Kelly Advice: Adjust Daily or Periodically

[HedgeHog]

If you follow Kelly to the letter, you should re-calculate your next bet following each known outcome. I've been using a 1/2 kelly method, but only adjust my bet size after the bankroll has changed 10% or after I take some profit. For example if my bankroll starts at 10k and goes to 11k, I adjust. Any other thoughts on this?

[donjuan]

Why would you base it on an arbitrary number? Use your current bankroll to calculate.

[HedgeHog]

DJ: So you obviously update your bet daily. Do you use Full Kelly or a fraction of Kelly?

[donjuan]

DJ: So you obviously update your bet daily. Do you use Full Kelly or a fraction of Kelly?

I update it whenever I place a wager or a wager wins, so you could say I update it minutely. As for Kelly, I use full Kelly (when possible) with a margin of error.

[Poker_Beast]

Can you point me to this Kelly formula, I have not heard of this before.

[HedgeHog]

On this page, under the SBR logo, click where it says tools.

[dogman]

Hedge, personally I adjust it day by day

[20Four7]

HH. The first thing I do when I get up every morning is punch into my spreadsheet all my balances at every book. I know where my roll is for the day. I do this every morning before I do anything. I use that figure.

[Poker_Beast]

On this page, under the SBR logo, click where it says tools.

Thanks

[Billy Baxter]

Correct money management is easy to explain. It is almost impossible to adhere to.

There have been about as many 'systems' for varying the bet size as there are systems to beat the craps table. None of them work and all of them do harm by giving false expectations. Double-up systems, star betting, the Kelly Criterion, etc. all have the same thing in common. They adversely affect the cash flow and they raise the breakeven percentage that must be accomplished.

Any time you vary your bet size by the slightest amount, you increase the breakeven percentage of 52.38. If you refigure your percentage after each bet, your breakeven jumps to over 55%. If you have 1 star and 5 star bets or bet twice as much on some plays, you should understand that only the large bets are meaningful to whether you will be a winner. You might as well not bet the smaller ones. In the end, they won't matter.

Some touts say you should increase your bet when you are on a winning streak and decrease it when you are on a losing streak. The key word is 'are'. If you won yesterday, you 'were' on a winning streak. But that was yesterday. If you know you will win today, why not bet it all? If you know you will lose today, you might consider not betting at all.

The problem with bet size is not the streaks, but the breakeven. If you have a 56% advantage on each bet, over 200 games you will win less than 50% about 17% of the time.You will win more than 60% about 17% of the time. Winning 100 and losing 100 with a 5% unit, you will lose 50% of your bankroll to the vigorish. (100 wins times $50 less 100 losses times $55 equals minus $500 on a $1,000 bankroll).You will lower your bet and never get even. I recommend playing no more than 2% of your bankroll. Anything over 2% is unacceptably risky, even for recreational bettors.

.

[TLD]

Any time you vary your bet size by the slightest amount, you increase the breakeven percentage of 52.38. If you refigure your percentage after each bet, your breakeven jumps to over 55%.

An interesting claim. Could you perhaps show the math with an example?

If you have 1 star and 5 star bets or bet twice as much on some plays, you should understand that only the large bets are meaningful to whether you will be a winner. You might as well not bet the smaller ones. In the end, they won't matter.

Why will they not matter? I would think the 1 star plays would matter, on average, 1/5 as much as the 5 star plays. Or for the system that doubles some plays, the regular plays would matter, on average, 1/2 as much as the double plays.

Why is it better to skip +EV plays that are 1/5 or 1/2 as meaningful as your biggest plays?

[Ganchrow]

Correct money management is easy to explain. It is almost impossible to adhere to.

-snip-

If you're going to quote JR Miller's foolish website word-for-word you should at least enclose your post in [QUOTE][/QUOTE] tags.

http://professionalgambler.com/rjmiller1.htm

[HedgeHog]

I don't understand Miller's logic in betting exactly 2% of bankroll regardless of your edge. You mean to tell me that someone who has isolated a 60% edge on a -105 play should just wager 2% of his 'roll each time. You're leaving money on the table, which is poor money mgmt IMO.

