BlackjackInstitute.com

MJ · Joined: 03 Jul 2004 Posts: 35

MIT Mike, David, Johnny C,

I ran a 1 BILLION round simulation of MIT Hi-Lo as outlined in the Home Training Course. All index numbers used are the CE numbers taken from the HTC, limited to the I-18Fab4. Most importantly, the bet schedule used is strictly in accordance with Bet = (TC-1) * Unit. Furthermore, rounding was used for the TC calculation along with 1/2 deck estimation.

Rules: 6D, DAS, DA2, LS, 75% penetration, play all, 5 players at table. 100 Rounds/Hr.
Bet spread: 1 to 32 (1/2 Unit to 16 Units).

Offset = 1 in light of the rules. Since 266 units = Full kelly as per the HTC, I decided to use a Bankroll of $26,600. Thus, 1 unit = $100 wagered at TC=2, 2 units = $200 wagered at TC = 3, etc.

The following 2 charts represent the SAME simulation, but just different tabs (screens). The first chart is TC from -1 to 12. On the second chart, you can see all TC from 4 through 17. The max bet of $1600 is placed at TC = 17. Notice a TC of 17 virtually NEVER occurs anyway.

Look at the MIT bet schedule (custom bet column) and the optimal bet schedule calculated by the software. They bear no resemblance. This explains the huge difference between the Hourly Earnings and CE between the optimal and custom row. Also take note of N0, this is the # of hands it requires to overcome 1 standard deviation.

MIT Bet Schedule: $77/Hr with CE = $36/Hr
Optimal Bet Schedule: $122/Hr with CE = $61/Hr

Clearly, the bet calculation used by the MIT Blackjack Team generates positive EV, but is far from optimal. As a consequence, it also requires about 15,000 additional hands to reach the long run. Perhaps Mike, David, or Johnny C can point something out that I am missing.

MJ
_________________
"Thou shalt NOT cheat. Thou shalt, however, take FULL advantage of dealer and casino errors. Thou shalt NOT tip dealers" -JP

RJT · Joined: 01 Jan 2006 Posts: 157

I don't have much time as i'm just heading out to work, but the MIT team NEVER used full Kelly. I believe they normally used 0.3 Kelly.

RJT.

mdlbj · Joined: 27 Sep 2006 Posts: 143

.3 kelly and 1/4 deck. Also a + 1 count with 1/2 unit. Norm does not quantify the offset.
_________________
Regards,

mdlbj

MJ · Joined: 03 Jul 2004 Posts: 35

Makes no difference what kelly factor you use. Whether it be 0.25, 0.3, 0.5, or full kelly, the optimal bet schedule always wins more $$$ all else being equal. The good thing is it uses a much lower top bet which will not attract as much heat.

As for 1/4 vs 1/2 deck estimation, it really does not make any difference whatsoever for a shoe game. This has been proven time and time again. Lets not reinvent the wheel gentleman.

Best,
MJ
_________________
"Thou shalt NOT cheat. Thou shalt, however, take FULL advantage of dealer and casino errors. Thou shalt NOT tip dealers" -JP

Mr. J · Joined: 03 Jun 2004 Posts: 28

MJ,

I just noticed a couple of things right off the bat. Your risk of ruin is unacceptable high. Also, if you follow the so called “optimal” betting strategy you will get kicked out of casino after casino, as the simulation doesn’t take things into account such as the act you need to play for long periods of time. Let me explain.

First, the risk of ruin we had on the team was almost nil based on the Kelly we used (which was slightly less than 0.3, as others have pointed out). Also note that the goal of blackjack is not only to generate EV, but to minimize variance AND play for a lot of hours at the table (i.e., have a strategy that will not get you kicked out). You will note that the standard deviation of the "optimal" betting strategy is higher than ours, and that the “optimal” betting strategy gets to its max bet at a relatively low TC. To take it to an extreme, if you bet table max anytime you had an advantage over the house (for your game, TC>1), you would maximize your EV. BUT, you would also have a HUGE standard deviation and you will get kicked out of the casino with your huge bet swings. This is a little of what they “optimal” betting strategy does. While in our betting strategy and in the “optimal” strategy, the ratio of max bet to min bet is 32 in both cases, the “optimal” betting strategy gets to the max bet at a TC of 6 or greater. We don’t get to our max bet until a TC of 17 or greater, which rarely happens as you pointed out. This means that the “optimal” betting strategy will have extreme bet swings fairly often, which is much more likely to get you kicked out of the casino even though it theoretically will make you more EV per hour. But, what is the EV per hour when you are walking home after getting kicked out? Zero, of course!

Our goal was (and still is) to generate EV, but at the same time try to do our best to get in a lot of play and also minimize the large peaks and valleys in your bankroll (i.e., minimize variance).

Remember, your value as a card counter is affected by your skill level and strategy, and how long you can play at the table.

I hope this helps a little bit.

Cheers,
Dave

QFIT · Joined: 27 Feb 2006 Posts: 7

Hi guys. A few points.

