Chemistry and stacks
NHL line chemistry: who plays better together, and who scores together
Shared ice time tells you who plays together. It does not tell you whether it works. For every pair of teammates we compare expected goals together with each player apart, and for DFS we measure how closely their DraftKings points move game to game.
Showing the full 2025-26 season. 2026-27 pairs do not have enough games together yet to fill these tables; the page switches over once they do.
2025-26 regular season · 6,134 pairs across 32 teams · typical DK correlation: forward linemates .28, a forward with a defenceman .09, defence partners .10; teammates on different lines .02.
Best DFS stacks: high-scoring linemates
Ranked by stack score: the two players' combined DraftKings points per game times the correlation typical for their pair type (shrunk toward it). The raw correlation column is this season's value with its 90% interval and games, shown as context: a pair's own correlation does not predict next season better than its type does.
| Pair | Team | 5v5 min | xG% lift | xGF/60 lift | DK corr. (raw) | Stack |
|---|---|---|---|---|---|---|
| Nathan MacKinnon + Martin NecasF–F · PP | COL | 1053 | −0.2 | +0.03 | .35.17 to .50, 75 GP | 9.4 |
| Leon Draisaitl + Connor McDavidF–F · PP | EDM | 318 | +7.2 | +0.49 | .37.03 to .63, 24 GP | 9.4 |
| Nikita Kucherov + Brandon HagelF–F · PP | TBL | 513 | +5.1 | +0.42 | .55.34 to .71, 43 GP | 9.1 |
| Zach Hyman + Connor McDavidF–F · PP | EDM | 725 | +8.1 | +0.93 | .32.10 to .51, 52 GP | 8.4 |
| Mark Stone + Jack EichelF–F · PP | VGK | 460 | +2.7 | +0.24 | .57.36 to .73, 40 GP | 8.3 |
| Nikita Kucherov + Jake GuentzelF–F · PP | TBL | 352 | −2.5 | +0.02 | .31−.00 to .57, 28 GP | 8.2 |
| Will Smith + Macklin CelebriniF–F · PP | SJS | 850 | +4.2 | +0.22 | .44.26 to .59, 66 GP | 8.0 |
| Jason Robertson + Wyatt JohnstonF–F · PP | DAL | 477 | +1.3 | +0.65 | .46.18 to .67, 31 GP | 7.9 |
| Artturi Lehkonen + Nathan MacKinnonF–F · PP | COL | 698 | +1.7 | +0.63 | .26.03 to .46, 53 GP | 7.7 |
| Nikita Kucherov + Anthony CirelliF–F | TBL | 415 | +2.5 | +0.37 | .55.32 to .72, 35 GP | 7.6 |
| Ryan Nugent-Hopkins + Connor McDavidF–F · PP | EDM | 548 | +3.2 | +0.63 | .32.07 to .52, 45 GP | 7.5 |
| Nikita Kucherov + Brayden PointF–F · PP | TBL | 492 | +1.1 | +0.11 | .43.20 to .62, 43 GP | 7.5 |
| Jesper Bratt + Jack HughesF–F · PP | NJD | 689 | +0.7 | +0.11 | .44.23 to .60, 54 GP | 7.5 |
| Nick Suzuki + Cole CaufieldF–F · PP | MTL | 1005 | +6.3 | +0.49 | .33.16 to .49, 81 GP | 7.4 |
| Mika Zibanejad + Artemi PanarinF–F · PP | NYR | 453 | +0.9 | +0.14 | .53.28 to .71, 32 GP | 7.3 |
| Mikko Rantanen + Wyatt JohnstonF–F · PP | DAL | 682 | +3.3 | +0.38 | .43.22 to .60, 52 GP | 7.1 |
