Baccarat is beatable statistically
Quote: chrisrAny system that could work would have to tell you something about the deck composition.
it's also my understanding that others have put quite a bit of thought into such systems and have determined them to be a waste of time.
Quote: wz60
Well without looking into any deep, -4.792% for baccarat is laughable, go back to check you algorithm.
laughable? Let's compare and contrast this laughability with the idea that past patterns matter in a random game....
Whoops, sorry. I miscounted ties initially. Here are the updated results, broken down in more detail.Quote: wz60Quote: endermikeMy point is that you can take random data and then find patterns in it. The value in finding such patterns is if they will repeat themselves moving forward. You took simulation 1 and found a pattern: "Bet B on the second hand after P on the first." Now for that pattern to be of value, it needs to be repeatable (statistically significant). Hence, this is why I'm encouraging you to further test your pattern to now demonstrate that it seems to hold in general. Without getting too deep into the philosophy of science, this is what is required for a claim to be accepted. It should be repeatably demonstrable by independent sources. That is why I tried it with the data from simulation 2.
Return = -4.792%, oh well, can't quit the day job yet
Well without looking into any deep, -4.792% for baccarat is laughable, go back to check you algorithm.
Banker wins on 1st hand =11,367
Ties on the first hand = 2,410
PP starts (loss for the bet) = 5,030
PB starts (win for the bet) = 5,151
PT starts (push for the bet) = 1,052
Return = -1.301%
Algorithms are not needed, a simple if statement in excel is enough. See, you found my error (improperly estimating how useless your pattern is), I found yours (this pattern finding nonsense), science works. Thanks, -4.8 seemed high.
Quote: rdw4potusLet's compare and contrast this laughability with the idea that past patterns matter in a random game....
Are you sure you understand what I am talking about? Did "past patterns" ever appear in my posts?
#2 do not tell anyone....do not come to a forum
#3 go to a casino and become rich.
Quote: endermikeWhoops, sorry. I miscounted ties initially. Here are the updated results, broken down in more detail.Quote: wz60Quote: endermikeMy point is that you can take random data and then find patterns in it. The value in finding such patterns is if they will repeat themselves moving forward. You took simulation 1 and found a pattern: "Bet B on the second hand after P on the first." Now for that pattern to be of value, it needs to be repeatable (statistically significant). Hence, this is why I'm encouraging you to further test your pattern to now demonstrate that it seems to hold in general. Without getting too deep into the philosophy of science, this is what is required for a claim to be accepted. It should be repeatably demonstrable by independent sources. That is why I tried it with the data from simulation 2.
Return = -4.792%, oh well, can't quit the day job yet
Well without looking into any deep, -4.792% for baccarat is laughable, go back to check you algorithm.
Banker wins on 1st hand =11,367
Ties on the first hand = 2,410
PP starts (loss for the bet) = 5,030
PB starts (win for the bet) = 5,151
PT starts (push for the bet) = 1,052
Return = -1.301%
Algorithms are not needed, a simple if statement in excel is enough. See, you found my error (improperly estimating how useless your pattern is), I found yours (this pattern finding nonsense), science works.
Good
But we are not still on the same page, here is what I mean by ignoring T:
TPB should be count as PB, TPP as PP, PTB as PB and PTP as PP
Quote: ontariodealer#1 discover how to beat a casino game
#2 do not tell anyone....do not come to a forum
#3 go to a casino and become rich.
But nobody will argue with you at the casino while becoming rich. That in itself is a bigger turn on than becoming a millionaire.....apparently :)
Quote: wz60Quote: endermikeWhoops, sorry. I miscounted ties initially. Here are the updated results, broken down in more detail.Quote: wz60Quote: endermikeMy point is that you can take random data and then find patterns in it. The value in finding such patterns is if they will repeat themselves moving forward. You took simulation 1 and found a pattern: "Bet B on the second hand after P on the first." Now for that pattern to be of value, it needs to be repeatable (statistically significant). Hence, this is why I'm encouraging you to further test your pattern to now demonstrate that it seems to hold in general. Without getting too deep into the philosophy of science, this is what is required for a claim to be accepted. It should be repeatably demonstrable by independent sources. That is why I tried it with the data from simulation 2.
