Login

awils13 · 04-18-2019, 10:38 AM

Quote:~1600 word, charts, research

Team corsis

First, let’s take a look at team corsis and find out if there is any correlation between winning and CF%.
I haven’t actually added team corsis to the scraper yet, so for this I just summed shot attempts of all players from each team, which I suspect should be close enough.

Platoon - 53.55
Stampede - 53.11
Specters - 52.46
North Stars - 52.08
Dragons - 51.95
Renegades - 51
Steelhawks - 50.85
Chiefs - 50.77
Rage - 50.77
-------------------------------------------
Blizzard - 49.83
Panthers - 49.56
Wolfpack - 48.93
Pride - 48.05
Barracuda - 46.55
Syndicate - 45.06
Jets - 44.1

*green – made the playoffs.

• As we can see, 7/10 teams that made the playoffs had a positive corsi differential.
• Eventual cup finalists Stampede and Specters ranked second and third respectively.
• Blizzard and Panthers made the playoffs and were fairly close to 50% corsi.
• Renegades missed the playoffs despite having sixth best corsi but they tied with Blizzard and Dragons at 53 points, so they were close to making it.
• Jets and Syndicate were very bad.
• Wolfpack finished with the worst goal differential (-35), having scored only 125 goals, but their corsi was at 48.93. They’re not a good team but they weren’t terrible either. Perhaps, this is just bad sim luck.

The big outliers here are Platoon and Barracuda.
• WKP finished the best corsi differential which suggests they were actually a great team. Indeed, Platoon scored 183 goals (second best in the league behind Chiefs - 184). They’re the only team with a positive goal differential (+11) to miss the playoffs, and with 55 points they would’ve made the playoffs if they were in the Western conference.
What went wrong? Poor goaltending sank this team. Beaujeaux Biscuit had the worst save percentage among starters (0.895), and his backup didn’t play well either (0.855 sv%).
Obviously free agency and regression can change things but right now it seems that if Biscuit can bounce back from his disappointing season, Platoon will be one of the top teams next season.
• The exact opposite for Tampa Bay Barracuda as Benjamin Blue carried them all season. This is the SHL version of Anaheim Ducks.

Corsi vs production

Another thing I wanted to research is correlation between corsi and points (individual).
It seems logical to assume that high corsi == good production and low corsi == low production. Let’s see if this is the case.

First, I excluded players with small sample sizes
< 20 games played
or < 500 shot attempts total

Second, we need to establish a baseline for good production. I suspect that forwards score more than defensemen on average, and indeed:

Defensemen: 20.82
Forwards: 23.72

But the difference is smaller than I thought it would be. However, if we exclude players with less than 10 points, the gap is more significant:

Defensemen: 26.96
Forwards: 33.02

It seems that 25 points for defensemen and 30 points for forwards is a good baseline.

Next, let’s draw some charts!

Forwards

[Image: DWSqw4U.jpg]

The x-axis is corsi for %, the y-axis is points scored.

Not shown here:
Andrew Hawkins (28.98 corsi, 2 points)
Vladimir Vaskov (29.33 corsi, 8 points)
Nolan Snipez (29.81 corsi, 5 points)

All players can be divided into 4 categories:
Good: > 50 corsi, > 30 points (first quadrant)
Overperformers: < 50 corsi, > 30 points (second quadrant)
Bad: < 50 corsi, < 30 points (third quadrant)
Underperformers: > 50 corsi, < 30 points (fourth quadrant)

As you can see, most players (70%) are either in the first or third quadrant. Here are the exact numbers:
146 forwards total
Good: 48 (33%)
Bad: 54 (37%)
Overperformers: 21 (14%)
Underperformers: 23 (16%)

Now let’s look at defensemen

[Image: MygrmL9.jpg]

Not shown here:
Chase Byron (34.14 corsi, 8 points)

96 defensemen total
Good: 29 (30%)
Bad: 28 (29%)
Overperformers: 17 (18%)
Underperformers: 22 (23%)

Again, most players are either in the first or third quadrant, though there is a higher percentage of under/over-performers compared to forwards.

And here’s a chart with all skaters

[Image: 8bBxxqr.jpg]

So yes, in general, positive corsi == good production and negative corsi == low production.
Without diving deeper, I can suggest that underperformers are either players benefiting from better linemates or are defensive defensemen.
Overperformers are a more curious case. In real life I’d assume these are guys like Patrick Laine who are great offensively but are liabilities in their own end. But we all know that defense is one of the most important attributes in simon t and it is one of the first things people focus on.
I will choose a few players and try to examine why they are under/over-performing.

Luke Atmey - F - Syndicate (47.73 corsi, 55 points) - 1434 tpe, 85 IPP
Daniel Smeb - F - Syndicate (48.98 corsi, 20 points) – 389 tpe, 33 IPP
Jack Kennedy - F - Syndicate (44.73 corsi, 18 points) - 639 tpe, 43 IPP
Fredrich Koenig - D - Syndicate (45.08 corsi, 34 points) - 925 tpe, 64 IPP
Mikhail Petrikov - D - Syndicate (45.42 corsi, 17 points) - 614 tpe, 34 IPP

This is the Syndicate top line & top pairing, at least according to their final lines.
Atmey and Koenig are high-ish TPE players and they can be considered overperformers with 55 and 34 points respectively.
If we look at the IPPs, Atmey is the main source of offense when he’s on the ice and Koenig contributes a lot offensively as well.
I think in this case they just couldn't drag their lower TPE teammates to respectable possession numbers.

