Betting on NHL Games

1
Betting on NHL Games

• Matthew S. Baker
• Principal
• Edge Research

2
Why Hockey?

• If you want to find bets based on doing your own
  analysis of existing data, I suggest staying away
  from the National Football League.
• Stanford Wong

• NHL the smallcap of sports betting

3
Game of Hockey

• Similar to soccer
• 3 periods of 20 minutes, most goals wins
• Each team has 3 offense, 2 defense, 1 goalie
• Substitutes on the fly
• 82 game regular season
• Baseball style series post season

4
Legal Betting

• Las Vegas
• good odds, safe, inconvenient

• Internet
• good odds, less safe, convenient

• Offshore
• less good odds, safe, inconvenient

5
Betting Fundamentals

• Money Line

• Negative is favorite -150 means betting 150
  wins 250 if you are right, else lose 150.

• Positive is underdog 150 means betting 100
  wins 250 if you are right, else lose 100.

6
The Data

• All NHL games 1997 - 2002
• Home team, visiting team
• Final score
• Regular season or playoff game
• Opening money line

7
Visitor Team Scoring
8
Home Team Scoring
9
Visitor Team Payoffs
10
Home Team Payoffs
11
Regular Season

• Average Visitor Score 2.57
• Average Home Score 2.82
• Visitor Win Rate 38
• Home Win Rate 48
• Tie 14
• Average Line 139

12
Playoffs

• Average Visitor Score 2.35
• Average Home Score 2.60
• Visitor Win Rate 47
• Home Win Rate 53
• Tie 0
• Average Line 149

13
Model Variables

• Trailing 20 Game Win Rate
• Trailing 50 Game Win Rate
• Trailing 20 Game Goals For / Against
• Trailing 50 Game Goals For / Against
• Playoff Indicator
• Visiting Team Indicator
• Consensus Expectation Opening Line

14
Solution Framework

• Chance Team A Beats Team B
• NORMSDIST( (MuA MuB) / SQRT(SigA2 SigB2) )

• MuA, MuB, SigA, SigB not observable

• Use historical data to estimate

15
Univariate Visitor Win Rate
16
Stages of Model Development

• Multivariate/Backward on Absolute Win Chance

• Multivariate/Backward/Ridge on Relative Win Chance

• Incorporation of Schedule Strength

17
Absolute Model Results
18
Relative Model Results w/SS
19
Simulation Results
20
Further Research

• Shots on net
• Goalie save percentage
• Out of sample (2002/2003 season)
• Live testing!

21
Conclusions

• Bet on visitors during playoffs
• Bet on underdogs
• Bet on team with worse recent record, but
  comparable or better goal scoring

22
References

• Wong, Stanford, Sharp Sports Betting
• Thorp, Ed, The Kelly Criterion in Blackjack,
  Sports Betting, and the Stock Market
• McCune, Bob, Insights into Sports Betting