By Ashkon Baharlou

Eight games on night three of the season, Thursday night, and the picks went five for eight: New Jersey, Columbus, Minnesota, Utah, and Edmonton won; Tampa Bay, Calgary, and Florida lost. Three of the misses came down to goaltending, and the Whiteboard's read on all three was right about the chances and wrong about the goalies.

New Jersey 3, Philadelphia 2 (OT): Jesper Bratt drove it with an 86 percent on-ice share and the assist on Jack Hughes's overtime winner. Joseph Woll kept it close with 41 saves. NY Rangers 5, Tampa Bay 1: Igor Shesterkin stopped 24 of 25 and scored an empty-net goal himself. Tampa Bay's power play went 0 for 7. Columbus 6, Buffalo 3: Right winner, different reason. Buffalo had the better chances; Columbus finished and went 2 for 3 on the power play. Minnesota 3, Nashville 1: The read, start to finish. Minnesota controlled 3.82 expected goals to 2.24. Seattle 6, Calgary 1: Seattle had the better chances and Dustin Wolf had a hard night, six allowed on 3.93 expected. Utah 6, Chicago 0: The surest pick, emphatically. A 71 percent share of 5-on-5 chances and a shutout from Karel Vejmelka. Edmonton 9, Vancouver 7: The wildest game of the young season. The pick came in; the goaltending on both sides was the story. San Jose 4, Florida 3 (OT): Florida had the better chances but lost the special-teams goals 2 to 1, missing Gustav Forsling and most of Aleksander Barkov.

A quick note on the numbers, since they carry this article. Expected goals (xG) is the value of every chance. Individual expected goals (ixG) is a player's own number from the shots he took. Goals saved above expected (GSAx) is what a goalie stopped beyond the chances he faced: a positive number means he saved more than expected, a negative one means fewer. And on-ice share is a player's slice of 5-on-5 expected goals while he was out there. Power play covers 5-on-4, 5-on-3, and 4-on-3; penalty kill covers the reverse. All of it is the Whiteboard's tracking of the eight games, built on the NHL's play-by-play.

New Jersey 3, Philadelphia 2 (OT)

The pick was New Jersey at 63.2 percent, and the chances say it should never have been close. New Jersey led 4.59 expected goals to 1.75 at all strengths and 3.22 to 1.14 at 5-on-5, a 74 percent share, with 9 high-danger chances to 1 and 44 shots to 25. The one thing the preview flagged as the pick's uncertainty was Joseph Woll in Philadelphia's net, and he was the whole game: 41 saves on 44 shots, 4.59 expected goals faced, plus-1.59 goals saved above expected, about a goal and a half better than those chances normally produce. Jake Allen faced 1.75 expected and allowed 2, minus-0.25.

Jesper Bratt drove New Jersey's play: an 86 percent on-ice share, on the ice for 1.18 expected goals for and 0.19 against, with 0.77 individual expected goals on 6 shots, the most of any skater in the game, and the assist on Jack Hughes's overtime winner. Hughes had 0.59 individual expected goals on 5 shots. Dawson Mercer and Evan Rodrigues were at 82 percent. Dougie Hamilton and Luke Hughes scored the regulation goals.

Philadelphia's points came from its defense: Travis Sanheim had a goal and an assist and Rasmus Ristolainen a goal. Up front, Alex Bump was at a 6 percent on-ice share, on the ice for 0.06 expected goals for and 1.02 against in 7.2 minutes, and Trevor Zegras at 11 percent; Matvei Michkov, at 49 percent, was Philadelphia's best.

Special teams: New Jersey went 0 for 4 on 0.90 expected goals in about 8.4 minutes; Philadelphia went 0 for 1 on 0.06. All five goals came at even strength.

The pick held, and the margin understated it. New Jersey carried the play; Woll made it a one-goal game.

