pm: Team RoR

[pm]

I post with extreme trepidation now, hoping I'm not missing something obvious, but whatever...

Bill C. and Marc Ingenoso mention in "Risk Formulas For Proportional Betting" (section 3.4) that k = 0.5 (half-kelly) is far too risky for a team.

I really don't get that. Say you start with a $25,000 BR, and you consider dropping to $50 as "going broke". According to the paper, your odds of hitting $50 when betting perfect full kelly = 50/25000 = a 1 in 500 chance. That's somewhat substantial. Your odds of hitting $50 when betting perfect half-kelly = (50/25000)^3 = a 1 in 125,000,000 chance. That's essentially zero.

And it shouldn't matter that you can't bet perfect half-kelly. Say your session bank is 20% of your BR, distributed amongst a few BPs, and each BP bets half-kelly. In the worst-case scenario, towards the end of the session, a BP might actually be betting 0.625 kelly. The odds of dropping from $25,000 to $50 when betting 0.625k are still only 1 in 866431. Again, essentially zero.

??? So my question is how could you ever have anything to worry about when you're betting half-kelly, even when you're on a team? As long as you have a decent game and you're not screwing up the count, you should be golden, right?

(The paper never clarified whether or not the teams were resizing; if they weren't, I can understand the < 0.25k figure, but why wouldn't a team resize?)

[Sun Runner]

> I post with extreme trepidation now, hoping
> I'm not missing something obvious, but
> whatever

Welcome to the club ...

[pm]

Nice, I'm feeling a little dense after that growth rate discussion..

[G Man]

> I post with extreme trepidation now, hoping
> I'm not missing something obvious, but
> whatever...

> Bill C. and Marc Ingenoso mention in
> "Risk Formulas For Proportional
> Betting" (section 3.4) that k = 0.5
> (half-kelly) is far too risky for a team.

Betting half Kelly yelds a RoR of 1.83% which is much too high for most professional teams.

[pm]

Christ, here we go again, I REALLY hope I'm not missing something obvious.

Betting half-kelly yields a 1.83% INSTANTANEOUS RoR, i.e. if you never resize, your RoR is 1.83%.

This is what I've understood from all these discussions, and I'm damn near positive about this: if you bet half-kelly and always resize after your session, RoR is near zero.

Bill and Marc prove in their paper that (assuming you can place fractional bets), if you bet perfect full-kelly, RoR is truly ZERO, and that betting full-kelly ensures that you always reach your bankroll goal (exit times are a different issue, of course).

[Don Schlesinger]

> Christ, here we go again, I REALLY hope I'm
> not missing something obvious.

We really need to end this going around in circles. NO ONE plays any form of Kelly in real life. You simply don't constantly resize. Hell, you resize VERY infrequently. It's much too much of a pain in the ass. Instead, you pick a fraction of Kelly with which you're comfortable, and you stick with it for a long time.

There is no sense explaining to us that half-Kelly has zero ROR if we resize: so does FULL Kelly! But, I repeat, we don't, for the most part, resize.

Finally, suppose a team has a dozen players all playing simultaneously. Do you want them all to play half-Kelly without communicating results to one another until their sessions are over?

You're really misunderstanding what terrible things can (and do!) happen when: 1) players don't resize mid-session, and 2) players don't communicate results until after one or more sessions have been completed.

Don

[pm]

> We really need to end this going around in
> circles.

Sorry, I tend to repeat myself until someone smacks me and shows me what I'm doing wrong. I'll keep myself in check from now on.

> Instead, you pick a fraction of Kelly with
> which you're comfortable, and you stick with
> it for a long time.

So you stick with it until you've either dropped enough or gained enough to move to another sensible bet level?

> Finally, suppose a team has a dozen players
> all playing simultaneously. Do you want them
> all to play half-Kelly without communicating
> results to one another until their sessions
> are over?

How do teams actually operate? Do they simply pick a fraction that's small enough to allow team members to play simultaneously? Or is it a lot more complex than that?

[Don Schlesinger]

> So you stick with it until you've either
> dropped enough or gained enough to move to
> another sensible bet level?

Exactly.

> How do teams actually operate? Do they
> simply pick a fraction that's small enough
> to allow team members to play
> simultaneously? Or is it a lot more complex
> than that?

It's basically as simple as that. You get a lot of capital, and you play at very, very safe levels.

Don

[pm]

[Lou Percalia]

If I were running a team, I would play at full Kelly, no resize. Why? Because everything I've heard about team play is that the biggest risk of ruin involves a team member(s) screwing you over. The more money in the pot the greater the temptation there is. Also the longer you are out there, the greater the risk a player is going to go bad, either a bad habit kicking in, or getting ripped off, or having to go home, whatever. So the quicker you make bank and get the hell out of there the better. Also if you are going to get screwed, better to get screwed with a full Kelly BR than a half Kelly one.

Not that I'm an expert on team play or anything, just my opinion, from a person who has a lot of experience with how easily well-planned things can go to hell in a handbasket in real life.

> Exactly.

> It's basically as simple as that. You get a
> lot of capital, and you play at very, very
> safe levels.

> Don

[Francis Salmon]

> If I were running a team, I would play at
> full Kelly, no resize.
Let's assume you have a 5-member team with a total bankroll of 50k. Now everyone has 10k at his disposal and they all play simultaneously in different casinos.
If they played full Kelly to their team bankroll,they would have to place bets of over a tousand bucks quite frequently.
Easy to foresee that many sessions would take an abrupt end that way.

Francis Salmon

[BillC]

Your chances of being reduced to half your bankroll at half Kelly is one eigth. That is too risky for most people, in my experience.

BJ teams resize, but obviously not continuously. A team with a 50k bank might round down to the nearest green chip at the end of each day or session.

Bill Chin

> I post with extreme trepidation now, hoping I'm not
> missing something obvious, but whatever...

> Bill C. and Marc Ingenoso mention in "Risk
> Formulas For Proportional Betting" (section 3.4)
> that k = 0.5 (half-kelly) is far too risky for a team.

> I really don't get that. Say you start with a $25,000
> BR, and you consider dropping to $50 as "going
> broke". According to the paper, your odds of
> hitting $50 when betting perfect full kelly = 50/25000
> = a 1 in 500 chance. That's somewhat substantial. Your
> odds of hitting $50 when betting perfect half-kelly =
> (50/25000)^3 = a 1 in 125,000,000 chance. That's
> essentially zero.

> And it shouldn't matter that you can't bet perfect
> half-kelly. Say your session bank is 20% of your BR,
> distributed amongst a few BPs, and each BP bets
> half-kelly. In the worst-case scenario, towards the
> end of the session, a BP might actually be betting
> 0.625 kelly. The odds of dropping from $25,000 to $50
> when betting 0.625k are still only 1 in 866431. Again,
> essentially zero.

> ??? So my question is how could you ever have anything
> to worry about when you're betting half-kelly, even
> when you're on a team? As long as you have a decent
> game and you're not screwing up the count, you should
> be golden, right?

> (The paper never clarified whether or not the teams
> were resizing; if they weren't, I can understand the

[pm]

> Your chances of being reduced to half your bankroll at
> half Kelly is one eigth. That is too risky for most
> people, in my experience.

> BJ teams resize, but obviously not continuously. A
> team with a 50k bank might round down to the nearest
> green chip at the end of each day or session.

> Bill Chin

I just reread my original question here and I'm amazed at how ridiculous it was.

Thanks for the response.

pm