Login

Slowpoke · 01-03-2018, 07:35 PM

Code:
Ready for grading. Written Word Count: 1319 Spreadsheet: 1642 Total Word Count 2961

Iâ€™m back with another overly detailed research project. Last offseason I took a look at hits, and how they are perceived. Some of the discussions sparked a bit of talk about what mechanics qualify a turnover, and possibly if good teams do this better than bad. I have also been a bit frustrated since I joined the league as there is no mechanism to track the turnover stat in the sim. This season I wanted to track the individual hits, success rate, and turnovers that resulted from the hits. I chose two players who I felt would be near the top, if not at the top of the hits list. I selected Jonathan Lundberg and Maximilian Wachter, as well as myself (because I like to know how to get better) and Zach Evans. I chose Evans because last season I labeled him as the prototypical physical player as he makes a large number of hits in comparison to the number of penalties he takes.

I tracked how many hits each player had for each game of the season, how many times the hit resulted in the skater losing the puck, and finally how frequently the loose puck was picked up by the defending team, or the team that made the hit. I had a couple ideas on what kind of information I would see. First, good teams would recover the puck more often. A discussion with my team in discord led me to make the assumption that Skating and Defense were the stats that determined who would recover a loose puck. With that it would be safe to say that the better team, or possibly the team with the better line on the ice would recover the puck more frequently.

Now to take a look at what I found. Jonathan Lundberg finished first in hits this year with 229 hits. Of those hits 131 resulted in the skater losing the puck, I called this a productive hit, this resulted in a 57.2% Productive Hit Rate. Of those loose pucks, 65 were recovered by the defending team. Using the same math here are the rankings for each stat:

<div align="center">Hits
Jonathan Lundberg 229
Vasily Horvat 203
Maximilian Wachter 186
Zach Evans 151

Productive Hit Rate
Zach Evans 65.6% (99)
Vasily Horvat 65.5% (133)
Maximilian Wachter 59.7% (111)
Jonathan Lundberg 57.2% (131)

Hits Resulting in Turnovers Rate
Vasily Horvat 37.4% (76)
Maximilian Wachter 36.6% (68)
Zach Evans 34.4% (52)
Jonathan Lundberg 28.4% (65)</div>

This next ranking is a look at how well the hitting team was at recovering the puck. Ranked in order is the percent of loose pucks recovered by the hitting team only for the hits performed by these players. With that, the numbers may be off a bit:

<div align="center">Free Puck Recovery Rate
Los Angeles Panthers 61.3%
Texas Renegades 57.1%
Winnipeg Jets 52.5%
Seattle Riot 49.6%</div>

With that said I would love to be able to show you how every team does with recovering loose pucks, but that would be a pain to do as I would have to sort through every sim. What I can say is that the results were a little surprising. Winnipeg was hands down my pick for the team that recovered the most pucks as they were the best team this season, in my opinion.
This is the point where I thought I was done, but it turns out I wasnâ€™t. Once I had this data I wanted to know how it measured up to the other mechanisms to create turnovers. As a recap, I consider hits, blocked shots, and intercepted passes as ways to create a turnover. There are loose pucks created after a shot misses or a goalie is unable to gather the rebound, but I don't consider that a turnover as a scoring chance is what created it, and a turnover for me is the ability to stop a team from having a scoring chance. With that said, I went back into the sims and looked at the three top defenders in terms of blocking shots. Jasper Clayton, Adam Kaiser, and Zander Rhys were those defenders who topped the list. I donâ€™t know if I could say that I had any idea what the data would show, but I did end up being surprised, so let's look at it.

<div align="center">Blocked Shots
Jasper Clayton 93
Adam Kaiser 91
Zander Rhys 90

Blocked Shots Resulting in Turnover
Jasper Clayton 19.4% (18)
Adam Kaiser 18.7% (17)
Zander Rhys 14.4% (13)</div>

I expected the loose puck generated by the block to be recovered by the defending team at a similar rate as hits, but I was wrong. Very wrong. I donâ€™t know what is creating this almost 30% difference, maybe the fact that a blocked shot is generated from the defensive zone where players are in position and less likely to go after the puck? Another weird thing I saw, but didn't recognize until to late was that a large portion of these loose pucks that were recovered by the defending team, were recovered by the defender who blocked it.

With that done I was left with how to calculate intercepted passes. What I ended up doing was to use a random number generator to provide 15 numbers between 1 and 350 which I used to pick 15 games. I then use the search function to find the number of â€œpass toâ€ and â€œpass intercepted byâ€ to find the total number of passes attempted in a game, and how many were intercepted. The totals were way beyond what I expected and way beyond the numbers you would see in a normal NHL game. In total between the 15 games there was a total of 830 passes completed with 223 being intercepted. This accounted for the lion's share of turnovers in your average SHL game, which in reality is not surprising and showed that 23.3% of all passes ended in an interception.
At this point I stopped looking at the type of turnover and just classified each type of hit, block, and pass as an event. With that done I was able to calculate the total number of events that occurred throughout the SHL season, and then figure out which is the highest percentage chance. At this point I went back and combined the data for hits and blocks and made a general percentage for turnovers. I know that I used different sample sizes for each, but I think that by using the top 3 in blocks, and 4 players that were a bit random for hits it gave me a good idea about the mean statistics for each. By looking at the team stats for hits and blocked shots I was able to apply the percentages that I found for hits and blocks to those totals.