Return = -4.792%, oh well, can't quit the day job yet
Well without looking into any deep, -4.792% for baccarat is laughable, go back to check you algorithm.
Banker wins on 1st hand =11,367
Ties on the first hand = 2,410
PP starts (loss for the bet) = 5,030
PB starts (win for the bet) = 5,151
PT starts (push for the bet) = 1,052
Return = -1.301%
Algorithms are not needed, a simple if statement in excel is enough. See, you found my error (improperly estimating how useless your pattern is), I found yours (this pattern finding nonsense), science works.
Good
But we are not still on the same page, here is what I mean by ignoring T:
TPB should be count as PB, TPP as PP, PTB as PB and PTP as PP
Or you can give us these numbers:
How many TPP
How many TPB
How many PB
How many PP
How many PTB
How many PTP
Fine, but this is the last one you get for free.Quote: wz60Or you can give us these numbers:
How many TPP
How many TPB
How many PB
How many PP
How many PTB
How many PTP
How many TPP = 509
How many TPB = 498
How many PB = 5,141
How many PP = 5,030
How many PTB = 459
How many PTP = 488
Return = -1.475%. Who else is excited to try this bet once every 2 shoes and end up almost exactly where the math predicts it should?
Quote: wz60Are you sure you understand what I am talking about? Did "past patterns" ever appear in my posts?
You're looking at data from shoes that have already happened and determining whether a certain betting strategy would be effective if a similar series of trials occurred in the future. The data already exists, so the hands must have happened in the past. Maybe your strategy fits, maybe it doesn't. The idea that the hands didn't happen in the past is just silly. But, I suppose we could pretend they're all happening in the present. Either way, this exercise has 0 predictive power in informing us about good bets to make in the future.
Quote: rdw4potuslaughable? Let's compare and contrast this laughability with the idea that past patterns matter in a random game....
hey, a lot of smart people thought they did in the 16th century.
Quote: ontariodealer#1 discover how to beat a casino game
#2 do not tell anyone....do not come to a forum
#3 go to a casino and become rich.
This^
It's fine to make a hypothesis from data, but then you must test it with new data. Without doing so, it's merely stating that a set of data had a particular feature NOT that all sets of data from the same source will have the same feature.
Without proposing and testing the hypothesis and the null hypothesis your method is flawed and the results are not useful. Engineering degree or otherwise.
Keep the fills coming because that chip rack is about to get slaughtered.
Quote: Lemieux66I love when there is a new gun in town trying to change the world.
Boot Hill is gettin very crowded. My reconoitering of the graveyard indicates a vast majority are college grads with more chips on their shoulders than on the table.
Quote: michael99000what would happen if gr8player, wz60, and Egalite were all playing at the same bac table at the same time?
And, as always, I wish it for all of you.
Quote: thecesspitEngineering degree or otherwise.
I sure hope that a real engineering degree didn't lead to this very flawed analysis. "World-class". Right. Of course, I am aware that there are some schools awarding "engineering" degrees that aren't valued as particularly "real" by the profession.
I want to share my findings and I want anyone who can do modeling to challenge me.
With ties ignored, B has an edge of 50.68%
DAMN so very very close I am a world renown mathematician with a PHDH degree. My modeling came in at 49.9999999999996
Quote: BuzzardQuote: wz60
I want to share my findings and I want anyone who can do modeling to challenge me.
With ties ignored, B has an edge of 50.68%
DAMN so very very close I am a world renown mathematician with a PHDH degree. My modeling came in at 49.9999999999996
In which case you need to return that PhD.
So, are you ready to admit that your findings have been invalidated?Quote: wz60As a world-class expert in one of engineering modeling fields with >30 years experience I don’t feel a need to argue with you about terminology.
I don’t claim that my findings are absolutely true since my sampling is limited (10000 shoes) and my algorithm may not be 100% correct although I am pretty sure they are correct. My purpose is to get those who CAN do modeling to confirm or invalid my findings.