| Mark Scheifele + Kyle ConnorF–F · PP | WPG | 1234 | +1.8 | +0.36 | .24.06 to .40, 82 GP | 7.0 |
| Nick Suzuki + Juraj SlafkovskýF–F · PP | MTL | 640 | +2.9 | +0.04 | .47.26 to .65, 45 GP | 6.9 |
| Sidney Crosby + Bryan RustF–F · PP | PIT | 768 | +1.5 | +0.33 | .41.22 to .58, 60 GP | 6.9 |
| Patrick Kane + Alex DeBrincatF–F · PP | DET | 743 | −1.2 | −0.11 | .39.19 to .55, 60 GP | 6.9 |
| Brady Tkachuk + Dylan CozensF–F · PP | OTT | 450 | +4.5 | +0.35 | .53.31 to .69, 41 GP | 6.8 |
| Auston Matthews + Matthew KniesF–F · PP | TOR | 427 | +0.4 | −0.11 | .39.11 to .61, 33 GP | 6.8 |
| Cutter Gauthier + Leo CarlssonF–F · PP | ANA | 283 | −2.2 | −0.20 | .16−.21 to .48, 23 GP | 6.8 |
| David Pastrnak + Morgan GeekieF–F · PP | BOS | 572 | +0.0 | +0.08 | .24−.00 to .46, 46 GP | 6.8 |
| Nick Schmaltz + Clayton KellerF–F · PP | UTA | 1025 | +8.3 | +0.66 | .38.21 to .53, 81 GP | 6.8 |
Best chemistry: more offence together than apart
Ranked by the shrunk change in expected goals for per 60 when the pair shares the ice at 5v5, against each player without the other. Lifts carry over only partly from season to season, so read them as this season's results, not fixed traits.
| Pair | Team | 5v5 min | xG% lift | xGF/60 lift | DK corr. (raw) | Stack |
|---|---|---|---|---|---|---|
| Zach Hyman + Connor McDavidF–F · PP | EDM | 725 | +8.1 | +0.93 | .32.10 to .51, 52 GP | 8.4 |
| Tommy Novak + Ryan SheaF–D | PIT | 361 | +9.4 | +0.82 | .14−.16 to .41, 33 GP | 1.1 |
| Logan Stankoven + Jackson BlakeF–F · PP | CAR | 900 | +7.5 | +0.72 | .34.16 to .50, 76 GP | 4.4 |
| Nathan MacKinnon + Devon ToewsF–D | COL | 591 | +7.8 | +0.72 | .30.10 to .49, 61 GP | 3.1 |
| Tom Wilson + Jakob ChychrunF–D · PP | WSH | 367 | +6.7 | +0.71 | .16−.13 to .42, 35 GP | 2.1 |
| Miro Heiskanen + Wyatt JohnstonF–D · PP | DAL | 481 | +4.3 | +0.70 | .12−.10 to .33, 57 GP | 2.2 |
| Roman Josi + Michael BuntingF–D | NSH | 241 | +9.3 | +0.69 | .02−.34 to .38, 22 GP | 1.6 |
| Robert Thomas + Dylan HollowayF–F · PP | STL | 316 | +5.2 | +0.69 | .42.07 to .68, 22 GP | 6.1 |
| Nick Seeler + Trevor ZegrasF–D | PHI | 345 | +3.7 | +0.69 | −.06−.39 to .27, 26 GP | 1.2 |
| Alex Turcotte + Andrei KuzmenkoF–F | LAK | 220 | +8.3 | +0.67 | −.15−.50 to .24, 20 GP | 2.4 |
| Nick Schmaltz + Clayton KellerF–F · PP | UTA | 1025 | +8.3 | +0.66 | .38.21 to .53, 81 GP | 6.8 |
| Jack Hughes + Luke HughesF–D · PP | NJD | 351 | +10.3 | +0.66 | .02−.23 to .26, 46 GP | 1.9 |
| Jakob Chychrun + Aliaksei ProtasF–D | WSH | 392 | +4.2 | +0.66 | .16−.11 to .41, 40 GP | 2.0 |
| Anthony Cirelli + Brandon HagelF–F | TBL | 629 | +6.8 | +0.66 | .27.05 to .46, 57 GP | 6.0 |
| Brady Tkachuk + Jordan SpenceF–D | OTT | 278 | +5.8 | +0.65 | −.06−.40 to .30, 23 GP | 1.6 |
| William Nylander + Matias MaccelliF–F | TOR | 298 | +3.4 | +0.65 | −.02−.36 to .33, 24 GP | 4.8 |
