questions for irishtimmy - zero vig line, line movement & simulation

[Seanie Mac]

Have a couple questions for you buddy. I read alot of what you say at eog, I post as the blueprint

1.) could you break down finding a no vig line for me, and the importance of it

2.) for football what kind of formula do you use. I read that you use a simulator, did you create it, or borrow one

3.) do you factor in line movemets and public percentages?

4. How many points do you need for an edge in football/basketball when compared to your simulation, and how many cents for baseball?


thankyou

[IrishTim]

Originally Posted by Seanie Mac

Have a couple questions for you buddy. I read alot of what you say at eog, I post as the blueprint

Thanks, blue, good to see you ever here. Hope early season bases are treating you well.

1.) could you break down finding a no vig line for me, and the importance of it

Romo described in depth how to do it here (analyzing a set of lines & determining the win probability implied by those lines) but I can show you again real quick. I'll use Pinnacle's lines for tomorrow's Boston/Texas game. Pinnacle has Boston -192 and Texas +180. Now you can find the implied winning percentage of a team whose line is negative by taking the absolute value of the number and dividing it by that number plus 100.

So Boston's implied winning percentage would be 192/(192+100) so 192/292 = .6575 and then you multiply that by 100 to convert the decimal to a percentage, so according to Pinnacle's line, Boston has a 65.75% chance of winning the game.

For teams with a price greater than zero (i.e. +101 or more) you determine the implied winning percentage by dividing 100 by the line plus 100.

So Texas' implied win percentage would be 100/(180+100) which is 100/280 = .3571 or 35.71%.

So according to Pinny's lines, the implied winning percentages are as follows:

Boston: 65.75%
Texas: 35.71%

Now the problem is that when you add these percentages, you get a number over 100%. Obviously this is a problem, as when you only have two possible events, the respective probabilities must add up to exactly 100% (or 1 in decimal form).

This small percentage over 100% is what is called the overround which in practical terms, is the book's theoretical hold percentage. That is, this is what % of every dollar wagered that the book would make ("hold") if there was equal action on both sides. It's theoretical because of course there won't always be exactly equal action.

In our case, 65.75% + 35.71% = 101.46% so Pinnacle can expect to make 1.46% of every dollar wagered. You'll see other full-juice books with a significantly higher expected hold, but Pinnacle makes up for this with their incredible volume. They take $30k on an NFL side where other books will only take $5k. This is the primary reason why their lines are so efficient (accurate) - because they cater to sharps with reduced juice, early lines, and high limits. So they have the sharpest bettors shaping their markets.

But obviously there is a problem with analyzing winning percentages that have the books hold (the vig) factored in to it. What we are after are the true probabilities of an event occurring, which we determine by the zero-vig line.

The way to factor out the book's hold is to divide each implied win percentage by the sum of the two implied winning percentages.

Boston: 65.75/101.46 = .6480 or 64.80%
Texas: 35.71/101.46 = .3520 or 35.20%

As you can see, 64.80% + 35.20% add up to exactly 100%. So these are the true or zero-vig probabilities of each team winning, according to Pinnacle's markets.

In a line: you calculate the zero-vig win probabilities by dividing each implied win percentage by the sum of the two WPs.

Now to go from the no-vig win percentages back to a no-vig line, you use the following equations, with P = no-vig win percentage in decimal form

*if the win probability is greater than 50% (decimal > 0.5):
-100/[(1/P)-1]

*if the win probability is less than 50% (decimal < 0.5):
[(1/P)-1] x 100

So for Boston, who has a no-vig WP of 64.80% it would be:
-100/[(1/0.648)-1] = -184.1

For Texas, whose no-vig WP was 35.20%:
[(1/0.352)-1] x 100 = +184.1

So the no-vig line is -184/+184.

Now to the second part of your question - how do you use this?

Obviously just knowing the no-vig win probabilities isn't going to make you money.

One method is to create a model that generates a more accurate win % than the ones the book produces (opening lines). If you make the fair line on a game -205/+205 and they have the game -192/+180 (no vig = 184), then you believe there is "value" in the -192. The implied WP of -192 is 65.75% and at -205, you believe the team has a 67.21% chance of winning so your edge is 1.46%.

There are other ways as well. The first thing I do when I sit down at my desk is take the no-vig line of every single Pinnacle market. I used to do this by hand until I came in contact with a programmer who wrote a program that scrapes every Pinnacle line in a matter of seconds for me. Then I go and look for books that are posting numbers off of Pinny's no-vig lines. Say Pinny's no-vig line on a baseball game is -184/+184. Since most places just copy their lines, you'll see a lot of lines -200 to +165 and of course none of these offer value.

