[VegasDave]

If I'm not mistaken, I've read somewhere that there is a formula for the likelihood of a team winning based on the point spread.

Example... a team that is a pick'em is 50%. a team that is -1 is say 53%, -2 56%, etc...

Obviously lines move all the time and this isn't a science, I was just curious if such a formula exists.

[VegasDave]

And another question... anyone know of a site that gives the past spreads to every superbowl? Thanks!

[Razz]

http://www.docsports.com/point-sprea...uper-bowl.html

[VegasDave]

One down, thank you

[Wheell]

In answer to your obvious question, yes, the point spread and the moneyline are not suggesting the same thing. However, the reason for this is that historically Superbowls are unusually variant. In fact, the average spread for a Superbowl is 7.6. The average margin of victory is 15.4. Favorites have scored 1061 points. Underdogs have scored 766. Counting the point spread underdogs have scored 1070.5 points. Favorites are 21-17-2 against the spread (although the spreads have fluctuated during the days before the game so those numbers aren't gospel). Favorites are 28-12 outright. Only 10% of all Superbowls have been decided by 3. 27 Superbowls have been decided by 10 or more. Half have been decided by 15 or more.

Random aside: Holy crap is the band "Guillemots" good live.

Anyway, I know it seems VERY tempting to either take the points or bet Indy on the moneyline, but historically that has not been effective in the Superbowl. It has worked a bit lately with the Patriots winning by 3 as 7 point favorites twice, but we've also been on a bit of a good run for the dogs.

[VegasDave]

Thank you for the info, but the info isn't for betting purposes.

The first Super Bowl I remember, I was 11 years old. That was 10 years ago. I was rooting for the huge dog Patriots to beat the Packers.

Since then, every team I have rooted for in the Super Bowl has lost. Every year.

I was actually just wondering what the odds of that were. So I was going to figure it out by multiplying all of the winning percentages together. Obviously it isn't going to be perfect, but I was just wondering ballpark.

Like, the Raiders were -3.5, so lets say they had a 57% chance of winning. The Patriots of 97 were a +14, so we'll say they had a 20% chance of winning. So... .80 X .43 = a 34% chance that BOTH teams won.

Curious what the odds on me losing 10 straight are.

[Wheell]

Let me get this straight: You were rooting for the Titans against the Rams but not the Patriots against the Rams. You were rooting for Green Bay against Denver, New England against Green Bay, and Atlanta against Denver.

Bullshit.

[Wheell]

And, if I might add, there is a push list on various numbers from the football forums. It should have you deduce moneylines, although again, for the Superbowl favorites are less likely to win than they would otherwise be.

[Wheell]

And last but not least, using the actual monleylines you are about 968-1 to have been brokenhearted each of the last 10 years.

[VegasDave]

Correct. I'm a Raiders Fan, for point of reference.

In XXXI, I was just rooting for the underdog. Neither team meant anything to me.
In XXXII, Hate the Broncos. Would pick anyone over them.
In XXXIII, Same story, hate the Broncos.
In XXXIV, was rooting for the underdog, plus I always felt as an LA fan you had to like the Raiders or the Rams (not saying that makes sense, but it did to me as a kid).
In XXXV, Siragusa injures Gannon, I root against Ravens
In XXXVI, Tuck Rule. I HATE Brady and Pats worse than a faux rivalry with the Rams. Rooted for Rams.
In XXXVII, Oakland of course...
In XXXVIII, Rooted against NE
In XXXIX, Rooted against NE, plus I like the Eagles
In XXXX, Rooted for the Seahawks

You don't have to believe me... but it is the sad truth.

[VegasDave]

968-1? Wow... whered you get those numbers? Thanks for crunching them for me.

[Wheell]

You were rooting for 7 dogs and 3 favorites, one of which was quite large:

0.6
0.75
0.75
0.4
0.142857143
0.58
0.75
0.75
0.2
0.82

Multiply those by each other and presto.

[Ganchrow]

If I'm not mistaken, I've read somewhere that there is a formula for the likelihood of a team winning based on the point spread. Example... a team that is a pick'em is 50%. a team that is -1 is say 53%, -2 56%, etc... Obviously lines move all the time and this isn't a science, I was just curious if such a formula exists.