| Jason Robertson + Wyatt JohnstonF–F · PP | DAL | 477 | +1.3 | +0.65 | .46.18 to .67, 31 GP | 7.9 |
| Jesper Bratt + Luke HughesF–D · PP | NJD | 396 | +7.0 | +0.64 | .13−.12 to .36, 46 GP | 1.6 |
| Adam Pelech + Mathew BarzalF–D | NYI | 465 | +7.9 | +0.64 | .20−.05 to .42, 47 GP | 1.6 |
| Thomas Harley + Mavrik BourqueF–D · PP | DAL | 402 | +5.7 | +0.64 | .14−.11 to .37, 46 GP | 1.5 |
| Connor McDavid + Evan BouchardF–D · PP | EDM | 913 | +7.0 | +0.63 | .33.16 to .49, 81 GP | 4.2 |
| Bo Horvat + Mathew BarzalF–F · PP | NYI | 293 | +5.1 | +0.63 | .57.23 to .78, 19 GP | 6.5 |
| Artturi Lehkonen + Nathan MacKinnonF–F · PP | COL | 698 | +1.7 | +0.63 | .26.03 to .46, 53 GP | 7.7 |
| Jack Eichel + Jeremy LauzonF–D | VGK | 221 | +3.3 | +0.63 | .29−.18 to .65, 15 GP | 1.9 |
| Ryan Nugent-Hopkins + Connor McDavidF–F · PP | EDM | 548 | +3.2 | +0.63 | .32.07 to .52, 45 GP | 7.5 |
Pairs that create less together than apart
| Pair | Team | 5v5 min | xG% lift | xGF/60 lift | DK corr. (raw) | Stack |
|---|---|---|---|---|---|---|
| Ian Moore + Tim WasheF–D | ANA | 222 | −3.7 | −0.90 | −.06−.40 to .30, 23 GP | 0.6 |
| Keegan Kolesar + Jeremy LauzonF–D | VGK | 269 | −5.3 | −0.80 | −.03−.44 to .39, 17 GP | 0.7 |
| Alexander Wennberg + Collin GrafF–F | SJS | 224 | −8.2 | −0.76 | .29−.19 to .66, 14 GP | 4.1 |
| Lars Eller + Nick CousinsF–F | OTT | 221 | −0.3 | −0.66 | .20−.15 to .51, 24 GP | 2.0 |
| Cody Glass + Luke HughesF–D | NJD | 248 | −7.8 | −0.65 | −.02−.41 to .37, 19 GP | 1.1 |
| Mario Ferraro + Collin GrafF–D | SJS | 363 | −1.8 | −0.65 | −.22−.53 to .13, 24 GP | 1.0 |
| Matt Roy + Martin FehérváryD–D | WSH | 259 | −8.1 | −0.65 | .07−.45 to .55, 12 GP | 1.4 |
| Dmitry Orlov + Alexander WennbergF–D · PP | SJS | 401 | −5.4 | −0.64 | .13−.15 to .39, 38 GP | 1.2 |
| Alexander Wennberg + Kiefer SherwoodF–F | SJS | 236 | −6.5 | −0.63 | .28−.11 to .60, 20 GP | 4.2 |
| Beck Malenstyn + Bowen ByramF–D | BUF | 239 | −3.5 | −0.63 | −.23−.88 to .73, 5 GP | 0.9 |
Stack leaders by pair type
Forward with a defenceman
| Pair | Team | 5v5 min | xG% lift | xGF/60 lift | DK corr. (raw) | Stack |
|---|---|---|---|---|---|---|
| Connor McDavid + Evan BouchardF–D · PP | EDM | 913 | +7.0 | +0.63 | .33.16 to .49, 81 GP | 4.2 |
| Nathan MacKinnon + Cale MakarF–D · PP | COL | 735 | −0.4 | +0.18 | .22.02 to .40, 72 GP | 3.6 |
| Leon Draisaitl + Evan BouchardF–D · PP | EDM | 461 | +7.5 | +0.47 | .31.08 to .51, 49 GP | 3.2 |
| Nathan MacKinnon + Devon ToewsF–D | COL | 591 | +7.8 | +0.72 | .30.10 to .49, 61 GP | 3.1 |
| Zach Werenski + Adam FantilliF–D · PP | CBJ | 495 | +3.9 | +0.56 | .37.16 to .54, 57 GP | 3.1 |
| Kirill Kaprizov + Quinn HughesF–D · PP | MIN | 465 | +2.4 | +0.47 | .36.12 to .55, 46 GP | 3.1 |
