math question and the effect on over/under percentages..please help

[DrSues02]

I'm working with an NHL database to come up with some over/under percentages for various periods. I have all of the games from this current season for every time broken down by period.

I then have these calculated to determine that teams Over or Under 1.5 percentage, a line that is commonly offered for each period.

Then, I have a summary page where I can average the values for two teams who will be playing that day:

Example: Detroit goes under 54.55% of the time and PHI goes under 43.64% of the time.
If they are playing each other, that becomes an average of 49.09% on the under.

I hope everyone can follow me so far. Everything seem logical up to this point?

Then, since many NHL teams use various goalies, I have an average of each teams goals allowed per game. Det might give up 2.61 goals per game...this number is an average of ALL the goaltenders they might use during the season.

I want to use the goaltender for THAT particular game. If Goalie #1 gives up 3.09 goals per game, he would be 118.39% ABOVE the average for the team.

Therefore, the team would theoretically give up more shots and the game would go OVER more often than average.

Everyone following this?

My problem: How do I incorporate this goalie number into the averages I have for the OVER/Under?

I've tried taking the average of the two goalies (say its 105% above their teams averages), then multiplying that by the current over/under percentage.

An example:

Goalies:
DET MACDONALD 118.39%
PHI LEIGHTON 100.18%
Average for the combined goalies:
109.28%

O/U:
49.09% 50.91%

Weighted by Goalies:
O/U:
53.65% 55.64%

Basically, multiplying 109.28% by the O/U numbers adds 9.28% to the total. However, as u can see, the %'s no longer add up to 100%.

Anyone have any ideas or thoughts on where I'm going with this?

Please let me know. I've spent a very long time working on this.

Thanks.

DrSues02

[136913691369]

[ QUOTE ]
I want to use the goaltender for THAT particular game. If Goalie #1 gives up 3.09 goals per game, he would be 118.39% ABOVE the average for the team.

Therefore, the team would theoretically give up more shots and the game would go OVER more often than average.

[/ QUOTE ]

This is not necessarily true. The goalie could (and most likely ) just gives up more goals with an equal # of shots because he is not as good. Your point on the game going over more is valid though.

This would be a VERY time consuming process, but can't you just look up the goalie for each game and use all those games as the average for the team, instead of saying that he gives up x% more then the total average.

I know you're using a variant of my system, but IMO you're making it WAY too complicated. You're on a slippery slope here, if you're going to break the games down so far as each goalie being used, why don't u consider the quality of the ice at game time, the referees in the game, and other IMO not so important factors that play into each game.

I understand that different goalies will effect the outcome, but for all that work you have to do to come up with these stats, I don't think it will raise your ROI enough to justify the extra work.

[Mediocre_Player]

I agree with 139 - The only exception might be teams that specifically employ a 60/40 split or smaller in goalie duties. I can only think of 1 such team: San Jose.

Perhaps Tampa as well.

[npc]

I'm not knocking the OP, but I REALLY HATE this forum's software. One misstep and a lot of writing is lost. This will probably be the last long post I'll reply to with proper context.

[ QUOTE ]
I'm working with an NHL database to come up with some over/under percentages for various periods. I have all of the games from this current season for every time broken down by period.

...

Example: Detroit goes under 54.55% of the time and PHI goes under 43.64% of the time.
If they are playing each other, that becomes an average of 49.09% on the under.

I hope everyone can follow me so far. Everything seem logical up to this point?

[/ QUOTE ]

As long as one opponent is above average and one is below average, this is fine. If both are above or below average this method is badly flawed. I hope you can see why.

[ QUOTE ]

My problem: How do I incorporate this goalie number into the averages I have for the OVER/Under?

[/ QUOTE ]

Here's how I'd do it (and how I've done it for baseball totals). I'd record who the goalie is for each game and treat each "team:goalie" combination individually. This eliminates your problem.

I haven't carefully analyzed your results, but I'm guessing that your math problem is due to the fact that each goalie hasn't played half of all games. So, no matter what, you have to track how many games are played by each goalie so you know how to weight your percentages, so you may as well break their stats out separately.

Make sense?

[MyTurn2Raise]

[ QUOTE ]
, if you're going to break the games down so far as each goalie being used, why don't u consider the quality of the ice at game time, the referees in the game, and other IMO not so important factors that play into each game.

I understand that different goalies will effect the outcome, but for all that work you have to do to come up with these stats, I don't think it will raise your ROI enough to justify the extra work.

