Looking for some help - Hockey winning margin models

**The General** · 04-15-09, 11:12 PM

I cannot help you, but Welcome to SBR.

**MrLuckyPants** · 04-15-09, 11:23 PM

hehe thank you sir.

**Pancho sanza** · 04-16-09, 06:23 AM

Are those hypothetical #'s?

Your "by 1's" are way too high.

**MrLuckyPants** · 04-16-09, 10:17 AM

Not my numbers no.... I dont want to mention the sportsbook that I got them from, I'm assuming I'd get in crap. odds of 2/1 for a 1 goal win is too big for sure bro.

**Justin7** · 04-16-09, 10:57 AM

I think you need to use the total also.

**MrLuckyPants** · 04-16-09, 01:03 PM

that makes sense Justin - perhaps also GF/GA home and away would be just as crucial to figure out something that is closer to a true %. Ah well... I'll crack on

**TomG** · 04-16-09, 03:28 PM

In general, one sided propositions such as these are almost always bad bets. The juice is too high. With the exception of gross incompetence on behalf of the bookmaker, there is almost never value in these types of lines. Work on a model that values puck lines and alt puck lines instead.

**Neil Nollidge** · 04-24-09, 06:30 PM

Tip-strength mechanics for line-probability sets.

I do not know of any comprehensive calculator/( computer program ) for this sort of problem, but there could be. This thread seems to come from a bookmaker perspective, but the punter/bookmaker line is becoming more blurred. The 1/2 point calculator is helpful. There are 2 key hurdles to jump over: 1. Some sports have significant variation in high/low scoringness expectation - ( weather conditions & some defensive/attacking playing styles ). 2. Some, perhaps most, sports have significant score-line probability trend skewing on and near the tie line - ( leading-team time-burning possession ploys ). However, the tip-strength concept can be used to easily calculate a foundation/guideline probability set. I think that it is fairly accurate with Australian rules football, and it might be good for fantasy sports and jockey challenges. A sufficient parameter set would be a pair of 1/2 pointed probability sets. If there is no skewing, a tie probability plus one 1/2 pointed probability set will suffice. Another no-skew parameter set: the tie probability plus the raw win market ( Pf = Du*( 1 - Pt )/( Du + Df ) ) ( tie/push = refunds ). Nitty gritty: Define all possible scorelines in terms of the fav. Pn: probability that the fav score + n points > dog score. T[a,b]: strength required to convert Pa into Pb. T[a,b+m] = ( T[a,b] )^( ( b + m - a )/( b - a ) ). T[a,b] = ( 1 - Pa )*Pb/( Pa*( 1 - Pb ) ). Pb = ( (1/Pa - 1 )/T[a,b] + 1 )^-1. Ok, so we have the tools - eg: Define Tt = T[-1/2,1/2]; P(n-1/2) = ( ( 1/Pf - 1 )/( Tt )^n + 1 )^-1. And, of course, P( fav wins by q ) = P(q+1/2) - P(q-1/2). A possibility for tie-skew games, is to use wide parameters, and use another method ( stat sample? ) to find scoreline probabilities close ( far out enough to avoid including significantly skewed line probabilities ) to and on the tie line. Hmm, this does not look as simple as I thought it would. But when one gets a feeling for tip-strength, these sort of calculations are a natural progression.

**Neil Nollidge** · 04-24-09, 06:37 PM

I suppose that I should have pointed out the parameter set involving the market, generates a market version probability set.

**MrLuckyPants** · 04-24-09, 11:08 PM

Thanks alot Neil

I appreciate the theory.

I'm not as gifted in math as you are, but it makes sense. Just takes me a little longer than most lol.

I will have to give it a shot when I have the time to mess with it. Might be able to take some bets on exchanges in the future on a winning margin market lol.

**Neil Nollidge** · 04-27-09, 05:18 AM

CRUCIAL AMENDMENT - see my main post on this thread. When you see the words " of course ", be wary. Where you see " fav wins by q ", replace " fav " with " dog ".

**Arilou** · 04-27-09, 06:35 PM

Justin's comment is the key one: The total matters and it matters a lot. For proof, look at the goal-lines for different games. Those lines can get you a good market price for 1-goal victories if you use OT-included to avoid ties. Similarly there are some lines of 1 the other way, or -2.5, that can help you with those odds. To do better getting some actual game data would be the best approach, but you can also apply some common sense. For example, if MON wins by 5 80% as often as by 4, and BOS wins by 5 40% as often as it wins by 5, that can't be right: Boston is a favorite! The distribution isn't going to look like that, quite the opposite. Of course with a 140% market there's probably not a single line there with any value.

**Data** · 04-27-09, 07:25 PM

Originally posted by MrLuckyPants

My excel skills arent the greatest, but I'm trying to figure out a way, some sort of excel model where I can input a money line, and get odds for a 10-way winning margin market for Ice hockey.

ex:

-230 Boston v Montreal +190

Boston by 1 31%
Boston by 2 25%
Boston by 3 13%
Boston by 4 10%
Boston by 5 4%

Montreal by 1 30%
Montreal by 2 17%
Montreal by 3 5%
Montreal by 4 3%
Montreal by 5 2.5%

I got those prices from a fairly reputable sportsbook... that's quite a bit of margin though. 140 % rip off

Does anyone have any idea how I should go about making a spreadsheet for this?

Here is how to make a spreadsheet for regulation time. If you need scores including overtime then you need to adjust this converting all the ties into wins-by-1s.

1. Create a list of all the possible scores after 3rd period.
2. Assign probabilities for every score using Poisson distribution as a function of "power of team A facing team B" or whatever you want to call it.
3. Make a list of probabilities of certain outcomes by adding probabilities of the exact scores (step 2) that fall into certain criteria such as
.....
and so on
Away Win by 2
Away Win by 1
Away win
Tie
Home Win
Home Win by 1
Home Win by 2
and so on
......
and so on
Under 4.5
Under 5
Under 5.5
Under 6
Over 4.5
Over 5
Over 5.5
Over 6
and so on
.......
4. Now, you have a spreadsheet that takes "team powers" for every team gives you the odds for every outcome above. Now, by trial and error (takes about 10 iterations - a few seconds) you can find "team powers" so that the total and home/away lines would match the market.

Again, you do not convert line and total into scores but find "team powers" that make given line and total and just see what are the fair odds for all those "home by 1" outcomes.

Creating this calculator should make for a very good exercise while the calculator itself is a fun to play with.

**calm** · 04-28-09, 10:04 AM

It may be fun to play with, but it won't make you any money.

**MrLuckyPants** · 05-05-09, 10:11 PM

Thanks Data, great stuff my man.