Flat betting versus the Kelly Criterion - a simulation

[specialronnie29]

wrecktangle

youre right. we cant ever know anything about anything. so go get a job and leave the betting to us

[Wrecktangle]

You guys really can't read. I've never said that Kelly is not a good BR management system, matter of fact, I use it.

What I have said, it is extremely difficult to use correctly especially when you can't calc your win% under every conditional that matters. Since you haven't exhibited much clue of this, or much evidence that you keep accurate enough records of your plays both long term and short term, I certainly assume you're doing it incorrectly. Thusly, my point has been, nubes (you guys) shouldn't play with nitroglycerin. In that case, flat bet and keep your BR for the season if you tend toward -EV bets, and build it if you can manage +EV.

The other thing I pointed out was that Kelly is probably not optimal.

For all the howls of pain, you'd think I'd said that Monkey and biztips are cross-dressing Obama-loving liberals. While they might be, I never said that nor that Kelly was not a theoretically good (but probably not optimal, and certainly difficult to manage) BR management system.

Grow up.

[bztips]

There you go again...

When you make grand pronouncments like "you can't calc your win% under every conditional that matters", that's not a particularly helpful (or practical) comment. But since you profess to use Kelly yourself, I guess you're the one person in the world who can properly consider "every conditional that matters".

Similarly, comments like "Kelly is probably not optimal" indicates that maybe you really have no idea at all what you're talking about. "Probably not optimal" is not a statement that any knowledgeable person would make when talking about the Kelly criterion. Given its assumptions, by definition it is optimal. If you want to argue the assumptions -- in particular, knowledge of the actual size of your edge -- fine that's legitimate. But it's a huge leap of faith to go from there and conclude that Kelly is therefore not useful for real-world betting.

[MonkeyF0cker]

You guys really can't read. I've never said that Kelly is not a good BR management system, matter of fact, I use it.

What I have said, it is extremely difficult to use correctly especially when you can't calc your win% under every conditional that matters. Since you haven't exhibited much clue of this, or much evidence that you keep accurate enough records of your plays both long term and short term, I certainly assume you're doing it incorrectly. Thusly, my point has been, nubes (you guys) shouldn't play with nitroglycerin. In that case, flat bet and keep your BR for the season if you tend toward -EV bets, and build it if you can manage +EV.

The other thing I pointed out was that Kelly is probably not optimal.

For all the howls of pain, you'd think I'd said that Monkey and biztips are cross-dressing Obama-loving liberals. While they might be, I never said that nor that Kelly was not a theoretically good (but probably not optimal, and certainly difficult to manage) BR management system.

Grow up.

LMAO. Are you part of BigDaddy's syndicate by chance?

[JustinBieber]

"kelly is probably not optimal" nice one.

[tim0402]

how do you figure one play has a 54% chance of winning and another play has a 57% chance of winning?

[Dark Horse]

how do you figure one play has a 54% chance of winning and another play has a 57% chance of winning?

The payout has to be factored in of course. Recently, I've been playing 2 team round robins for NHL underdogs. The first time the dogs went 5-1 SU and the average ML was +142. The second time the dogs went 4-2 SU and the average ML was +149. Because I parlayed these games, the payouts were very sweet (where's Kelly where it comes to parlays?).

If I take a winning expectation of 55%, and enter a +142 line into the Kelly calculator, the full Kelly bet size is 23% of BR for one game. lol That would give me about four shots, and if I went 0-4, which is just a matter of time, that would be it. (also, counter-intuitively, Kelly gives me more bets as the moneyline comes down). This is simple proof that Kelly is not seamlessly compatible with sports betting. Of course, proponents of Kelly will then suggest fractional Kelly. Fine. But remember that the whole idea behind Kelly is to maximize profit. Which brings me back to parlays...

Kelly is not the final answer. It is undoubtedly superior to flat betting -as long as there's no gap between theory and personal ability-, but there are more profitable ways open to those who have closed that gap.

[bztips]

The payout has to be factored in of course. Recently, I've been playing 2 team round robins for NHL underdogs. The first time the dogs went 5-1 SU and the average ML was +142. The second time the dogs went 4-2 SU and the average ML was +149. Because I parlayed these games, the payouts were very sweet (where's Kelly where it comes to parlays?).

