1. #1
    roasthawg
    roasthawg's Avatar Become A Pro!
    Join Date: 11-09-07
    Posts: 2,990

    Variance?

    Any benefit in attempting to smooth out variance by playing favorites on the ml and taking the points with dogs? If a play is +ev then generally it'll be +ev money line and spread... is it best to simply choose the play that gives the largest +ev or should you look to lower variance if possible?

  2. #2
    Pancho sanza
    Pancho sanza's Avatar Become A Pro!
    Join Date: 10-18-07
    Posts: 386

    Depends on what your edge is for each of ML and spread.

  3. #3
    Wrecktangle
    Wrecktangle's Avatar Become A Pro!
    Join Date: 03-01-09
    Posts: 1,524
    Betpoints: 3209

    Roast, this is a good point. All of the theory is in Kelly strategies, which assumes essentially constance variance, which we all know does not happen. I've worked on this some, but I haven't any answers.

  4. #4
    BeatingBaseball
    It's all about the price
    BeatingBaseball's Avatar Become A Pro!
    Join Date: 06-30-09
    Posts: 904
    Betpoints: 70

    Seems to me that it would be most efficient to make the play that you calculate as offering the greater +EV and rely on your staking formula to mitigate the variance.

  5. #5
    MonkeyF0cker
    Update your status
    MonkeyF0cker's Avatar Become A Pro!
    Join Date: 06-12-07
    Posts: 12,144
    Betpoints: 1127

    Quote Originally Posted by Wrecktangle View Post
    Roast, this is a good point. All of the theory is in Kelly strategies, which assumes essentially constance variance, which we all know does not happen. I've worked on this some, but I haven't any answers.
    How do you come to the conclusion that Kelly "assumes essentially constan(t) variance?"

  6. #6
    MonkeyF0cker
    Update your status
    MonkeyF0cker's Avatar Become A Pro!
    Join Date: 06-12-07
    Posts: 12,144
    Betpoints: 1127

    Quote Originally Posted by roasthawg View Post
    Any benefit in attempting to smooth out variance by playing favorites on the ml and taking the points with dogs? If a play is +ev then generally it'll be +ev money line and spread... is it best to simply choose the play that gives the largest +ev or should you look to lower variance if possible?
    I guess that depends on your staking. If you're utilizing Kelly, there is an opportunity cost associated with betting heavy favorite ML's as large portions of your bankroll will be appropriated to single wagers.

  7. #7
    Wrecktangle
    Wrecktangle's Avatar Become A Pro!
    Join Date: 03-01-09
    Posts: 1,524
    Betpoints: 3209

    Quote Originally Posted by MonkeyF0cker View Post
    How do you come to the conclusion that Kelly "assumes essentially constan(t) variance?"
    You claim to be the smart guy, why don't you figger it out?

  8. #8
    Pancho sanza
    Pancho sanza's Avatar Become A Pro!
    Join Date: 10-18-07
    Posts: 386

    This is easy enough to set up in excel, just input the percentages for the 3 outcomes (fav wins and covers spread, fav wins but does not cover spread, fav loses game), and solve.

  9. #9
    MonkeyF0cker
    Update your status
    MonkeyF0cker's Avatar Become A Pro!
    Join Date: 06-12-07
    Posts: 12,144
    Betpoints: 1127

    Quote Originally Posted by Wrecktangle View Post
    You claim to be the smart guy, why don't you figger it out?
    You're the one making the claim. Care to support it?
    Last edited by MonkeyF0cker; 12-13-10 at 06:49 PM.

  10. #10
    BeatingBaseball
    It's all about the price
    BeatingBaseball's Avatar Become A Pro!
    Join Date: 06-30-09
    Posts: 904
    Betpoints: 70

    It seems to me that Kelly is essentially assuming the mathematically expected variance that would be a function of the specific inputs to the Kelly formula and it is doing so "constantly." But we all know that in any given series the actual variance experienced is unlikely to perfectly mirror the mathematically expected variance. So in that sense Wreck would be correct. No?

    It also seems academic since I don't how you could improve upon Kelly's variance projection. It will probably prove inaccurate for any given seres - but how could you do better?

    It also seems to me that betting all favs on the ML and dogs on the spread would certainly minimize the variance in outcomes relative to doing it the other way around - but not necessarily minimize the variance in bankroll since one is laying and losing the long money prices for those ML favs that will sometimes lose outright and foregoing the long money pay offs you would get on the dogs who will sometimes win the game. Those outcomes will occur less often but, like everything else in the real world, the bottom line effect on bankroll would depend on how the actual results fall within a given series.
    Last edited by BeatingBaseball; 12-13-10 at 08:42 PM.

