Learn
Short answers to the questions people ask about hockey numbers and betting, in the sense this site uses them. The first sentence of each answer stands on its own; the rest is detail.
What is the Grit & GF rating?
Grit & GF is a 0 to 100 NHL skater rating built from shift data for the things box scores skip: physicality, goals and chances from the hard areas (GF), intimidation, chemistry with linemates, and late-game play in one-goal games.
Each part is measured per season from the NHL's play-by-play and shift charts, pulled toward average when the sample is small, and turned into a z-score against forwards or defencemen. Four of them are weighted into the rating (physicality 38%, GF 38%, chemistry 13%, intimidation 11%) and the result is reported as a percentile, so 50 is the middle of the league. Clutch is still shown, but since October 2026 it counts for nothing, because it didn't repeat from one season to the next.
Separately, the parts with a measured link to scoring get a goals-above-average value. Hits, blocks and takeaways count toward the rating and are valued at zero goals.
The methodology requires 15 to 20 games to be accurate. Before that, a player's rating leans on the league average for his position, so early-season numbers are provisional.
It describes how a player plays. In our test on 2024-25 and 2025-26, the rating didn't predict game results beyond expected goals; one part, chemistry, did.
Does the Grit & GF rating predict wins?
Not the rating as a whole. Across 2,624 games in 2024-25 and 2025-26, last season's Grit & GF ratings of the dressed lineups explained no goal differential beyond our expected-goals game model. The chemistry part did, about 0.15 goals a game per standard deviation of the gap between lineups.
The rating repeats from season to season about as well as points per 60 (r of 0.80), and its top 25 is mostly not the league's top scorers. Clutch doesn't repeat at all, so it no longer counts in the rating. Physicality has no measurable effect on goals, which is why hits and blocks are valued at zero.
Chemistry held up after controlling for how talented the lineup was, but it was much stronger in 2024-25 than in 2025-26. It is now in a forward test: each morning a shadow version of the game model adds it, using that day's projected lineups, and the shadow prices are scored against the live ones at the end of the season under rules written down in advance.
What are expected goals (xG) in hockey?
Expected goals is the probability that a shot becomes a goal, based on where it was taken and how it happened. Add up a team's shot probabilities and you get its xG for a game.
Our model is trained on several seasons of NHL unblocked shots and uses distance, angle, shot type, rebounds, rush chances and the game situation. Empty-net shots are excluded from team totals.
Over a few weeks, a team's xG predicts its future goals better than its actual goals do, because finishing and goaltending swing a lot game to game.
What does xGF% mean?
xGF% is a team's or player's share of the expected goals in the minutes being measured. Above 50% means creating more scoring chances than allowing.
For players we usually quote it at 5-on-5, where special teams don't distort it. A player's on-ice xGF% depends heavily on his linemates and deployment.
What is goals saved above expected (GSAx)?
GSAx is the expected goals a goalie faced minus the goals he allowed. A positive number means he stopped more than an average goalie would have on the same shots.
It is noisy over a season, which is why our game model shrinks goalie ratings hard toward average before using them.
What is the puck line in NHL betting?
The puck line is hockey's point spread, almost always 1.5 goals. The favourite at -1.5 has to win by two or more; the underdog at +1.5 covers with any win or a one-goal loss.
An overtime or shootout win is always by exactly one goal, so the favourite at -1.5 can only cover in regulation. Late empty-net goals make two-goal wins more common than a simple scoring model expects, and our game model accounts for that.
What is a no-vig or fair probability?
A no-vig probability is what a sportsbook's odds imply once its margin is taken out. If both sides of a market imply 52.4%, the fair split is 50/50.
We remove the margin from each book's two-way prices, then take the consensus across books as the market's fair probability.
What is an edge in sports betting?
An edge is the gap between your probability for an outcome and the probability the price implies. On this site it is our model's probability minus the market's no-vig probability, priced at the best available line.
We blend the model with the market before measuring an edge, because the market knows things the model can't see, like injuries and morning-skate news. Edges that come from one book alone, or that look too good, stay off the board.
What is closing line value (CLV)?
Closing line value measures whether you got a better price than the final odds before the game started. Consistently beating the closing line is the best early evidence that a bettor has an edge.
Wins and losses take seasons to separate skill from luck. CLV takes a few hundred bets, because the closing line is the sharpest estimate the market produces.
How are NHL shots on goal props priced?
We project a player's shots on goal from his expected ice time, his shot rate and the opponent's shots allowed, then turn that into a full distribution so any line (1.5, 2.5, 3.5) gets a probability.
The model is tested on a season it wasn't fitted on before changes go live. For three or more shots it beats a constant-rate guess by a wide margin, and the live calibration is on the report card.
Why are same-game parlays priced differently from regular parlays?
Legs from the same game are correlated, so multiplying their odds gets the price wrong. A winger scoring makes his team winning more likely, so that pair hits more often than two independent bets would.
Our parlay builder simulates the game 40,000 times with the legs linked and shows the fair price next to the independent price. Books know about correlation too and build a large margin into same-game prices.
What is the Kelly criterion?
The Kelly criterion sizes a bet as a fraction of your bankroll based on your edge: (b × p − q) / b, where b is the net decimal odds, p your win probability and q = 1 − p.
It assumes your probability is right. Most bettors use a quarter of the Kelly stake or less, because overestimating your edge at full Kelly drains a bankroll quickly.
Do back-to-backs matter in NHL betting?
Yes, mostly through the goalie. In our 2024-25 and 2025-26 test games, teams started the previous night's goalie again less than one time in ten, and home teams won 62% when only the road team was on the second night.
Once the starting goalie is guessed correctly, a separate fatigue adjustment fitted on 2024-25 made 2025-26 predictions slightly worse, so our game model doesn't use one yet. The schedule page shows rest for every team so you can judge it yourself.
What is NHLe (NHL equivalency)?
NHLe converts a player's scoring in another league into the NHL points it would roughly equal, using a factor for each league. A point in the KHL is worth far more than a point in junior hockey.
Our prospect model goes further: it learns from what actually happened to the 2010 to 2017 draft classes, and adjusts for age, draft slot, how much his team scores and how good his linemates are.
How are prospect tiers decided?
A prospect's tier is the highest outcome he has at least an even chance of reaching, based on past draft picks with similar numbers at the same age. Elite means a top-30 forward pace, about 75 points per 82 games.
Each prospect page shows the full ladder of odds, from becoming an NHL regular (150+ games) to elite, with a scouting report written from his numbers.
How accurate is the Grit & GF game model?
On two seasons of games it was never fitted on, the side the model favoured won 56.6% of the time, against 54.3% for always taking the home team. Its probabilities run about a sixth too confident, so we blend it with the market before posting edges.
It shows no skill on totals, so totals edges follow the market. The tests rerun every Monday and the live numbers are on the report card.
Where is today's free NHL pick?
Each game day the biggest edge on the board that passes our rules goes out free, with the reasoning, in the daily post on the blog and on the home page. It posts by noon Eastern once the morning odds are in.
Every pick is stamped before puck drop and graded the next morning, wins and losses both.
Last updated 2026-10-15. How each number is built is on the methodology page.