Any algorithm to properly calculate this could not be determined from first principles alone but would rather need to be derived from a scoring distribution for specific spread/total combinations. In other words, for every possible spread/total combination we'd need to determine the probability of the favorite winning or losing.

In practice however, by jointly considering spreads AND totals what you'd find is that you'd lack sufficient time relevant data to produce statistically meaningful results. So as a simplification, you might choose to ignore totals and limit your domain to spreads or even perhaps just spread tranches^* (in other words, for a spread of 13½ you may choose to look at scoring distribution of spreads from 12½-14½). It is this latter tack which Stanford Wong follows in his book Sharp Sports Betting. However, Wong only derives with NFL spread equivalences for spreads of 3½ points or less. Still SSB would nevertheless show you the process of coming up with your own spread equivalency chart.

I'll also point out that historically Super Bowls have tended to be much higher scoring affairs that regular season, so as such it's conceivable that a regular season spread equivalency chart might not be strictly applicable for a Super Bowl. Anyway as a first order approximation, here's an NFL spread/money line equivalency chart I just put together. One note of caution: as the spread increases, the accuracy will tend to decrease. (There's a reason why Wong only included spreads of 3½ or less.)

At Even Odds: Spread Prob. ML 0 EV 50.00% ±100.0 0½ EV 50.05% ±100.2 1 EV 51.08% ±104.4 1½ EV 52.08% ±108.7 2 EV 53.02% ±112.9 2½ EV 53.86% ±116.7 3 EV 58.78% ±142.6 3½ EV 62.22% ±164.7 4 EV 63.84% ±176.5 4½ EV 64.75% ±183.7 5 EV 65.94% ±193.6 5½ EV 66.56% ±199.0 6 EV 68.53% ±217.8 6½ EV 69.44% ±227.2 7 EV 73.20% ±273.1 7½ EV 74.58% ±293.3 8 EV 75.86% ±314.3 8½ EV 76.27% ±321.4 9 EV 77.05% ±335.7 9½ EV 77.28% ±340.1 10 EV 80.64% ±416.6 10½ EV 81.45% ±439.0 11 EV 83.28% ±498.2 11½ EV 83.65% ±511.7 12 EV 84.64% ±551.1 12½ EV 84.82% ±558.8 13 EV 86.83% ±659.1 13½ EV 87.13% ±677.1 14 EV 90.34% ±934.9 14½ EV 90.68% ±972.9 15 EV 91.86% ±1128.0 15½ EV 91.96% ±1144.0 16 EV 93.63% ±1468.9 16½ EV 93.74% ±1497.3 17 EV 96.61% ±2850.2 17½ EV 96.71% ±2940.5

At -110: Spread Prob. Fave ML Dog ML PK -110 50.00% -110.0 -110.0 0½ -110 50.05% -110.2 -109.8 1 -110 51.08% -115.1 -105.1 1½ -110 52.08% -120.1 -100.8 2 -110 53.02% -125.0 +103.2 2½ -110 53.86% -129.5 +106.9 3 -110 58.78% -160.2 +131.5 3½ -110 62.22% -187.2 +152.7 4 -110 63.84% -201.9 +163.9 4½ -110 64.75% -210.9 +170.8 5 -110 65.94% -223.4 +180.3 5½ -110 66.56% -230.3 +185.4 6 -110 68.53% -254.5 +203.3 6½ -110 69.44% -266.9 +212.3 7 -110 73.20% -328.9 +256.1 7½ -110 74.58% -357.1 +275.4 8 -110 75.86% -387.2 +295.4 8½ -110 76.27% -397.5 +302.2 9 -110 77.05% -418.5 +315.9 9½ -110 77.28% -425.2 +320.1 10 -110 80.64% -544.4 +393.1 10½ -110 81.45% -581.5 +414.5 11 -110 83.28% -684.2 +471.0 11½ -110 83.65% -708.7 +483.9 12 -110 84.64% -782.8 +521.5 12½ -110 84.82% -797.7 +528.9 13 -110 86.83% -1006.4 +624.6 13½ -110 87.13% -1046.9 +641.8 14 -110 90.34% -1765.2 +887.8 14½ -110 90.68% -1899.1 +924.2 15 -110 91.86% -2553.3 +1072.2 15½ -110 91.96% -2632.6 +1087.4 16 -110 93.63% -5121.0 +1397.6 16½ -110 93.74% -5466.1 +1424.7 17 -110 96.61% 17½ -110 96.71%