1. Yes the risk is unacceptably high in the example sim; but simply changing the Kelly setting lowers that risk to the correct level and does not change the point being made.

2. No one is saying the goal is to maximize EV. CVCX doesn't do that. It maximizes SCORE which balances EV and variance. Optimal betting certainly does NOT maximize EV.

3. By betting optimally, you can lower your betting unit reducing your likelyhood of a barring.

4. Yes it's hard to make money if you keep getting thrown out. But it's also hard if you never get a chance to bet your max bet Very Happy

regards,
norm

Tbonez · Joined: 12 Jul 2006 Posts: 4

I'm sorry but this seems like a typical computer jockey vs real player debate that you often see about playing advantage blackjack. The simulations are great, but there is a human element involved in actual play that can't be dismissed. Thats why in sports they have the saying "thats why we play the game". The so called optimal outcome figured through stats does not figure certain human intangibles. So many things on paper look to be optimal but we and our opponents are human and in such case are susceptible to results not figured by machine. I do not discount the computer simulations by any means, but if Derek Jeter was to explain to me how to hit and play baseball I would listen very intently, just as I would listen the same to Mike and Dave on the best way to play blackjack. (Sorry Dave about the Jeter analogy, I just couldn't think of a Red Sox player worthy of comparing your talents to. lol)

QFIT · Joined: 27 Feb 2006 Posts: 7

The Derek Jeter comment is OK with me. He lives directly across the street from me.

As for the 'typical computer jockey' snipe: CV Blackjack software is fully capable of realistic casino play including preferential shuffling, dealer errors, playing errors, casino heat, extremely complex cover betting, cover play, etc. Simulation is a realistic as you wish. All that aside, you might be surprised at how many pros play according to optimal betting.

Mr. J · Joined: 03 Jun 2004 Posts: 28

Tbonez,

I just want to say that Mike and I really like the simulation software Norm puts out and we encourage people to use it. It is the best on the market, bar none. Simulations give you a lot of great information and allow blackjack players to make informed decisions.

We like it so much, in fact, that Norm and the Blackjack Institute are collaborating on creating a BJI branded version of CVBJ. It should be available shortly, so all our customers should keep their eyes open for it. I guess this is as good a time as any to announce it!

-Dave

QFIT · Joined: 27 Feb 2006 Posts: 7

Thanks Dave. I'll add that TBonez' analogy is of course quite valid. Learning Baseball from Derek Jeter or BJ from Mike and Dave is miles better than reading a book. I was barred from three casinos in my first trip to LV. It was obvious that I couldn't continue this way. So, I spent an afternoon at Revere's house learning camouflage techniques. Didn't get barred again for six months. When you can, go to the source.

MJ · Joined: 03 Jul 2004 Posts: 35

David,

As I stated before, regardless what kelly factor we use the "optimal" bet schedule will produce greater hourly earnings than the schedule used by the MIT Team.

While the standard deviation is certainly greater with the optimal bet schedule, the CE (risk adjusted variance) is ALSO greater. The goal is not to maximize EV/hour, but to maximize CE/Hr..

Finally, the Risk or Ruin is lower and it requires 15,000 fewer hands to get to the long run when betting optimally.

MJ
_________________
"Thou shalt NOT cheat. Thou shalt, however, take FULL advantage of dealer and casino errors. Thou shalt NOT tip dealers" -JP

AndyBloch · Joined: 08 Mar 2007 Posts: 4

You should use the same min and max bet. You can't fairly compare the two betting systems if one has a min bet of 20 and the other 50. Also, you don't specify when you would leave the shoe. The MIT team would usually leave at -1 true (earlier when doing call-ins). If you never leave, having a lower minimum is very important.

Also, we used a .25% "casino factor" (playing errors, casino mistakes, theft, lawsuits, etc.) which meant we would bet half a unit less than what would be "optimal" without the factor. If you don't account for the casino factor, then of course the betting system would do worse than a betting system that does not subtract 1/2 a unti.

Andy
_________________
Visit my home page at andybloch.com
Play poker with me at FullTiltPoker.com

mdlbj · Joined: 27 Sep 2006 Posts: 143

Thanks for taking the time of putting that together MJ. Good stuff.
_________________
Regards,

mdlbj

QFIT · Joined: 27 Feb 2006 Posts: 7

Andy, I disagree that min and max bet must be the same for a valid comparison. What matters is that RoR and bankroll remain the same. For example, if you want to compare backcounting against play all, you wouldn't use the same min and max bet in the comparison. You usually want to keep Spread fixed also; although not in the case of backcounting vs play all.

As far as "casino factor," surely you should take it into account for both or neither. But for playing errors, I think this is best done by reducing EV. Theft, lawsuits, etc should be taken off as a percentage of results. It has nothing to do with betting calculations. Casino mistakes should be in your favor.

regards,
norm

QFIT · Joined: 27 Feb 2006 Posts: 7

I should add, a good example is the 73 pages of comparative sims in Blackjack Attack Chapter 10. RoR is fixed and bankroll is fixed; but the min and max bets change.