Quote: wz60Nothing like that in the simulation. Let me re-frame how to get these numbers:
1 Ignoring all T, you just collect all BPB and BPP ---- BPB is a win and BPP is a loss
2 First 2 hands are exceptions, in this case just collect PB and PP for the first 2 hands, PB is win and PP is loss
3 Total win /Total bets would be the winning rate for this selection, 51.28% is the winning mark, 50.65% is the quoted rate for flat B betting, any number larger is an improvement.
You shouldn't collect simulation data and only consider the mean of the results. You have to consider the standard error and compute a confidence interval for your results to have any significance. The coin flip analogy is apt: you are sampling from a small set of data and drawing an erroneous conclusion.
There is simply not enough evidence you have found a winning system. (surprise)
If you're doing half as well as the latter, you will be more than content.Quote: michael99000I just had a thought, what would happen if gr8player, wz60, and Egalite were all playing at the same bac table at the same time?
Keep the fills coming because that chip rack is about to get slaughtered.
Quote: evoqueIf you're doing half as well as the latter, you will be more than content.
Welcome back Johno!
Quote: evoqueIf you're doing half as well as the latter, you will be more than content.
Half as well as three guys who have baccarat beating systems? Pretty sure if I played the game, all 4 of us would have similar results.
Quote: wz60Are you sure you understand what I am talking about?
I think I understand.
Your claims are that:
(a) most of the talk about the probability of the player winning, or the bank winning, assumes a full shoe. or at least a case where the remaining cards in the shoe have the same number of Aces through 9s, and four times as many 0-value cards;
(b) there may be cases where a particular set of consecutive results may be more likely in cases where the remaining cards would have a more favorable result - either where the probability of the player winning > the probability of the bank winning, or the probability of the bank winning multiplied by 0.95 > the probability of the player winning;
(c) since there are something like 10376 permutations of an 8-deck shoe without taking suits into account, and counting all 0-cards the same (e.g. two shoes are not different if their only difference is that the first appearance of the 10 of hearts is switched with the first appearance of the King of diamonds), we can't do an exhaustive analysis, but a simulation / Monte Carlo approach can find patterns where the condition in (b) is met.
Something you may wish to consider: there are 5500 distinct sets of 6 values from 0 to 9 ("distinct" where, for example, {0 0 1 1 2 3} and {1 0 3 0 1 2} are not distinct), and 720 permutations of each of the 6-card sets with those values. Do the player wins "use up" more of a certain rank of card than others? How about the bank wins? Don't forget to take into account that not every one of the deals will use six cards (for example, 0 5 8 2 will only use 4 cards as the player has a natural 8).
1584080 player wins
0.500 Aces per player win
0.498 2s
0.491 3s
0.495 4s
0.493 5s
0.489 6s
0.490 7s
0.487 8s
0.490 9s
0.497 0-value cards
1622464 bank wins
0.486 Aces per bank win
0.480 2s
0.475 3s
0.475 4s
0.492 5s
0.501 6s
0.502 7s
0.500 8s
0.495 9s
0.494 0-value cards
"On average," there will be slightly more high cards used in a bank win, and slightly more low cards used in a player win, but I don't know how significant these differences would be when you're still early into a shoe.
Unless these 250000 shoes data is still not large enough, my simulation of a very simple “pattern” gives a different number:
BPB vs BPP 50.74% Banker wins (2256297/4449616)
Well 50.74 vs 50.68 is still a quite improvement.
Quote: wz60The status-quo about Baccarat is that no matter what you do your odds are unchanged ----- Banker has 50.68% advantage over Player (this number is from ignoring all T’s).
Unless these 250000 shoes data is still not large enough, my simulation of a very simple “pattern” gives a different number:
BPB vs BPP 50.74% Banker wins (2256297/4449616)
Well 50.74 vs 50.68 is still a quite improvement.
Not when you lose 5% commission.
Quote: RSNot when you lose 5% commission.
No but clearly if you can ignore ties, you can ignore commission and you can ignore mathematical certainties.....
This is the way the mind of the system player works :)
Quote: wz60The status-quo about Baccarat is that no matter what you do your odds are unchanged ----- Banker has 50.68% advantage over Player (this number is from ignoring all T’s).
