Good morning my friend.
Thank you the short resume but I was speaking to @Sassy_Sarah.
Good to know thou 
1 Like
I was watching an interview with Edward Snowden & Neil DeGrasseTyson, and 1 part touched on a RNG. Snowden said basically , while there are many ways/algorythms to get the numbersā¦ it ALL hinges on that 1st āseed numberā, and askād if there was a cosmological number that could be used as this seed number. He was told not reallyā¦
For this exact reason, I said that instead of the way Replay āshufflesā ( from a clean deck each time )ā¦ that they should use a way to have a dead pile, influenced by table play and burn cards ect ect, then addād back to the rest of the āshuffled cardsāā¦ then use that as the deck they pull from, to build the shuffled deckā¦ ( not ever a clean deck after 1st hand @ table )
What it does is add in randomness, each hand ā¦all thru playersās playā¦ ie- who folds/when and who goes till showdown. 
2 Likes
Hello spivak,
I know Iām replying to your post a bit late but I hope youāre still checking and you can reply.
I have read your initial post and itās very interesting that you analysed the whole system and came up with these conclusions.
I wonāt argue with any of your initial findings as they are logic and could be very true. But I only have one more situation to ask you about. It occurs a lot and not just because āit feels that wayā. At the same table (say 6 or 9 players), and in the same hand, the board would have a flush potential, and at least 3 players would have a flush. If thereās 1 (or 2) pair on the board, at least 4 players would have a full house (one would even have quads).
What Iām saying is, itās one thing to have too many strong hands, but a totally different thing to have so many players have the stronger cards in the same hand. Where thereās a flush, 2 players would have good flushes, and a 3rd would have the ace high. Where there are straights, 3 players would have the lower straight, and 1 would have the higher one. Where there are full houses, 3 players would have lower full houses, a 4th player would have a higher full house, and a 5th player would beat them all with quads.
I have played on this site long enough to notice that these incidents happen very frequently and Iām pretty sure they donāt happen as much in real life, or even on other poker sites Iāve used in the past.
Could you maybe tell me what you think about that?
Thank you
Maya
1 Like
Maya,
I have not seen that many , strong hands in the same hand, as you are describing. Yes it happens, but I see more where you have 1(ppl)-trips, 2(ppl)-flush, 1(ppl)-boatā¦
Let me play Devilās Advocate here :
Since there will always be some patterns in picking random numbersā¦ picking from a clean deck, and only doing this once = there will then be patterns transferred into the hand results. Which is why, even if a more exotic( more code ) solution is not usedā¦ the simple fact of doing this 3-6 times (randomly) might just be enough provide better ārandomnessā.
Let me show how this is doneā¦
A small section of code randomly picks cards from deck 1 and fills deck 2 till all cards have been picked. Deck 2 becomes the deck used to deal from/with. Before this section of code is excecuted, the deck ( deck 1 ) is reset to a clean deck.
NOT using a more sophisticated shuffleān methodā¦ All Iām saying is that you do this 3-6 times not once, and each time you simple replace deck 1 with deck 2 rather than a clean deck ( after the 1st shuffle ) ā¦ So basically instead of ā¦ grab new deck, shuffle once ā¦ its grab new deck, shuffle 3-6 timesā¦
This is such a small change, that If I was on staff as a programmer, I should be able to make this change in just a few lines of code, and it shouldnāt take me longer than 10-15 minutesā¦
Yes, The more complete solution is to make the previous hand relevantā¦ The ādead pileā of cards from any hand, and the rest of the cards once the hand is overā¦ SHOULD be where Deck 1 comes from going into the next hand, not a clean deckā¦ Then you still āshuffleā that deck 3-6 times, then deal the handā¦ Yes that involves more code to build that deck as the previous hand is played, But the above solution doesnāt.
If Replay continues to start with a āclean deckā every hand, and only āshufflesā once, then there WILL be patterns that will show thru, when sample size of hands reaches critical mass. ( if we look @ enuff hands )ā¦ and it all stems from them using a āclean deckā to start with each handā¦
Would anyone here, go to a bar/casino/homegame ā¦ where EVERY hand, they crack a new deck, and only shuffle it once ? my guess is 99.5% of us would NOT play in that environment.