Ganch has provided the answer of how to use Kelly for parlays:

http://forum.sbrforum.com/handicapper-think-tank/521569-simple-closed-form-solution-unconstrained-simultaneous-bet-kelly-staking.html

[donjuan]

The payout has to be factored in of course. Recently, I've been playing 2 team round robins for NHL underdogs. The first time the dogs went 5-1 SU and the average ML was +142. The second time the dogs went 4-2 SU and the average ML was +149. Because I parlayed these games, the payouts were very sweet (where's Kelly where it comes to parlays?).

If I take a winning expectation of 55%, and enter a +142 line into the Kelly calculator, the full Kelly bet size is 23% of BR for one game. lol That would give me about four shots, and if I went 0-4, which is just a matter of time, that would be it. (also, counter-intuitively, Kelly gives me more bets as the moneyline comes down). This is simple proof that Kelly is not seamlessly compatible with sports betting. Of course, proponents of Kelly will then suggest fractional Kelly. Fine. But remember that the whole idea behind Kelly is to maximize profit. Which brings me back to parlays...

Kelly is not the final answer. It is undoubtedly superior to flat betting -as long as there's no gap between theory and personal ability-, but there are more profitable ways open to those who have closed that gap.

So much fail in one post. Congrats.

[u21c3f6]

The payout has to be factored in of course. Recently, I've been playing 2 team round robins for NHL underdogs. The first time the dogs went 5-1 SU and the average ML was +142. The second time the dogs went 4-2 SU and the average ML was +149. Because I parlayed these games, the payouts were very sweet (where's Kelly where it comes to parlays?).

If I take a winning expectation of 55%, and enter a +142 line into the Kelly calculator, the full Kelly bet size is 23% of BR for one game. lol That would give me about four shots, and if I went 0-4, which is just a matter of time, that would be it. (also, counter-intuitively, Kelly gives me more bets as the moneyline comes down). This is simple proof that Kelly is not seamlessly compatible with sports betting. Of course, proponents of Kelly will then suggest fractional Kelly. Fine. But remember that the whole idea behind Kelly is to maximize profit. Which brings me back to parlays...

Kelly is not the final answer. It is undoubtedly superior to flat betting -as long as there's no gap between theory and personal ability-, but there are more profitable ways open to those who have closed that gap.

Dark Horse, I generally like your posts but the above is off base.

With valid inputs (edge/odds) Kelly is always the optimal answer but there are also valid reasons to use a fractional Kelly. The problem with your scenario above is that just because you are wagering 23% of your bankroll per wager, it does not mean that you are wiped out when you have 4 losses in a row. With 4 losses in a row you would still have a little more than a third af your bankroll left because for the next 3 losses after the first loss you would be wagering 23% of smaller bankrolls. In fact, after 9 losses in a row you would still have a little more than 10% of your bankroll left. Of course, after 6 losses in a row I would begin to doubt my 55% win rate.

I find that the ones that tend to argue against Kelly either

Don't use it,
Don't understand it,
Don't apply it properly,
Don't have an edge or
Don't know how to quantify the necessary Kelly inputs.

Kelly is very much worth understanding whether you can actually use it or not. Unfortunately there is much misinformation written about Kelly that can and does confuse the issue. This makes it more difficult for someone to understand Kelly when they add the misinformation they have read into the mix.

Joe.

[sharpcircle]

Kelly wins , flat betting loses... game over

[durito]

Those damn 55% plays at +143

[donjuan]

Those damn 55% plays at +143

But if you parlay them you get a "sweet" payout. How could Kelly ever account for "sweet" payouts?!?

[Dark Horse]

Joe, four losses in a row, at 23% of BR per bet, result in a 35% BR. If you have no problem with that, don't let me talk you out of it.

dj, if only everything was as predictable as you.

[That Foreign Guy]

I'd just love to find a genuine 23% betting opportunity.

[Dark Horse]

I'd just love to find a genuine 23% betting opportunity.

A favorite of -300 that you think has a 81% chance of winning gives you a Kelly bet size of 24% of BR.

But if that chance is down to 75%, the Kelly bet size is 0%. No bet.

6% difference in winning expectation translating into 24% of BR. Good luck with that.