  11. #11
    Data
    Data's Avatar Become A Pro!
    Join Date: 11-27-07
    Posts: 2,236

    Quote Originally Posted by Wrecktangle View Post
    All of the theory is in Kelly strategies, which assumes essentially constance variance
    I'd be interested (and surprised) to hear a reasoning behind this too. This sounds very much as a brain fart. Kelly does not assume any variance, it accounts for variance and, unlike for a financial investor, for a sport bettor, variance is known because it is easily calculated. Even for the financial investor, the variances are of much lesser importance comparing to the means (true odds, for the sport bettor).

  12. #12
    Wrecktangle
    Wrecktangle's Avatar Become A Pro!
    Join Date: 03-01-09
    Posts: 1,524
    Betpoints: 3209

    Quote Originally Posted by MonkeyF0cker View Post
    You're the one making the claim. Care to support it?
    No. Proofs for posters with attitude are not what I spend my time on.

    I had you on my ignore list up until a few months ago. You're back on it.

  13. #13
    MonkeyF0cker
    Update your status
    MonkeyF0cker's Avatar Become A Pro!
    Join Date: 06-12-07
    Posts: 12,144
    Betpoints: 1127

    Quote Originally Posted by Wrecktangle View Post
    No. Proofs for posters with attitude are not what I spend my time on.

    I had you on my ignore list up until a few months ago. You're back on it.
    With good reason too. How dare someone ask another poster to back up their assertions with facts...

    LOL.

  14. #14
    mathdotcom
    mathdotcom's Avatar Become A Pro!
    Join Date: 03-24-08
    Posts: 11,689
    Betpoints: 1943

    Wrecktangle is a classic case of a poster who tries to come across as knowing something, but every post proves he knows absolutely nothing. His game theory posts are unbelievably ridiculous.

    The other classic case is of course me.

  15. #15
    uva3021
    uva3021's Avatar Become A Pro!
    Join Date: 03-01-07
    Posts: 537
    Betpoints: 381

    Kelly just applies the variance that happens to be there to a formula for minimizing risk and maximizing growth. Its not a conscious being that systematically filters in variance for each event lol

  16. #16
    roasthawg
    roasthawg's Avatar Become A Pro!
    Join Date: 11-09-07
    Posts: 2,990

    Quote Originally Posted by MonkeyF0cker View Post
    I guess that depends on your staking. If you're utilizing Kelly, there is an opportunity cost associated with betting heavy favorite ML's as large portions of your bankroll will be appropriated to single wagers.
    I don't agree with this... whether you betting the favorite -7 -110 with a 53% chance of winning or ML -350 with a 78% chance of winning Kelly's gonna call for roughly the same percentage of your bankroll. Wager size is not an issue here as they will be roughly the same whether betting ml or spread.

    My thinking was by playing ML on favorites and spread on dogs you may be able to more evenly disperse bankroll growth so that you are seeing more of a steady small increase (assuming you are making +ev wagers) rather than the large ups and downs associated with playing strictly spreads or even moreso with dogs on the ml. All can be +ev wagers but win/loss percentages are significantly different... possibly lowering variance while keeping expected bankroll growth roughly the same?

  17. #17
    MonkeyF0cker
    Update your status
    MonkeyF0cker's Avatar Become A Pro!
    Join Date: 06-12-07
    Posts: 12,144
    Betpoints: 1127

    Quote Originally Posted by roasthawg View Post
    I don't agree with this... whether you betting the favorite -7 -110 with a 53% chance of winning or ML -350 with a 78% chance of winning Kelly's gonna call for roughly the same percentage of your bankroll. Wager size is not an issue here as they will be roughly the same whether betting ml or spread.

    My thinking was by playing ML on favorites and spread on dogs you may be able to more evenly disperse bankroll growth so that you are seeing more of a steady small increase (assuming you are making +ev wagers) rather than the large ups and downs associated with playing strictly spreads or even moreso with dogs on the ml. All can be +ev wagers but win/loss percentages are significantly different... possibly lowering variance while keeping expected bankroll growth roughly the same?
    Your wagers will generally be larger especially with your larger edges. Try a -1 NBA away favorite that has a FV of -3. Your stake is more than double.

    You may want to plug your numbers into the Kelly calculator and look at how much expected profit and expected growth you give away for the sake of variance. It is not a trivial amount. It's fivefold less.

