Handicapping: Questions for the statistically inclined

I have a question about Sports handicapping.

What is more significant a guy who can go 65 for 100 against the spread or someone who can go 550 out of 1000 spread bets? And what's the odds these are just chance results? And at what point would you presume a handicapper has skill (i.e. if he could beat the spread for 100 or 1000 or 5000 picks)?

I've seen some pick tracking sites and rarely do I see guys near 55% after 500+ picks. Are those 55% guys just lucky survivors in a quarter flipping contest?

Quote: bazooooka

And what's the odds these are just chance results? And at what point would you presume a handicapper has skill (i.e. if he could beat the spread for 100 or 1000 or 5000 picks)?

If they are beating the closing number consistently over 1000 bets it is because they are a winning bettor. If they're not beating the closing number much more likely it's luck

Yeah I've heard it like 1 in 10M+ to get 600 heads out of 1000 if flipping quarters? However assuming the closing line provide a near 50/50 proposition to someone choosing sides randomly then what the odds of someone even get 550 out of 1000 by luck alone? If anyone wants to get technical with SD and variance that's fine. I'm just curious to see where significance is when talking about 100 or 500 or 1000+ picks.

Once you get close to the 500 sample size mark, you shouldn't see any significant changes beyond that. 100 is still enough to have strong confidence in the results though.

Assuming the guys who report 55% are really more like 53% percent since there always seems to be some reporting inaccuracy. How significant is 530 out of 1000 picks?

And I'm still curious if the guys who have gone 65 out of 100 have done so with luck. Seems like a few each year do that well in the nfl Vegas Super Contest. Anyone that can show me the math breakdown - that'd be cool.

Looking at spread betting, it's close to a 1:1 payout. We essentially have a binomial distribution where without knowing any expert information, you would expect to win 50% of the time. Excel can give you some numbers with formulas for each scenario. The one I used for this example is the BINOMDIST(x, n, p, c).

Scenario 1 of 550 correct choices out of 1000 outcomes: The chance of "randomly" obtaining 550 wins would be .00069 so this would be seen as significant. The threshold is around 535 wins where your probability of "randomness" increases to .01235 and the significance is debatable at different alpha levels.

Scenario 2 of 65 correct choices out of 100 outcomes: Chance of randomness equals .000894, very significant. Threshold is around 60 wins where your probability increases to .0176.

Now keep in mind, significance does not always equate to success with spread betting. In most cases, you bet $110 to win $100. Winning 535/1000 bets under these circumstances awards you a grand total of $235! Ex: .535*100-(.465*110). You can see how scale really plays a role here (you see news stories about people betting thousands, even tens of thousands on individual games).

The math may be relatively rudimentary here and I am couple years removed from really diving into stuff like this, so anyone else can jump in and correct me if necessary. Hopefully this helps though!

As far as I know, those tracking sites are all bull.

When you look at a handicapping contest, you need to udnerstand that they are allowed to bet stale lines. Like, the numbers they are picking against might come out on tuesday, and the picks might be due by Friday. So everyone should have much better results than they would if they had to bet against current lines.

Also, edges are not necessarily forever. Even if someone does have that advantage over 1000 picks, they might lose it going forward for any number of reasons.

As for beating the closing line, I generally think in those terms, but a friend of mine made an interesting post about it on SBR.

http://www.sportsbookreview.com/forum/handicapper-think-tank/716057-misunderstanding-beating-closing-line.html

The probability of 65 or more correct picks out of 100 is 0.001758821, which is about 1 in 569.

The probability of 550 or more correct picks out of 1000 is 0.000865268, which is about 1 in 1,156.

So, I'd respect the 550 out 1,000 record more. I would exercise proper skepticism about any boasts about handicapping records. I should write a whole article about it.

Charles, I think the more pertinent question is what are the odds of going x out of y, or better.

Quote: Wizard

So, I'd respect the 550 out 1,000 record more. I would exercise proper skepticism about any boasts about handicapping records. I should write a whole article about it.

Winning 55% against the spread on -110 lines should lead to the bankroll multiplying 15 times for every 1000 bets, assuming no limits, according to the Kelly Criterion calculator I use. Which means if we start with $20,000 it would be over $300,000 after 1000 bets. And close to $5 million after another 1000. Given the number of games available in all sports, professional and college, sides, totals, first-half, props, etc., it shouldn't be hard to find a few bets per day to get you to that $5 million in only a few years.

Even if we do look at limit betting, it would take until a 55% winner had $1 million before he reached limits on NFL sides.

If someone hasn't yet won millions from picking winners in 55% of games they haven't been doing it very long

Quote: TomG

Winning 55% against the spread on -110 lines should lead to the bankroll multiplying 15 times for every 1000 bets, assuming no limits, according to the Kelly Criterion calculator I use. Which means if we start with $20,000 it would be over $300,000 after 1000 bets. And close to $5 million after another 1000. Given the number of games available in all sports, professional and college, sides, totals, first-half, props, etc., it shouldn't be hard to find a few bets per day to get you to that $5 million in only a few years.

Even if we do look at limit betting, it would take until a 55% winner had $1 million before he reached limits on NFL sides.

If someone hasn't yet won millions from picking winners in 55% of games they haven't been doing it very long

This statement assumes that the gambler would be playing with maximum efficiency. I will go out on a limb and say that a great many gamblers including winning gamblers do not play with maximum efficiency. While they may win big in one arena they may lose in others. Also, when they begin to win very large bets there are many saying that the books will begin playing games with them; changing their lines when they recognize them or refusing their action. These talented gamblers may also be unwilling to shovel large sums of money back and forth offshore into websites of questionable legality and questionable trustworthiness. If they do do this; at least in the U.S.; they may be suspected of money laundering. If they reported earning large sums as taxable, which was won from offshore books; they may have to reveal where they earned this money which may be in effect admitting to the IRS and therefore any other interested Government agency that they committed a crime. So, making big bucks, even if you have this talent, may not be near as easy as it may seem.

