Simulation Hockey League

Ready for grading (2420 words)

Now that preseason is done, that means it’s once again time for me to bombard everyone with graphs like I do. I’ve made a lot of graphs regarding team and individual player performance in the SMJHL during the preseason, and I wanted to do a quick recap of how the league did. I mostly plan on showing the graphs without going into excessive detail explaining them or the results, because my hope is that the graphs will do a decent job of being self-explanatory, while also leaving the results and analysis open for interpretation with what the graphs show.
 
One huge preface that I want to get out of the way before I explain the data, is that yes, I know it’s preseason and teams are still experimenting around with different strategies, as well as being nice to rookies and giving them perhaps more play time than they’ll get in the regular season. But I can’t analyze the data based on hypothetical optimal rosters and strategies, I just have to go with what we’re given. So this analysis will be based purely on the success and outcomes of teams and players, but that does not mean it’s representative of what their successes will look like in the regular season. I also know that with some teams and some players, there was certain issues with build imports into FHM. But again, I just have to go with what’s on the index. With that aside, without further ado, let’s move on to the first section of graphs.
 


Team points per game:


                     



These first two graphs simply show each team’s place in the standings after each game they played. The first graph shows the cumulative preseason points, while the second graph shows how far ahead/behind a team is relative to the league average points per game. Remember, because of overtime and shootouts, the league average is not 1 point per game, it ended up being 1.06 for the preseason. In my opinion, the interesting takeaways from these graphs are the success of the Berserkers, and the lack of success of some teams I thought may be favorites. You’ll see in a few graphs that overall, the Berserkers have a ton of rookies, and thus low TPE across the board in all stats, so it was surprising to see them sitting in 2nd, and that’s coming from a member of said team. On the flip side, Detroit, and Colorado both underperformed from what I was expecting, which was top 5 for both teams. But in the end, both ended up below the league average points per game pace. Carolina was another powerhouse last year, however I believe lost a good amount of high TPE players, and thus struggled heavily this preseason, finishing in dead last. Anchorage had a great showing in the preseason, winning all but their first game against the aforementioned struggling Kraken, and they’ll hope the preseason is an accurately indication of what’s the come in the regular season.
 



Team ratings:

 
                    


These two graphs depict how highly rated each team is in certain categories compared to the rest of the league. These ratings aren’t based on TPE purely, but rather how each player spent their points and upgraded each ability, basically their player build. The first graph, the polar chart, shows how highly rated each team is on average in the 4 skater categories on the player builder, as well as average goalie ratings from their builds. There’s a lot to digest in this graph, so I have to address a few points. First and foremost, after I averaged every player’s build stats for each team, I ranked each team by this average value. The bigger the slice of the ‘pie’ is, the more highly ranked they were. So teams like Detroit, Kelowna, and St. Louis all have very highly ranked players in all categories across the board. Second, if you remember from your build, there was only like one or two stats you could actually upgrade in the ‘mental’ category, so I view that ranking as perhaps not as important as the other 4. Third, this isn’t taking into account position at all, it’s averaging every skater on each team; because some teams may have offensive minded defenseman, or forwards that put a lot of points into defensive categories. And finally, the average goalie ratings is normalized based on goalie playing time. I felt like it would be unfair to rank a team with a 425 TPE and 155 TPE goalie lower than a team with two 300 TPE goalies, because the 425 TPE goalie would be playing as many games as possible. So goalie ratings are normalized to goalie play time. Based on team ratings, Detroit, Kelowna, and St. Louis look to be the most talented rosters in the league. I think Kelowna’s point distribution might be the most favorable of the three, because they have high physical and goalie rankings, while only missing out on mental ratings. Detroit has an incredibly talented roster, but one of the worst goalie ratings, and a bad goalie is sometimes all it takes to lose games. St. Louis looks to be very well rounded, however I’ve heard physical stats are proving to be very important in FHM, so lacking in that category may hurt them. Although you’d think LPLL alone would make up for it. On the other end of the spectrum, Vancouver, Newfoundland, and Carolina are all filled with rookies and low ranked rosters, and may be off to a slow start while their rookies slowly update and continue to improve. The other graph in this section breaks down team’s skaters by position. Each skater is given an average rating, based on all their individual stats averaged together. The curve on the graph are colored by position for each team, and represents the distribution of these average rating values. Basically, a taller curve means that a team has a higher percentage of players at that corresponding rating of the x axis. So a team like Anaheim, has such a huge abundance of defenseman rated around 9.5 average overall, that they actually messed up the y axis for the rest of the league. The reason I made this graph as well as the previous was to separate teams out by who’s unbalanced offensively or defensively, or which teams are even. It looks like Vancouver and Anaheim have the biggest discrepancy between the talent of their forward and defensive cores, while the previous mentioned Knights have a really well balanced and high rated team, furthering their case for one of the favorites this year.


