Some Super Bowl stat projections

[Justin7]

New England:
Passing yards: 293
Rushing Yards: 101
Brady: 0.65 interceptions
3.05 TD passes
3d down conversions: 6.5

New York Giants:
Passing yards: 196
Rushing yards: 118
Manning: 1.15 INTs
3rd down conversions: 4.5

1.5 Total fumbles

"Fair line": New England -10.5/52
Projected score:
New England 31.25 NYG 20.75

It may seem random, but this is some of the "core" work I used in betting props this year.

[ritehook]

Ok, NE wns by only 10.
Great, that means I'll collect on my little Giant bet

[Justin7]

I hope so...

The idea of prop betting is to bet a lot of things that are independent. Don't risk too much on any one set of props that are correlated.

I like the Giants (I think I'll get +13 or 13.5 on Sunday) spread bet... but if I take that, I have to be careful taking any other "pro giants" prop like over on Giants rushing.

[pico]

i like to put a little parlay on giant ML and under

[pat venditto]

giant moneyline way to throw money away sir lol

[moses millsap]

The idea of prop betting is to bet a lot of things that are independent. Don't risk too much on any one set of props that are correlated.

If you have a strong confidence level in one particular angle that naturally correlates several props together, wouldn't it be counter productive to hold yourself to a rule like that?

I understand your general comment, but I don't think it holds universally to all prop betting. It seems to be analogous to somebody advising another not to put all of one's eggs in a basket, but if you have strong confidence on a certain angle/play(s), I think it's imperative to try to take the shot to win big instead of spreading things out in hopes that one doesn't lose all the wagers. Basically lowering your potential ceiling for profit in exchange for lowering the # in regards to potential loss.


Thanks for your # projections.

[Ganchrow]

If you have a strong confidence level in one particular angle that naturally correlates several props together, wouldn't it be counter productive to hold yourself to a rule like that?

I understand your general comment, but I don't think it holds universally to all prop betting. It seems to be analogous to somebody advising another not to put all of one's eggs in a basket, but if you have strong confidence on a certain angle/play(s), I think it's imperative to try to take the shot to win big instead of spreading things out in hopes that one doesn't lose all the wagers. Basically lowering your potential ceiling for profit in exchange for lowering the # in regards to potential loss.


Thanks for your # projections.

This needs to be considered in light of the player's chosen system of risk management.

Now obviously if all bets were perfectly correlated (assuming no restrictions on maximum bet size) then the optimal strategy would be a wager on only that bet with the highest payout odds.

Of course in reality the only truly fungible bets are identical bets offered across books (as well as certain "sucker" bets created through rephasing bet terms). While certain prop bets may be highly correlated with one another we know from basic portfolio optimization theory that we can reduce risk by diversifying across even correlated bet classes provided the correlation is less than perfect.

This implies that a rational risk averse bettor may well choose to allocate non-zero funds to a wager correlated with another higher EV wager. This effect can become even more apparent when the lower EV wager is offered at shorter payout odds.

In general, wagering limits aside, when dealing with positive EV, non-parlayable wagers that are less than perfectly correlated, optimal strategy will involve betting less than the amount that you would wager were the two bets independent, but (typically) more than what would be wagered were the two bets perfectly correlated (where you'd only be wagering on the higher bet with better payout odds).

So don't be afraid to wager on correlated events -- just be mindful of that correlation and adjust your bet size accordingly.

[donjuan]

If you have a strong confidence level in one particular angle that naturally correlates several props together, wouldn't it be counter productive to hold yourself to a rule like that?

I understand your general comment, but I don't think it holds universally to all prop betting. It seems to be analogous to somebody advising another not to put all of one's eggs in a basket, but if you have strong confidence on a certain angle/play(s), I think it's imperative to try to take the shot to win big instead of spreading things out in hopes that one doesn't lose all the wagers. Basically lowering your potential ceiling for profit in exchange for lowering the # in regards to potential loss.

Read about Kelly and understand Kelly. Then think about why you would prefer two +EV anti-correlated (not sure what the correct term is so I'll use this until corrected) wagers to two +EV correlated wagers assuming they both have the same expected value and the same payout odds.