1 Like
Based on my experience on another site, the biggest problem with any computer RNG program is simple human nature. When youāre playing a live game, you can see the dealer shuffle the cards and you know there is every effort to have an honest and random distribution of the cards. Just check with your favorite dealer on a slow day and he will tell you the standard rifles and rips that makes a shuffle. People cannot see that happening on a computer and that leads to an innate sense of distrust.
An RNG has to live up to the mathematics of the game. Not an exact number for each player but a feel of reasonableness. Let me give an example. Math will tell you any specific pocket pair, be it pocket aces or pocket twos, has a probability of hitting once every 221 deals. On another site I traced my top five pockets over more than 100K hands and my results surprised me. Three were within a 5% deviation from the anticipated results, one was high and one low, but not by a major percentage.
Let me use a number near and dear to every poker player. Pocket aces have an 85% probability of winning pre flop. Now there are a lot of factors that will affect a personās personal statistics but the mathematics act as a guideline or benchmark. My numbers over that review showed an 86.8% winning percentage. That is a variance of 1.9% from the expected, which is a reasonable difference.
On the other site I believed the operators cheated players by having good hands loose. Funny thing happened, I proved the opposite to my personal satisfaction. This site doesnāt provide the tools necessary to perform such a review. Although the logical side of me realizes the site has no reason to prefer one player over another, the emotional side will doubt since I can neither confirm nor deny the randomness of the RNG to my personal satisfaction.
1 Like
Hello Sarah,
Thanks for your reply.
Do you think the same rules apply to the Omaha games? The incidents I was describing could occur much more in Omaha games than in Holdāem (where they happen as well but less frequently).I know in Omaha each player is dealt 4 cards which makes the odds higher, but weāre still talking about 6 players (24 cards dealt) or 9 players at most, and repeated incidents of many players having what should be a rare winning hand, in the same hand.
Fresh example: Iām playing an Omaha tourney right now, and 2 hands ago 3 of us had flushes and got beaten by a 4th who had a full house, followed by another hand where 3 of us had flushes as well, and one won with the higher flush, and the current hand where 2 players had full houses and one took the other out.Thatās 3 consecutive hands in this tourney where such events occur, and thatās one of many sequences in the same game.
I would love to hear your opinion on this. And thanks again for replying.
Maya
1 Like
In Omaha, yes the simple fact that everyone gets 4 cards will increase alot the drawing power of any hand. While I do not have a good reason why the river seems so cruel around here, I have had tables where when ppl are not playing stupid poker, you see far less bad beats/river beatsā¦
While I donāt think Replay is riggād, I just think the current implementation may lend itself to certain ātrendsāā¦ yes Omaha is worse, and Royal might as well just be bingo, nothing there seems to ever hold up.
(edit) plus good Omaha players know what cards to play, and HiLo is even more cruel it seems.
2 Likes
Hi CairnDestop
I know what you mean. But just to make one thing clear, I donāt think the site favors any player over another. While there might be some sort of programming meant to produce more great hands than regular ones, in order to keep the game fun and attractive to online players, Iām sure it would be completely random and any player can win any hand at any point.
1 Like
I hope you are not an official for the company since this is the very definition of a rigged game. It does not matter that it is a one-time, one hand deal and that the winner is random. It removes all the skill in poker by having the RNG decide who will win. It does tend to explain why hands seem to have multiple sets and a high number of quads over the course of a tournament.
1 Like
Wow, lots of replies here all of a sudden. I donāt have time to read them all. I am at work again, so my time for this kind of stuff is much less than it was a month ago. The first post, however, asks a very good question.
Getting data to answer this question is pretty tough. The issue is that I donāt see anyoneās cards except mine, so the only way I ever find out if there is more than one big hand is if all of the big hands are shown.
I agree that seeing more than one big hand at once is pretty common on RPP, but I think this has to do with the number of hands that make it to showdown. Lots of people here just never fold, and they will play their 5-9 off suit. Then it turns into a double inside straight draw because a single 7 comes on the flop, and they just call anyways. When the turn and the river complete the straight, they suddenly have a great hand. This is terrible play, and in a physical poker game you will never know they had the 5-9 to begin with because they will fold that hand before you get a chance to see it.
Now, I could very easily be wrong here. This particular kind of meddling would not be too hard to set up, and is a lot harder to detect.
Setting up a good statistical test for this is quite difficult because unless every single player at the table shows their cards at showdown, I will never find out how many flushes there were. Comparing how many flushes make it to showdown sounds like it might work, but this depends heavily on player behavior, and it is pretty obvious that player behavior on RPP is very different from what you typically see at a physical game.
I will give this some thought in my spare time and maybe I can come up with some way to figure this out. I guess maybe I can take notes on my own winning ratios, for instance with the question: āWhen I have trips and there are two other people in the pot, how often do I win?ā. In this sort of situation though, I would need some sample from physical play to compare it to, which means that I have to play a lot of online poker and also a lot of physical poker and take notes on both. The last time I played any physical poker was probably at least 6 or 7 years ago, but I might be up for trying this out for science.
3 Likes
OK, I thought of the following just now. This is not a formal test, but it should be a decent sanity check. I will describe it and if you approve I will do it:
It is not too hard to calculate the mean number of flushes accross n hands when there are three or four to a flush on board. This can be done for all values of n from 2 to 9 (these are 18 different calculations, but the 4 to a flush cases are easy. The 3 to a flush cases are much harder, but I can still do them).
We do not know how many flushes make it to showdown, but we do know that the number of flushes that make it to showdown is not greater than the total number of flushes. Hence, if the average number of flushes we see at showdown is greater than my calculations of expected values we can conclude that the system is likely to be rigged. Even if it is equal or close, this is a sign that the system is not trustworthy. If the number of flushes at showdown is, however, less than the numbers I calculated, the RNG can be said to have passed this test.
Hence, I can perform these calculations and then observe (not play, since complete knowledge of any single players cards does not add anything to this test) about a thousand hands, take the average number of flushes in each situation, and compare to my calculations.
This is NOT a particularly good test, but I think it would satisfy my own curiosity; would it satisfy yours?
2 Likes
You are not contradicting me, weāre both saying the same thing. The only difference is what I already said, I donāt believe it favors a certain player over another, but it still randomly gives unrealistic hand which raises questions when it occurs so many times.
1 Like
Hi spivak
Thanks for your replies.
Being the best option we have at testing the theory, yes I believe it would give an idea if youāre able to do it.
Thanks again
Maya
1 Like
Spivak,
When we all talk proving odds, we also talk about sample sizeā¦ as in # of hands sampled. Noone in thier right mind would take 10 hands, then gleam stuff from thatā¦ it takes more like 10,000 hands to see the patterns I believe mightā¦ yes I said might exsist.
Now, the same exsists in life ā¦ the path we walk is filled with optionsā¦ everytime we make a choice, our life takes a distinct pathā¦ the more times we āchooseā the more ādegrees of separationā from the starting point there is. But if we continue to start each hour, from the same placeā¦ then the options from there are static.
Because of both of these reasons/examplesā¦ what Iām saying is by starting from a āclean deckā every hand, and then only shuffleān onceā¦ there is a finite amount of randomness that can occurā¦ when you start each hand with a different dirty deck, then there are far more possibilities for the randomness of the deckā¦ Just like you wonāt just sample 10 hands, you also shouldnāt shuffle once, especially when its a clean deck.
ā Replay Poker Is Not Riggād ā not with the available information I have, and I certainly am not a part of the staffā¦ but really how hard is it to add 1 freakān āfor-next loopā around the shuffle subroutine ?
1 Like
Calculating sample sizes etc. is my bread and butter. All of the testing I have done so far is pretty informal, but the sample size is indeed a common variable in all statistical tests.
My opinion about multiple shuffles has been posted a few times. The same goes for the patterns which you say might exist.
1 Like
Best I could think of off the top of my head, hehe. Anyways, I will give this a shot. You will have to give me a couple of weeks to do this. I need the spare time to do the calculations and then I need to collect data, which will also take a while.
1 Like
As you said, we are talking sample size. There is no way I as an individual can verify the validity of the RNG for the site as my sample is too small. Even the 142,368 hands I recorded on the other site was less hands than played there in five minutes. You are also right that a minimum sample would have to be no less than 10K hands.