| Nikita Kucherov + Darren RaddyshF–D · PP | TBL | 468 | +4.1 | +0.51 | .21−.03 to .43, 48 GP | 3.0 |
| Mark Scheifele + Josh MorrisseyF–D · PP | WPG | 728 | +1.3 | +0.13 | .32.14 to .48, 76 GP | 2.8 |
| John Klingberg + Macklin CelebriniF–D · PP | SJS | 416 | +1.6 | +0.35 | .37.14 to .57, 45 GP | 2.8 |
| Dylan Larkin + Moritz SeiderF–D · PP | DET | 518 | −2.8 | −0.07 | .29.08 to .48, 58 GP | 2.7 |
Defence partners
| Pair | Team | 5v5 min | xG% lift | xGF/60 lift | DK corr. (raw) | Stack |
|---|---|---|---|---|---|---|
| Filip Hronek + Quinn HughesD–D | VAN | 348 | +4.9 | −0.15 | .47.15 to .70, 24 GP | 2.4 |
| Quinn Hughes + Brock FaberD–D | MIN | 817 | +5.3 | +0.62 | .14−.10 to .37, 47 GP | 2.4 |
| Damon Severson + Zach WerenskiD–D | CBJ | 394 | +12.2 | +0.52 | .23−.15 to .56, 21 GP | 2.3 |
| Zach Werenski + Ivan ProvorovD–D | CBJ | 404 | −2.6 | −0.05 | .04−.33 to .40, 21 GP | 2.3 |
| Esa Lindell + Miro HeiskanenD–D | DAL | 1015 | −0.5 | +0.08 | .23.04 to .40, 76 GP | 2.2 |
Lists use pairs with 200+ minutes together at 5v5 (lift) or 20+ games with 5+ minutes together (stacks).
Chemistry by team
How chemistry is measured
- Together vs apart (the lift)
- For every pair of teammates with 50+ minutes together at 5v5 we compare the pair's expected goals for and against per 60 with each player's own numbers when the other is off the ice. The lift is the rate together minus the average of the two rates apart. "Together" is every shift the pair shared; "apart" is each player's 5v5 time for the team minus that.
- Shrinkage
- Raw lifts are noisy: 200 minutes is about ten expected goals. Each lift is shrunk toward zero by its own sampling variance against the spread of true lifts across the league (empirical Bayes), so a pair needs both a big gap and a big sample to rank high. We rank on the xGF/60 lift because it repeats season to season better than the xG-share lift; both are shown.
- DraftKings correlation
- For the fantasy side we score every game on DraftKings classic (goal 8.5, assist 5, shot 1.5, block 1.3, plus the hat-trick, 5+ shot, 3+ block and 3+ point bonuses; the shorthanded and shootout bonuses are left out because the box score lacks them) and correlate the two players' points across the games they spent 5+ minutes together. Most of that correlation is structural: linemates are credited on the same goals. What is specific to one pair does not carry over: a pair's own record predicts next season's correlation no better than the league value for its type (forward pair, forward with defenceman, defence pair). So the tables show each pair's raw value, with its 90% interval and games, as context only, and the stack score uses the pair shrunk on the Fisher-z scale toward its type with 400 games of prior weight, which in practice is close to the type value.
