Mid-Season Team Statistics Reviewby Eric Schneider | February 1, 2018, 3:46 PM ET
As we reach the mid-season point and collectively hold our breath for the announcement of the 2017-18 SICHL All-Star Game, one undeniable trend reveals itself: teams that score more goals than they allow tend to win more games.
I know, stop the presses.
So, looking beyond that groundbreaking observation, what else can we learn from our myriad of team statistics? I've made tremendous use of my working hours (if I do say so myself) to compile league-wide statistics, adjusted for games played and relative league position, to try and decipher which stats matter and which are just balogna.
The single best predictor of team succes: Goal Differential. Duh. The standard deviation (SD) for teams ranked by goal differential vs. their current place in the standings is just 2.49, which makes it far and away the most accurate judge of a team's abilities. The Norsemen and Canons buck the trend most by sporting goal differentials 5 positions lower than their current place in the standings; on the flip side, the Mustangs are 9 spots lower than they ought to be based on their goal differential. Tough break, Matt. Most teams though are dialed in right where you would expect, and as the season stretches on you can probably expect the standings to stratify even more based on the ol' GF/GA.
The next best predictors: Goals Against. With an SD of 5.48, GA relative to league position pegs more than half of the league in roughly the right spot. This will make a bit more sense later when you see how unreliable shooting percentage is as a predictor, but the gist is that defence is a more reliable predictor of success this season than offence.
This brings us to the last clump of reliable stats, and they're worth talking about as a group: Shot Differential, and Goals For. With SD's 5.66 and 6.37, respectively, you start to see the value in controlling play compared with simply trying to outscore your opponent. The teams that get more shots on net and allow fewer against, regardless of the actual goals, are more consistent in the standings than those that score in bunches.
Now for special teams. Special Teams % is the key figure here at 8.34 SD, while Power Play % and Penalty Kill % both varied wildly. It's clear that even strength is what really matters. Boston, Vancouver Island, Las Vegas, Reykjavik, and Ottawa are all double-digits ahead of their ST% in the standings, while a number of bubble teams and bottom-feeders are shockingly adept.
Lastly, we have the outliers: Penalty Minutes and Shooting %. Neither of these seem to matter independently as standard deviations are around 10, but both become more interesting when used in combination with other statistics. I averaged team rankings in PK% with inversed PIM rankings and the league's top-10 teams were right there at the top, just as you would expect. Basically, PIMs don't matter if your PK is good. Likewise, SH% doesn't matter as long as your shot generation is strong. Where it becomes interesting is if you take an Adjusted GF based off SF/GP, then multiply using the current league-average SH% of 10.30% -- the top half of that range is generally competitive, but teams like Kansas City and Edmonton are clearly on the wrong end of the equation. If those underperforming teams had even league-average SH%, they'd be 10 spots higher in the standings. By that token, teams riding high SH% with relatively few SF, like Acadia, Calgary, and Denver, may be due for a correction. The SD on this AdjGF statistic is much stronger at 5.8 -- it'll be interesting to see if that levels out towards the end of the season.
The takeaway from all this? Goal differential is just as influential as one would expect and shot differential slightly less so; special teams don't really matter unless you take a ton of penalties and likewise PIMs don't really matter unless your PK sucks; and shooting percentage is super-wacky and barely matters at all unless you're on the extreme ends of the range or your SF stinks. That said, half a season is probably too small a sample size to get meaningful results, and the league changes so drastically from year-to-year that it may be pointless to expand it. It's also worth noting the other potentially significant flaw is that I used overall position in the standings to check relative strength instead of working off win percentages. League position changes dramatically every day, whereas win% is a bit subtler and more accurate. ROW % would be even better, as SO's unfairly skew statistics. I just couldn't figure out a way to make use of that without it becoming an apples & oranges comparison.
Thanks for reading! Check out this Tweet to see how the teams were ranked in each category as well as the SD breakdown: