iiinova — where your digital self keeps evolving

Aaron K @ladyotreesyrup.bsky.social · 29/09/2026

Same thing for a fenwick-based 5v4 xG model Log Loss: 0.299, AUC: 0.677, Calibration: 1.001 Power Play xG is much harder, lol

density charts for xG density distributions for each of the last 4 seasons

001

Aaron K @ladyotreesyrup.bsky.social · 28/09/2026

Density distribution for fenwick-based 5v5 xG model trained each new day of games on the previous 1,000 game played to dynamically incorporate changing tactics and event recording behavior Log Loss: 0.203, AUC: 0.777, Calibration: 1.001 Looking forward to tracking performance this season

010

Aaron K @ladyotreesyrup.bsky.social · 22/09/2026

More changes in PBP event recording: starting in 2019-20, the number of events recorded outside of the rink's dimensions drops to almost zero suggesting at minimum some validation for coordinates. Most outside of rink events from previous seasons were from the corners

line chart of share of full coordinate run of play events recorded outside of rink

rink map of events recorded outside of rink

120

Aaron K @ladyotreesyrup.bsky.social · 18/09/2026

2023-24 is a weird outlier in a lot of ways. Over 10% of goals (10/68) from below the goal line being scored from inside the net seems is nearly double any other season and over 3x in raw count. Also completely explains the spike in goals from below the goal line compared to surrounding seasons

line chart of percent of goal line fenwicks recorded inside net

line chart of percent of below line goals recorded inside net

table of goals scored from below the goal line and goals scored from in the net

010

Aaron K @ladyotreesyrup.bsky.social · 17/09/2026

Don't think this is breaking any ground, but some interesting context for behind the net shots/shooting percentages. From spot checking, some behind the net goals are recorded from inside the goal, as in dekes or jam plays, but those aren't filtered here)

110

Aaron K @ladyotreesyrup.bsky.social · 17/09/2026

what did john gibson do to deserve this

Table of own goals against a goalie, John Gibson has the most with nine, next closest are Jake Allen and Cam Talbot with six each

030

Reposted by Aaron K

Aaron K @ladyotreesyrup.bsky.social · 11/09/2026

I've spent a fair amount of time looking into recording changes for PBP shot locations with the introduction of puck and player tracking, but those aren't the only events that have seen changes in recent seasons. Giveaways, takeaways, and hits have also been affected

Density map of takeaways for the 2020-21 to 2025-26 NHL regular seasons

Density map of hits for the 2020-21 to 2025-26 NHL regular seasons

184

Aaron K @ladyotreesyrup.bsky.social · 11/09/2026

184

Aaron K @ladyotreesyrup.bsky.social · 04/09/2026

Sort of an open question here: experimenting with an RAPM model and almost every context adjustment I add makes to year-to-year autocorrelation for the results worse. There are a few cases where it makes one direction (off/def) or one position slightly better, but nothing unequivocally better

bar chart showing differences in autocorrelation for offensive, defensive, and net RAPM outputs accounting for various contextual factors, almost all the bars are negative indicating worse autocorrelation

410

Reposted by Aaron K

Micah McCurdy @hockeyviz.com · 04/09/2026

Here is a curiosity with a lot of graphs:

1174

Reposted by Aaron K

SlimCliffy @slimcliffy.bsky.social · 02/09/2026

I wrote about Nashville trading Evangelista, teams paying for their sins, and the inanity of prospect 'depth' cliffys-hockey-analytics-department.ghost.io/luke-evangel...

cliffys-hockey-analytics-department.ghost.io

Luke Evangelista and Prospect Depth

Nashville traded Luke Evangelista to New Jersey. Why they traded him matters.

053

Reposted by Aaron K

Louis Boulet @louisboulet.bsky.social · 01/09/2026

Last Friday we dropped an article on the various defensive dimensions that go into a player's shutdown kit. And it helped us see a trend within the top pairs of recent Cup winners. Based on this here are the current pairs that have been deemed defensively strong enough to win the next Stanley Cup:

194

Reposted by Aaron K

Aaron K @ladyotreesyrup.bsky.social · 31/08/2026

Need some help crowd-sourcing this: what are some instances of NHL defensemen playing forward for a game or vice versa from the last few seasons? For example, Brendan Smith played a few games at forward in the 2024-25 season. I'm sure there are more cases, but I don't know them

1122

Aaron K @ladyotreesyrup.bsky.social · 31/08/2026

1122

Reposted by Aaron K

Louis Boulet @louisboulet.bsky.social · 28/08/2026

New article up on LB-Hockey! 🚨🏒 We identify & weight defensive skills to quantify a player’s shutdown ability, supplement Selke voters’ analysis, and figure out what it seems you NEED from a top pair to contend for the Stanley Cup. Out Now 🔗: lb-hockey.com/2026/08/28/b...

