Zone Starts in Hockey (OZS% and DZS%)

Zone starts came out of the first wave of public hockey analytics, and nobody's quite sure who thought of them. War-on-Ice's glossary guessed "probably Gabe and Vic" and noted that the number had been "found at behindthenet.ca for the longest time." Rob Vollman's player usage charts, which War-on-Ice credited too, made it famous by putting zone starts on one axis and quality of competition on the other, with each player's bubble colored by relative Corsi.

The research that followed shrank it. Eric Tulsky, who was named the Carolina Hurricanes' general manager in 2024, estimated that an offensive-zone start was worth about 0.31 unblocked attempts (roughly 0.4 Corsi events), about half of what analysts had assumed, according to a 2014 Hockey-Graphs summary. In 2015 Matt Cane measured the effect in goals at Hockey-Graphs and found that even after an offensive-zone faceoff win and 45 seconds of play, it came to less than 0.02.

Micah Blake McCurdy added the biggest piece the same year at HockeyViz. Around half of all shifts start on the fly, he found, and once he reweighted every shot for where the skaters' shifts began, only 5% of the players with 100 five-on-five minutes in 2014-15 saw their shot share move more than a percentage point. Two moved more than two.

The NHL publishes two versions today (OZ Start% and OZ Ratio, both at five-on-five and both available back to 2009-10), Hockey-Reference has oZS% and dZS%, and Natural Stat Trick and MoneyPuck count shift starts by zone.

There are two ways to count a zone start, and the sites split on them. Natural Stat Trick and MoneyPuck count shift starts, so a faceoff counts only if the player's shift began with it. The NHL and Hockey-Reference count every offensive- and defensive-zone faceoff a player was on the ice for, whether his shift started there or not. Either way, the pieces are the same:

• Offensive-zone starts, the shifts or faceoffs in the attacking zone
• Defensive-zone starts, the same in his own zone
• Neutral-zone starts, which most versions leave out of the percentage
• On-the-fly starts, the shifts that began during play, which no faceoff-based version can see
• The game state, almost always five-on-five

Here's a worked example with invented round numbers. A center starts 1,000 five-on-five shifts, 600 of them on the fly, 160 in the neutral zone, 140 in the offensive zone and 100 in his own end. His OZS% is 140 ÷ 240, or 58.3%, which reads as a sheltered player. Only 24% of his shifts started with a faceoff in either end, though, so the label describes about a quarter of his night.

The two counting methods can tell different stories about the same player. Moritz Seider started 34.5% of his end-zone shifts in the offensive zone by MoneyPuck's count of shift starts, while the NHL's on-ice faceoff count has him at 42.8%, because the NHL's count includes every draw he was on the ice for, the ones in the middle of a shift included.

OZS% measures deployment. It tells you where a coach chose to start a player when the whistle gave him the choice. At five-on-five last season the median forward with 500 minutes started 53.1% of his end-zone shifts in the offensive zone and the median defenseman 46.3%.

Most players live in the middle. More than half of those 385 forwards sat between 40% and 60%, and the 10th and 90th percentiles were 36.1% and 67.6%. The extremes belong to specialists. Washington's Alex Ovechkin started 305 shifts in the offensive zone and six in his own end (98.1%), and Pittsburgh's Blake Lizotte, Noel Acciari and Connor Dewar were the league's three most buried forwards, with Lizotte's 6.6% at the bottom.

Detroit's forwards were deployed evenly. The eight with 900 five-on-five minutes all landed between 46.1% and 56.2%. The defense was where the choices showed, and the top pair took the hard starts. Moritz Seider (34.5%) and Simon Edvinsson (34.8%) were 16th- and 17th-lowest among the league's 213 defensemen with 500 minutes, and Axel Sandin-Pellikka was the most sheltered Wings defenseman at 50.8%.

How much does a tough deployment cost? Less than you'd think. Run Tulsky's rate on Seider's season and his 117 more defensive starts than offensive ones were worth about 36 unblocked attempts of differential, which would have moved his Fenwick share from 55.3% to about 56.2%. That's a rounding error next to how well he played.

Grading a buried defenseman

Seider's results look better once you know where he started. He began 247 shifts in his own end and 130 in the offensive zone, and Detroit still had 55.3% of the expected goals and 54.1% of the high-danger shots with him on the ice at five-on-five. A defenseman who wins that share from those starts is driving play, and the deployment makes the results more impressive. How rare was his deployment? Only fifteen of the league's regular defensemen started a smaller share in the offensive zone.

Checking a sheltered scorer

Ovechkin's 305 offensive starts against six defensive ones were the most lopsided deployment in the league, and Washington still had only 49.6% of the unblocked attempts with him on the ice at five-on-five. Lizotte, buried at the other extreme, was on the ice for 52.5%. Zone starts tell you what a coach asked of a player. His shot share tells you what he did with it, and the two don't always line up.

Adjusting shot numbers

McCurdy's adjustment weights each shot by how rare it was given where the shifts began. Shots for after an offensive-zone start count 0.80 and shots against 1.23, and after a defensive-zone start it flips (1.33 for, 0.78 against). Evolving-Hockey's RAPM goes a step further and regresses zone starts out alongside teammates, competition and score state, so you don't have to adjust by hand.

Most shifts start on the fly

Six in ten five-on-five shifts began during play last season, and only about one in five began with a faceoff in either end. OZS% describes a minority of a player's shifts and says nothing about the rest.

The effect fades fast

Cane found that a defensive-zone faceoff win stops mattering after about 8 seconds and a neutral-zone win after 14, and that even an offensive-zone win takes close to a minute to pay out in full. A player who changes 15 seconds after an offensive-zone win gets about 75% of what a full 45-second shift would've given him.

It ignores the faceoff result

A defensive-zone start after a won draw is a different shift from one after a lost draw, and OZS% counts them the same. Cane's expected faceoff goal differential was built to account for both the result and the shift length that follows.

Where a shift ends is on the player

McCurdy's framing is the cleanest. Where and how a shift starts is the coach's decision, and what happens next, including winning the draw and getting the puck out, is the player's. A line that can't get the puck out keeps producing defensive-zone faceoffs, and somebody's got to take them.

The sites don't count the same thing

Shift starts and on-ice faceoffs give different answers, as Seider's 34.5% and 42.8% show. Compare zone starts from one source at a time, and check whether it's counting shifts or faceoffs.

Do zone starts count power-play faceoffs?

Not in the versions most people quote. The NHL calculates OZ Start% and OZ Ratio at five-on-five, Hockey-Reference labels its zone starts even strength, and MoneyPuck and Natural Stat Trick let you pick the situation. Special-teams draws describe a different job and stay out of the deployment numbers.