Randomly Varying Bets at +TC

[vonQuux]

I have an idea for disguising advantage play based largely on bluewhale's earlier post...

As I understand it, raising bets as the count goes up and lowering bets when the counts go down is a dead-giveaway for a CC/AP. If a TC of one or greater means a net advantage over the house, why not incorporate a random multiplier/divisor into the bet?

Normally, when it comes time to bet, you determine your RC, calculate your TC, subtract 1 and that gives you the bet unit multiplier. If your bet unit is $10 and your multiplier is 3, you'd be $30, etc, etc.

Just for visualization purpose, suppose you had a 7-sided with you but instead of 1-7 on the dice, you'd have -3 through +3 (-3, -2, -1, 0, 1, 2, 3). At any multiplier >=1, you'd roll the dice and adjust your bet accordingly.

This means that sometimes you'd start with unit multiplier of 4, roll the dice, get -2 and wind up placing a $20 bet. Other times you'd start with a unit multiplier of 2 and roll +3 and be betting $50.

How does this effect the game in total?

1) Your betting appears chaotic during the short-term (ie, the time during which you're being observed for AP). Sometimes you're at an advantage and betting flat, sometimes you're at barely a positive count and betting 3x your unit. Over time, however, your average roll will be zero and therefore your bets will be unchanged.

2) Of course this will have the undesired effect of amplifying standard deviation like a b*tch. Yes, you'll wind up with the same +/- over the very long haul but in the short term, it'll be a roller coaster.

Then there is the practical matter of how to perform a dice roll at the table. That's not too hard. During each round of play, remember the first card that hits the table between, say, 1-7 (count ace as a "1") and subtract 4 from that number. 1 would become -3, 7 would become +3. If no A-7 cards come into play on the round proceeding your bet, just use zero and leave your bet unchanged.

My thinking here is to break up the somewhat linear relationship between count and bets.

OK, I put on my asbestos jammies, flame away!

vQ

[EasyRhino]

Originally Posted by vonQuux

As I understand it, raising bets as the count goes up and lowering bets when the counts go down is a dead-giveaway for a CC/AP.

True, but also bear in mind that this is the fundamental tenant of card counting. If you're not raising your bets at high counts and lowering them at low counts, then you're not card counting.

2) Of course this will have the undesired effect of amplifying standard deviation like a b*tch. Yes, you'll wind up with the same +/- over the very long haul but in the short term, it'll be a roller coaster.

Do not underestimate the impact of this. Most of us like to avoid ruining an entire bankroll.

I can't sim this right, but for giggles I fooled around in powersim.
Playing a 6D game, play-all, with a $10-$100 (roughly linear) spread, with a max $100 bet at TC +5, and a $10,000 bankroll, you end up with a win rate of $8 per 100 hands, variance of 997, and a risk of ruin of almost 20%. (I'll admit this "base game" is super-weak, and not recommended anyway)

But, sim the same game, with a $10-100 spread, but some of the high bets are in the "wrong" place at different + counts. (in the sim, these wrong bets ALWAYS done at the same count). Win rate drops to $6 per 100 hands, variance is 1060, and risk of ruin is now 32%.

[vonQuux]

Originally Posted by EasyRhino

True, but also bear in mind that this is the fundamental tenant of card counting. If you're not raising your bets at high counts and lowering them at low counts, then you're not card counting.

I'd still be raising and lowering bets to reflect the count but it would appear chaotic and erratic during an observable span of play, not methodical like a CC'er. I can't do a rote example since this would play out long-term but you'd be overbetting (in proportion to what the TC dictates) sometimes, you'd be underbetting sometimes, but on average you'd be betting the correct amount.

Suppose you're at TC +5 or whatever. Minus one gives you 4x your bet unit. If I'm doing $10 units, that means the count would call for a bet of $40. Sometimes my random card would push me down to $20. Sometimes I'd lose that bet (yay!, I lost less) and other times I'd win it (boo!, didn't win as much). On the other hand, sometimes the random card would push me up to $40 and the reverse logic would occur.

But if my random modulator is truly random, I'm speculating that after thousands of hands, I'd come out to the same amount.

The upside is that I'd decrease heat. The downside is that I increase short-term volatility.

Originally Posted by EasyRhino

But, sim the same game, with a $10-100 spread, but some of the high bets are in the "wrong" place at different + counts. (in the sim, these wrong bets ALWAYS done at the same count). Win rate drops to $6 per 100 hands, variance is 1060, and risk of ruin is now 32%.

To sim this accurately, one would have to insure that the +/-3 adjustment is randomly determined, allocated randomly and with enough frequency that the average deviation from TC asymptotically approaches zero.

vQ

[QFIT]

Originally Posted by vonQuux

Normally, when it comes time to bet, you determine your RC, calculate your TC, subtract 1 and that gives you the bet unit multiplier. If your bet unit is $10 and your multiplier is 3, you'd be $30, etc, etc.