<div align="center">Total SHL Hits: 13507
Total Productive Hits: 8326 (61.6%)
Hits Resulting in Turnover: 4584 (33.9%)

Total SHL Blocked Shots: 6514
Blocked Shots Resulting in Turnover: 1141 (17.5%)

Total Passes Attempted: 379031
Passes Completed: 290500 (76.6%)
Passes Intercepted: 88550 (23.4%)

Total Events: 399052
Total Turnovers: 94276
Turnover %: 23.6%
Turnover by Hit: 4.9%
Turnover by Block: 1.2%
Turnover by Interception: 93.9%</div>

This is the most impressive information that I have found, and in the end gives merit to the current state of the league which places and emphasis on the Defense stat, almost over every other stat. With only 4.9% of all turnovers being created by a stat other than Defense it's easy to see why. With that said, I am open to finding out if anyone has any questions regarding the data, or even if someone has something they want me to look into as I try to do one of these every offseason because nobody wants to hire me as a banker.

<div align="center">
Data Here</div>

Atlas · 01-03-2018, 08:22 PM

Very good analysis, more of this would be great

Do you have mean standard deviation and etcetera

***WannabeFinn*** · 01-03-2018, 08:26 PM

Less than 5% of turnovers are caused directly by body checks.. nice

Slowpoke · 01-03-2018, 10:30 PM

Quote:Originally posted by Atlas@Jan 3 2018, 07:22 PM
Very good analysis, more of this would be great

Do you have mean standard deviation and etcetera

No not really, I thought about it, but got lazy

Beaver · 01-03-2018, 10:49 PM

This is really interesting. I've been doing some some basic tracking in the SMJHL and there's a pretty strong correlation between a player's Defense attribute and their interceptions per 60. I thought about expanding this to all turnovers but looking at those numbers I'm glad I was too lazy to since interceptions account for over 90% of them. Nice work.

Slowpoke · 01-03-2018, 11:31 PM

Quote:Originally posted by Beaver@Jan 3 2018, 09:49 PM
This is really interesting. I've been doing some some basic tracking in the SMJHL and there's a pretty strong correlation between a player's Defense attribute and their interceptions per 60. I thought about expanding this to all turnovers but looking at those numbers I'm glad I was too lazy to since interceptions account for over 90% of them. Nice work.

Thanks! How are you doing the correlation?

Atlas · 01-04-2018, 12:08 AM

Quote:Originally posted by Slowpoke@Jan 3 2018, 09:30 PM

No not really, I thought about it, but got lazy

You're killin me smalls

dankoa · 01-04-2018, 12:10 AM

Dam slowpoke so smart

Beaver · 01-04-2018, 12:44 AM

Quote:Originally posted by Slowpoke@Jan 3 2018, 10:31 PM

Thanks! How are you doing the correlation?

1. Someone way smarter than me held my hand in creating a scraper to pull all interceptions,
2. Exported those as a csv,
3. Translated from a list of turnovers by game to a count of turnovers by player,
4. Dumped player data (attributes to find correlations and minutes played to make the interceptions into a rate stat - TA/60) into the spreadsheet,
5. Compared players' TA/60 to various attributes and plotted that with regression lines.

Unfortunately I didn't have a huge sample for the correlation since I was only interested in Militia players so I'm not sure how reliable it is but two potentially interesting things I found were:
1. Tracking giveaways was a waste of time without pulling all pass attempts and turning it into a success rate stat instead of a counting stat. The best players had high GA/60 rates because they had the puck a lot and passed the puck a lot. This scuttled the original intent of pulling all this data which was to find a turnover ratio for players.
2. The correlations are wildly different by position. For the Militia forwards the relationship between TA/60 and Defense attribute had an R-squared of 0.775 and defensemen had 0.594 but all players in the same sample were 0.434 since the takeaway rates were so much higher for defensemen (a 56 DF defenseman had more TA/60 than an 80 DF forward) that combining all positions fucked up the data. Unfortunately splitting the already small sample into even smaller samples isn't optimal but it suggests that defensemen start out decently at intercepting pucks due to their position but forwards have an easier time improving if they want to invest in it.

Slowpoke · 01-04-2018, 05:46 AM

Quote:Originally posted by Beaver@Jan 3 2018, 11:44 PM

2. The correlations are wildly different by position. For the Militia forwards the relationship between TA/60 and Defense attribute had an R-squared of 0.775 and defensemen had 0.594 but all players in the same sample were 0.434 since the takeaway rates were so much higher for defensemen (a 56 DF defenseman had more TA/60 than an 80 DF forward) that combining all positions fucked up the data. Unfortunately splitting the already small sample into even smaller samples isn't optimal but it suggests that defensemen start out decently at intercepting pucks due to their position but forwards have an easier time improving if they want to invest in it.

This is really interesting. I really want to know how the sim accounts for positioning, and it seems you can back up the idea that positions have inherently set skill values and that the stats users put in only enhance them.

I had thought about building a program to pull in the turnover data, but its a little beyond me and its a ton of work.

Slowpoke · 01-04-2018, 05:51 AM

Quote:Originally posted by Atlas@Jan 3 2018, 11:08 PM

You're killin me smalls

:(

Navigation

Extra Menu

About us