Optimal bet size question

[Darkside Magick]

Darkside to me if everything is as random and chaotic as you say it is then every now and then we’d have an NBA final score of like 30-27.

It is not what I say.. It is what our science say.. That everything is chaos and that everything in our universe comes from chaos

Let's throw random away.. I use it to simplify things for people.

Now you said there should be a score like 30-27 should pop up... Only if that out come is in the the millions maybe billions of outcomes that are run simultaneously and can be collapsed of the wave function by humans

I

[d2bets]

It is not what I say.. It is what our science say.. That everything is chaos and that everything in our universe comes from chaos

Let's throw random away.. I use it to simplify things for people.

Now you said there should be a score like 30-27 should pop up... Only if that out come is in the the millions maybe billions of outcomes that are run simultaneously and can be collapsed of the wave function by humans

I

Magic mushrooms or LSD?

[d2bets]

Variance is a killer if :

A: You've miscalculated your edge

B. You over play your bankroll because you believe you have that edge.

Happens to a lot of gamblers.

But his point is even if the edge is calculated exactly, variance can kill you too easily if you bet full Kelly.

[d2bets]

Kelly criterion is Theory for Sports Betting.. It is not practical because again you cannot know your edge for random(Chaotic) events

Now the sample size can be use a Monte Carlo Simulation of sort where one can infuse Bayes Theorem and get a more optimal bet size

So with 44% win percentage and avg odds of +170 and taking the ROI as and using that as your conditional probability(Bayes Theorem)

The optimal bet size is 4.14%

Dude, set that aside. Let's say it's not sports and it's just a literal coin flip and you are being given +110 or +120 or something. You can know your edge precisely, bet it based on Kelly, and go broke on an unlucky run.

Kelly depends on how willing you actually are to lose your bankroll. If you don't want to take that chance, bet less.

[semibluff]

Extreme variance is an inevitable certainty that will eventually bankrupt a Kelly bettor. Kelly Criterion is designed to maximise rewards. It's not designed to minimise risk. At some point you will lose 8 coin flips in a row. Your staking system has to be able to withstand that carnage. With an unlimited number of profitable odds coin flips i'd recommend maybe 10-15% of Kelly. If the profitable odds coin flips are going to be shut down very quickly it comes down to how much greater a risk you're willing to absorb.

[Gaze73]

It's funny because I actually read that kelly is designed to never make you go broke since you always bet in percentages. But you will have swings like $1k to $10k in 3 months, then down to $2k in 5 days. You might want to chase with $500 bets to get back up and tilt away the rest.

[Optional]

Kelly criterion is Theory for Sports Betting.. It is not practical because again you cannot know your edge for random(Chaotic) events

Sport is not a random chaotic event.

You ignored it last time, but how do you explain every single sports team not trending to a .500 record if it is truly random?

If you can't, then it is not.

[Optional]

It's funny because I actually read that kelly is designed to never make you go broke since you always bet in percentages. But you will have swings like $1k to $10k in 3 months, then down to $2k in 5 days. You might want to chase with $500 bets to get back up and tilt away the rest.

I don't think that was a design concern at all.

Kelly came up with the theory to help design the most efficient telephone network possible.

It's conversion to sports betting was all about maximizing bankroll growth, not avoiding going broke.

As I understand it anyway.

[Gaze73]

From a kelly calculator website: "What do you think about simple Kelly betting? Even though it is designed to never let you go bankrupt, Kelly still allows wild volatility swings."

It makes sense because when your even if you lose 90% of the BR you simply reduce bet size by 90%. But in practice a gambler is more likely to deposit again.

[Optional]

From a kelly calculator website: "What do you think about simple Kelly betting? Even though it is designed to never let you go bankrupt, Kelly still allows wild volatility swings."

It makes sense because when your even if you lose 90% of the BR you simply reduce bet size by 90%. But in practice a gambler is more likely to deposit again.

For Kelly purposes you are supposed to count all your assets as your bankroll, so there is no such thing as being able to redeposit if following the theory properly.

That website is making an amateur assumption.

Sure, it is supposed to reduce your risk of ruin as well as maximize profit, but just because it wont ever get to zero due to % bet sizes, it does not mean you can't go broke using it.

As many people point out, using Full Kelly is dangerous for most.

[TommieGunshot]

I was just about to make a thread about Kelly because I find it stupidly dangerous.

I found this guy I'd like to follow, he has a 11.6% ROI with a very large sample size of almost 5k bets. The win rate is 44% at avg. odds of 2.7/+170.