NY Rangers 5, Tampa Bay 1

The pick was Tampa Bay at 69.1 percent, and the preview's premise was that the goalies would cancel out. They did not. The chances were close to even: Tampa Bay had 3.29 expected goals to 2.99 at all strengths, 5-on-5 was 2.32 to 2.26, though New York had 7 high-danger chances to 4. Igor Shesterkin stopped 24 of 25 shots on 3.29 expected goals, plus-2.29 goals saved above expected, about two and a half goals better than expected. He also scored an empty-net goal himself. Andrei Vasilevskiy allowed four on 2.91 expected, minus-1.09.

The other half of the miss was Tampa Bay's power play, the unit the Whiteboard rated highest of the night: 0 for 7, on 0.57 expected goals in about 13.6 minutes, a rate of 2.52 per 60. New York went 1 for 2, Mika Zibanejad scoring on 0.22 expected goals in about 2.6 minutes.

For New York, Pavel Dorofeyev had a goal and an assist on 0.60 individual expected goals and 5 shots, Zibanejad a goal and an assist, and Adam Fox two assists. Eeli Tolvanen scored at a 69 percent on-ice share, and Will Cuylle, also at 69 percent, generated 0.60 individual expected goals without a point. For Tampa Bay, Anthony Cirelli was at 72 percent and Brandon Hagel 69 percent, and Nikita Kucherov scored. Pontus Holmberg generated 0.67 individual expected goals, the most of any skater in the game, without a point.

The chances were close to even, and the night turned on Shesterkin and on Tampa Bay's power play coming up empty seven times.

Columbus 6, Buffalo 3

The pick was Columbus at 59.6 percent, and it came in, but not on the read the preview gave. The preview said Jet Greaves, the night's highest-graded goalie, would decide it. Greaves was even: 23 saves on 26 shots, 2.95 expected goals faced, minus-0.05. Buffalo had the better chances, 2.95 to 2.25 at all strengths and 1.84 to 1.44 at 5-on-5. The difference was Ukko-Pekka Luukkonen: six allowed on 2.24 expected, minus-2.76 goals saved above expected, nearly three goals worse than expected.

Columbus won on finishing and the power play, which went 2 for 3 on 0.67 expected goals in about 2.9 minutes. Matthew Knies scored twice, once on the power play, with an assist, on 0.56 individual expected goals, despite a 31 percent on-ice share. Charlie Coyle had three assists and Zach Werenski two. Emil Andrae, on the second pair with Carson Soucy, who replaced the injured Ivan Provorov, was at 25 percent.

Buffalo's best line carried play: Jack Quinn, who scored, and Zach Benson, both at a 79 percent on-ice share, and Noah Ostlund, with two assists, at 76 percent. Tage Thompson generated 0.57 individual expected goals without a point. Konsta Helenius scored but was at 17 percent. Buffalo's power play went 0 for 5 on 0.98 expected goals in about 6.9 minutes, including 2.0 minutes at 5-on-3.

Right winner, different reason. Buffalo outplayed Columbus; Columbus outscored it on Luukkonen's night and a 2-for-3 power play.

Minnesota 3, Nashville 1

The pick was Minnesota at 63.8 percent, and it played out the way the Whiteboard read it. Minnesota controlled the game: 3.82 expected goals to 2.24 at all strengths, 2.26 to 1.45 at 5-on-5, a 61 percent share. Jesper Wallstedt stopped 19 of 20, plus-1.24 goals saved above expected, and Juuse Saros 27 of 30, plus-0.82.

Kirill Kaprizov scored on 0.69 individual expected goals, the most in the game, and Matt Boldy had 0.61 on 5 shots without a point. The fourth line was Minnesota's best at 5-on-5: Marcus Foligno at 78 percent and Nick Foligno, who scored, at 76; Zach Bogosian, on the third pair the preview flagged as the weakest, led the team at 79 percent. Blake Coleman scored on the power play, which went 1 for 4 on 1.10 expected goals in about 7.4 minutes, including 1.4 minutes at 5-on-3.

For Nashville, Roman Josi scored but was at 29 percent, on the ice for 1.06 expected goals against. Matthew Wood was at 27 percent. Nashville's power play went 0 for 4 on 0.67 expected goals in about 10.4 minutes.