But there are books (and locals) out there where you will see something like -210/+190 which would offer an advantage on the dog or -175/+155 which would give you an edge on the favorite.

The assumption, of course, is that Pinnacle's lines are sharper than the other books which I believe to be so because of the high volume and amount of sharp action that shapes their lines.

2.) for football what kind of formula do you use. I read that you use a simulator, did you create it, or borrow one

Borrowed. Well, really shared. Sports betting is a lot about teamwork, a lesson I'm just starting to learn.

It doesn't make much sense to me to spend hours and hours trying to create a profitable model for football for a few reasons.

1) Football, especially NFL, sees the most volume by the books so they put more time into their openers. This is different from other sports like baseball where many of them are in a race to just get lines up for the next day (Olympic/Greek has lines up at 3 PM for the next day and Pinnacle, Bookmaker/Cris aren't far behind). These lines are just dart throws. It's cheaper for them to just put up an estimate of what the line might be and then let the sharps who hit the openers at lower limits shape the market so it's efficient by the time the limits open up.

2) Sample size. The kind of modeling I do works best where each team plays one another ("closed circuit"). This is great in sports like WNBA where there are only 12 teams and a fairly long season to collect data. What kind of data can you get in something like college football where there are 120+ teams (so adjusting for strength of schedule is a nightmare) and only 12 games?

3) Volume. With only a 16 game regular season in NFL and 12 game regular season in college football, sample size becomes an issue. Even if you make a line that's say 4 points better per game than the lines you are betting into. The variance of such a small sample size could be such that you still have a losing or break-even season. Additionally, what's the point of slaving trying to create a +EV model in NFL that already has efficient markets when there are only 16 games a week for 4 months? In my estimation, it's much more worth it to spend that time trying to make a model where the books have week openers and you can get 15 games a night for 6 months in baseball?

3.) do you factor in line movemets and public percentages?

I probably would factor in public percentages if I could find reliable numbers but I don't really trust the ones that are out there. But I don't care who the public, I care who the sharps are on. Which brings me to line movements.

In my opinion, a book like Pinnacle isn't going to pull "head fakes" with their lines because the second they do put up an off number to try and "trap" people into a line, they could get hit with a 6 figure position by the sharps on the other side. So when Pinnacle moves a line, I believe that they are getting hit with action by people who 1) have an opinion they believe is credible/people whose accounts they have marked and 2) a lot of money on one side, and since sharps bet the most money it's usually them.

If I see Pinnacle move from -2.5 to -3 and I can get -2.5 at say -105 at another book, I will take it blindly so long as the push probability of the 3 outweighs the juice I am betting into. If it does, then I have a +EV position.

4. How many points do you need for an edge in football/basketball when compared to your simulation, and how many cents for baseball?

It depends but I usually start looking when I think I have a 3% edge on the ML (typically 10-15 cents). In basketball, usually 1 to 1.5 points (which is the same edge since the spread is simply a function of the moneyline which is a function of the win%). In football, it depends on the number. Obviously the difference between my line of 2.5 vs the market's 3.5 and my line of 19.5 to the book's 20.5 is significant. Generally speaking, about 1 to 2 points. Things get tricky with the larger spreads though.

Hope that helps. I may have more to add later and anyone else who feels that they have something to add or notices anything I got wrong, feel free to jump in. Any more questions, I'll try to answer them. There are also plenty of other sharp guys in here who I'm sure could give better explanations than me.

[Seanie Mac]

boom! i copy and pasted all this so I never lose it. ur a great addition to this forum, I look forward to working with you from here on out. ur someone I really respect as I have stated on EOG.

not really doing bases, im on the other side of things during baseball season. im trying to get a hold of a simulator for football. gonna miss shrink alot, he gave me ALOT of info on who had what during football last year via AIM.

I'll PM you at EOG and we can exchange contact info

[IrishTim]

Sounds good, blue. Shrink definitely had the contacts and the info, probably as good as anyone else in the biz. The summer/MLB season is a good period to take some time off. Enjoy it.

Any other questions, fire away.

[Romanowski]

good stuff Tim, great info :)

[uva3021]

Good stuff, just got around to reading this, well written. I like your philosophy on how the opportunity cost is exhausted in trying to find a +EV model. No sense in competing with Vegas, let the linesmakers do all the work for you.

[IrishTim]

Thanks UVA. You've contributed a lot of great material in here. I was planning on going through it all as soon as final exams got through this week.

[Romanowski]

tim is your wnba in a csv by chance?

[IrishTim]

My model or data? I absolutely cannot share the model or my primary db, but I do have some other line data that I would be willing to trade. You know how to reach to me.