Bitcoin Forum

Now don't lecture me about the Gambler's fallacy problem

This is the dumbest strategy to date.

This can be done in dice sites with 1%/0.5% house edge, but you will need a bot to stop on win and switch, as with what I know none of the dice sites has this option on their autopilot

Sorry but I did simulations with 200,000 rolls, and it works.  I either end up with -0.01% loss or with 7.7% gain, so I think its a perfect strategy for low risk/ low reward gamblers.

The average consecutive loss/win there is ~ 15-17, so a 20 consecutive trigger is rare, so the risk to bet on that is low.

I am interesting to know so how would you get recover when you lost amount at any losing streak,

Rule 2 in the book is never chase your losses.  8)

Every bet is independent.

you use your strategy to the simulator? Don't use to real gambling site? Your stat are only "simulated" ? Why don't test directly to one dice with 1 satoshi minimum bet?

I put out my strategy simulator I made, you can try it out if you want, I rolled out simulations on easily 200,000- 2,000,000 bets and it does hold up to those wins.

I also think the gambler fallacy is not what it looks like, it is certainly misinterpreted.

Yes because with any type of recovery strategy, the risk increases faster than the reward and you will end up losing more.

You have to let the losses go in order to stop losing more! :)

Individually the bets are random and independent, but collectively they tend to built up a pattern.

Yeah, coz of variance it could show anywhere from a massive profit to a massive loss.

LOL, you did. xD

But apply it in this case. With this strategy too the risk is 1.9% more than the reward.

This method takes too long to win a little bit of satoshi and every bet is independent of the previous.
In theory, you can lose 30 times in a row and you cannot blame that the gambling site is rigged.

Thanks for sharing anyways.

These are nothing new though.... it's the same as the old double bet strategy on Roulette.

Good luck with your gambling

Even though if you roll 20 heads in a row, it is more likely that the 21st will be tails. Of course next time it might be 21 heads in a row, but its all random.

It's not just the individual outcome that is random, but the cluster too, so even though if you have 20 heads in a row, it is morel likely that the 21st will be tails, it is not true always.

By betting tails after the 20th, you might win once, but you might lose another time. But it gives a bigger certainty anyway than just naked betting randomly.

[...] is the mistaken belief that, if something happens more frequently than normal during some period, it will happen less frequently in the future [...]

No I was explaining that the official explanation of the fallacy is short sided, its more to it just than the independent outcomes. Because the cluster is randomized too, so weather the odds are set to any amount, the distribution tends to follow the variance.

Or in other words , except a few black swans, the distribution is also contained to a local variance and it can be exploited to control risk better.

Example: if you flip a coin 100 times, the odds of being tails 10 times is 0.09765625% and yet the 11th roll  is less likely to be tails, but not impossible.So eventually the heads have to come, so we can say that the local probability can vary wehereas the overally probability remains constant. Now this is no way a guarantee, because somehow the 12 consecutive tails have to come too, but then the local probabilities just reshuffle, randomly.

This is a common mathematical phenomena. I think its called local probability distribution:
http://pitt.edu/~druzdzel/psfiles/flairs06.pdf

Hey its random anyway, but you have to understand that based on odds, somebody can come out a net winner even after 100,000 rolls. It all depends on the local distribution , how it gets shuffled.

And those who are losers, at least lose less than they normally would with other strategies.

I don't think people understand the gambler's fallacy.

Even if you flip a coin and get tails 100 times, the next coin flip has a 50% chance to be tails and a 50% chance to be heads. Not anything different.

However, from the BEGINNING, there is a 3.9443045261050590270586428264139e-31% chance to roll tails 101 times.

But your very next bet has no relation with your previous bets.

If the winning chance is 50%, it will always be 50%. No more, no less.

So your betting strategy is basically the definition of gamblers fallacy?  How would you expect us not to mention it?  There is no way you studied stats and came up with this, I studied stats for 1 year and I can tell you this is pretty much what you learn on day 1, "Are the events independent on one another or not"

In your case, they are not, so you are doing nothing with stats at all.

I can't believe your charging for the demo

This is a sales thread  after all....

