strength" because those with the highest faceoffs have the highest TPE, leading to more goals or assists.

[eaglesfan036 4,603]

15 minutes ago, flyersfan1453 said:

Theory - faceoffs is a "strength" because those with the highest faceoffs have the highest TPE, leading to more goals or assists.

Agreed

[Fire Tortorella 2,653]

Just now, eaglesfan036 said:

Agreed

 

Correlation, causation, pirates, global warming, etc...

[eaglesfan036 4,603]

2 minutes ago, flyersfan1453 said:

 

Correlation, causation, pirates, global warming, etc...

That sounds like a freakonomics reference or is that from something else?

[Fire Tortorella 2,653]

1 minute ago, eaglesfan036 said:

That sounds like a freakonomics reference or is that from something else?

 

[BOOM 8,757]

Probably Jay-Z.