Are computers capable of remembering..

the order of the cards played, say in a 10 million hand sim? I'm sure they could produce a hard copy of a stream of cards of whatever length, but could memory handle a stream of 25 million cards? My thinking is that if one could play against the same simulated hands using different systems, it might give us better control of/on variance.

If the simulated cards can become a permanent programme against which we can all play whatever system we choose, it might give us a good tool for comparison.

10 M hands would be about 54 million cards assuming 2.7 cards per hand for player or dealer. The laptop I am running on has 2 gigabytes of RAM. Storing a single card only requires 4 bits (1/2 byte) because the suits are irrelevant in 21. I could use 1/2 my RAM to store 2 billion cards. Or I could write the stuff to disk every now and then and use the 80 gig hard drive to store 160 billion cards, or will over 20 billion rounds...

The numbers are enormous...

Naturally, computers are capable of remembering digital data like this much better than people. But computers really don't need the kind of memory you're talking about.

To generate shoes, a computer program uses a Random Number Generator (RNG). The RNG used by PowerSim spits out over 18 quintillion random numbers for a given seed. That's enough random numbers to generate shoes for thousands of years.

All you have to do is write down the seed you input to the random number generator. Or you can store it on the computer. (It shouldn't have any trouble remembering one number.) Then when you rerun your sim with a different count system, or a different basic strategy, or WHATEVER, you use the same seed. In this way you are guaranteed the second system is facing the exact same set of shoes faced by the first system. This is an extreme example of something called "stratifying" a sim.

"My thinking is that if one could play against the same simulated hands using different systems, it might give us better control of/on variance.

If the simulated cards can become a permanent programme against which we can all play whatever system we choose, it might give us a good tool for comparison."

I do it all the time.

ETF

COMPUTER POWER TO THE PEOPLE

If the simulated cards can become a permanent programme against which we can all play whatever system we choose, it might give us a good tool for comparison.

One could develop a system which performs better for a specific sequence of cards, but underperforms in the general case.

by tested against random simulations to see if the improvements are real or not.

... a random sequence of cards where millions of shoes are involved. When you use two different seeds, you have twice as much opportunity to fix results. The first seed could be chosen to make system 1 look good, and the second seed chosen to make system 2 look bad. If you just pick a random seed for sim #1 and re-use the same seed for sim #2, you are practically guaranteed to get more significant results for any prespecified number of rounds.

"One could develop a system which performs better for a specific sequence of cards, but underperforms in the general case."

I'm always amazed about the almost religious belief of many people that simulations can produce some miraculous system where profits will be abundant and fluctuations will be rooted out.
Even with the enormous progress in computer performance, a system should still be designed in an analytical way.Simulations are only there for the testing, especially in view of fluctuations.
Basic strategy has been calcucaled decades ago for any set of rules and modern sims with billions of hands won't change a iota about it.
You probably expect some improvements for index plays but even there your hope is vain. A card counter has usually just one number in his head while playing and he knows approximately how many cards have been played. This allows him to calculate the TC which will be his only help for finding the right decision in close shots.The right decision depends on the frequency of each card rank remaining in the pack and thats precisely the reason why we are counting cards. The TC gives us a pretty good idea of the card distribution and a good player knows the precise point (calculated or memorised beforehand) at which to operate a deviation from BS.
It is pointless to predetermine (or reproduce) the card order of a sim since this is supposed to be completely random.By the way, a sim is just another a bit more complicated way to arrive at proper card rank frequencies and that's why they need to be endless. If you don't get 1/13 for every rank,there is something wrong with the RNG and the sim is worthless.

Francis Salmon

have occurred recently. Doubling 11 against a face was always considered iron clad by most players; now we are told not to double against a face. Not splitting Aces against an Ace is another one.

"Basic strategy has been calculated decades ago for any set of rules and modern sims with billions of hands won't change a iota about it."

Basic strategy for 11 v. 10 hasn't changed; the correct play has always been to double, and any reputable source would make this recommendation.

A,A, v. A hasn't changed either; the correct play has always been to split. The exception to this is ENHC, where it's correct to hit A,A v. A, but this "change" isn't the result of any new analysis, but simply a variation in play in response to a variation in rules.

ENHC, 4 decks, hit soft 17, double any two cards,double after splits except Aces and early surrender except against an Ace.

the argument put forward was usually that because you don't see the dealers hole there is no difference between the two methods. Whether the card is under the up card or still in the deck, either way you don't know what the card is or will be. Dealer looking to see if he has a bj in the hole card method is offset by the dealer taking only the original bet if the hand results in a BJ.

"ENHC, 4 decks, hit soft 17, double any two cards,double after splits except Aces and early surrender except against an Ace."

ENHC has been understood to be different from regular blackjack, since ENHC came into existence. ENHC means if you split or double down, and the dealer has a blackjack, you lose all bets wagered instead of just the original bet.

This is different from regular NHC as played in a few casinos. In that case, there is no hole card, but when the dealer turns over a blackjack (after everyone has played) nobody loses more than their original bet.

In ENHC, there are a few changes to basic strategy. With regular NHC, there are no changes. It may be that this particular BS engine only recently added ENHC to its rule set. Or it may be that you just never noticed it there before. But these things really have been undertood for decades.

Now, there are some interesting things you can do with simulators. And there are some newer wrinkles that come under the rubric of basic strategy (e.g. counter's basic strategy). But I sincerely doubt you're going to make any new discoveries about the basic-basic strategy, via simulations.

ETF

"I'm always amazed about the almost religious belief of many people that simulations can produce some miraculous system where profits will be abundant and fluctuations will be rooted out.
Even with the enormous progress in computer performance, a system should still be designed in an analytical way.Simulations are only there for the testing, especially in view of fluctuations."

I see simulators as time savers. If I have an idea, I have to physically play each hand to get some results. Very expensive and time consuming. I try to be as analytical as I'm capable off.

Another purpose: The programme suggested would be a good example of a controlled experiment.

You can't make improvements if you're not looking for them.

It's not about the seed, but the number of trials. A greater number of trials justify or more closely models the expectation. Choosing 2 different seeds just doubles your trial sample.

Please excuse my absence, I was in an abstract coma.