You're a cheeky monkey

Do you believe your strategy is positive EV? ;)

You are talking about a distribution with trillions of trillions of trillions...... of rolls tending towards infinity. The concept you are talking of doesn't apply here. The sample size even if you make a trillion rolls is too small and not representative

I do appreciate the effort. Anyway, you will learn some new stuff. ;D

And what if there are not enough same rolls? Let's say I decide to change the bet after 5 same rolls (5x or 5y), start betting and end up with a pattern like 1x 3y 4x 1y 1x and so on. I could end up gambling whole day and never really getting to use the strategy but losing all my money if I choose wrong.

I am a pretty old gambler, and I know how gambling works, you play against the odds of the house. Yet, there is still a possibility to win in the long term, all based on luck of course, but with a good gambling strategy , atleast we can manage the risk.

So this is all based on luck, i`m not claiming that you will win guaranteed, because that is impossible, but if you add some basic risk management to the equation, you can play much safer!

Here is what I thought:

This strategy only works for dice/coin flipping type games where you got a 50% winrate.


So there aren't many gambling strategies out there, let's analyze them each:

1) Martingale
The problem with martingale is that the profits increase linearly, but the risk increases exponentially with every bet, so martingale really empties your account fast, and the profit is not worth the risk, even though it gives the illusion of loss-protection in the short term, but the risk is not worth it

2) Naked betting
So this strategy is when you just bet the same amount, or varying amounts, without any type of pattern, each dice roll. This strategy is more safe than the martingale believe it or not, it might have a less win rate, but it doesn't empties out your account, plus you could win more.
The problem here is that, usually there is a small percent that 1 player could win at all in the end, and you have no way of knowing.


3) Variance hunter (my strategy)

Ok so we know the problems with the 2 above strategies, so I thought i make my own observations. My strategy plays off the variance, and it exploits the consecutive rolls. Now don't lecture me about the Gambler's fallacy problem , because I know that, I`ve studied mathematical statistics many years, plus I`m not claiming you will get any edge with this strategy, because you won't. The only advantage you will get by using my strategy is that you can manage your risk and have a low risk , low reward type gambling experience!

Ok so here it is:

Requirements:

• You need a gambling game with 50% winrate  (dice,coin flip game), this is a must, but the payout ratio is obviously gonna be <x1 due to house edge, so that doesn't matter. Ex:  50% winrate with 0.98x payout rate = 2% house edge game
• The lower the house edge is the better , and the house edge must be constant, it doesnt work in varying house edge casinos
• The casino must be provably fair, and must use real random numbers or atleast pseudo-random with ultra-high entropy
• The casino should have an API system, because this system is hard to do manually, you will probably need a programmer to do this, because it might involve betting 5000 times in a row
• The casino should have a low minimum betting amount, the lower the better
• The casino should not have any betting size restrictions, you should be able to bet 5-10% of your balance in a bet
• The casino should be very fast to roll, and not have any restrictions on rolling because you will probably have to bet many thousands of times

The strategy:

-You will bet on 1 outcome the minimum betting size, until X amount of consecutive events happen, then switch to the other outcome, that X should be set according to how many bets you will going to play (more on this later)

Ex.: bet on heads 1 satoshi, and after 10 consecutive heads happen, then switch to tails.
The same if you bet on heads and 10 consecutive tails happen, then you stay at heads, so you only look at what happened consecutively X times, and switch to the opposite outcome next bet.

-After the X amount of consecutive bets happened, you switch to the other bet, and increase the betting size to Y% of the balance, presumably < 10% / bet to have moderate risk, but more on this later.

-After the Y% bet happened, regardless of you won or lost, you resume to bet the minimum amount , until another consecutive event row happens.

EXAMPLE SETUP:

This is what I use personally:

I use Satoshidice (https://www.satoshidice.com/) and it has a 1.9% house edge and it has nice API platform so you can make a gambling bot. Since satoshidice doesnt have changeable outcomes, I use a slightly modified version of my strategy, which is essentially the same but it's easier to do. Instead of waiting for heads/tails to switch, I just look at consecutive wins or losses. So I bet 100 satoshi, the minimum , until I get X consecutive losses, then I increase the bet on the next roll.

I usually gamble 0.5 BTC balance.
Bet size: 100 satoshi
Trigger: after 20 consecutive losses I increase the bet to 8% of my balance so that is 0.04BTC
Risk %:  8% of my balance after the 20 consecutive events, then reset to base bet size of 100 satoshi.