- Stack score
- Stack score = the pair's combined DraftKings points per game × its correlation shrunk toward the typical value for its type (about .28 for forward linemates, about .09 when a defenceman is involved). In practice it ranks high-scoring linemates: the correlation part is mostly set by the pair type, not by the pair. The Stack Finder shows the same correlation on each line and PP1 stack as a neutral chip; it does not change the expected-goals ranking there.
- Does it repeat?
- Every season we check whether the same pairs show the same chemistry the next season (year-over-year correlation of the metric). The table below shows the latest numbers. Lifts that do not repeat are mostly noise, and we say so instead of calling them an edge.
What the data says (2017-18 to 2025-26)
- Typical game-to-game DraftKings correlation, every season: forward linemates .27 to .30, a forward with a defenceman .07 to .10, defence partners .05 to .12, teammates who rarely share a shift .02 to .03. Stacking linemates is a real, stable edge.
- One pair's own deviation from its type barely repeats: season-to-season r of .07 for forward pairs, .05 for forward-defence and .12 for defence pairs. Forward pairs in the top quarter by shrunk correlation averaged .30 the next season, the bottom quarter .26.
- Shrinking each pair toward its type cut the error in predicting next season's correlation by about 45% against the raw value, every season. But the shrunk value did no better than the type average alone (squared error .0279 vs .0281 for forward pairs, .0299 vs .0301 forward-defence, .0220 vs .0223 defence pairs, all eight seasons pooled): a pair's own correlation adds essentially nothing beyond its type.
- With-you xGF/60 lift repeats at r of about .14 to .28 season to season; xG-share lift at about .06 to .21. Treat a big lift as this season's result with some carry-over, not a fixed trait.
| Measure | Split-half r | Sample | Verdict |
|---|---|---|---|
| xGF/60 lift (shrunk)Year over year, 2024-25 → 2025-26. The number we rank chemistry on. | 0.27 | 1272 pairs with 150+ min together both seasons | Moderate |
| xGF/60 lift (raw) | 0.25 | 1272 pairs with 150+ min together both seasons | Moderate |
| xG share lift (shrunk)Shown, not ranked on: it repeats less than xGF/60 lift. | 0.13 | 1272 pairs with 150+ min together both seasons | Noise |
| Pair DK correlation vs its type (raw)What is specific to the pair beyond "forward linemates" or "defence partners". Why each pair gets 400 games of prior weight. | 0.06 | 408 pairs with 25+ games together both seasons | Noise |
Latest season-over-season check; r is the correlation of the same pairs' values in consecutive seasons. Shrinkage priors this season: typical DK correlation forward pairs .28, F–D .09, D–D .10, with 400 games of prior weight.
FAQ
What is line chemistry in hockey?
Here it means two teammates producing better results together than apart. We measure it as the change in expected-goal share and expected goals for per 60 when they share the ice at 5v5, compared with each player's numbers without the other, shrunk for sample size.
Which NHL players should I stack in DFS?
Players whose DraftKings points rise and fall together. Across the league the typical game-to-game correlation is forward linemates .28, a forward with a defenceman .09, defence partners .10, against .02 for teammates who rarely share a shift. Stacking linemates, especially those who also share the first power-play unit, captures most of that. The stack score ranks pairs by their combined DraftKings scoring times the correlation typical for their pair type.
Is a specific pair's correlation reliable?
Not much beyond its type. Across eight seasons, what is specific to one pair (its correlation minus the typical value for forward pairs, forward-defence or defence pairs) repeated the next season at only r ≈ .05 to .12, and .06 in the latest check. Predicting next season from the shrunk pair value did no better than using the type average alone. So treat a pair's raw correlation as a description of this season, not a forecast; the stack score effectively ranks linemates who both score a lot, and that part is reliable.
Why does a famous duo show a small lift?
Lift compares the pair together with each player apart. Two stars who are excellent with anyone can have a small lift even though both are great: they do not need each other. Lift also shrinks toward zero when the time apart is short, because then there is little to compare against.
How often is this updated?
Nightly during the season, after the previous night's games are processed. Early in a season the pages show last season until enough pairs have 50+ minutes together.