Unless these 250000 shoes data is still not large enough, my simulation of a very simple “pattern” gives a different number:
BPB vs BPP 50.74% Banker wins (2256297/4449616)
Well 50.74 vs 50.68 is still a quite improvement.
So, how many hands per shoe or preferably per hour do you get to take advantage of .04%? And how big do you have to bet to make anything significant from that edge, those hands you do bet? Seems like you're chasing pennies with Benjamins at best, and still -EV, but I'm the first to say I don't know baccarat.
I'm having a hard time thinking this is anything worth pursuing.
Quote: wz60, first post in this thread51.28 so a statistically win can be achieved?
The answer is yes and some selections do provide much better edges. Here are some examples:
1 You should not bet B after B --- continuation is always bad bet in Baccarat statistically, although each one has slightly different statistical expectation, e.g. bet 5B to continue is much better than 6B to continue (50.18 vs 49.28).
2 As you can see from above example, we actually have an edge even bet P after 6B ---- 100-49.28=50.72
3 Lets back to bet B strategy, here is a winning one: bet B every time after first P appears, this will give us a 51.41% edge, it is a winner since it is >51.28.
If you bet B after 2P, you will get a 51.24% edge
If you bet B after 3P, you will get 51.27% ---- very close to even
If you bet B after 8P, you have a big edge at 52.5%
Do you understand why on on 10,000 flips we don't expect EXACTLY 5,000 head and 5,000 tails? (given a 50/50 coin of course)Quote: wz60Unless these 250000 shoes data is still not large enough, my simulation of a very simple “pattern” gives a different number:
BPB vs BPP 50.74% Banker wins (2256297/4449616)
Well 50.74 vs 50.68 is still a quite improvement.
Edit (for the more 8 deck baccarat minded): or exactly 5,068 heads and 4,932 tails on 10,000 flips of a 50.68/49.32 coin
No one (who understands dependent composition) has ever said that. We have only said that the value of dependent composition in baccarat is not worth enough to exploit for practical means. (FYI: .0006 is not practically significant even if it is statistically significant. And further, you don't give standard errors of p values so we can't judge statistical significance either)Quote: wz60The status-quo about Baccarat is that no matter what you do your odds are unchanged ----- Banker has 50.68% advantage over Player (this number is from ignoring all T’s).
Quote: beachbumbabsSo, how many hands per shoe or preferably per hour do you get to take advantage of .04%? And how big do you have to bet to make anything significant from that edge, those hands you do bet? Seems like you're chasing pennies with Benjamins at best, and still -EV, but I'm the first to say I don't know baccarat.
I'm having a hard time thinking this is anything worth pursuing.
Really? How many "rules" a BJ player needs to use to bring his odds down, just looking at the basic strategy table and ask how much improvement that any single rule can bring . I just gave one simple patter that it already down 0.06%, how do you know that there aren't many other different selections that can also bring it down and collective effects will lead to win baccarat?
If you would like to try your system and only bet after a BP sequence over 100,000 made bets or more, I will happily accept your action. In fact, I will even give you a cut in the commission rate.
This is so funny!Quote: wz60The status-quo about Baccarat is that no matter what you do your odds are unchanged ----- Banker has 50.68% advantage over Player (this number is from ignoring all T’s).
Unless these 250000 shoes data is still not large enough, my simulation of a very simple “pattern” gives a different number:
BPB vs BPP 50.74% Banker wins (2256297/4449616)
Well 50.74 vs 50.68 is still a quite improvement.
Stop, it hurts!
2256297/4449616 = what??
it is NOT an improvement.
It is ONLY random variation. Here I will spell it out for you.
V A R I A T I O N
Come on here.
just Test all 416! possible Baccarat shoes and then get back to us with your data.
While that is going on...
the standard deviation of P = .506825
over 4,449,616 trials = SQRT ((p*q)/n) = 0.0237%
Here is the 3 standard deviation range of what you have posted
3sd: 50.7536%
2sd: 50.7299% <<<50.74% a little over 2sd. is your value accurate?
1sd: 50.7062% <<< 50.707679% a little over 1sd. NEITHER one is statistically significant. period.