  18. #18
    Wrecktangle
    Wrecktangle's Avatar Become A Pro!
    Join Date: 03-01-09
    Posts: 1,524
    Betpoints: 3209

    Quote Originally Posted by Data View Post
    I'd be interested (and surprised) to hear a reasoning behind this too. This sounds very much as a brain fart. Kelly does not assume any variance, it accounts for variance and, unlike for a financial investor, for a sport bettor, variance is known because it is easily calculated. Even for the financial investor, the variances are of much lesser importance comparing to the means (true odds, for the sport bettor).
    Brain Fart?

    OK, I'll tell YOU where this came from.

    A few years ago, a blackjack buddy and I were exploring a math concept we called the "ratchet" on a card progression. We got into a lot of simulation work using Kelly. The Kelly distributions all were very Gaussian looking, so we assumed for our work it was close enough. Gaussian is very nice because of lot of statistical tests work for it, while they won't for other dists. As you know, Gaussian dists are defined only by a Mean and Standard Deviation.

    Now most of our Kelly fractions were 0.05 or less your fractional advantages in BJ are all very narrow, usually much less than 1%. For larger Kelly fractions approaching 1.0, I can see where the variance can be skewed, but for both BJ and Sports betting, we'll never get there as a win% of 55% is rare enough.

    So this is were my term "essentially" came from. If you have more quibbles, go work it out yourself. If you choose to beleve I'll full of shit, also fine. This is the internet, you're just some knuckle-head in the aether.

    Now having said that, let me mention something else. We have two well know Trolls in the Tank: Monkey and Mathdocom.

    You are well on your way to becoming a third.

  19. #19
    acw
    acw's Avatar Become A Pro!
    Join Date: 08-29-05
    Posts: 576
    Betpoints: 1564

    Long time ago I had to use Kelly. Usually I want to bet more (at a certain price) than I can. Anyhow Kelly should give you the betting size for both the ML and spread, then simply multiply this with the advantage of the two different bets and the one that gives the highest return should be chosen.

  20. #20
    Data
    Data's Avatar Become A Pro!
    Join Date: 11-27-07
    Posts: 2,236

    Quote Originally Posted by Wrecktangle View Post
    Brain Fart?

    OK, I'll tell YOU where this came from.

    A few years ago, a blackjack buddy and I were exploring a math concept we called the "ratchet" on a card progression. We got into a lot of simulation work using Kelly.
    From Kelly bettor's standpoint there is a principal difference between BJ and sports betting. The latter has fixed payoffs while in the former payoffs vary. This is why their respective Kelly bet formulas differ. What you are referring to is true for BJ but untrue for sports betting.

    Yes, brain fart. There is no reason to get cocky, we all make mistakes and you sound like you think you never do.

  21. #21
    roasthawg
    roasthawg's Avatar Become A Pro!
    Join Date: 11-09-07
    Posts: 2,990

    Quote Originally Posted by MonkeyF0cker View Post
    Your wagers will generally be larger especially with your larger edges. Try a -1 NBA away favorite that has a FV of -3. Your stake is more than double.

    You may want to plug your numbers into the Kelly calculator and look at how much expected profit and expected growth you give away for the sake of variance. It is not a trivial amount. It's fivefold less.
    So your take is that variance is indeed reduced but ml bets for favorites generally have a lower eg than spreads do?

  22. #22
    Flying Dutchman
    Floggings continue until morale improves
    Flying Dutchman's Avatar Become A Pro!
    Join Date: 05-17-09
    Posts: 2,467
    Betpoints: 759

    Quote Originally Posted by Data View Post
    From Kelly bettor's standpoint there is a principal difference between BJ and sports betting. The latter has fixed payoffs while in the former payoffs vary. This is why their respective Kelly bet formulas differ. What you are referring to is true for BJ but untrue for sports betting.

    Yes, brain fart. There is no reason to get cocky, we all make mistakes and you sound like you think you never do.
    Bullshit. Have you read Kelly's proof? It's about win percentage ONLY. AND there is only ONE formula. There is not one for Blackjack, and one for football, and one for baseball, etc. Get your shit straight.

    ...now as for folks getting cocky, you never do, do you?

  23. #23
    Data
    Data's Avatar Become A Pro!
    Join Date: 11-27-07
    Posts: 2,236

    Quote Originally Posted by Flying Dutchman View Post
    Bullshit. Have you read Kelly's proof? It's about win percentage ONLY. AND there is only ONE formula. There is not one for Blackjack, and one for football, and one for baseball, etc. Get your shit straight.