Leshaun King - F - Pride (49.3 corsi, 35 points) - 805 TPE, 58 IPP
Richard Metcalf Jr. - F - Pride (43.74 corsi, 20 points) - 740 TPE, 48 IPP
Dionyz Vyskoc - F - Pride (48.82 corsi, 34 points) - 1118 TPE, 65 IPP
Charlie Schieck - D - Pride (46.25 corsi, 37 points) - 1020 TPE, 62 IPP
Isak Odegard - D - Pride (46.54 corsi, 37 points) - 1653 TPE, 64 IPP

SF Pride's top line & top pairing. 3 players over 1000 TPE, offense distributed more evenly.
I'm guessing Metcalf didn't play on the top line all season since his corsi is vastly lower.
Odegard had 18 secondary assists. His 5v5sh% is 10.6 which is higher than average but not significantly. Same thing with Leshaun King.
Perhaps there is some sim luck involved here but that's only around 3-5 extra points.
Ugh I'm not sure. These guys played against other top lines and didn't do well but somebody on this team had to score, I guess?

-------------

It seems that a lot of overperformers are good players on bad teams. Which makes sense.
Now let’s take a look at some of the underperformers.

Nat Emerson - D - Platoon (55.31 corsi, 10 points) - 461 TPE, 18 IPP
Emerson is an easy choice here as he finished with one of the highest corsi in the league but only scored 10 points.
He was a rookie so his TPE was pretty low. And his IPP was extremely low which means he didn't create a lot of offense.
It's pretty obvious that Emerson benefited from playing with high TPE teammates and his corsi is inflated.
Though it's worth noting that he had a 1.46 STL/TO ratio so he did pretty well in his own end, just didn't do much on offense.
But he’s not really an underperformer and his offensive production probably won’t improve much next season.

Quick Mafs - D - Stampede (55.06 corsi, 9 points) - 650 TPE, 35 IPP
Another good defensive defenseman. Put a player like Mafs or Emerson on a bad team, and their corsi will plummet. But they’re definitely valuable assets on a contending roster, they just need to be surrounded with good offensive players.

So far it seems that majority of players in this category are defensive defensemen on good teams. Let’s take a look at a couple of forwards.

Hippo Passamus - F - Stampede (55.02 corsi, 25 points) - 921 TPE, 33 IPP
Passamus definitely benefited from playing with 2 50+ point players in Marius and Konig but he was a good compliment to their builds.

Alex Andani - F - Specters (53.91, 17 points) - 738 TPE, 63 IPP
This is an interesting case. Could be the first actual underperformer on this list.
Didn't play great defensively (0.73 stl/to) but Specters' 3rd pairing limited offensive chances well, which is probably the reason for Andani's high corsi.
But still, his on ice sh% was pretty low and I won't be surprised if he scores 20-25 points next season.
He didn't play on the PP at all, though, so his upside is capped if that doesn't change.

-----------

In conclusion,
It seems that In most cases corsi correlates pretty well with individual production.
High corsi/low production = usually means mediocre players on good teams and/or defensive defensemen.
Low corsi/high production = usually means good players on bad teams.

But I don’t think corsi can be used to predict future production. I’m also not sure if luck is as much of a factor in simon t as it is in real life. We've seen Nick Foligno score 73 points out of nowhere or William Karlsson score 43 goals on 23.4 sh%, which are a couple of examples of extreme puck luck, but in simon t it seems that players mostly perform as expected, at least there aren't any major fluctuations.

I also realized by the end of the article that I should've excluded power play points because corsi is calculated at even strength only. Well, too late. But I don't think it skewed the results since more points are scored at even strength

UPDATE: without PP points

[Image: eAuTL9h.png]

.bojo · 04-18-2019, 10:44 AM

i didn't need an article to tell me i suck, i know that already.

Mayuu · 04-18-2019, 11:43 AM

Yeah, somehow Marius and König managed to carry a Hippo

RETIRED

Tomasnz · 04-18-2019, 11:56 AM

I think for the production scatters I'd love to see some regression lines added.. to my eye the forwards has a greater degree of collerlation than the defensman. So maybe is more predictive for forwards?

Also would be interesting to see if removing pp points removes some over performers.

Great article.

awils13 · 04-18-2019, 04:11 PM

04-18-2019, 11:56 AMTomasnz Wrote: I think for the production scatters I'd love to see some regression lines added.. to my eye the forwards has a greater degree of collerlation than the defensman. So maybe is more predictive for forwards?

Also would be interesting to see if removing pp points removes some over performers.

Great article.

Added at the end of the article

Definitely a greater correlation for forwards because a lot of defensive defensemen are in the underperformers category.

Tomasnz · 04-18-2019, 04:45 PM

04-18-2019, 04:11 PMaaronwilson Wrote:
04-18-2019, 11:56 AMTomasnz Wrote: I think for the production scatters I'd love to see some regression lines added.. to my eye the forwards has a greater degree of collerlation than the defensman. So maybe is more predictive for forwards?

Also would be interesting to see if removing pp points removes some over performers.

Great article.

Added at the end of the article

Definitely a greater correlation for forwards because a lot of defensive defensemen are in the underperformers category.

Wow .that's amazing how much a difference pp makes to the scatters

FlappyGiraffe · 04-18-2019, 05:59 PM

04-18-2019, 10:38 AMaaronwilson Wrote: Not shown here:
Andrew Hawkins (28.98 corsi, 2 points)
Vladimir Vaskov (29.33 corsi, 8 points)
Nolan Snipez (29.81 corsi, 5 points)

hey thats the Jets third line Surrender

Navigation

Extra Menu

About us