1209

Aaron K @ladyotreesyrup.bsky.social · 23/08/2026

We should really be using @louisboulet.bsky.social / @lb-hockey.com 's SPAR model more when talking about individual player contributions. Better year-to-year repeatability at the player level and prediction of future team results than other public spreadsheet-based models

Scatter plots of evolving hockey's WAR/xWAR, HockeyStat's WAR, and LB-Hockey's SPAR models converted to per 60 minutes and z-scores with year 0 on the x axis and year+1 on the y axis. NHL regulars (top 384 forwards/192 defensemen by ice time) only, 2022-23 to 2025-26

Scatter plots of the sum of the previous season's SPAR/60 rates (regulars only, top 384 forwards/192 defensemen across both seasons) multiplied by next season's ice time (x-axis) compared to standings points won by their team in the next season with contributions from goalies and non-regulars removed (y-axis) comparing Evolving-Hockey's WAR/xWAR models and LB-Hockey's SPAR model.

0156

Aaron K @ladyotreesyrup.bsky.social · 20/08/2026

me, pressing the crossing button: please wait 30 seconds so I can cross the street safely drivers, their eyes enormous: you STOP the driver? you destroy the driver's life? oh! oh! jail for the pedestrian! jail for the pedestrian for One Thousand Years!!!!

051

Reposted by Aaron K

Ian Fleming @ianmfleming.bsky.social · 20/08/2026

New to woso.live: Bayesian shotstopping talent estimates for goalkeepers

161

Reposted by Aaron K

David Johnson @hockeyanalysis.com · 20/08/2026

The Calgary Flames are also hiring a Hockey Analytics Specialist. A good opportunity for anyone wanting to get into hockey analytics. recruiting.ultipro.ca/CAL5002/JobB...

177

Reposted by Aaron K

Aaron K @ladyotreesyrup.bsky.social · 18/08/2026

One of the challenges of using the NHL EDGE animation data is the mismatch between the locations in the data and the PBP. For example, the nearest EDGE location to the PBP location in the final second before the goal is scored is greater than one foot more than half the time (diffs > 5ft excluded)

bar chart of distance minimum distance between the PBP shot location and the nearest EDGE animation location in the final second before the puck enters the net

212

Aaron K @ladyotreesyrup.bsky.social · 18/08/2026

212

Aaron K @ladyotreesyrup.bsky.social · 16/08/2026

Interesting development in the NHL goal animations, during the 2023-24 season, there were usually ~8 seconds shown before the goal, in subsequent seasons it's usually 9 or increasingly 10 seconds. (second chart removes an outlier that shows 19 seconds before the goal) www.nhl.com/ppt-replay/g...

bar charts of number of frames before goal is scored broken down by season showing the 2023-24 centered around 80 frames, while the 2024-25 and 2025-26 seasons are bimodal around 90 and 100 frames preceding the goal

230

Reposted by Aaron K

Aaron K @ladyotreesyrup.bsky.social · 11/08/2026

New post: Goaltending Is Not Voodoo Looking at near misses instead fenwick or shots improves year-to-year repeatability. Moreover, SV% adjustments/standardization seem to be improving year-to-year repeatability across a variety of metrics puckovertheglass.substack.com/p/goaltendin...

puckovertheglass.substack.com

Goaltending Is Not Voodoo

Improved on-net recording may be removing the noise

3177

Reposted by Aaron K

Aaron K @ladyotreesyrup.bsky.social · 11/08/2026

Something else that did not make it in, if you accept @evolving-hockey.com 's goalie GAR as a good descriptive measure, near-miss metrics were better at predicting the *next* season's GAR/FA z-score than any other metrics

Goalie metric z-score correlation with next season GAR/FA z-score, 2018-19 to 2022-23

Goalie metric z-score correlation with next season GAR/FA z-score, 2023-24 to 2025-26

Goalie metric z-score correlation with next season GAR/FA z-score, 2018-19 to 2025-26

001

Aaron K @ladyotreesyrup.bsky.social · 11/08/2026

puckovertheglass.substack.com

Goaltending Is Not Voodoo

Improved on-net recording may be removing the noise

3177

Reposted by Aaron K

Prashanth Iyer @prashanthiyer.bsky.social · 04/08/2026

Wanted to share a new, super cool hockey analytics site that’s been built by a member of the Wings’ community. They’ve got some really cool features not available on other sites and are still adding more! @hockeyalpha.bsky.social www.hockeyalpha.com

84911

Reposted by Aaron K

Aaron K @ladyotreesyrup.bsky.social · 26/07/2026

Some quick thoughts: - It's a bit surprising that snap shots pull more similarly to slap shots than wrist shots, though in most other respects they profile more like wrist shots. I'm still not sure whether I think they should be their own category of shot in an xG model or lumped in with wrist shots