[donjuan]

Joe, four losses in a row, at 23% of BR per bet, result in a 35% BR. If you have no problem with that, don't let me talk you out of it.

dj, if only everything was as predictable as you.

What's really predictable is you making a fool of yourself in the Think Tank anytime you open your mouth. For you, it clearly doesn't matter whether you go full Kelly, half Kelly or any variant of Kelly. The optimal stake is the same: 0.

[Dark Horse]

What's really predictable is you making a fool of yourself in the Think Tank anytime you open your mouth. For you, it clearly doesn't matter whether you go full Kelly, half Kelly or any variant of Kelly. The optimal stake is the same: 0.

dj in one sentence: 'there could be nothing beyond the little that I know'.

[donjuan]

I'd just love to find a genuine 23% betting opportunity.

It's not unheard of in the prop market. Of course if you have any bankroll to speak of limits wouldn't let you get near 23% down.

[That Foreign Guy]

A favorite of -300 that you think has a 81% chance of winning gives you a Kelly bet size of 24% of BR. But if that chance is down to 75%, the Kelly bet size is 0%. No bet. 6% difference in winning expectation translating into 24% of BR. Good luck with that.

Yeah, if you're not estimating your win% accurately then Kelly isn't going to work. I think it should be possible to estimate your winning chance a bit more accurately than +/- 6% though.

It's not unheard of in the prop market. Of course if you have any bankroll to speak of limits wouldn't let you get near 23% down.

Yeah, I don't have a particularly large bankroll but would definitely struggle to get 23% down on a prop.

[Snowball]

kelly is a bad system because it relies on odds as an indicator of likelihood.
therein is the flaw.

[u21c3f6]

kelly is a bad system because it relies on odds as an indicator of likelihood.
therein is the flaw.

I honestly don't know how to respond to this statement.

Joe.

[LUSabres]

Flat betting is a nightmare in the long run.

[Sawyer]

U kidding? Flat betting is good

[FourLengthsClear]

kelly is a bad system because it relies on odds as an indicator of likelihood.
therein is the flaw.

No it does not.

(In case you misunderstood that, I will rephrase.)

The odds are NOT used as an indicator of likelihood.

[Dark Horse]

Yeah, if you're not estimating your win% accurately then Kelly isn't going to work. I think it should be possible to estimate your winning chance a bit more accurately than +/- 6% though.

I gave an extreme example to show how sensitive Kelly is.

If you allow me to use live betting as example, a difference between 81% and 75% may be as little as one first down (or less). I've seen plenty of games with a >80% winning expectation, late in the game, swing the other way. Far too common to risk 23% of BR on. And I don't see why live games, and betting, would bend the rules of math. One late hit in baseball, one first down in football, and that 6% (near the extremes) is gone like a snowball in hell. Of course, the math sect is not bothered by such realities. If Kelly says 23%, that's all there is to it.

I tried to point out the value of parlays, as often being superior to Kelly bet sizing (both in terms of payout and exposure), and illustrate it with a personal example, but even the simplest form of out-of-the-box thinking would draw out the sect with their predictable expressions of misplaced arrogance and superiority. These guys may know math well. But they do not know how to think. To them that is a foreign concept that should be tanked. And they conveniently forget that the math they use was discovered for them by those who did think. They take great pleasure in chasing people away from here, most recently longtime posters Flying Dutchman and Wrecktangle (who has since returned sporadically). That is their heritage; the one thing they take pride in. Hey, to each their own...

[donjuan]

I gave an extreme example to show how sensitive Kelly is.

If you allow me to use live betting as example, a difference between 81% and 75% may be as little as one first down (or less). I've seen plenty of games with a >80% winning expectation, late in the game, swing the other way. Far too common to risk 23% of BR on. And I don't see why live games, and betting, would bend the rules of math. One late hit in baseball, one first down in football, and that 6% (near the extremes) is gone like a snowball in hell. Of course, the math sect is not bothered by such realities. If Kelly says 23%, that's all there is to it.

I tried to point out the value of parlays, as often being superior to Kelly bet sizing (both in terms of payout and exposure), and illustrate it with a personal example, but even the simplest form of out-of-the-box thinking would draw out the sect with their predictable expressions of misplaced arrogance and superiority. These guys may know math well. But they do not know how to think. To them that is a foreign concept that should be tanked. And they conveniently forget that the math they use was discovered for them by those who did think. They take great pleasure in chasing people away from here, most recently longtime posters Flying Dutchman and Wrecktangle (who has since returned sporadically). That is their heritage; the one thing they take pride in. Hey, to each their own...

