turtleknife_poker_blog

Low limits at Full Tilt may get a call no matter how weak they are...

Should you call slow or fast? Consistent or mix it up?

I am more hesitant to call with time taken. To me it looks like someone strong pretending to be weak. If you are weak why would you raise?

A quick re-raise to me (again) seems to imply I want to throw money out here, and see if you will fold. So, I guess my answer as answer would be I would be more likely to call a faster bet than a slower 1.

As a general question, this is my general answer, not taking into account any reads I have developed on a given player. However at the $1.00 level at Full Tilt Poker, you could probably re-raise quick or slow, and type in the chat box "I have the nuts" and still get a call. Nothing against smaller stakes games, it just goes back to my theory that bad players lose their money calling, not betting.

There are many more reasons to expect poor play than your opponent simply not understanding. Reasons like, not caring, being too tired, drinking and playing, teenage sillyness..  etc. Its not simply a question of poker odds calculations or strategy. If it were you might be able to figure them out, but I wouldn't bother with it at this level. You never know why your opponent may be making horrendous plays, but the more they make the better. Even if some of those plays work out for them!!

Full Tilt is Tough.

CrazyThing has something to say in the FTForums:

I have to agree that FT does not use a random number generator. The losses I've had for a while now cannot be coincidence. If you know anything about poker at all, you know that mathematically you can’t loose over and over with the best hand, especially not to ridiculous beats. Just watch guys like Gus Hansen and Phil Ivey duke it out nightly at the omaha hi-lo tables, but I do not hear them crying bad beats....ever!!!

Don’t get me wrong, I know bad beats (or just simple bad luck) happens, more often than any of us would like. But I play in live tourneys all the time, and I am very good at matching odds… when this site says “tag, you’re it”, you might as well sit back and forget about it because the odds are no longer fair and even. Maybe I believe a little too much in statistics, but nevertheless, they’ve always proven correct for me so I’ll continue.

You have likely got to know when to quit, when to NOT shove it all in, and when you are really ahead or just marginally ahead. Most of the bad beats I see, really aren't that bad.

And statistically, if a player plays every day, plays very well and time after time has the best hand – that same player doesn’t get beat by basically nothing every time they play for weeks at a time. Not randomly anyway! Don’t get me wrong – I’m not speaking as a ‘sore loser’, I play for money that matters against people I can see, not against a computer; but it is still very disheartening to play on here for enjoyment and practice only to have the computer decide it’s not your “turn” to win this month (or however long they keep you “tagged”).

Nothing personal against Full TilT however, I’ve been playing for years, and all of the sites are doing it more and more. But for those of you that may be fairly new at playing online – it wasn’t always like this… that is what is so sad about it!! And it doesn’t bother me if you copy my post into yours so you can call me an idiot or make fun of what I’ve said, I don’t come in here and read these things enough to know it if you do. After all, it is my belief that anyone who actually believes FT, or any online gaming site, is here for you and your enjoyment and not for your money doesn’t have anything to say that I would want to hear.

The Full Tilt RNG

Posted by Mathemagical in the poker forum:

I've been reading through some old threads on the subject of skewing in Full Tilt Poker, and I saw a disturbing mix of solid information and misinformation about random number generation and statistics, so I thought I'd take a moment to address some of the issues raised and correct a few errors.

Why should you listen to me? I have a degree in Pure Mathematics and Computer Science, and one of my primary research interests is number theory, so I know (more or less) what I'm talking about.

I'll start out talking about random number generation in general, and then move on to how it applies to card shuffling algorithms.


1. What is a random number?

When we talk about generating 'random numbers' in a computer, we generally want to generate very large (say, 32- or 64-bit) random numbers. However, from a theoretical standpoint the problem is no different from randomly generating a 0 or 1 -- i.e., a coinflip -- so we'll talk about unbiased coinflips instead of numbers in this section, just until we get a grip of exactly what it is we mean when we talk about 'random numbers.'