So for a spread of ±7 EV, the equivalent ML would be approximately ±273, and for a spread of ±7 -110, the equivalent ML would be approximately +256.3/-329.2. This is fairly close to Bet Jamaica's current market of ±7 -110 and +260/-320.

(In fact, the Bet Jamaica money line is offered at only 3.82% vig. which is equivalent to about a -108.3 line set. At ±7 -108.3, the equivalent ML from the chart above would be about +259.1/-318.7).

---

^*Another possibility would be to divide the data into 2 or more subgroups based on the magnitude of the totals and then analyzing each group separately. I’ve yet to actually try this.

[VegasDave]

Thanks for the chart, very cool.

[Arilou]

Does it bother anyone else that given access to Pinnacle, Matchbook or any other book with low juice on moneylines, these numbers are saying the dogs will always have ML value at every point? Not only that, he is saying you can profit by playing the dog on the moneyline at Pinnacle on every game with a spread of 3 or more. Given that, I would strongly suspect bias in these numbers.

[Ganchrow]

Does it bother anyone else that given access to Pinnacle, Matchbook or any other book with low juice on moneylines, these numbers are saying the dogs will always have ML value at every point? Not only that, he is saying you can profit by playing the dog on the moneyline at Pinnacle on every game with a spread of 3 or more. Given that, I would strongly suspect bias in these numbers.

Could you please clarify your point here? I'm not sure what you mean by "the dogs will always have ML value at every point".

[Arilou]

What I mean is, in a 'normal' NFL game, you can get, for example, +275 or so on a dog in a game with a flat 7. You will almost always be able to get more than +265. If that line corresponds to a neutral 7, then this chart says you will win as long as the dogs cover half the time. At 10, you can get north of +400, and at +14 you can get at least +800. It seems highly unlikely that there aren't math players willing to drive a truck through that kind of hole, if it was really that easy.

[Ganchrow]

What I mean is, in a 'normal' NFL game, you can get, for example, +275 or so on a dog in a game with a flat 7. You will almost always be able to get more than +265. If that line corresponds to a neutral 7, then this chart says you will win as long as the dogs cover half the time. At 10, you can get north of +400, and at +14 you can get +800. It seems highly unlikely that there aren't math players willing to drive a truck through that kind of hole, if it was really that easy.

Well that's why I noted that these are only approximations and why I further cautioned that "as the spread increases, the accuracy will tend to decrease."

If you look at the raw historical data from the past 17 years, you'll see that the SU win rate for 10 point favorites is 75.7%, and the SU win rate for 14 point favorites is 84.6%. These correspond to no-vig MLs of about +311.5 and +550.5 respectively which are even lower than my numbers. It's typically the case that in practice when doing statistical analysis of historical data model, figures will be biased towards what's actually happened in the past even though that only represents a single realization of an entire distribution of actual outcomes. As such I view what you've noted above as evidence for the robustness of the estimation method I used (that is to say that the modeling picked up on what the sportsbooks seem to be saying, namely that the historical frequencies for 10 and 14 point spreads might indeed be anomalous).

The truth is though, that the numbers from the chart above are still in reality fairly close to the numbers you quoted. In all likelihood it's the case that the market Dog MLs are probably a bit too high and the chart and historical ML figures are probably a bit too low.

[VegasDave]

Damn, using Ganchrow's chart, my odds of successfully rooting for the losing team all 10 years is 1954 to 1.

Not bad.

If you all believe in trends, you should unload on the Bears moneyline, because I love Peyton and the Colts and will be rooting for them whole-heartedly.