I keep seeing this posted. Unfortunately, for play-all, shoes, the most common circumstance, it is a very poor way to bet. You need to raise your bet at TC +1 and place your max bet at +4 or +5, and that max bet has to be substantially higher than your min bet.

[vonQuux]

Originally Posted by QFIT

I keep seeing this posted. Unfortunately, for play-all, shoes, the most common circumstance, it is a very poor way to bet. You need to raise your bet at TC +1 and place your max bet at +4 or +5, and that max bet has to be substantially higher than your min bet.

Since a little knowledge is a dangerous thing (precisely what I have), I'm guessing you're right. But absent any numbers, any sim data, I can't rule this out. All I have to go on is people saying "won't work."

Frankly, I'd be happy if someone *could* destroy this idea. Then I could stop wondering if it has potential.

Remember, I'm not suggesting other that what you're telling me here. The substantive question, I think, is "if you vary your bets but your average bet for a given TC adheres to the ramp precisely, will the results be similar (or similar enough)?

Hm. I might be able to write a one-off script to sim this...

vQ

[QFIT]

I'm not talking about the cover betting that you are suggesting. I'm talking about the "correct" method that you stated in the first post in the thread. It isn't correct. The optimal bet at a specific TC is:

bet=(KEB)*TCEV/TCSD

Where KEB is the Kelly Equivalent Bank, and TCEV and TCSD are the advantages and standard deviations at each true count. This results in a far faster ramp up than TC-1 and substantially higher bankroll growth.

[golfnut101]

Maybe you should read up on opposition betting. Also, 'progressing' when you approach nuetral counts, can give some cover. Yes, the variance can be extreme, and one would need a solid br to use this technique, but, it seems to have some merit. I used a modified form of this in Omaha this past week(golf trip, NCAA hoops...heaven !)my unit was $10, spreading 1-20, sometimes 2x10. I was not always betting according to count. I know, some are cringing as they read this. I would gladly give up minimal loss of ev for longevity, not that is probably a concern at my betting level. But, I had no idea what to expect having never played their, and, if I am only giving up pennies/hr, it seems worth it. Not like none of the seasoned vets here dont know any of this, but, just some thoughts to share.

best of luck

[sagefr0g]

Originally Posted by QFIT

I'm not talking about the cover betting that you are suggesting. I'm talking about the "correct" method that you stated in the first post in the thread. It isn't correct. The optimal bet at a specific TC is:

bet=(KEB)*TCEV/TCSD

Where KEB is the Kelly Equivalent Bank, and TCEV and TCSD are the advantages and standard deviations at each true count. This results in a far faster ramp up than TC-1 and substantially higher bankroll growth.

maybe this isn't one of the best examples. but look at the w/l % for the TC=1. it's substantially higher than one might expect using the TC-1 offset idea.
an advantage of 0.47 which is higher than the casino has over you at the outset of the game.

[mdlbj]

Originally Posted by QFIT

I'm not talking about the cover betting that you are suggesting. I'm talking about the "correct" method that you stated in the first post in the thread. It isn't correct. The optimal bet at a specific TC is:

bet=(KEB)*TCEV/TCSD

Where KEB is the Kelly Equivalent Bank, and TCEV and TCSD are the advantages and standard deviations at each true count. This results in a far faster ramp up than TC-1 and substantially higher bankroll growth.

Agreed, you should always be betting according to your advantage. But without the table offset Norm, the TCSD will have greater fluctuations?

Trying to justify the TC -1 myself; all I know is that is the correct way according to the MIT methodology.

[sagefr0g]

Originally Posted by mdlbj

Agreed, you should always be betting according to your advantage. But without the table offset Norm, the TCSD will have greater fluctuations?

Trying to justify the TC -1 myself; all I know is that is the correct way according to the MIT methodology.

i'm ashamed that i didn't predict that you or RJT would pipe in here, lol.
but yeah it's interesting isn't it this confluence of two virtual schools of thought on this issue.
the bang for your bucker's say it ain't there at tc=1 and the sim's say wait one let's look at the numbers. thing is as Qfit has pointed out the idea that the advantage climbs in steps linearly as the tc progresses integrally isn't exactly accurate. borrowing again from the 'good ole book' http://www.blackjackincolor.com/truecount2.htm .
then on the other hand who plays perfectly and what would that mean at say tc=1 . but then again hopefully most of us would nail tc=1 fairly accurately. i would suspect RJT and Bojack would. lol . and please note this is not meant as a cheap shot. i know where i stand in the food chain relative to those two.
thing is for a player like me (gambler) lookin for an edge these sort of wiggly lines in the sand offer wiggle room.