    ...now as for folks getting cocky, you never do, do you?
    What an epic fail!

    While sitting in PT and talking to JJ, look up what is (bankroll*ev)/var, you may learn something.

  24. #24
    Wrecktangle
    Wrecktangle's Avatar Become A Pro!
    Join Date: 03-01-09
    Posts: 1,524
    Betpoints: 3209

    Quote Originally Posted by Data View Post
    From Kelly bettor's standpoint there is a principal difference between BJ and sports betting. The latter has fixed payoffs while in the former payoffs vary. This is why their respective Kelly bet formulas differ. What you are referring to is true for BJ but untrue for sports betting.

    Yes, brain fart. There is no reason to get cocky, we all make mistakes and you sound like you think you never do.
    I have mentioned the difference between BJ and sports in terms of Kelly Criterion several times here in the Tank. If you were a little more widely read, you would know this vice throwing around careless, provocative statements.

    As to mistakes, I'm sure I'm not the only one to see the irony of your statement. As to my statements, I'm pretty careful to only post on items I know something about; something you might want to think about.

    Thirdly, if you had read Kelly's paper, you would have seen that it indeed only does just cover expected winning percentage as it covers how much of a bankroll to bet optimally. Any expansion beyond that only exists in your "holo deck" imagination. BTW, it was a most unusual document since he was working at Bell Labs.

    I am afraid you have reached Troll level now. My communications with you are over.

    Wrecktangle out.

  25. #25
    Flying Dutchman
    Floggings continue until morale improves
    Flying Dutchman's Avatar Become A Pro!
    Join Date: 05-17-09
    Posts: 2,467
    Betpoints: 759

    Quote Originally Posted by Data View Post
    What an epic fail!

    While sitting in PT and talking to JJ, look up what is (bankroll*ev)/var, you may learn something.
    I'm not surprised a humorless, supercilious dweeb like you doesn't understand the comic genius of JJ and his ability to keep Player's Talk roiling with all sorts of inane shit. You're the type of pimply-faced dullard who won't watch The Big Bang Theory because the characters strike a little too close to home. Not only that, you don't get the humor.


    So you troll the Tank picking fights with well respected posters like Poker Joe after he posted a very nicely written NBA analysis in another thread, just because it hasn't the exact equations in it you like, or even worse: no equations at all.


    What happened? Did those mean jocks shove you into a locker a few too many times in high school so you want to take it out on folks now that you can hide behind a keyboard and internet anonymity?



    Game on asshole, I'm your huckleberry.

  26. #26
    Flying Dutchman
    Floggings continue until morale improves
    Flying Dutchman's Avatar Become A Pro!
    Join Date: 05-17-09
    Posts: 2,467
    Betpoints: 759

    BTW "Data," answer the question: did you read Kelly's Proof?

    Do you even know anything about the topic other than an equation that the Quants spout all the time?

    You've gotta be one of the back room troglodytes; a real Quant wouldn't even be messing about in forums like this, they've real money to make.

    ...unless you're just another broken-down, gambolholic, Quant-wannabe.

  27. #27
    MonkeyF0cker
    Update your status
    MonkeyF0cker's Avatar Become A Pro!
    Join Date: 06-12-07
    Posts: 12,144
    Betpoints: 1127

    Quote Originally Posted by Flying Dutchman View Post
    Bullshit. Have you read Kelly's proof? It's about win percentage ONLY. AND there is only ONE formula. There is not one for Blackjack, and one for football, and one for baseball, etc. Get your shit straight.

    ...now as for folks getting cocky, you never do, do you?
    Actually, there is. The fact is, if you are a modeler and have scoring distributions, you can add variance to your Kelly staking.

  28. #28
    MonkeyF0cker
    Update your status
    MonkeyF0cker's Avatar Become A Pro!
    Join Date: 06-12-07
    Posts: 12,144
    Betpoints: 1127

    Quote Originally Posted by Flying Dutchman View Post
    BTW "Data," answer the question: did you read Kelly's Proof?

    Do you even know anything about the topic other than an equation that the Quants spout all the time?

    You've gotta be one of the back room troglodytes; a real Quant wouldn't even be messing about in forums like this, they've real money to make.

    ...unless you're just another broken-down, gambolholic, Quant-wannabe.
    I'm not sure if you realize this, but some formulas are expanded upon after the initial presentation.