The Whiteboard's read, start to finish.

Seattle 6, Calgary 1

The pick was Calgary at 61.0 percent, and the preview's stated risk is exactly what happened. The preview said the pick rested on Dustin Wolf plus home ice, with Seattle's forwards and power play rated higher. Seattle had the better chances, 3.93 expected goals to 2.95 at all strengths and 3.09 to 2.55 at 5-on-5, and Wolf allowed six on 3.93 expected, minus-2.07 goals saved above expected, about two goals worse than expected. Joey Daccord stopped 24 of 25, plus-1.95, and assisted on a goal.

Jordan Eberle scored on the power play with 0.95 individual expected goals on 4 shots, the most of any skater on the night, at a 72 percent on-ice share. Jared McCann had three assists and Vince Dunn a goal and two assists. Seattle's scoring spread wide: six different goal scorers. Matty Beniers and Kaapo Kakko each had a goal and an assist despite on-ice shares of 25 and 28 percent.

Seattle's power play went 1 for 5 on 0.79 expected goals in about 8.9 minutes. Calgary's went 0 for 1 without a shot. Ryan Strome scored Calgary's only goal; Maxim Tsyplakov was at 25 percent and Adam Klapka at 26.

The pick was Wolf plus home ice, and Wolf had a hard night.

Utah 6, Chicago 0

The most confident pick of the night, Utah at 79.3 percent, and the most one-sided result. Utah led 4.23 expected goals to 1.74 at all strengths and 3.54 to 1.46 at 5-on-5, a 71 percent share, with 9 high-danger chances to 1 and 32 shots to 15. Karel Vejmelka stopped all 15 shots on 1.74 expected goals, plus-1.74 goals saved above expected.

The preview named Spencer Knight as the one thing keeping Chicago close, but Arvid Soderblom started instead and allowed six on 4.23 expected, minus-1.77.

Dylan Guenther scored twice on the power play, on 0.57 individual expected goals. Barrett Hayton was at an 89 percent on-ice share and Clayton Keller, with two assists, at 86. Utah's power play went 2 for 3 on 0.52 expected goals in about 3.0 minutes. Fourteen Utah skaters had a point.

For Chicago, Nick Lardis was at a 6 percent on-ice share and Alex Vlasic at 7 percent, on the ice for 1.76 expected goals against; Bowen Byram was on for 2.17 against, the most in the game. Chicago's power play went 0 for 3 with one shot. Nobody on Chicago had a point.

The Whiteboard's read, emphatically, with the goalie it feared on the bench.

Edmonton 9, Vancouver 7

The pick was Edmonton at 73.7 percent, and it came in, in the night's wildest game. Edmonton led 3.66 expected goals to 3.10 at all strengths and 11 high-danger chances to 9, while Vancouver had the edge at 5-on-5, 2.48 to 2.09. The goaltending on both sides was the story. Devon Levi, whose grade the preview said rests on almost no NHL data, allowed seven on 3.10 expected, minus-3.90 goals saved above expected. Kevin Lankinen allowed eight on 2.65 expected, minus-5.35, before Leevi Merilainen replaced him and allowed one on 1.01.

Connor McDavid had a goal and four assists, his goal shorthanded, at a 70 percent on-ice share. Leon Draisaitl scored twice, once on the power play, with an assist, on 0.62 individual expected goals, and Evan Bouchard had a goal and two assists. Vasily Podkolzin, on McDavid's line with Zach Hyman out, scored on the power play with an assist. Edmonton scored twice shorthanded, McDavid and Kasperi Kapanen, on 0.30 shorthanded expected goals, and went 2 for 5 on the power play on 1.27 expected goals in about 8.1 minutes.

For Vancouver, Marco Rossi scored twice with an assist at a 69 percent on-ice share, and Elias Pettersson, the center, had a power-play goal and two assists at 59 percent. Jonathan Lekkerimaki was at 78 percent. Vancouver went 1 for 3 on the power play on 0.46 expected goals.