Statistics are for the individual player and this site doesnāt allow such a review as the necessary data is withheld from the player.
This means one must use the mathematics of poker. These numbers will not change when the game is honest. You need not trust me. Instead, investigate the many books and web sites devoted to the math of the game. When your statistics hold close to the math of poker, you can rely on the honesty of the RNG.
When I did my study on the other site, I did not compile statistics on all 169 holdāem starting hands. My focus centered on the group one and two hands, a total of ten starting combinations. These are the top hands in holdāem. Let me list the ones just for the pocket aces.
wins = 545
loss = 082
fold = 001
win percentage = 86.8%
frequency = once ever 227 hands
Now compare that to the math. Pocket aces have an 85% win rate expectation. This means if you get pocket aces and hold through to the showdown, you should win 85 of every hundred hands. Statistics realized an event hitting perfect stats is just as crooked as a rigged deck, which is why there is a dead band to that numerical norm. 86.8 has a 2.1% variance. That difference can be explained by playing style and I can feel confident that the RNG works as intended.
Math tells me any pocket pair should hit once every 221 hands. My frequency is 6 higher, which is under a 3% variance. Again, close enough to be reasonable. Some players should have higher than the math and some should be below, but the larger the sample, the closer to the math the results should reflect.
Again, as Iāve noted, the site omits the vital information that would allow people to check out their results. If they did, the questions regarding the RNGās validity could be put to rest as people would than have the necessary tools to prove to themselves its accuracy.
1 Like
Cairn & Spivak,
I am not trying to re-argue something, nor am I talking about something thats un-proveableā¦
Many times I talk in analagies, but I have now tried to ask basically a damn yes/no question, 5 or 6 times, over multiple topicsā¦ and I still donāt get yes/noā¦ I get a response to a question I never askād, and no response to the question I did ask.
Just like you wonāt accept 10 hands as an acceptable data set, you also wonāt accept 1 shuffleā¦ that was the tact I took this time, and even that didnāt workā¦
I have a crisp 100$ bill here, and Iām guessing not 1 person will answer the following question, truthfully, with any answer other than NO.
Will you, live ( bar, casino, homegame ), allow every handā¦ for the dealer to un-seal a new deckā¦ Not do a washā¦ shuffle them onceā¦ no cut ā¦ then dealā¦ and accept the results.
People specifically ask for a wash, when they seem to be getting the same crap all the timeā¦ When they do so, the dealer does the wash and then usually shuffles the cards 4-5 times, but @least 2-3 timesā¦
In this whole discussion, Shakeraise posted exactly how Replay ādoes a shuffleā. That is not in dispute hereā¦ it didnāt take going over 10k hands to understand thatā¦ but noone in thier right mind, liveā¦ will put up with using a new deck every hand, and only with 1 shuffleā¦ hell we wouldnāt put up with just 1 wash & no shuffle either ( from a clean deck )ā¦
What I was trying to also describe, was the whole ādegrees of separationā from a certain spot in time. Showing that if you constantly start from the same beginning, that will limit the # of possible outcomes.
1 Like
I will answer your question directly: It depends on what kind of shuffle the dealer does. If the dealer does a single riffle shuffle then the answer is no. This is because of the fact that riffle shuffles must be performed several times to approximate uniform randomness.
HOWEVER, I would indeed be willing to accept the results if the dealer shuffles as follows: He takes a 52 sided die, rolls it once per card, and assigns that card to the position indicated by that die roll (rerolling if that position is already taken). This form of shuffling does not need to be performed several times in order to approximate uniform randomness (in fact, this is uniformly random automatically after a single iteration).
I would in fact prefer a single iteration of the second shuffle over several hundred riffle shuffles. The uniform distribution is the ergodic distribution of the first stochastic process, but it is not reached in finite time. In the second process the distribution is reached in one iteration exactly.
Spivak,
@least I got the āNoā somewhere in there, but yet againā¦ I askād a specific question, and you changed the question. I have no problem with additional info after your repsonseā¦ but I did not give you 2 choices of types of shuffles, therefore I did not want 2 different answers depending on criteria I did not give you.