132

Aaron K @ladyotreesyrup.bsky.social · 26/07/2026

132

Aaron K @ladyotreesyrup.bsky.social · 26/07/2026

In the 23-24 season, the NHL started including left/right (shooter's perspective) detail for misses (wide left/right, high and wide left/right, and hit left/right post). Using that, we can see if shooter/goalie handedness or side of ice influence which side of the net shooters tend to shoot towards

130

Aaron K @ladyotreesyrup.bsky.social · 17/07/2026

Net schedule impact on overall team xG. Estimates based on outputs of an RAPM model for the 2025-26 season. Included with and without time zone changes, because I'm less sure I'm capturing the impact correctly

Bar chart of 2026-27 estimated net travel effects on NHL team xG looking at travel between games and rest between games. In descending order: Anaheim, Calgary, Detroit, Chicago, Nashville, New Jersey, Los Angeles, Washington, St Louis, Montreal, Tampa Bay, Colorado, Vancouver, Vegas, New York Islanders, New York Rangers, Columbus, Dallas (change to negative impact after this), San Jose, Ottawa, Utah, Seattle, Edmonton, Carolina, Minnesota, Buffalo, Philadelphia, Florida, Winnipeg, Pittsburgh, Toronto, and Boston.

Bar chart of 2026-27 estimated net travel effects on NHL team xG looking at travel between games, time zones changed between games, and rest between games. In descending order: Anaheim, Nashville, Detroit, Los Angeles, Chicago, St Louis, New York Rangers, Washington, New Jersey, Colorado, Montreal, Dallas, New York Islanders, Vancouver, Tampa Bay, Calgary, Vegas, San Jose (switch to negative impact after this), Seattle, Minnesota, Columbus, Ottawa, Carolina, Philadelphia, Edmonton, Utah, Florida, Winnipeg, Pittsburgh, Toronto, Buffalo, and Boston.

251

Aaron K @ladyotreesyrup.bsky.social · 16/07/2026

13 back-to-backs for Dallas (T-7th in the league) 15: PIT 14: BOS, BUF, FLA, OTT, WSH 13: CBJ, DAL, NJD, NYI, SEA, SJS, TBL 12: CAR, CHI, DET, PHI, STL 11: ANA, ADM, NSH, NYR, TOR, VAN 10: COL, LAK, MIN, MTL, UTA 9: WPG 8: CGY, VGK

263

Reposted by Aaron K

hockeyskytte.bsky.social @hockeyskytte.bsky.social · 29/06/2026

The GM Mode is now live! You can access it in the NHL App: nhl.hockey-statistics.com/gm-mode With just a few days until free agency starts you can now see what a signing will do for your team's point projection and run simulations. I will do a write-up on the tool in the coming days.

062

Aaron K @ladyotreesyrup.bsky.social · 26/06/2026

The Stars‘ recent social media posts have been very bizarre in light of current events

030

Reposted by Aaron K

NHL Goal Tracking Data Analysis @nhl-goal-tracking.bsky.social · 25/06/2026

I used tracking and shot location data to estimate the shot speed of goals: danielglin.github.io/jekyll/updat...

danielglin.github.io

Estimating Shot Speeds for Goals

Introduction

121

Reposted by Aaron K

Aaron K @ladyotreesyrup.bsky.social · 19/06/2026

A couple years back I had the idea of penalizing goalies for posts hit with the thinking being that a post is a shot near enough the goalie should save it if they can (I cannot be convinced goalies intentionally let those shots go because they know it will hit the post)

121

Aaron K @ladyotreesyrup.bsky.social · 19/06/2026

121

Aaron K @ladyotreesyrup.bsky.social · 19/06/2026

Since the NHL started re-classing missed shots that were contacted by the goalie as misses, the share of missed shots that were frozen has ballooned. This doesn't look settled to me either. Something interesting to monitor next season will be whether this stabilizes or continues to increase

line chart of share of 5-on-5 missed shots frozen by the goalie from the 2019-20 season to the 2025-26 season. The shar is fairly steady from 2019-20 to 2022-23 before increasing in each of the three most recent seasons. OZ shots from above the goal line only, short and failed bank attempt misses excluded

000

Reposted by Aaron K

Eric Schumacher @pockecbj.bsky.social · 17/06/2026

Dimitri let me yap BIG on this one. Links to deeper logic and some of the fun armchair interactive plots: pocketcbj.substack.com/p/armchair-g... pocketcbj.substack.com/p/armchair-g... pocketcbj.substack.com/p/armchair-g...

pocketcbj.substack.com

Armchair GM: Setting Offseason Objectives and Assessing Market Potential

The Stanley Cup Finals has been a wicked affair up to this point, though I am finding it thoroughly enjoyable.

142

Aaron K @ladyotreesyrup.bsky.social · 16/06/2026

Cilantro. It doesn't taste like soap, it just tastes bad

010