The winner was right, and both of the preview's flagged risks showed up: Levi's thin grade, and goaltending low on both sides.

San Jose 4, Florida 3 (OT)

The pick was Florida at 67.9 percent, and it missed. Florida had the better chances, 3.88 expected goals to 3.19 at all strengths and 1.74 to 1.30 at 5-on-5, with 10 high-danger chances to 4. Most of Florida's edge came on the power play: 2.08 expected goals in about 10.3 minutes, 12.07 per 60, for one goal in six chances. Yaroslav Askarov, the lowest-graded starter of the night, stopped 23 of 26, plus-0.88 goals saved above expected; Akira Schmid allowed four on 3.19 expected, minus-0.81.

Florida played without Gustav Forsling, one of the Elite drivers the preview counted on, with Uvis Balinskis in his place, and Aleksander Barkov played only 9:15 on 11 shifts before leaving with an apparent left-leg injury. Maurice said postgame the injury is unrelated to last season's knee reconstruction, with more expected from the team Friday afternoon. Eetu Luostarinen scored twice, once on the power play, at a 73 percent on-ice share. Matthew Tkachuk scored, and Sam Reinhart generated 0.64 individual expected goals without a point.

San Jose's top line won it: Will Smith scored twice, once shorthanded and once on the power play, with an assist on Luca Cagnoni's overtime winner; Macklin Celebrini had three assists on 7 shots; Mason Marchment scored at an 84 percent on-ice share. Smith's shorthanded goal came on 0.26 San Jose shorthanded expected goals.

Florida had the better of it, but ran the power play into Askarov and lost the special-teams goals 2 to 1 while missing Forsling and most of Barkov.

The standouts

  • Igor Shesterkin, Rangers: 24 saves on 3.29 expected goals, plus-2.29 goals saved above expected, and an empty-net goal of his own.
  • Joseph Woll, Philadelphia: 41 saves on 4.59 expected goals in a loss.
  • Connor McDavid, Edmonton: a goal and four assists, the goal shorthanded.
  • Tampa Bay's power play: 0 for 7 on 0.57 expected goals, from the Whiteboard's highest-rated unit.
  • Utah: a 71 percent share at 5-on-5, 9 high-danger chances to 1, and fourteen skaters with a point.

What the crowd sees vs. what the model sees

The crowd sees five winners and three upsets and reads a mixed night. The Whiteboard's counter-read is that the misses were all goaltending. It had the chances right in all three: Tampa Bay's chances were even with the Rangers', Calgary's were worse than Seattle's the way the preview warned, and Florida's were better than San Jose's. Shesterkin was two and a half goals better than expected, Daccord two goals better, and Askarov nearly a goal better from the lowest-graded starter of the night. The process was right; the goalies rewrote the results.

It had the chances right in all three misses; the goalies rewrote the results.

The fair concession, and it matters: its ratings are still built on last season's data. Nothing from these games is in them yet. Three nights will tell us more than one did.

The verdict

Five of eight, with the three misses all landing on goaltending. New Jersey, Minnesota, Utah, and Edmonton won the way the Whiteboard read them; Columbus won while the chances said otherwise, on finishing and a 2-for-3 power play. Tampa Bay, Calgary, and Florida lost games where the Whiteboard's read on the chances held up and the goalie across the ice beat it.

The real risk, named plainly: it is three nights. None of these games are in its ratings yet, and one night of goaltending does not make Shesterkin, Daccord, or Askarov the story of the season. The chances tell you what happened while the scoreboard tells you who won, and on night three the Whiteboard read the chances right and the goalies decided three games anyway.

Want to check the work? Open the Whiteboard, the free line tool I built, and pull up any of these games yourself. Build tonight's lineups while you are there and watch the ratings move.

All expected-goals numbers from the Whiteboard's tracking of the eight games, built on NHL play-by-play. Scores, scorers, and goalie lines via the NHL's official data.

Follow @lastchangehockey on Instagram for more Whiteboard breakdowns.