Next off , how many times do they start with a new deckā¦ 1 when they start the tables, and maybee 1 more sometime during the nightā¦ and how the dead pile is created, 100% depends on 1 thingā¦ who folds, who doesnāt, and in what order that happens. that is the ādeckā that should be shuffled, not a new deck every handā¦ Its simple, because human input now is adding to the randomness or un-randomness, and it does directly effect the next hand.
If I owned Replay Poker, and I heard so many complaints about āriggādā as we do around here, I would be trying to actively solve this, by going outta my way to show thereās no way it can be riggād. And if that means I go to my head programmer, and tell them to add 3 lines of codeā¦ either he/she can do that in 15 minutes, or they get fired !
Nowā¦ Shakeraise has posted what a āshuffleā is to Replayā¦ with this many complaints over such a long period of timeā¦ when someone gives Replay a āgood/better/bestā solution, and the āgoodā one takes 10-15 minutes topps, I certainly would try it ( if for no other reason than my own curiosity )ā¦ But, if my income depended on revenue from Replay, your damn right I will go to any lengths to increase trust in the site ( including the ābestā solution )ā¦ so hopefully that translates into more revenueā¦ and more $$$ in my pocketā¦
Almost all of the propriatary nature of this site, is its appearance and the organization of the site. Iām sure that the 1st person that created the code for a ābubble-sortā is NOT getting royalties from it, its open sourceā¦ the same goes for shuffling a deck of cards, so if it was meā¦ Iād be " " close to just giving out the exact source code of the shuffle, just to show its not riggād.
My gift to Replayās playerbase :
999 REM ** SHUFFLE ROUTINE **
1000 FOR X=1 TO 52
1010 D1$(X,1) = D0$(X,1) : D1$(X,2) = D0$(X,2)
1020 NEXT X
1030 FOR X=1 TO 52
1040 Y=INT(RND(1)*52)+1
1050 IF D1$(Y,1)=" " THEN GOTO 1040
1060 D2$(X,1) = D1$(Y,1) : D2$(X,2) = D1$(Y,2)
1070 D1$(Y,1) = " " : D1$(Y,2) = " "
1080 NEXT X
1090 RETURN
picture 3 decksā¦ Deck 0 , Deck 1, and Deck 2
Deck 0 is a clean deck āA23456789TJQKA23456789TJQKA23456789TJQKA23456789TJQKā, and somewhere else in the program D0 was already defined for us as a clean(new) deck.
Deck 1 represents the ātemporary working deckā before its shuffled.
Deck 2 represents the shuffled deck.
Also the array used is 52x2 ā¦ where 1 is the card #, and 2 is the suit #, for all 52 cards.
Lines 1000,1010,1020 just refresh Deck 1 to a clean ( new ) deck.
Lines 1000 and 1030 represent a loop, where everything inside is done 52 times
Line 1040 picks a random whole number between 1-52
Line 1050 checks to see if we already picked that number, if soā¦ go pick another number.
Line 1060 puts the random card into the Deck we will deal from.
Line 1070 blanks out the picked card from the āworking deckā so cannot be picked again.
Line 1030 & 1080 represents a loop, again done 52 times, because we need to pick 52 cards @ random.
Line 1090 returns the flow of the program to wherever this subroutine was called from " like GOSUB 1000 "
Deck 2 is now āshuffledā and ready to deal. ( if you only wanted 1 shuffle )
Should you want more than 1 shuffle, let me give you the new lines of code to add.
1023 A=INT(RND(1)*4)+3
1027 FOR Z=1 TO A
1084 FOR M=1 TO 52 : D1$(M,1) = D2$(M,1) : D1$(M,2) = D2$(M,2) : NEXT M
1088 NEXT Z
Line 1023 determines how many shuffles ( picks a number 3, 4, 5, or 6 )
Line 1027 & 1088 is just the loop to do the shuffle multiple times
Line 1084 makes Deck 1 identical to the shuffled Deck 2, ready to do the next shuffle
Ok, so I said 3 lines, and it took me 4 ā¦ so sue me ā¦
But as described above, thoseb 4 lines represent the āgoodā solutionā¦ not the ābetterā or ābestā solution.
I wrote this in 1 of the easiest languages to understand āBASICā which is very simmilar to āVisualBasicā, a newer language. If any programmer out there can say I am wrong, bring it.
1 Like