However, correlation in parlays is great.

[Ganchrow]

two +EV anti-correlated (not sure what the correct term is so I'll use this until corrected)

"Negatively correlated" would be the preferred nomenclature.

[donjuan]

Thank you. I obviously spent too much time in the Antipodes last summer where they seem to attach "anti" to everything, such as anti-clockwise instead of counter-clockwise.

[Justin7]

If you have a strong confidence level in one particular angle that naturally correlates several props together, wouldn't it be counter productive to hold yourself to a rule like that?

I understand your general comment, but I don't think it holds universally to all prop betting. It seems to be analogous to somebody advising another not to put all of one's eggs in a basket, but if you have strong confidence on a certain angle/play(s), I think it's imperative to try to take the shot to win big instead of spreading things out in hopes that one doesn't lose all the wagers. Basically lowering your potential ceiling for profit in exchange for lowering the # in regards to potential loss.


Thanks for your # projections.

If I have enough independent events, I can bet my entire bankroll (or more than that) on Super bowl props. This year, I'll have close to 100% of my bankroll invested.

Notice that my line is smaller than the "retail" line. My total is lower also. each of these will cause a lot of props to "favor" the giants and the under. I could easily bet half my money on props relating to "giants +10.5 is more accurate than +13".

Assume my numbers give me a 10% edge against the market prices for props. Assume that a set of props has a 40% correlation to "giants cover +13".

How much can I safely risk on each prop alone? And how much on the set, all combined?

This is NOT a hypothetical.

[donjuan]

How much can I safely risk on each prop alone?

This depends on your % chance of winning. A 10% edge at +200 would have a different Kelly stake than a 10% edge at -200.

[Ganchrow]

Assume my numbers give me a 10% edge against the market prices for props. Assume that a set of props has a 40% correlation to "giants cover +13".

How much can I safely risk on each prop alone? And how much on the set, all combined?

This is NOT a hypothetical.

If we assume:

1. a push on NYG+13 occurs with probability 1.3%
2. the 10% edge figure for NYG is conditioned on not pushing
3. each bet is at -110 off a -110 line set
4. the prop bet is win/lose only

Then given a 10% conditional edge for each bet and a correlation coeff. of 40% the following outcome probability matrix would be implied:

WIN	WIN	42.48%
LOSS	LOSS	27.42%
WIN	LOSS	14.39%
LOSS	WIN	14.41%
PUSH	WIN	0.73%
PUSH	LOSS	0.57%

The full Kelly stakes would then be 7.7751% on NYG+13 and 7.8457% on the prop bet. To determine optimal stakes for more than just 1 prop bet you'd need to further define the covariance matrix.

[superquad07]

Did Any One Fine Out Brandol Langs Call

[Justin7]

The full Kelly stakes would then be 7.7751% on NYG+13 and 7.8457% on the prop bet. To determine optimal stakes for more than just 1 prop bet you'd need to further define the covariance matrix.

Assume I have 20 props that I described - each with a 10% edge (and all at about -110 as you assumed). Assume that each of the 20 possible bets is 30% correlated to the spread. Other than this correlation, assume the remaing correlation is purely random...

You could easily get yourself in trouble, especially if you're betting 2x your bankroll on props

[Ganchrow]

Assume I have 20 props that I described - each with a 10% edge (and all at about -110 as you assumed). Assume that each of the 20 possible bets is 30% correlated to the spread. Other than this correlation, assume the remaing correlation is purely random...

You could easily get yourself in trouble, especially if you're betting 2x your bankroll on props

When you refer to the cross correlations between the prop bets as being "random" do you just mean they're uniformly distributed across allowable values or that no correlation exists?

Are you trying to determine an exact Kelly solution here?

[Justin7]

Ganch,

Yes, that works.

What I am trying to show is that it is easy to overbet your bankroll with 20 monster plays due to correlation.

[Ganchrow]

What I am trying to show is that it is easy to overbet your bankroll with 20 monster plays due to correlation.

That's why it's important to appropriately manage risk.

I'm still not entirely certain I understand the precise problem you're looking to solve here. (Or are you not actually looking for an exact solution but rather trying to illustrate a point?)