What is the SHL penalty rate in this sample?

10,240 penalties in 1,854 regular-season games, 2020/21–2024/25 — 5.52 penalties per game.

Which high-sample referee sits at the high end?

Tobias Björk, 5.82 penalties per game across 250 assignments. Christoffer Holm is also 5.82 (219 games).

Which high-sample referee sits at the low end?

Johan Wedin, 4.94 penalties per game across 165 assignments. Mikael Sjöqvist is 5.08 across 250 games.

Does this prove some referees cause more penalties?

No. Two referees share every game, we do not know who called each penalty, and there are no team or score-state controls yet. What it does is reject the lazy version of “it’s all noise” on raw rates.

Where does the data come from?

Official SHL regular-season assignments and penalties in MotherDuck (swe.game_referees_json, game_penalties, swe.main.games), exported as referee_penalties_data.json.

SHL referees are not interchangeable: penalty rates for 23 officials over five seasons

The hockey version of “the refs decided it” is usually a feeling. The counter-feeling is that officials are interchangeable noise.

We can test the lazy form of that second claim. Not “who called the 2+10 in minute 58.” Just this: when a named referee is one of the two officials on an SHL ice, how many penalties does that game produce?

Same window as the timeout study: SHL regular season 2020/21–2024/25, 1,854 games.

The short version

League average: 5.52 penalties per game (10,240 penalties / 1,854 games).
23 referees with at least 40 assignments. High-sample band (≥150 games): Tobias Björk 5.82 vs Johan Wedin 4.94.
Björk and Mikael Sjöqvist each worked 250 games — 5.82 vs 5.08. That is not a 12-game illusion.
This is assignment rate, not who-called-it. Two referees share every game. No team or score controls yet.

What we measured

For every SHL regular-season game in the window we took the two Referee names (not linesmen) and the penalty count in that game. Each official gets the game’s penalty total — because we cannot see which referee raised the arm.

Then we kept anyone with at least 40 assignments. That leaves 23 officials.

If referees were interchangeable noise, those 23 rates should crowd the 5.52 league mean. They do not.

High-sample officials (≥150 games)

This is the table that matters. Small samples bounce; 150+ games do not bounce as much.

Referee	Games	Pens / game	PIM / game
Tobias Björk	250	5.82	12.84
Christoffer Holm	219	5.82	12.69
Andreas Harnebring	220	5.77	12.63
Johan Magnusson	151	5.77	12.35
Linus Öhlund	244	5.74	11.94
Daniel Eriksson	155	5.65	12.45
Mikael Holm	243	5.55	12.29
Kristoffer Folkstrand	164	5.55	11.88
Richard Magnusson	232	5.40	11.92
Marcus Linde	253	5.32	11.57
Mikael Nord	155	5.13	11.10
Mikael Sjöqvist	250	5.08	10.98
Johan Wedin	165	4.94	10.17

Björk vs Wedin is 0.88 penalties per game. On a 52-game season that is on the order of forty-five extra penalty events in the games one of them works, before you even talk about power plays.

Sjöqvist vs Björk is the cleaner twin: 250 games each, 5.08 vs 5.82.

The edges (smaller samples)

Johan Nordlöf sits at 6.14 (95 games). Daniel Wessner sits at 4.08 (51 games). Those are the headline-friendly ends of the 23-official list, and they are also the ones you should trust less. Forty to ninety games is enough to notice, not enough to convict.

We are not ranking character. Some officials work more chippy matchups; some work more special-teams chaos. The point for the claim we keep on file — “referee effects are noise” — is simpler: raw assignment rates are not flat.

What this is not

Not “Referee X called more minors.” We do not have caller identity.
Not a control for home/away, score state, team fighting style, or which partner they stood with.
Not HockeyAllsvenskan (yet). SHL only.
Not a product. The league-facing referee report is still an opportunity in our catalog, not a sellable HTML pack.

The honest next step is the falsification test we already wrote down: do differences survive after those controls? Until then the claim stays contested, not killed.

Source

MotherDuck swe: game_referees_json (role = Referee) × game_penalties × swe.main.games (SHL regular, season codes 2021–2025). Artifact referee_penalties_data.json from scripts/analyze_referee_penalties.py.

Source: The Hockey Analytics · MotherDuck SHL · Stockholm.