I'm working on a couple models for hockey and was looking for some help, ideas or feedback
Method 1
The first one is similar to a baseball model I got where the basics are you take a teams goals scored divide by the league average then mulitpy by 100 and then divide by the goals against.
So lets say the averge is 2.75 and Pittsburgh averages 3 goals per game while giving up 2.5 goals per game
You would divide 3 by 2.75 = 1.09 then mulitply by 100 = 109 then divide by 2.5 = 44
Then do that with the opponent and lets say they're final number is 30
You divive 44 by 30 = 1.46 and then multiply by 100 = 146
Pittsburgh is a -146 favorite
And then if Pittsburgh is at home add 30 cents -176 (I know this is crude but its just too early in the season to get an accurate home ice advantage for each team)
The baseball model I got is much more complex then that it has many more factors and weights each of them differently but its the same concept and it uses league min not league average I'm just using league average for my hockey model because its easier to combine the current form number.
My deal with this method is you can use a bunch of factors for the offense number but for defense your just using goals against or runs against.
An example from tonights game
Calgary +129
Phoenix -129
actuall line from pinny
Calgary +111
Phoenix -120
I did it for the other games tonight and they all landed in between the juice apart from Tampa but they played 3 games in 4 nights so I dont know how much to adjust the line for that and the Jersey game which I know the is'nt even going to come close becuase the line is kind of retarded having the worst team in hockey as a -170 fav but they are really due to win they're first home game of the season.
Method 2
My other model is just converting a football yards per point system to hockey by using goals and shots instead of points and yards. The problem with this method is I get a good predicted score which could be usefull for totals but cant make a usefull line out of it.
Heres an example from tonights games
Calgary 2.73
Phoenix 2.83
As you can see it could be usefull for game or team totals but getting a usefull line not so much.
Method 3
My other thought was nothing really matters in hockey apart from winning the game if a team wins who cares about any other stats. So I was thinking if you take a overall win% plus they're home/away win% and current form last 5 or last 10 games win% and combine them you should in theory get a good idea on how often that team should win.
The problem with this method is combining win% for each team to get a predicted outcome for a game. If you use log 5 formula you dont get a reasonable answer. Last season the best team was Washington at 67% and the worst team was Toronto at around 33% without adding in home ice advantage with log 5 Washington would be -412 and inreality the game was only in the low -300's with Washington at home.
So I was just looking for some feedback or thoughts of either of those 3 methods
Method 1
The first one is similar to a baseball model I got where the basics are you take a teams goals scored divide by the league average then mulitpy by 100 and then divide by the goals against.
So lets say the averge is 2.75 and Pittsburgh averages 3 goals per game while giving up 2.5 goals per game
You would divide 3 by 2.75 = 1.09 then mulitply by 100 = 109 then divide by 2.5 = 44
Then do that with the opponent and lets say they're final number is 30
You divive 44 by 30 = 1.46 and then multiply by 100 = 146
Pittsburgh is a -146 favorite
And then if Pittsburgh is at home add 30 cents -176 (I know this is crude but its just too early in the season to get an accurate home ice advantage for each team)
The baseball model I got is much more complex then that it has many more factors and weights each of them differently but its the same concept and it uses league min not league average I'm just using league average for my hockey model because its easier to combine the current form number.
My deal with this method is you can use a bunch of factors for the offense number but for defense your just using goals against or runs against.
An example from tonights game
Calgary +129
Phoenix -129
actuall line from pinny
Calgary +111
Phoenix -120
I did it for the other games tonight and they all landed in between the juice apart from Tampa but they played 3 games in 4 nights so I dont know how much to adjust the line for that and the Jersey game which I know the is'nt even going to come close becuase the line is kind of retarded having the worst team in hockey as a -170 fav but they are really due to win they're first home game of the season.
Method 2
My other model is just converting a football yards per point system to hockey by using goals and shots instead of points and yards. The problem with this method is I get a good predicted score which could be usefull for totals but cant make a usefull line out of it.
Heres an example from tonights games
Calgary 2.73
Phoenix 2.83
As you can see it could be usefull for game or team totals but getting a usefull line not so much.
Method 3
My other thought was nothing really matters in hockey apart from winning the game if a team wins who cares about any other stats. So I was thinking if you take a overall win% plus they're home/away win% and current form last 5 or last 10 games win% and combine them you should in theory get a good idea on how often that team should win.
The problem with this method is combining win% for each team to get a predicted outcome for a game. If you use log 5 formula you dont get a reasonable answer. Last season the best team was Washington at 67% and the worst team was Toronto at around 33% without adding in home ice advantage with log 5 Washington would be -412 and inreality the game was only in the low -300's with Washington at home.
So I was just looking for some feedback or thoughts of either of those 3 methods