What is an 'random coinflip generator,' then? Think of it as a black box (let's call it HAL) that sits idly in the corner of your room until you ask it to perform a coinflip. Then it replies with either 'HEADS' or 'TAILS' and nothing else. HAL will give you a response each and every time you query it, until the end of the time. However, here's the kicker: without opening HAL and looking at its internal states, you cannot predict what its next answer will be based on its previous answers. You may know how it is programmed, and what its thought process is WHEN it is in a certain state (i.e., you may know its algorithm) but you do not know what all of its internal 'variables' are set to at any point in time.

This is a tough criteria to impose on our machine. In reality, we don't have any true random coinflip generators, because given enough processing time we CAN predict what result will be generated. However, it would take roughly the lifespan of the universe to break our best random coinflip generators, so we still say its results are "impossible" to predict without knowing its internal states.


2. Can HAL output "HEADS" ten times in a row? A thousand?

Yes! Think about what would happen if it were impossible for HAL to answer "HEADS" ten times in a row. Then if it ever answered "HEADS" nine times in a row (which is bound to happen even with a physical coin, eventually) then you would know that its next answer has to be "TAILS", which means you've predicted HAL's next answer without looking at its internal states. This is a no-no, if HAL really is a random coinflip generator.


3. After a thousand answers, is HAL guaranteed to have given 'close to' 500 TAILS replies and 'close to' 500 HEADS replies?

No. Again, see above. If we knew that HAL would output roughly 500 TAILS and 500 HEADS, perhaps with some allowed error, then again we'd be able to predict to some certainty what its outputs would be as we get close to its thousandth answer.

(Here's the rub: if, over a large enough sample size, we CONSISTENTLY observe that a machine is answering TAILS 51% of the time and HEADS 49% of the time, then we can become reasonably certain that this machine is not truly a random coinflip generator.)


4. Is Full Tilt Poker's random number generation biased?

According to the email quoted in BBB's thread, Full Tilt Poker uses three, independent random number generators and combines their results to create a 32-bit random number.

Using XOR to combine the three numbers simply ensures that an attacker (or bored mathematician) would need to break all three generators in order to predict the next number to pop out of the machine. There's not much more to be said about this.

So, what about the three methods themselves?

i) ISAAC

The ISAAC method is very fast, well-known, and relatively well-researched RNG.

It is known to be imperfect: if you initialize ISAAC with certain starting states, it will provide biased outputs. However, this flaw has been known for several years and it is quite rare to begin with, so I'm willing to bet that it is taken into consideration.

ii) OpenSSL

OpenSSL has been around for over a decade and is incredibly well-researched. Used properly, it is incredibly secure, and its random number generators are the basis for much of your privacy online. In short, literally thousands of mathematicians and computer scientists have an interest in its reliability, and any significant flaws have gone over all of their heads.

If there's a weakness in Full-Tilt's RNG, it doesn't lie here.

iii) Hardware RNG

Implemented properly, the output from one of these things is truly random. Wikipedia has a decent article on how they work. I can't say much else about this, since Full-Tilt didn't disclose exactly what kind of RNG they use.

The beauty of XORing the three results together is that ALL THREE methods need to be biased in order for a bias to show in the result. The chances of this seem slim.


5. But how do we KNOW?

Here's the thing: a biased RNG is easy to detect. There are several algorithms for testing RNGs and determining, to any desired degree of accuracy, whether or not their outputs are unbiased. These tests aren't arbitrary. They work and are mathematically sound.

The question is whether Full-Tilt tests its RNG results, and to what accuracy. That's obviously something I don't know.


6. Is Full Tilt Poker's card shuffling skewed?

Okay... let's say we have a reliable, unbiased RNG and wish to use it for shuffling a virtual deck of cards.

There are two, classic algorithms for doing so, both of which produce a random shuffling as long as the underlying RNG is unbiased:

i) The simplest way is to assign every card in the deck a random number, and sort the cards based on their assigned numbers. This method is simple, but not particularly efficient, as sorting lists of numbers isn't particularly quick.

ii) A more efficient method is to move through the deck one card at a time, and swap each card with a random card from the remainder of the deck.

The problem with the second method is that it is very easy to screw up by an amateur programmer. It requires a fair bit of care to avoid errors that make the shuffling biased. However, if you're a multi-million dollar company and can afford to hire competent programmers, you'll probably do so. But who knows.

If Full-Tilt's card shuffling is indeed skewed (which I think is unlikely), I wouldn't be surprised if the error is here. However, it is at best unclear and at worst impossible (I suspect the latter) for such a skewing to consistently give underdogs an edge in poker games, so it seems irrelevant to the discussion.


7. Are decks skewed to fit 'typical' distributions?

JP says the following:

   

Quote:

For algorithms to run, it's mathematical, poker deal is "fixed" based on the program to deal specific "typical" statistical averages over time. The program knows what it's dealt and what should be dealt. How it evolves from that I don't have a clue. Perhaps the program picked out that its deal was missing some statistic point, switched to maintain "XY" cards in "X" hands dealt. Once point reached, it went to it's normal "routine".

This should ABSOLUTELY NEVER BE DONE and NEVER would be in a setting like this. EVER. If any one mathematician of the dozens surely employed by Full-Tilt did this, they would be fired on the spot. Skewing random results to make them look more 'typical' is incredibly dangerous to any random number generator, as it actually GIVES AWAY information about the results. (See questions 2 and 3.)

There is a story in Cryptonomicon by Neal Stephenson about how during World War II, middle-aged women in top secret facilities would pick out lotto balls from a chamber, day-in-day-out, in order to generate the random keys the Allies to use to encrypt their transmissions. Often, the ladies would notice what seemed to be anomalies: they'd pick out a Q or an X 'more often than they should' and would put back 'unlikely' letters like these that seemed to appear too often.

The Germans realized this, and were able to crack the Allies' transmissions, using the knowledge that certain letters in the encryption keys would always appear at a certain frequency.


8. What possibility are we left with?

The only remaining possibility is that Full-Tilt deals fixed decks on purpose, and reshuffles them during a hand, to let the n00bs suck out on you. But why would they? Why risk such massive lawsuits to skew the game in certain players' favour? If you have a theory, feel free to let me know. I can't think of one.


8. Any conclusions?

Not really. I just hope this was informative to someone, and helped to clear up some misconceptions about this subject.

Chatting too much can be costly

I am not one to chat much when playing online poker tournaments. I do however go out of my way to say hi if somebody makes an effort and recognizes my player ID. Overall though, you will never hurt your game or your chances by never uttering a word in the chat box.

Chatters tend to get distracted, emotionally involved, and open themselves up to ridicule and arguments with others who are willing to participate. What you should be using the chat box for is monitoring your opponents, not engaging them. By not participating in chat you accomplish many positive things. Not the least of which will be that you will always have your opponents guessing no matter how hard they try. Silence will also keep you focused on your game, while others waste time trying to divulge information from you.

Even the best of players with subdued temperaments can be tempted to get involved in conversations that will likely become aggravated exchanges. You have to admit yourself that most conversations just turned ugly when it comes to online poker. Most players that participate in these types of exchanges are angry, jealous, and heading for tilted landscapes. You're just not going to feel any better arguing with them, even if you are right. And what would that help anyway? By arguing something that is so right in poker, are you not just giving free tournament training to your weaker opponents?

Negative words can affect your play, no matter what type of idiot in its shooting them do you in the chat box. But they will only affect your play if you get involved. By choosing to not respond to chat, you are more likely to send your opponent to the rail by encouraging his poor play, and holding your head above the fray. If you engage in vicious online chat with these players, you may actually be giving them some advantage over you, while distracting you from your own poker game.

Chat box silence, you may find will be a very calm in fact upon you are decisions during the game. Next time you play a sit and go tournament, try restricting your total comments to "gg" after your last hand of the game, whether you win or lose.