Quote: TomG

Winning 55% against the spread on -110 lines should lead to the bankroll multiplying 15 times for every 1000 bets, assuming no limits, according to the Kelly Criterion calculator I use. Which means if we start with $20,000 it would be over $300,000 after 1000 bets. And close to $5 million after another 1000. Given the number of games available in all sports, professional and college, sides, totals, first-half, props, etc., it shouldn't be hard to find a few bets per day to get you to that $5 million in only a few years.

Even if we do look at limit betting, it would take until a 55% winner had $1 million before he reached limits on NFL sides.

If someone hasn't yet won millions from picking winners in 55% of games they haven't been doing it very long

I think part of the problem(?) is IF you're able to pull off a 55%+ winrate in sports betting the -110 lines, you're not going to be making a significant amount of bets, since most offered lines likely won't meet the 55%+ confidence criteria. And if you're that good with it, I'd think you'd be specializing in a specific sport, like the NFL. It's not like a good NFL bettor can automatically win 55%+ at basketball and baseball and golf and boxing and everything else.

I guess your best chance is either college football, any basketball or any baseball, since there are so many games. But NFL, I'd say no way Jose, no way can you get that many bets in.

Thank you to everyone for the great responses. Seeing the numbers behind everything really helped. Someone that can do 60% over a smaller sample might have skill but that doesn't mean that will continue over a larger sample (i.e. skilled handicapping can't be assumed across sports and seasons). And yes I would question why anyone who can do 53-55% over thousands of picks would sell or promote such a thing since surely making bets would be more lucrative. The fact that handicapping skill is transient and there is such huge variance really would make it hard to tail someone's picks versus coming up one's own. I do think data scientists likely are the ones with the real edge in this game. But it is indeed fun to try to beat the book on a handful of NFL games. To try to do it across sports and thousands of picks is almost a sure path to ruin for all but the most astute bettors.

I'd think that if one was betting 20K a game they'd have to have a HUGE bankroll to taken into account variance. What would be safe bet if one had 100k bankroll and an assumed 55% ability to pick spread winners? Would 3% per bet (i.e. 3k) be too aggressive? 3k a week net sounds good if come can lay a few bets per day or so? Those that can handicap college Bball might be able to pull this off.

Quote: TomG

Winning 55% against the spread on -110 lines should lead to the bankroll multiplying 15 times for every 1000 bets, assuming no limits, according to the Kelly Criterion calculator I use. Which means if we start with $20,000 it would be over $300,000 after 1000 bets. And close to $5 million after another 1000. Given the number of games available in all sports, professional and college, sides, totals, first-half, props, etc., it shouldn't be hard to find a few bets per day to get you to that $5 million in only a few years.

Even if we do look at limit betting, it would take until a 55% winner had $1 million before he reached limits on NFL sides.

If someone hasn't yet won millions from picking winners in 55% of games they haven't been doing it very long

Quote: bazooooka

What would be safe bet if one had 100k bankroll and an assumed 55% ability to pick spread winners? Would 3% per bet (i.e. 3k) be too aggressive?

5.5% of your bankroll per the Kelly Criterion.

I likely have been under betting then. I've down close to 55% over hundreds of picks but I doubt 55% is my true skill level. Harder to do know that regular season NBA is over. MLB I like to play underdogs and try to hit near 40% and often play the 150-200 dogs.
Always gets harder as season progresses since talent becomes more prevalent and the odds react.

Quote: Ace

5.5% of your bankroll per the Kelly Criterion.

My math says 0.05/1.091475 ~ 4.58% of BR.

Wager = Edge/Variance * BR

EV = 100*0.55 - 110*0.45 = $5.5 per $110 wagered
Edge = 5.5 / 110 = 0.05 = 5%
Var = 1.091475 (https://www.mathsisfun.com/data/standard-deviation-calculator.html)
BR = 100k

Wager = (0.05/1.091475) * 100000 = $4580....although $4400 would make more sense, as that would pay an even $4000.


But IMO, you shouldn't play full kelly, as it's very volatile and you'd have to continuously change your bets around. 1/2 kelly makes the most sense, to me. You also knock your ROR way down (from 13.5% to 1-2%).

Quote: RS

My math says 0.05/1.091475 ~ 4.58% of BR.

Wager = Edge/Variance * BR

EV = 100*0.55 - 110*0.45 = $5.5 per $110 wagered
Edge = 5.5 / 110 = 0.05 = 5%
Var = 1.091475 ml)
BR = 100k

The Kelly formula is simply your advantage divided by the payoff amount (expressed as "to 1").

5% / (10/11) = 5.5%.

Incidentally the variance is .55 x .45 x (21/11)^2 = 0.902.

Best cappers I know go 2% to 3% per bet and hit between 53% to 57%. Problem is whenever these guys scale up and try to do 1000 picks instead of a few hundred then hit rates drop toward 50%. I think one can beat the house but only if they are very careful to pick their spots and they can bet all sports. That kind of capper can just play one ot two per day and hopefully after a year. They have made 20-40 units on 500 picks. Doesn't add up too much unless you play high limits. 200 dollar players playing 10 games a day always will get ground to dust. But 2000 dollar players can make some nice side change. But those that can swing that big don't normally need side income. Thus its rare to find skilled players who can afford to play online book max or higher in town.