[Ganchrow]

For updated data please see my Half-Point Calculator.

Total	N	Std. Err.	Freq.
5	789	1.556%	25.602%
6	2,296	0.630%	10.148%
7	847	1.424%	21.960%

---

Methodology:

• All NHL regular season final scores and closing over/unders (from http://wwww.cover.com/) from the 2005/06 season through October 23^rd of the 2007/8 season were analyzed (2,584 games in total) for the push frequencies of various totals.
• The push frequency for a given over/under was determined by the percentage of games with a closing over/under within 0.5 points of the over/under in question ending with a total equal to that over/under. For example, the push frequency of an over/under of 6 was determined from all games with a closing over/under between 5.5 and 6.5.
• This is based on Stanford Wong's methodology as described in Sharp Sports Betting.

[Munson15]

Are you saying that playing for a side with an under 5.5 goals and an over 5 goals would be profitable? Assuming that the juice is less than what %?

[Ganchrow]

Originally Posted by Munson15

Are you saying that playing for a side with an under 5.5 goals and an over 5 goals would be profitable? Assuming that the juice is less than what %?

It could be profitable ... it depends on the payout odds offered on each bet (as well as your estimates of the probabilities of the game going over or under 5 -- unless you structure your bets irrespective of these probabilities). The juice is irrelevant.

Assuming you bet to equalize losses if the total lands other than 5 (thus rendering further probability estimates moot), then for a profitable betting opportunity to exist the decimal odds on the over and under must satisfy the following inequality:

du5½  >	do5
	do5 - 74.398%

or equivalently

do5  >  74.398% ×	du5½
	du5½ - 1

where d_o5 and d_u5½ correspond to the decimal odds on the over 5 and under 5½, respectively.

So, for example, if the odds on the over were -110, the odds on the under would need to be at least / ( - 74.968%) ≈ for a profitable half-middle to exist.

Conversely, if the odds on the under were -110, the odds on the over would need to be at least 74.968%* / ( - 1) ≈ for a profitable half-middle to exist.

[chemist]

Originally Posted by Ganchrow

Total   N  Std. Err.  Freq.
5      789    1.556%   25.602%
6    2,296    0.630%   10.148%
7      847    1.424%   21.960%

I know nothing of NHL but it seems strange that 5 and 7 should be so much more common totals given the spread total than is 6. What is the explanation for this?

Your text doesn't seem to quote very well.

[Ganchrow]

Originally Posted by chemist

I know nothing of NHL but it seems strange that 5 and 7 should be so much more common totals given the spread total than is 6. What is the explanation for this?

With the advent of the shootout NHL games can no longer end in a tie score.

Originally Posted by chemist

Your text doesn't seem to quote very well.

It's actually an HTML table. For future refernce you can just copy and paste the table data into a pair of [code][/code] tags.

[3put]

Ganchrow, I would like to know if it is possible to calculate the expected total given the o/u lines.
Examples from NHL today:

Boston vs. Carolina 5.5 -109/-101
Philadelphia vs. Ottawa 6 -131/121

[Ganchrow]

Originally Posted by 3put

Ganchrow, I would like to know if it is possible to calculate the expected total given the o/u lines.
Examples from NHL today:

Boston vs. Carolina 5.5 -109/-101
Philadelphia vs. Ottawa 6 -131/121

No, not without further defining the probability distribution beyond those values given for the 5, 6, and 7.

But just to be clear, is it really the expected total for which you're looking?

[3put]

I am interested in the connection between the puckline and the total in a game.
Obviously in a highscoring game the favorite has a greater probability of winning by at least 2 goals.
Ottawa is the favorite today and the o/u line is 6 -131/121 indicating that the book expect that there will be scored ca. 6.2 goals on average were this game played many times.
That is what I mean by expected total.

I wondered if the push probabilities and the o/u line in combination was enough to calculate this expected total.

[Ganchrow]

Originally Posted by 3put

Ottawa is the favorite today and the o/u line is 6 -131/121 indicating that the book expect that there will be scored ca. 6.2 goals on average were this game played many times.
That is what I mean by expected total.

Actually an over/under defines a percentile (which would be the median in the case of a balanced market) not an expectation. A line of 6o-131/6u+121 implies that about 55.62% of non-pushed totals will go over 6 and 44.38% will go under, but does not speak to the goals scored on average. In general the average total should be higher than the median total due to the fact that totals can never be less than 1 but have no theoretical maximum value (in other words, scoring distributions are positively skewed).

Originally Posted by 3put

I wondered if the push probabilities and the o/u line in combination was enough to calculate this expected total.

As I said, you'd first need to fully define the set of outcome likelihoods (which can be approximated with the push probabilities adjusted for the current line -- this is how the HPC estimates these figures). Once you have a set of outcome likelihoods with which you're satisfied, calculating the expected total becomes a trivial matter of taking the dot product of the vector of totals and the vector of associated outcome probabilities.

[rjp]

There's a reason why this kind of info isn't released...

[Obioha]

So, Sir, if I have a sportsbook that likes to post "unconventional" totals on Playoff Hockey games, according to your push frequency calc, I just need better than 3-1 in order to make money?

[Ganchrow]

Originally Posted by Obioha

So, Sir, if I have a sportsbook that likes to post "unconventional" totals on Playoff Hockey games, according to your push frequency calc, I just need better than 3-1 in order to make money?

Please illustrate your question with a specific example.

[Obioha]

If one sportsbook hangs up over 4' at -195 and another posts under 5 at -105, that is still a profitable wager?

The reason I ask is because I only take the wager when the odds are approx 4-1, but you might be giving me a fine tuning.

[Ganchrow]

Originally Posted by Obioha

If one sportsbook hangs up over 4' at -195 and another posts under 5 at -105, that is still a profitable wager?

The reason I ask is because I only take the wager when the odds are approx 4-1, but you might be giving me a fine tuning.

According to the above probabilities, no, it would not be