 
Team performance:





Many of you have already seen these next graphs since I made them a lot last season. The first graph is an attempt at all all-encapsulating performance and possession review for the teams. The x and y axis represent corsi for and against, the color of the dots represent PDO (shooting percentage + save percentage), and the size of the dots correspond to the team’s goal differential. The two teams on opposite ends of the corsi spectrum are Detroit and Carolina. The Kraken really struggled in the preseason, with 475 more shot attempts against than for, in just 7 games. On the other end, the Falcons put on 469 more shot attempts than their opponent throughout the preseason. However, like I mentioned before, low rated goalies could be a killer for some teams, and even though they had the best corsi ratio in the league, they were reworded with the lowest PDO and 3rd lowest goal differential, despite dominating the possession metrics. On the other hand, the Berserkers found themselves in 2nd place in the preseason, despite a negative corsi differential, because our goalie single handily stole us a few games (more on him later). Colorado looks like they’ll have fun games to watch, as both teams in those matchups will put an above average amount of shot attempts on goal, possibly making for high paced affairs, while Maine is the opposite, with both teams in those matchups taking a below average amount of shot attempts per game. Moving on, the bar graph displays the scoring for each team, and the scenario in which their goals came from. Anchorage led the preseason in points and did the same for goals, with Newfoundland coming in second. The difference between the two came from powerplay scoring, as both scored an equal amount of even strength goals. Last year Detroit eliminated the Berserkers in 4 games with something like a 43% powerplay percentage, so it seems like they used up all their special teams luck as they finished the preseason with the least amount of powerplay goals scored. Another team whose numbers jump out are the Knights. I’ve said Kelowna should be favorites a few times so far in this piece already, however they seem to be scoring a much larger percent of their goals on the powerplay than even strength. They could have an elite powerplay, but the relatively low even strength scoring could be a concern in their man advantage stats ever regress to the league average. Lastly, the final graph is this section shows the team goal differential on a game by game basis.  There’s not much use of going over team by team, but one thing I noticed is that it looks like there’s a pretty big abundance of 3+ goal victories. 4 out of 7 of the Scarecrow’s, Whalers, and Kraken’s games were decided by a difference of 3+  goals or more. A crazy 6/7 of Maine’s game were as well. It’s likely due to transitioning to FHM and figuring it out optimal strategies, as well as maybe some rookies goalies getting preseason starts, but for now we get to enjoy some blow outs.
 


Individual player rankings:
 
The second part of this media article will just be throwing a ton of ‘top 10’ graphs at you, for a bunch of different skater and goalie stats. For skater stats, the graphs are the top 10 forwards, and top 10 defenseman for each category. There won’t really be too much to explain for each graph so I’ll keep it brief in each section. As a note, the list is top 10, but I didn’t make a case for if the #10 is a tie with other people. For now, if there is a tie for number 10, only 1 lucky player of that tie gets their name on the graph.
 


Player production:




 
Only thing I think worth noting in this section is that Jukka Timonen guy looks like a pretty good offensive defenseman. Looks like Anchorage, Kelowna, Main, St. Louis, and Newfoundland were all pretty equally represented in the goals, assists, and points top 10 rankings.   
 


More player production:


Corsi:



 
No surprises here, when a team has such a dominate corsi relative to the league, they’ll have a lot of individuals with the best corsi as well, which is why Detroit has such a high presence in the corsi top 10 list, for both forwards and defenseman. On the other hand, good skaters on teams with lower team corsi have a strong potential to be highly rated in relative corsi, which is why we see a good amount of Vancouver, Carolina, and some Colorado and Maine of these charts.


 
Miscellaneous stats:


I wasn’t sure if hits and shots blocked both warranted their own graph, however I do feel like they’re important and unsung stats, so I combined them into one rating that I called ‘grit.’ I don’t think anyone is surprised to see (clean) Andrei run away with this one for the forwards. I mean the guy had 14 hits in one game. Also no surprise that LPLL is highly ranked for defenseman, probably from hits alone, however Kalashnikov’s balanced 24 hits with 17 shot blocks ends up with him being first in the league. The second stat I created was a possession stat, net takeaways. It’s a rating of takeaways – giveaways. Surprisingly, no one in the league ended up with more than a difference of 8. Seems like players are stealing and giving away the puck at an equal rate across the board.
 