  29. #29
    MonkeyF0cker
    Update your status
    MonkeyF0cker's Avatar Become A Pro!
    Join Date: 06-12-07
    Posts: 12,144
    Betpoints: 1127

    Quote Originally Posted by roasthawg View Post
    So your take is that variance is indeed reduced but ml bets for favorites generally have a lower eg than spreads do?
    For the most part, yes.

  30. #30
    Data
    Data's Avatar Become A Pro!
    Join Date: 11-27-07
    Posts: 2,236

    Quote Originally Posted by MonkeyF0cker View Post
    The fact is, if you are a modeler and have scoring distributions, you can add variance to your Kelly staking.
    What do you mean? How and why would one do this? What variance you are talking about?

  31. #31
    Data
    Data's Avatar Become A Pro!
    Join Date: 11-27-07
    Posts: 2,236

    Quote Originally Posted by Wrecktangle View Post
    I have mentioned the difference between BJ and sports in terms of Kelly Criterion several times here in the Tank. If you were a little more widely read, you would know this vice throwing around careless, provocative statements.
    False, you always talked about differences in assessing EV, the part that exists in both types of betting. Here you started talking about variances, the subject what has never been brought up in this way, and, unlike EV, addressing variance is neither common nor comparable issue. In fact, it is NOT an issue in sports betting due to fixed odds betting.

    Quote Originally Posted by Wrecktangle View Post
    As to mistakes, I'm sure I'm not the only one to see the irony of your statement. As to my statements, I'm pretty careful to only post on items I know something about; something you might want to think about.

    Thirdly, if you had read Kelly's paper, you would have seen that it indeed only does just cover expected winning percentage as it covers how much of a bankroll to bet optimally. Any expansion beyond that only exists in your "holo deck" imagination. BTW, it was a most unusual document since he was working at Bell Labs.

    I am afraid you have reached Troll level now. My communications with you are over.

    Wrecktangle out.
    I got it, you are good, I am bad, your shit don't stink, I am not too good to even realize this.

    It could be funny when an ignorant troll fails like in example a few messages above, it is rather sad when a good poster fails and starts talking exclusively about personalities and their stories.

    Here are a few points on the subject matter which is variance and Kelly:
    1) the OP question is clearly about sports betting, you bring up an issue irrelevant to sports betting
    2) where variance is an issue (BJ, finance), any assumptions about variance are not within Kelly theory but rather are the shortcuts on the bettor's part, the bettor can use any approach to assessing variance while staying within Kelly framework
    3) the issues with variance (where exist) are not an argument against using Kelly because the errors caused by mistakes in assessing variance are of much lesser significance comparing to the errors caused by mistakes in assessing the means.

  32. #32
    MadTiger
    Wait 'til next year!
    MadTiger's Avatar Become A Pro!
    Join Date: 04-19-09
    Posts: 2,724
    Betpoints: 47

    Quote Originally Posted by Flying Dutchman View Post
    Game on asshole, I'm your huckleberry.


    Couldn't resist.

  33. #33
    MonkeyF0cker
    Update your status
    MonkeyF0cker's Avatar Become A Pro!
    Join Date: 06-12-07
    Posts: 12,144
    Betpoints: 1127

    Quote Originally Posted by Data View Post
    What do you mean? How and why would one do this? What variance you are talking about?
    Simply bankroll variance. It would involve changing your multiplier to fit a specific volatility threshold. It would certainly not be optimal, but would ensure that you were not overstaking. It's a very poor idea. However, it could be done.

  34. #34
    Flying Dutchman
    Floggings continue until morale improves
    Flying Dutchman's Avatar Become A Pro!
    Join Date: 05-17-09
    Posts: 2,467
    Betpoints: 759

    Quote Originally Posted by MonkeyF0cker View Post
    I'm not sure if you realize this, but some formulas are expanded upon after the initial presentation.
    OK I'm game, where's your citation? Kelly wrote more?

    Show a poor dumb-shit like me your golden equation that mods Kelly for variance, dweeb. Remember, we need proof that is it optimal or you go into the locker again.

  35. #35
    Flying Dutchman
    Floggings continue until morale improves
    Flying Dutchman's Avatar Become A Pro!
    Join Date: 05-17-09
    Posts: 2,467
    Betpoints: 759

    Quote Originally Posted by MadTiger View Post


    Couldn't resist.
    Thanks Tiger, I love that movie. Probably Kilmer's best role...after his as Jim Morrison.

12 Last
Top