RESULT:

You will most likely end up either at -1-10% or win the risked %. In rare cases you will lose more than 20%, it is possible but not that probable.
So it is more like a small reward 7-8% gain/session, with the same risk, and a long term betting strategy.



Well I made a simulator for my strategy that simulates the exact strategy I described above, and I share it with you. If you want the source code too, then PM me.

http://i60.tinypic.com/2ql60xz.png

Download link: http://s000.tinyupload.com/?file_id=62215764482435873397
SHA 256 hash: 22ca3bb65893eb4b9637bd4aea6cacacfa3ea7be27a989310e9efd32b599908a       (https://md5file.com/calculator)
Virus Scan: https://www.virustotal.com/en/file/22ca3bb65893eb4b9637bd4aea6cacacfa3ea7be27a989310e9efd32b599908a/analysis/1443089362/


What do you think guys?

Disclaimer: Use this strategy at your own risk, the strategy doesn't guarantee anything, gambling is risky anyway. I`m not responsible for any of your losses if you lose.

after 20 consecutive losses I increase the bet to 8% of my balance so that is 0.04BTC

Of coure, Low risk, low rewards
High risk high return
But i'm not sure about this

As you know you can lose 2-3 times after that increasing, which is 0.08-0.12BTC and to get that amount back will take lots of time and what is more important lots of risk too. Most likely you'll lose everything trying to recover. Of course you can win with that strategy if you are lucky, but you can also win betting 0.5 BTC just one time on 0.98x payout.

Yeah that is so true, but sometimes people with a great luck can hit jackpot and win big and of course this is exception, mostly people will earn like that way but you need some good balance for you to sustain your loss or you will never hit it back again

If you'd care to read my posts before replying it would be better.

I didnt deny that the rolls are independent of another, but overall they are confined to a cluster that is dictated by the local probabilities.

After 10 heads the tails has a higher probability of occuring (locally). Yes this is true.

However if the overall cluster is reshuffled, to give room to 11,12,13... heads in a row, then obviously that will happen eventually too.

So we must look for a exploitation in the cluster until it is intact , but that is impossible to predict because that is random too, it can reshuffle anytime.

So yes you need to be lucky too, I dont give any edge here, i`m just trying to exploit a pattern until it exists, after its gone we can lose too, so it is gamble.


That was not my intention at all, i just didnt found a free file hosting service.

Here i found one, so now you can download it for free and see for yourself.

No, and i never said I will give any edge, I just said that this strategy can help gamble in a "safer" way.

THis is not true, this is the gambler's fallacy.

The odds of a fair coin being head or tails is 50/50, every time, no matter what the previous x flips were.

If you ALREADY flipped 10 heads, NOW the odds of getting 11 in a row are 50/50 because you ALREADY did the hard part of getting 11 in a row, getting 10 in a row.

You are stating the EV is not positive and is the same. (correct)

You are stating ROI % that implies a positive EV (contradictory and incorrect)



Suppose you make 1,000,000,000,000 rolls and got all heads, for the next head assuming the game to be a fair one, the odds are 50%, and for the tails the odds are 50%.

I think you are taking it in a way where in a probability distribution of say 1000 rolls, ~500 are heads and ~500 are tails, and if you get 100 heads, the next flip would most likely land a tail. This concept DOES NOT apply in this case. Because the distribution you are talking about is INFINITE. Even if the probability distribution happens to be evenly distributed (is it?) say in centillions (tending towards infinity to be stated properly) of hash combinations, it still gives you an astronomically insignificant effect (1/infinity tending to 0)

I understand what you are saying guys but you fail to see the perspective.

Let's look at "life". All life expectancy is negative because everybody will die eventually.

So by your logic nobody can be a billionaire in his lifetime because his life has a negative expectancy, but that is obviously not the case.

So you can have local positive expectancy, while the overall one is negative.

I tend to see probability as a collective of local ones, so the total negative expectancy of a gambling game, is composed of local positive expectancies + local negative expectancies, where the local negative dominates, but that doesnt mean that the positive cannot happen.


Take the inverse for example, if you win 1000 times in a row in a 50% probability dice game:
- did you played agains  overall negative expectancy: Yes because the payout ratio was < 1x
- did you had a local positive expectancy: Yes, because you won 1000 in a row. Otherwise you could not have the trend. It was lucky, but it was still a trend.