Goalie Stats:


 
These stats are filtered by any goalie that played more than 100 minutes in the preseason, basically more than one and a half games. It might be a little too inclusive, but oh well. For those of you who don’t know what goals saved above average is, it’s a measure of how many goals a goalie saved their team, compared to if they had the league average save percentage. Goalies with high save percentages, or on bad teams that let up a lot of shots will have high GSAA. Overall, I’d say that Kavanagh and Doyle were the best goalies in the preseason by pretty much every metric, but I also want to give a well deserved shoutout to Newfoundland rookie goaltender Admunsen. He got immediately thrown into the starter role on a team with overall terrible defense, and performed incredibly admirably, and I would personally and biasedly put him as the number 3 goalie this preseason.
 


Team preseason ranking:
That wraps up all my graphs. The very last thing I did was to try and make a formula for ranking the teams. It’s based mostly on performance, and weighted by different categories. It’s weighted most heavily by team points, followed by team corsi, goal differential, the latter of the categories are weighted much lighter than points. The goal was to use additional context as tiebreakers. For example, if team ‘A’ had just 1 more point than team ‘B’, but team ‘B’ has a much better corsi and goal differential, you might think team ‘B’ is better despite the 1 less point. It also takes into account the teams average player ratings, to account for teams that are highly rated but for some reason struggled. The rating value is slightly arbitrary, but will mostly range from -2 to 2, with -2 being pretty terrible and 2 being a lot better than the other teams. It also is not a straightforward ranking from 1 to 10. If the leading team had 14 points, but the second best team had 8, the leading team will end up with a higher rating score than if the second team had 12 points. Again, this ranking is mostly based on team performance, which is why Detroit is so low, and it’s also a new formula that I’ll still be tweaking and playing around with. Regardless, here’s the list:

1.  Anchorage (0.96)
   
2.  Kelowna (0.64)
   
3.  Anaheim (0.55)
   
4.  Newfoundland (0.36)
   
5.  St. Louis (0.35)
   
6.  Detroit (-0.14)
   
7.  Colorado (-0.18)
   
8.  Maine (-0.46)
   
9.  Vancouver (-0.77)
   
10.  Carolina (-1.30)
    

A few edited goalie graphs:

                     

Jukka with the most goals and the biggest brain.

excellent work as always

Good article, we for sure over performed and were carried by the first line more or less. I didnt know we had that many even strength goals relative to the league which is also interesting. One thing you might consider, especially on the goalie graphs is using a log scale to actually have the bars mean something since you cant really differentiate the bars with the scale and way you presented the data. There are other ways to plot this that also might be useful visually.

03-28-2020, 02:21 PMgolden_apricot Wrote: [ -> ]Good article, we for sure over performed and were carried by the first line more or less. I didnt know we had that many even strength goals relative to the league which is also interesting. One thing you might consider, especially on the goalie graphs is using a log scale to actually have the bars mean something since you cant really differentiate the bars with the scale and way you presented the data. There are other ways to plot this that also might be useful visually.

Appreciate the feedback! I thought of doing log scales but I figured log(save percentage) would mean even less to people since it would end up losing the values that people are used to with usual save percentage. I added a few edited goalie graphs that I'm hoping do a better job. I think moving forward, the best way is to just edit the scales to be focused around just the 10 goalies in the graph, as opposed to being from 0 to 1. As well as do that for all the graphs. Stuff like the PDO and corsi ones could definitely benefit from that as well.

Nice work right there, Im still trying to paste a picture!

03-28-2020, 02:54 PMSmalinowski7 Wrote: [ -> ]

03-28-2020, 02:21 PMgolden_apricot Wrote: [ -> ]Good article, we for sure over performed and were carried by the first line more or less. I didnt know we had that many even strength goals relative to the league which is also interesting. One thing you might consider, especially on the goalie graphs is using a log scale to actually have the bars mean something since you cant really differentiate the bars with the scale and way you presented the data. There are other ways to plot this that also might be useful visually.

Appreciate the feedback! I thought of doing log scales but I figured log(save percentage) would mean even less to people since it would end up losing the values that people are used to with usual save percentage. I added a few edited goalie graphs that I'm hoping do a better job. I think moving forward, the best way is to just edit the scales to be focused around just the 10 goalies in the graph, as opposed to being from 0 to 1. As well as do that for all the graphs. Stuff like the PDO and corsi ones could definitely benefit from that as well.

ya that was my other thought. or graphing the values relative to the average

Always love seeing these pieces that you put out! Good stuff.

Cool stuff as always.

Fantastic article, really enjoyed the read.

Enjoy reading these. Great work!

Manhattan prospects are smart

Thanks everyone! Glad you enjoy the graphs and found them intuitive enough. The goal is to update a few times throughout the season to see how everyone is doing, and I'm of course open to more feedback of better ways to summarize the results!

Super informative, thanks for the charts and info!

2 things
1. in anchorage we now stan scoochie stratton as goalie
2. i'm surprised that even though i led the smjhl in preseason scoring that my advanced stats weren't that good, i guess i was bad at puck possession or something idk