[/ QUOTE ]


because goalie in net probably has a significant effect of decent magnitude and is easily obtained
whereas the other factors mentioned don't line up that easily.

[DrSues02]

[ QUOTE ]

I'm not knocking the OP, but I REALLY HATE this forum's software. One misstep and a lot of writing is lost. This will probably be the last long post I'll reply to with proper context.

[ QUOTE ]
I'm working with an NHL database to come up with some over/under percentages for various periods. I have all of the games from this current season for every time broken down by period.

...

Example: Detroit goes under 54.55% of the time and PHI goes under 43.64% of the time.
If they are playing each other, that becomes an average of 49.09% on the under.

I hope everyone can follow me so far. Everything seem logical up to this point?

[/ QUOTE ]

As long as one opponent is above average and one is below average, this is fine. If both are above or below average this method is badly flawed. I hope you can see why.

[ QUOTE ]

My problem: How do I incorporate this goalie number into the averages I have for the OVER/Under?

[/ QUOTE ]

Here's how I'd do it (and how I've done it for baseball totals). I'd record who the goalie is for each game and treat each "team:goalie" combination individually. This eliminates your problem.

I haven't carefully analyzed your results, but I'm guessing that your math problem is due to the fact that each goalie hasn't played half of all games. So, no matter what, you have to track how many games are played by each goalie so you know how to weight your percentages, so you may as well break their stats out separately.

Make sense?

[/ QUOTE ]

Honestly, I don't see why where my math is flawed in the original average. If there is a better way to average them both..not taking into account the goalie factor, how would u go about the math?

[adanthar]

[ QUOTE ]
Honestly, I don't see why where my math is flawed in the original average. If there is a better way to average them both..not taking into account the goalie factor, how would u go about the math?

[/ QUOTE ]

If DET goes under 55% of the time and so does PHI, you assume that a DET-PHI game will be under 55% of the time, as well. But chances are that this is already factored into the line.

[136913691369]

[ QUOTE ]
[ QUOTE ]
, if you're going to break the games down so far as each goalie being used, why don't u consider the quality of the ice at game time, the referees in the game, and other IMO not so important factors that play into each game.

I understand that different goalies will effect the outcome, but for all that work you have to do to come up with these stats, I don't think it will raise your ROI enough to justify the extra work.

[/ QUOTE ]


because goalie in net probably has a significant effect of decent magnitude and is easily obtained
whereas the other factors mentioned don't line up that easily.

[/ QUOTE ]

you are right and I guess I exaggerated a little when I compared it to those other factors. The goalie in net has a much greater impact on over/under totals then the quality of ice or referees. I was just trying to make a point that there are a basically unlimited # of factors that go into each and every game and to try to analyze them all is very unlikely.

[DrSues02]

[ QUOTE ]
[ QUOTE ]
Honestly, I don't see why where my math is flawed in the original average. If there is a better way to average them both..not taking into account the goalie factor, how would u go about the math?

[/ QUOTE ]

If DET goes under 55% of the time and so does PHI, you assume that a DET-PHI game will be under 55% of the time, as well. But chances are that this is already factored into the line.

[/ QUOTE ]

What I'm looking for is when it ISNT factored into the line. If they are both going under 55% of the time, then the average between the 2 teams is 55% U. The fair line for this is -122.

If i can find a line at +100 for this bet, wouldn't there be value there?

Please let me know if you think my math is flawed.

DrSues02

[dsteff]

You're making a couple of important math mistakes throughout.

The first is that you're not recognizing the difference between median and mean. When working with small numbers (such as goals in a period) this can be a significant mistake. For example, a team that scores an average of 4 goals a game may seem like a good bet to score more than 2.5 goals in their next game. However if that team has scored 2, 2, 2, 2, and 12 goals in the five games they've played they have been under the 2.5 line four out of the five times.


This is a median based calculation:
[ QUOTE ]
Detroit goes under 54.55% of the time and PHI goes under 43.64% of the time.

[/ QUOTE ]

This is a mean based calculation:
[ QUOTE ]
If Goalie #1 gives up 3.09 goals per game, he would be 118.39% ABOVE the average for the team.

[/ QUOTE ]

You cannot calculate a weighted O/U % from these two statistics alone.


The second math mistake is related. In your final calculation you attempt to multiply % O/U and % goals. These two units don't match. You're attempting to factor in the goalkeeper way to late in your calculation. As npc suggested a good solution would be too keep your data separated by team:goalie then do the calculations from there.