P: 50.6825%
-1sd: 50.6588%
-2sd: 50.6351%
-3sd: 50.6114%
your next 5 million plays can easily produce a value of 50.6351%
easily
just random variation
wz60 are you an expert Baccarat player?
from what you have posted sounds to me, just my feeling, you just wanna be.
What syndicate do you belong to?
Asians rule in Baccarat. Just a fact of life.
you need to become one to be able to break on through to the other side
(the winning side)
Sally Oh
For all Bank/Player Win patterns of lengths 1-9, the best Player result was 49.401%, and the best Bank result was 50.779%.
For all possible values of (number of Player wins, number of Bank wins) in the current shoe up to that point, it depends on how many times a result has to happen before it is statistically significant. The most likely result in the 20 million shoes where the player won more than 50% of the time (excluding ties) was after 15 player wins and 2 bank wins into the current shoe (the player wins 50.138% of the non-tied hands), but this only happened in 17,037 of the shoes. The most likely result where the bank won more than 55% of the time (again, excluding ties) was after 30 player wins and 6 bank wins into the current shoe (the bank wins 55.952% of the non-tied hands), but this was in only 420 shoes.
UPDATE: After a new round of 30 million shoes, the 15 player / 2 bank result dropped to 49.70513% in 25,774 shoes, and the 30 player / 6 bank result is now in the player's favor (52.703%) in 592 shoes.
I think you found a winning system!
I'm off to the casino, will update.
Quote: ThatDonGuyI did a Monte Carlo on 20 million, count them, 1, 2, 3, is that ThatDonGuy in a RENTAL CAR?, 20 million, 8-deck shoes (with stop card 16 from the bottom), and got these results:
For all Bank/Player Win patterns of lengths 1-9, the best Player result was 49.401%, and the best Bank result was 50.779%.
For all possible values of (number of Player wins, number of Bank wins) in the current shoe up to that point, it depends on how many times a result has to happen before it is statistically significant. The most likely result in the 20 million shoes where the player won more than 50% of the time (excluding ties) was after 15 player wins and 2 bank wins into the current shoe (the player wins 50.138% of the non-tied hands), but this only happened in 17,037 of the shoes. The most likely result where the bank won more than 55% of the time (again, excluding ties) was after 30 player wins and 6 bank wins into the current shoe (the bank wins 55.952% of the non-tied hands), but this was in only 420 shoes.
nicely done.
Quote: ThatDonGuyI did a Monte Carlo on 20 million, count them, 1, 2, 3, is that ThatDonGuy in a RENTAL CAR?, 20 million, 8-deck shoes (with stop card 16 from the bottom), and got these results:
For all Bank/Player Win patterns of lengths 1-9, the best Player result was 49.401%, and the best Bank result was 50.779%.
For all possible values of (number of Player wins, number of Bank wins) in the current shoe up to that point, it depends on how many times a result has to happen before it is statistically significant. The most likely result in the 20 million shoes where the player won more than 50% of the time (excluding ties) was after 15 player wins and 2 bank wins into the current shoe (the player wins 50.138% of the non-tied hands), but this only happened in 17,037 of the shoes. The most likely result where the bank won more than 55% of the time (again, excluding ties) was after 30 player wins and 6 bank wins into the current shoe (the bank wins 55.952% of the non-tied hands), but this was in only 420 shoes.
UPDATE: After a new round of 30 million shoes, the 15 player / 2 bank result dropped to 49.70513% in 25,774 shoes, and the 30 player / 6 bank result is now in the player's favor (52.703%) in 592 shoes.
This is very good stuff. I like open-mined people and those who engage constructive discussions.
"For all Bank/Player Win patterns of lengths 1-9" ---- you mean the streak length, right?
Can you run a simulation for following 2 cases to see if my results on 250000-k shoes are in line with yours:
a) Bet 5P to stop (BPPPPPB vs BPPPPPP) I got B has only 50.62% edge
b) Bet 7P to stop (BPPPPPPPB vs BPPPPPPPP) I got B has 51.10% edge
Quote: wz60"For all Bank/Player Win patterns of lengths 1-9" ---- you mean the streak length, right?
