Mathematical Based NHL PredictionsThe one thing that we have never really done is look at a prediction model for NHL games, so we'll give it a shot here. This is one of those articles where I really have no idea how successful the method will be, but am basing it on models that have had a fair amount of success in predicting other sports.
What I am hoping, is much like the Statistical Baseball Betting System it's a case of believing it's harder than it actually is to come up with a model-based NHL system.
This method closely follows the Predicting NBA Games With Home-Away Averages method that uses home and away scoring averages, so it may be best used at the tail end of this season, or possibly during a season that has a full schedule, as the sample size that we have for the current season is extremely small, with most teams having played between eight and 10 home or away games.
The MethodThe first thing you will need are the home and away median scores, unless you're more comfortable using averages, in which case you'll want to calculate the home and away average scores. As I've often said, I prefer medians, as they are not influenced nearly as much by one or two extremely high-scoring or low-scoring teams, not to mention they are much faster to calculate.
Through games played Wednesday, Feb. 20, the median score for home teams was 2.75 goals and they allowed 2.54 goals. If you're looking at it from the road team's point of view, then it's obviously 2.54 goals scored and 2.75 goals allowed per game.
Next, look at the home or away scoring for the two teams involved in the game you want to handicap. For this example, we'll use Vancouver at Dallas and see the Canucks are averaging 3.0 goals in road games and allowing 2.3, while the Stars are averaging 3.2 goals in home games and allowing 3.0.
The next step is to divide each team's scoring averages into the league median so that you get a percentage that tells you how each team does compared to an average team.
Vancouver's 3.0 goals scored divided by the median goals scored by road teams of 2.54=1.181 and the Canucks' 2.3 goals allowed divided by median goals allowed by road teams of 2.75=.836.
For Dallas, we'll do the same thing, but this time be dividing into the median scores of home teams, so that Dallas' 3.2 divided by 2.75=1.164 and the Stars' goals allowed of 3.0 divided by 2.54=1.181.
The next step is to multiply each team's offensive percentage by the opposition's defensive percentage and then multiply that by the median number of goals scored for teams on the road, for the visiting team, or at home, when you are calculating the predicted number of goals scored by the home team.
For Vancouver, you will multiply its offensive percentage of 1.181 by the Stars' defensive percentage of 1.181 by the median number of goals scored by road teams, which is 2.54. So for the Cauncks you have 1.181*1.181*2.54=3.543, which is the team's predicted number of goals for the game.
For the Stars, you have 1.164*.836*2.75=2.676, which is Dallas' predicted goals in the game.
Obviously teams aren't going to score 2.676 goals, but what we can do is create our own line and to do that we need to go back to our old pal Pythagoras, which means a bit more math is involved.
So take each team's predicted goals and square that number. Vancouver's 3.543 becomes 3.543*3.543=12.553 and Dallas' 2.676*2.676=7.161.
If you remember back to Pythagorean Formula and Basketball Betting the next step is to divide the larger of the two numbers into the sum of the two numbers, or divide 12.553 by (12.553+7.161=19.714) 19.714 and 12.553/19.714=.637, meaning Vancouver has a 63.7% chance of winning the game.
To convert a percentage into a money line, subtract your percentage from 1 and you get .363. Divide the larger number by the smaller number and .637/.363=1.754 or -175. Therefore our line on the game is Vancouver -175 with a total of 6, or 6-over, if you prefer.
The posted odds were Vancouver -150 with a total of 5.5-under, so you would have a lean to the over, but probably not enough of a cushion to make a play.