End profit is not the answer!

[Sonny]

Many people choose to test their betting system through actual play instead of using computer simulations. They feel that actual results will give them a better idea of how well their system really works. They think that as long as their system continues to win money that is must be effective. Unfortunately, this can not only lead to wildly unreliable results but it can also be completely inaccurate if the numbers are misunderstood.

For example, let’s say our friend Smitty has a new system he thinks might work. He decides to play against a coin flip and see if his system can get an advantage. He bets $1 per flip and after 10 flips he manages to win $4. He is very happy, but he also knows that it could just be luck so he decides to play 10,000 flips before making up his mind.

After 100 flips he is ahead $10. Things are looking good, but he isn’t celebrating just yet. After 1,000 flips he is ahead by $30 and his system still seems to be working. After the final 10,000th flip he finds himself ahead by $100. His system has been consistently winning money this whole time. He is now completely convinced that his system is a winner. He goes on all the blackjack websites and posts that he has an amazing system that gives you a small advantage over the house and is much easier than card counting. Is he right?

Unfortunately, he wasn’t looking hard enough. Even though his profits were rising his win rate was plummeting. After 10 flips he had a 4/10=40% win rate. After 100 flips his win rate was down to 10%. At 1,000 hands it had dropped to 3%, and by 10,000 flips he was all the way down to 1%! If he had been looking at his win rate instead of his profit then he would have seen his advantage slowly diminishing the more he played. That should have been his first clue that the system was slowly failing, not winning. Also, the fact that his win rate was still fluctuating so wildly, even after 10,000 flips, should have indicated that he was nowhere near an accurate result.

Unfortunately, this is a trap that many players fall into. That is why profit alone cannot be used to determine advantage. It is also why so many people prefer computer simulation over "kitchen table" tests. The results are much more reliable and it can often take less than an hours to see the results of billions of hands. A computer simulation can also give you valuable information like bankroll requirements and profit probabilities for any number of hours of play. With the speed and power of today's software there really is no reason to test a system the old fashioned way.

-Sonny-

[ChefJJ]

Originally Posted by Sonny

Unfortunately, this is a trap that many players fall into. That is why profit alone cannot be used to determine advantage. It is also why so many people prefer computer simulation over "kitchen table" tests. The results are much more reliable and it can often take less than an hours to see the results of billions of hands. A computer simulation can also give you valuable information like bankroll requirements and profit probabilities for any number of hours of play. With the speed and power of today's software there really is no reason to test a system the old fashioned way.

-Sonny-

With many of these computer sims, is there a way to take a statistical "slice" of those billion hands? Those slices could represent, say, a year's worth of playing that system...that would give that individual an idea of how the "intermediate run: may behave.

Just curious.

[Sonny]

The standard deviation will tell you what range your results will fall into for any given length of time. It can also be used to find the probability of being ahead (or behind) by $X after any given amount of time. That should give you a pretty good idea of what your “slice” will taste like.

-Sonny-

[ChefJJ]

All of this "slice" talk is making me hungry for some pizza! Thanks for the info Sonny.

[fwb]

"But I'll never play billions of hands..."

[Sonny]

Originally Posted by fwb

"But I'll never play billions of hands..."

But doesn't it make sense that the system with the biggest advantage and smallest variance will give you the biggest chance of being ahead in the short run? The less you rely on luck the more likely you are to win, the more money you expect to win, and the sooner you will expect to get that money. Unless you have very accurate results you won’t really know which system is better than the other. That is why computers run billions of hands even though humans never do. If you don’t expect accurate results, why waste your time testing a system?

Obviously professional players don’t play a billion hands every year. They may only need to play a few hundred hours to ensure a profit and less than 1,000 hours to have a good chance of reaching the long run. That’s a measly 20 hours per week. That’s why so many people use blackjack as a form of supplemental income. You don't need to play billions of hands.

-Sonny-

[Marsha]

You are reading my mind! No wonder this forum is called voodoo. Now i need to come up with a whole new set of questions.