Not just streaks, but all player/bank win permtations - for example, a "pattern of length 8" includes BPBPBPBP, BBPPPBBB, and BPPPPBBB.
Quote: wz60Can you run a simulation for following 2 cases to see if my results on 250000-k shoes are in line with yours:
a) Bet 5P to stop (BPPPPPB vs BPPPPPP) I got B has only 50.62% edge
b) Bet 7P to stop (BPPPPPPPB vs BPPPPPPPP) I got B has 51.10% edge
This time, with 200 million 8-deck shoes:
BPPPPP is followed by a player win (BPPPPPP) 49.301% and by a bank win (BPPPPPB) 50.699% of the time
BPPPPPPP is followed by a player win (BPPPPPPPP) 49.300% of the time and a bank win (BPPPPPPPB) 50.700% of the time
Note that percentages exclude ties
In fact, all of the sequences have player win percentages ranging from 49.279% to 49.333%. This is in line with the 49.318% chance of a player win (disregarding ties) with a full 8-deck shoe.
Quote: wz60Quote: ThatDonGuyI did a Monte Carlo on 20 million, count them, 1, 2, 3, is that ThatDonGuy in a RENTAL CAR?, 20 million, 8-deck shoes (with stop card 16 from the bottom), and got these results:
For all Bank/Player Win patterns of lengths 1-9, the best Player result was 49.401%, and the best Bank result was 50.779%.
For all possible values of (number of Player wins, number of Bank wins) in the current shoe up to that point, it depends on how many times a result has to happen before it is statistically significant. The most likely result in the 20 million shoes where the player won more than 50% of the time (excluding ties) was after 15 player wins and 2 bank wins into the current shoe (the player wins 50.138% of the non-tied hands), but this only happened in 17,037 of the shoes. The most likely result where the bank won more than 55% of the time (again, excluding ties) was after 30 player wins and 6 bank wins into the current shoe (the bank wins 55.952% of the non-tied hands), but this was in only 420 shoes.
UPDATE: After a new round of 30 million shoes, the 15 player / 2 bank result dropped to 49.70513% in 25,774 shoes, and the 30 player / 6 bank result is now in the player's favor (52.703%) in 592 shoes.
This is very good stuff. I like open-mined people and those who engage constructive discussions.
"For all Bank/Player Win patterns of lengths 1-9" ---- you mean the streak length, right?
Can you run a simulation for following 2 cases to see if my results on 250000-k shoes are in line with yours:
a) Bet 5P to stop (BPPPPPB vs BPPPPPP) I got B has only 50.62% edge
b) Bet 7P to stop (BPPPPPPPB vs BPPPPPPPP) I got B has 51.10% edge
You dont mean edge, you mean chance.
Quote: ThatDonGuyNot just streaks, but all player/bank win permtations - for example, a "pattern of length 8" includes BPBPBPBP, BBPPPBBB, and BPPPPBBB.
This time, with 200 million 8-deck shoes:
BPPPPP is followed by a player win (BPPPPPP) 49.301% and by a bank win (BPPPPPB) 50.699% of the time
BPPPPPPP is followed by a player win (BPPPPPPPP) 49.300% of the time and a bank win (BPPPPPPPB) 50.700% of the time
Note that percentages exclude ties
In fact, all of the sequences have player win percentages ranging from 49.279% to 49.333%. This is in line with the 49.318% chance of a player win (disregarding ties) with a full 8-deck shoe.
Can you run the same 250000k data to see what you get? Link is in page 4. This would confirm why we got such big different results.
Here is my code for the k-th shoe:
do 300 kk=7,numHand(k)
KWIN = 0
if(K25Data2(kk-6:kk-6) .ne. "B") goto 300
if(K25Data2(kk-5:kk-5) .ne. "P") goto 300
if(K25Data2(kk-4:kk-4) .ne. "P") goto 300
if(K25Data2(kk-3:kk-3) .ne. "P") goto 300
if(K25Data2(kk-2:kk-2) .ne. "P") goto 300
if(K25Data2(kk-1:kk-1) .ne. "P") goto 300
NBET=NBET +1
if(K25Data2(kk:kk) == "P") then
KWIN = -1
NPP = NPP + KWIN
endif
if(K25Data2(kk:kk) == "B") then
KWIN = 1
NBB=NBB+KWIN
endif
300 continue
Quote: wz60Can you run the same 250000k data to see what you get? Link is in page 4. This would confirm why we got such big different results.