So if you can jump on a local positive expectancy train, you can win locally, but if you "force" your luck, you will eventually lose.

I didn't say no one could be a billionaire, I didn't say that you couldn't flip 1 million heads in a row. I said that if you flipped 10 heads in a row, the odds the next one being heads is 50/50. In your nonsense you said that tails was more likely on the 11th roll. That is flat out wrong. If you don't want people to call you on your bullshit don't post bullshit.

There is no way to PREDICT trends among independent events. You can look back and find them, but you cannot predict them because each roll is independent and has nothing to do with previous rolls/flips.

Ok, but what if your personal luck variable, and the casino's luck variable are independent.

We know the casino has negative expectancy, but we dont know if the person has. And if we add that together, if you have a higher perosnal expectancy than the casinos negative expectancy (his personal luck), then you can come out a winner in the end.

Many people did that, they played roulette, baccarat, dice or whatever and they ended up net winners.


There is no other way to find out, but only if you gamble and see for yourself.

It is more or less as the advice: 'stop when you are winning', or, in others words, stop when the variance is working in your favor

The only way to win at dice is to play 3  strategies at the same time.

Example:

start at 1 satoshi at 10x payout doubling the bet after 5 losses in a row. Reset the bet back to 1 satoshi after a win.

FOR BETTING LOW
At the same time keeping track of any rolls\results of less then 1 and how many bets you have placed. Once you reach 500 rolls without a roll/result being under 1, start betting at 99x payout and double the bet after 48 losses in a row. Reset the bet back to 1 satoshi and 10x payout.

FOR BETTING HIGH
At the same time keeping track of any rolls\results of greater then 99 and how many bets you have placed. Once you reach 500 rolls without a roll/result being over 99, start betting at 99x payout and double the bet after 48 losses in a row. Reset the bet back to 1 satoshi and 10x payout.


repeat

Each bet is independent. After 500 rolls without <1, the chance to have <1 does not change but is still the same at exactly 1%, so the "switch" part doesn't quite work. Also, even if a very low base bet at 1 satoshi, you will quickly run out of fund to continue doubling your bet when you hit a long losing streak.
In short, your strategy would not work in the long run. Indeed, no strategy would work in the long run for the luck based dice game with negative EV.

Each bet is not independent of each other. Each bet is determined by a algorithm where a nonce is incremented each bet. Which means each bet is dependent on all previous bets, since the bet with nonce=10 would never exist without the previous 9 nonces.

The algorithms that determine the outcome are pseudo random and not truly random.

Each bet is not independent of each other. Each bet is determined by a algorithm where a nonce is incremented each bet. Which means each bet is dependent on all previous bets, since the bet with nonce=10 would never exist without the previous 9 nonces.

The algorithms that determine the outcome are pseudo random and not truly random.

The result is independent.

If I give you 10 rolls, can you predict the next? No. That is referred to as independence here, of the result.

The 11th roll is dependent on the server seed, client seed, and nonce and the previous rolls.

The 11th roll is dependent on the server seed, client seed, and nonce and the previous rolls.

Example:

roll 1: 0.65
roll 2: 0.23
roll 3: 0.99
roll 4: 0.55
roll 5: 0.09
roll 6: 0.13
roll 7: 0.33
roll 8: 0.67
roll 9: 0.04
roll 10: 0.97

I would be willing to bet that the next roll will be higher then 1. But without the knowledge of the previous 10 bets I wouldn't.

At the same time I would like to see a server seed, client seed, nonce produce rolls anywhere close to what I listed.

I would be willing to bet that the next roll will be higher then 1. But without the knowledge of the previous 10 bets I wouldn't.

At the same time I would like to see a server seed, client seed, nonce produce rolls anywhere close to what I listed.

The gambler's fallacy, also known as the Monte Carlo fallacy or the fallacy of the maturity of chances, is the mistaken belief that, if something happens more frequently than normal during some period, it will happen less frequently in the future, or that, if something happens less frequently than normal during some period, it will happen more frequently in the future (presumably as a means of balancing nature). In situations where what is being observed is truly random (i.e., independent trials of a random process), this belief, though appealing to the human mind, is false. This fallacy can arise in many practical situations although it is most strongly associated with gambling where such mistakes are common among players.

thank you for that and Chapeau!

there will be many user who will disagree with you but a few will understand the point you made

Each bet is not independent of each other.