According to Kelly, with a $1k bankroll you should bet $110.59. He's currently on a run of 2 wins and 11 losses in the past 5 days. If you bet 11% of your BR per bet, it turns out that going 2-11 is not very good for your finances. So I think anything above half Kelly is insane.

If someone was able to use fully Kelly on bets that paid +170 with a 44% chance of winning, they would be turning $1,000 into $1 billion in less than 2,100 bets. (1000 x 1.01022^2039 - using expected growth from sbr Kelly calculator, if that math is wrong, someone please explain why).

While Kelly betting is correct to maximize profits, if anyone is able to consistently find bets like this, they will never bet using Kelly, because even betting the limit is so far under the optimal bet size. In that case, just bet whatever the limit is and move on. For every 13 games, there is a 7% chance of winning 2 or fewer. That will happen on occasion, and they won't care at all.

[semibluff]

Hitting a bottom 7% sequence shouldn't devastate a bankroll. Hitting a bottom 0.4% sequence will devastate your bankroll unless you're super cautious with staking. Encountering that +25000 sequence approaches inevitability if you continue long enough. I encountered a 58-92 sequence last year when I should have been hitting at 51.5%. With a staking reset it would theoretically take over 2000 bets/choices just to win back losses incurred on that 150 outcome sequence. If I were continuing with a devastated bankroll it would take much longer.

[TommieGunshot]

Hitting a bottom 7% sequence shouldn't devastate a bankroll. Hitting a bottom 0.4% sequence will devastate your bankroll unless you're super cautious with staking. Encountering that +25000 sequence approaches inevitability if you continue long enough. I encountered a 58-92 sequence last year when I should have been hitting at 51.5%. With a staking reset it would theoretically take over 2000 bets/choices just to win back losses incurred on that 150 outcome sequence. If I were continuing with a devastated bankroll it would take much longer.

Sequences at the top 0.04% also approach inevitability. In this example, if the odds offered were +100, betting full Kelly, this 0.04% to the bad example means losing 68% of the bankroll; the 0.04% to the good side means the bankroll nearly quadruples.

So long as you experience the positive swings at an equal rate as the negative ones, you will come out further ahead using Kelly than any other bet sizing. And when someone insists they experience the negative swings more than the positive ones, they have either miscalculated probabilities, or they are a troll.

[semibluff]

Sequences at the top 0.04% also approach inevitability. In this example, if the odds offered were +100, betting full Kelly, this 0.04% to the bad example means losing 68% of the bankroll; the 0.04% to the good side means the bankroll nearly quadruples.

So long as you experience the positive swings at an equal rate as the negative ones, you will come out further ahead using Kelly than any other bet sizing. And when someone insists they experience the negative swings more than the positive ones, they have either miscalculated probabilities, or they are a troll.

I'm not disputing the math or the upside to using Kelly. The problem is that you can't just say 'So long as you experience the positive swings at an equal rate as the negative ones' because there isn't a real World balancing out mechanism. At every point probability resets itself. After a negative +25000 probability sequence you are just as likely to hit another negative sequence as hit a positive sequence. Go long enough and you'll hit multiple massively negative sequences before you encounter a similarly massively positive sequence. Full or high proportion Kelly is for those who accept the downside risks and those who can reset whenever necessary.

[TommieGunshot]

I'm not disputing the math or the upside to using Kelly. The problem is that you can't just say 'So long as you experience the positive swings at an equal rate as the negative ones' because there isn't a real World balancing out mechanism. At every point probability resets itself. After a negative +25000 probability sequence you are just as likely to hit another negative sequence as hit a positive sequence. Go long enough and you'll hit multiple massively negative sequences before you encounter a similarly massively positive sequence. Full or high proportion Kelly is for those who accept the downside risks and those who can reset whenever necessary.

It seems like you are disputing the math. Go long enough and you will hit multiple massively positive sequences before you encounter a similarly massive negative sequence. Go long enough and if it hasn't balanced out to a certain extent, it means the probabilities were not right.
https://en.wikipedia.org/wiki/Conver...ndom_variables and https://en.wikipedia.org/wiki/Proofs...ndom_variables
Reminds me a good line I heard from a blackjack player. Put enough hours so that even being three standard deviations below expectations means earning a profit. Why not also do that with sports bets?

[TommieGunshot]

Hitting a bottom 0.4% sequence will devastate your bankroll unless you're super cautious with staking.

Play long enough and even a bottom 0.04% sequence means winning money. In your example of 51.5% chance of winning each bet, even on the 0.04% negative side, you would have still won 1000 more bets than you lost after 50000 bets, at even money that is a good profit even with extremely unlucky variance. Even with extreme negative results, anything less than Kelly betting would mean earning less money.