Actually thinking, rather than pretending to think would be a start for you. Optimal Kelly staking includes parlays. And if you are betting just round robin parlays, you're probably overbetting. You really shouldn't lecture others when you haven't even taken the time to try to understand Kelly.

[tomcowley]

I gave an extreme example to show how sensitive Kelly is.

If you allow me to use live betting as example, a difference between 81% and 75% may be as little as one first down (or less). I've seen plenty of games with a >80% winning expectation, late in the game, swing the other way. Far too common to risk 23% of BR on.

So what exactly are you saying? That more than 20% of games with an 80% win expectancy at point A end up losing, or that more than 20% of games with an 80% win expectancy at point A drop below 75% win expectancy at some point during the rest of the game?

[and24]

U kidding? Flat betting is good

Flat betting is good if you have a win percentage above %52 ATS which is really hard in the long term.

[bztips]

I gave an extreme example to show how sensitive Kelly is.

That's one of the reasons many people use fractional Kelly. Using your example, quarter-Kelly would suggest 6.4% of bankroll as your bet.

[That Foreign Guy]

If you make a bet with 81% chance of winning and it then reduces to 75% chance of winning, you haven't lost your entire bet. In fact that bet is still worth (75% * odds * stake). This may be less than you paid for it, such is life, but it definitely is not worthless.

Let's say I bet 23% of bankroll at 81% chance of winning at -300 (I use 1.33 decimal so there's a little rounding error). Conveniently, 23% of bankroll is 3000.

If you want a made up live betting example, the Steelers are 4th and 3 and need to convert for a First Down or the Packers will kneel out the clock. If they get the First down they will become 30% likely to win and they have whatever % chance of converting it that gives a bet on the Packers 81% chance of winning right now (can't be assed doing the math for a made up situation).

Expected Value of the bet is (81% * 1.33 * 3000) = 3231.9 minus the 3000 we paid for it = +231.9

Now a the Steelers convert and our chance of winning plummets to 70%. What is the bet worth now?

I'll give you a clue, it's not zero.

Value of the bet is (70% * 1.33 * 3000) = 2793 minus the 3000 we paid for it = -207. OK, so it's now a bad bet, but at the time we made it, it was a good one, and even now it is far from having zero value.

[Dark Horse]

I'm the one lecturing dj? lol

There once was a pond that was very crowded with frogs. One day a frog hops up the hill and discovers an ocean on the other side. When he tells the others about this, they don't believe him. "It's impossible that there is a bigger body of water than our crowded pond!! How can you even suggest such a thing?!"

Have fun in the think pond, tanking any thought that doesn't 'belong'.

(For your information, new ideas seldom come ready-made and prepackaged. You must have heard of 1% inspiration/99% perspiration... Who are you to work with ideas that others worked out for you, yet criticize those who are tilling the soil of new ideas?)

[MonkeyF0cker]

So what exactly are you saying? That more than 20% of games with an 80% win expectancy at point A end up losing, or that more than 20% of games with an 80% win expectancy at point A drop below 75% win expectancy at some point during the rest of the game?

Could you please respond to this, DH?

And are you really calling flat-betting a "new idea?"

Perhaps, you, Wrecktangle, and/or Flying Dutchman could create a +EV simulation which shows how flat-betting outperforms Kelly. I would trust that you've already done the legwork on this anyway.

[Dark Horse]

And are you really calling flat-betting a "new idea?"

Perhaps, you, Wrecktangle, and/or Flying Dutchman could create a +EV simulation which shows how flat-betting outperforms Kelly. I would trust that you've already done the legwork on this anyway.

Where did I advocate flat betting?

Sect, right? Figures. Who else could be so full of themselves that they would ask others to respond to something, without even taking the trouble to read what was said. lol

[MonkeyF0cker]

Then what exactly are you promoting here, DH? Simply anti-Kelly rhetoric? Random wager amounts? Progressive wagering? What?!?

And why are you avoiding the question about your own live-betting scenario?