Here is my code for the k-th shoe:
do 300 kk=7,numHand(k)
KWIN = 0
if(K25Data2(kk-6:kk-6) .ne. "B") goto 300
if(K25Data2(kk-5:kk-5) .ne. "P") goto 300
if(K25Data2(kk-4:kk-4) .ne. "P") goto 300
if(K25Data2(kk-3:kk-3) .ne. "P") goto 300
if(K25Data2(kk-2:kk-2) .ne. "P") goto 300
if(K25Data2(kk-1:kk-1) .ne. "P") goto 300
NBET=NBET +1
if(K25Data2(kk:kk) == "P") then
KWIN = -1
NPP = NPP + KWIN
endif
if(K25Data2(kk:kk) == "B") then
KWIN = 1
NBB=NBB+KWIN
endif
300 continue
Have you ever thought you just got noise. I mean it is entirely possible in your dataset you were getting close to 51% or so that however does not mean you had any statistically significant deviation from the norm. Upon further examination it was found you probably did not find any statistically significant deviation.
Quote: wz60Can you run the same 250000k data to see what you get? Link is in page 4. This would confirm why we got such big different results.
Not at the moment, but I did run blocks of 250,000 random shoes, and after "only" 40 blocks, on B5P to stop, the bank percentage ranged from 50.4486 to 50.9663 (which includes the 50.62% you got), while on B7P to stop, the bank percentage ran from 50.3591 to 51.1366 (which includes the 51.10 that you got).
Quote: ThatDonGuyNot at the moment, but I did run blocks of 250,000 random shoes, and after "only" 40 blocks, on B5P to stop, the bank percentage ranged from 50.4486 to 50.9663 (which includes the 50.62% you got), while on B7P to stop, the bank percentage ran from 50.3591 to 51.1366 (which includes the 51.10 that you got).
Can you provide a table (or better a graph) of the results for sampling size of:
10000
20000
50000
100000
.......
250000
500000
1000000
1500000
2000000
This is what we need to see how the mean is converging
Quote: wz60Can you provide a table (or better a graph) of the results for sampling size of:
10000
20000
50000
100000
.......
250000
500000
1000000
1500000
2000000
This is what we need to see how the mean is converging
Why would it matter what the rate of convergence for the mean is? Also that could be calculated directly by just looking at the variance of hands and how many hands you will play per shoe. But again that hardly matters since if the expected value is negative it cannot be used as a better system. I mean there will be times you will be ahead with it for some string of shoes but you are more likely to be down.
Quote: TwirdmanWhy would it matter what the rate of convergence for the mean is? Also that could be calculated directly by just looking at the variance of hands and how many hands you will play per shoe. But again that hardly matters since if the expected value is negative it cannot be used as a better system. I mean there will be times you will be ahead with it for some string of shoes but you are more likely to be down.
You don't need to flip a coin 200 million times to know that your statistical expectation is 0.5, may be 100000 is enough to converge to a practical close-enough value (say 0.49995) and any more sampling will not make a meaningful difference.
Quote: wz60Can you provide a table (or better a graph) of the results for sampling size of:
10000
20000
50000
100000
.......
250000
500000
1000000
1500000
2000000
This is what we need to see how the mean is converging
I am confused as to what you want here.
Are you asking for the percentage of bank wins following BPPPPP (and BPPPPPPP) after 10,000 shoes, 20,000 shoes, 50,000 shoes, and so on? If you are, then the numbers from just one set of 10,000 shoes, for example, are meaningless. If you want the lowest and highest values over all of the blocks of a particular size (e.g. the 20,000 blocks of 10,000 shoes in my 200 million), even that might not have a large enough population to be particularly statistically significant.
Besides, I don't use the same random seed each time, so each run is going to be different.