[Justin7]

I'm not impressed with JR Miller.

As mentioned earlier, you need to be sure of your edge.

If you use Kelly betting, it doesn't hurt you much if you underbet - the important thing is to avoid overbetting. I'd adjust my bet size anytime I lost 10% of my bankroll.

[HedgeHog]

I'm not impressed with JR Miller.

As mentioned earlier, you need to be sure of your edge.

If you use Kelly betting, it doesn't hurt you much if you underbet - the important thing is to avoid overbetting. I'd adjust my bet size anytime I lost 10% of my bankroll.

We're in the same camp. Thanks, Justin.

[newb411breaker19]

I used to think that JR Miller was a legitimate handicapper after I tracked him for around 6 months and he was up 55 units. Since that time in late 2005 he is now down over 80 units.

He's a loser and shows how inept he is in his "debunking the Kelly Criterion" article

[chemist]

I update it whenever I place a wager or a wager wins, so you could say I update it minutely. As for Kelly, I use full Kelly (when possible) with a margin of error.

I assume this means you discount your estimate of the edge. A wise precaution. How do you account for multiple concurrent wagers? just use the currently available bank as the bank?

[Ganchrow]

How do you account for multiple concurrent wagers? just use the currently available bank as the bank?

That would of course be suboptimal. The proper methodology (although not necessarily the most computationally efficient) would be to fully specify the utility function and maximize from there.

I'll note that my Kelly calculator properly handles simultaneous wagers (as long as said wagers are parlayable).

[chemist]

That would of course be suboptimal. The proper methodology (although not necessarily the most computationally efficient) would be to fully specify the utility function and maximize from the

Where dat text go?

Of course the method I proposed is suboptimal. It leads to understaking, mild unless the bets are at short odds or very numerous. Most Kelly practitioners apply expectations discounting, fractional Kelly, or other ad hoc adjustments that reduce stakes, anyway.

I'll note that my Kelly calculator properly handles simultaneous wagers (as long as said wagers are parlayable).

I presume by "parlayable" you mean independent. AFAIK multiple Kelly doesn't have an algebraic solution so your calculator would be doing numerical optimization. Did you base it on published work? Using it probably isn't practicable when you're racing to beat the steam for the fifth time in ten minutes.

[Ganchrow]

Sorry for taking so much time to get back to you on this ... but I'm on vacation and have only been checking the forum sporadically.

I presume by "parlayable" you mean independent.

No. By "parlayable" I literally just mean able to be parlayed (although I should have specified parlayable and uncorrelated), which is neither a necessary nor a sufficient condition for uncorrelated. For example if Book A is the only book offering odds on the New York Mets to win the 2012 World Series and Book B is the only book offering odds on the Kansas City Chiefs to win the 2013 Super Bowl then while these bets are essentially (there is Armageddon risk, after all) independent and hence uncorrelated (although the reverse is not true -- while a necessary condition, uncorrelated doesn't imply independence), they aren't parlayable.

AFAIK multiple Kelly doesn't have an algebraic solution so your calculator would be doing numerical optimization.

In fact there is an algebraic solution for events that are parlayable and uncorrelated. See this post for an overview.

[chemist]

Sorry for taking so much time to get back to you on this ... but I'm on vacation and have only been checking the forum sporadically.

No. By "parlayable" I literally just mean able to be parlayed (although I should have specified parlayable and uncorrelated), which is neither a necessary nor a sufficient condition for uncorrelated. For example if Book A is the only book offering odds on the New York Mets to win the 2012 World Series and Book B is the only book offering odds on the Kansas City Chiefs to win the 2013 Super Bowl then while these bets are essentially (there is Armageddon risk, after all) independent and hence uncorrelated (although the reverse is not true -- while a necessary condition, uncorrelated doesn't imply independence), they aren't parlayable.

In fact there is an algebraic solution for events that are parlayable and uncorrelated. See this post for an overview.

Thanks. I see I'll have to think about that for some time.