How this is measured
Every number on this site comes from the NHL's own shift charts and play-by-play, resolved to the second. Nothing is projected, nominated, modelled or bought.
What a power-play unit is here
Membership is derived from shared power-play ice, not from power-play minutes. That distinction is the whole product: a skater can accumulate four minutes of power-play time by taking thirty seconds with each unit and belong to neither of them. The five with the most shared power-play time are the first unit. The unit is on the ice whenever its members are most of the club's skaters out there -- three of the five on a five-on-four -- so every power-play second, shot and goal belongs to one unit or to neither, and the two units' shares never add to more than the club's power play. How often the exact five were all out together is its own figure, same five.
What a run is
A run counts consecutive games with a power play, never consecutive games. A night in which a club drew no penalty is not evidence about anybody's power-play unit, so it neither breaks nor extends a run. A game a skater did not dress for is treated the same way and is disclosed separately rather than bridged silently.
Every threshold on this site
| what | value | why |
|---|---|---|
| Unit credit floor | 25% of the unit's power-play ice, or 30 seconds, whichever is lower | A player is credited to the unit where he took most of his power-play ice, and only if that share clears this floor. Not yet calibrated against hand-checked games. |
| Run | games with a power play | A game in which the team drew no penalty neither breaks nor extends a run. A game a player did not dress for neither breaks nor extends it either, and is disclosed. |
| Confidence tiers | locked 20+ · established 8+ · emerging 3+ · new 1+ | Read off the run length. |
| Story: the gap | ranks disagree by 7 or more of 32 | Below this, two ranks are the same rank with noise on them. |
| Story: the streak | run of 10 or more, or a run ending | Length is the trend; the site publishes it as one integer. |
| Story: the non-story | all ranks within 5 of each other | Published deliberately. An engine that speaks only when surprised is an engine that manufactures surprise. |
| Unit minimum | 30 minutes as the same five | Below this a five is an accident of one power play, not a unit. |
| Top-line minimum | 100 minutes together at 5v5 | Applies to the forward-trio tables only. |
| PP1 player | 40% of his games on the first unit | Used only where a page needs to call somebody a first-unit player. |
The unit-credit floor has not yet been calibrated against hand-checked games. It is stated here because a threshold nobody can see is a threshold nobody can argue with.
What produced this build
- toi_pp
- shift charts intersected with situationCode, per second
- pp_toi
- sum of the team's individual power-play seconds, / 5
- pp_opps
- penalty spans against the opponent, counted
- unit
- shared power-play ice, seeded on the top trio and grown to five
Ranking pages not published in this build, for want of their inputs: best_power_play_units, what_is_a_pp1_spot_worth, which_clubs_draw_the_most_power_plays, which_clubs_lean_hardest_on_pp1, most_stable_power_play_units, longest_power_play_runs, pp1_players_with_no_power_play_time, defencemen_power_play_time, top10_pp_points, top10_pp_goals, top10_pp_shots, top10_pp_attempts, top10_pp_time, top10_pp_hdc, top10_best_pp_clubs, top10_worst_pp_clubs, top10_pp1_time_no_points.
What this site does not track, and why
| not tracked | why |
|---|---|
| High-danger chances | Computed from the home-plate area but never externally calibrated, because the source named for calibrating it returns HTTP 403 here. The comparison could not be run and was not faked. Never used to rank anything on this site. |
| Expected goals | No fitted model exists in this project. A column of measured high-danger chances wearing the name xG would be a worse lie than having no column, so there is none. |
| Zone entries, exits and passes | Manual-tracking statistics. Not in the NHL's public data, and this project takes no other input. |
| Nominated and projected lines | Excluded deliberately. Measuring the unit that was deployed rather than the one that was announced is the entire product. |
| Market prices of any kind | No feed. This site can never call a number mispriced or claim an advantage; it reports a deployment fact and stops there. |
| Practice and morning-skate lines | Requires beat reporters. Coverage was measured at 6.1% of club-days, with 19 of 32 clubs at zero. |
| Power-play shots on goal, as such | Total shots validate exactly and power-play state assignment validates at 99.3%, but their intersection has never been checked. It is therefore not published as a column. |
The one chart
The deployment strip is the only visualization on this site. Each game is one cell, and every state has its own shape as well as its own tone, so the strip reads in greyscale too: a solid ink cell is a game on the first unit, a solid grey cell a game on the second, an ink outline a game he played but not on the power play, a small grey dot a game his club drew no penalty in, a red box with a red slash a game he did not play, and a dashed box a night the data failed its checks. The key under every strip uses the same marks. Percentages are printed rather than drawn. Where a figure carries a direction it carries a glyph and a signed number too, so hue is never the only thing saying which way it went.