[Canceler]

Originally Posted by Marsha

Now i need to come up with a whole new set of questions.

Most of those new questions have already been answered in the Card Counting forum. That's where you need to start hanging out, instead of here in Voodoo!

[Preston]

My win rate is a really good 83%.

i have 63 wins for the year 5 break evens and 8 losses. Trouble is a couple of the losses were pretty significant and some of the wins were pretty small.

I'm still up about $6000. (since June 1) took a $1500 loss tonight. My unit is $10 -> I had four hands in a row with $100 bets out that were unsuccessful double downs... which is what most of the hit came from.

I don't know the exact amount I am up since my old laptop died and that's where I was tracking everything. I was up around $1500 last I remember. So $7500 for the year so far -> Not bad for a "part time" job.

[SilentBob420BMFJ]

Originally Posted by Sonny

Many people choose to test their betting system through actual play instead of using computer simulations. They feel that actual results will give them a better idea of how well their system really works. They think that as long as their system continues to win money that is must be effective. Unfortunately, this can not only lead to wildly unreliable results but it can also be completely inaccurate if the numbers are misunderstood.

For example, let’s say our friend Smitty has a new system he thinks might work. He decides to play against a coin flip and see if his system can get an advantage. He bets $1 per flip and after 10 flips he manages to win $7. He is very happy, but he also knows that it could just be luck so he decides to play 10,000 flips before making up his mind.

After 100 flips he is ahead $65. Things are looking good, but he isn’t celebrating just yet. After 1,000 flips he is ahead by $550 and his system still seems to be working. After the final 10,000th flip he finds himself ahead by $5,100. His system has been consistently winning money this whole time. He is now completely convinced that his system is a winner. He goes on all the blackjack websites and posts that he has an amazing system that gives you a small advantage over the house and is much easier than card counting. Is he right?

Unfortunately, he wasn’t looking hard enough. Even though his profits were rising his win rate was plummeting. After 10 flips he had a 7/10=70% win rate. After 100 flips his win rate was down to 65%. At 1,000 hands it had dropped to 55%, and by 10,000 flips he was all the way down to 51% - only 1% above his expectation! If he had been looking at his win rate instead of his profit then he would have seen his advantage slowly diminishing the more he played. That should have been his first clue that the system was slowly failing, not winning. Also, the fact that his win rate was still fluctuating so wildly, even after 10,000 flips, should have indicated that he was nowhere near an accurate result.

Unfortunately, this is a trap that many players fall into. That is why profit alone cannot be used to determine advantage. It is also why so many people prefer computer simulation over "kitchen table" tests. The results are much more reliable and it can often take less than an hours to see the results of billions of hands. A computer simulation can also give you valuable information like bankroll requirements and profit probabilities for any number of hours of play. With the speed and power of today's software there really is no reason to test a system the old fashioned way.

-Sonny-

its amazing how often you hear counters and smart people on here say "ya, i tried the system out, it works pretty good" or you will hear somebody say "have you tried that system? does it work"?.. sometimes people will agree such and such is ploppy logic, but then they will be using that same logic later on.. its like they are fast to admit that AP logic is correct, except they dont use it or really think about it

Originally Posted by Preston

My win rate is a really good 83%.

i have 63 wins for the year 5 break evens and 8 losses. Trouble is a couple of the losses were pretty significant and some of the wins were pretty small.

I'm still up about $6000. (since June 1) took a $1500 loss tonight. My unit is $10 -> I had four hands in a row with $100 bets out that were unsuccessful double downs... which is what most of the hit came from.

I don't know the exact amount I am up since my old laptop died and that's where I was tracking everything. I was up around $1500 last I remember. So $7500 for the year so far -> Not bad for a "part time" job.

83% win rate? you mean session win rate? also, i sure as hell hope your counting and spreading nicely, because if not, im jealous! (of results, not approach, if u are indeed a bs player only)