Each bet is determined by a algorithm where a nonce is incremented each bet.

Which means each bet is dependent on all previous bets

The algorithms that determine the outcome are pseudo random and not truly random.

If you would run a simulation of bets/rolls (same server seed and client seed) the outcome over many bets should roughly show equal distribution of numbers. It might happen in 1000 bets, 10000 bets, or 100000 bets, but it will happen. But knowing this will happen allows you to change the way you bet in your advantage.

Now if you say there is not equal distribution of numbers, that's even better. You can determine where most of the roll/bets hit and play that to your advantage.

Either way, not looking to the past results to determine the way you play dice is why most call it a luck game.

This is 1000 percent wrong.

Lets just look at flipping a coin to make things simple. The equal distribution happens from having the correct probability happen over and over and over again over a large sample.

If you flip 10 heads in a row, the odds of the next one being heads is still 50/50. If you keep flipping a coin, eventually your results will get to 50/50, not because tails is more likely on future events, but just because you keep doing it.

If you get 10 heads out of 10 flips you got heads 100% of the time.

If the next 10 are 50/50, you have 15 out of 20, 75% heads

If the next 10 are 50/50, you have 20 out of 30, 66 2/3% heads

If the next 10 are 50/50, you have 25 out of 40, 62.5% heads

If the next 10 are 50/50, you have 30 out of 50,  60% heads.

And so on and so on until you get back down to 50.

You are comparing something that is truly random(flipping of a coin) to a algorithm. Have you run a simulation with the algorithm from a dice site?

I also notice that in your example that your approaching 50% as more flips happen, which is what I'm saying.

This is 1000 percent wrong.

Lets just look at flipping a coin to make things simple. The equal distribution happens from having the correct probability happen over and over and over again over a large sample.

If you flip 10 heads in a row, the odds of the next one being heads is still 50/50. If you keep flipping a coin, eventually your results will get to 50/50, not because tails is more likely on future events, but just because you keep doing it.

If you get 10 heads out of 10 flips you got heads 100% of the time.

If the next 10 are 50/50, you have 15 out of 20, 75% heads

If the next 10 are 50/50, you have 20 out of 30, 66 2/3% heads

If the next 10 are 50/50, you have 25 out of 40, 62.5% heads

If the next 10 are 50/50, you have 30 out of 50,  60% heads.

And so on and so on until you get back down to 50.

Stats are a crazy thing.  Looking at things in a short run, getting 10 heads in 10 flips is .5^10.  Not good odds.

However after 10 flips, the odds of flipping the next coin is still 50/50.  In your example after the 10 ones are head, the next 100 rolls are still supposed to be closer to 50 heads 50 tails than anything else. 

That's what I was showing. It is unlikely to get 10 heads in a row. But if you do, the next one is still 50/50, AND you will get to 50/50 over the long term without changing the odds.

If you would run a simulation of bets/rolls (same server seed and client seed) the outcome over many bets should roughly show equal distribution of numbers. It might happen in 1000 bets, 10000 bets, or 100000 bets, but it will happen. But knowing this will happen allows you to change the way you bet in your advantage.

Now if you say there is not equal distribution of numbers, that's even better. You can determine where most of the roll/bets hit and play that to your advantage.

Either way, not looking to the past results to determine the way you play dice is why most call it a luck game.

Yeah I was mostly agreeing with you, but sort of.  I was saying the dice don't know that they rolled 10 heads in a row, so even after the 10 rolls it won't start to go more toward 50/50, from that point on it would be closer to 50/50, but after I suppose 100000 rolls if you had 10 more heads than tails it still would be close to 50/50.  

Which is called gambler's fallacy. :P
The funny thing is you are the first guy I have seen who have that fallacy for select (may be only for pseudo random) cases but does not have it for the rest.

LOL, your evening out strategy doesn't work. Pseudo random is kind of pre-determined random (in this case) numbers generated using a certain method. It is not predictable and does not even out (esp. in finite rolls).




To clarify, if you make infinite rolls (10^100000000000000000000000000000000000000000000000000000000000000000000000000...)
the heads and tails tend to even out. But when you take 10 or a trillion rolls, there is no such concept.

Yeah, it is mostly correct, but if it was to even out, as the rolls go by, the tails would have more odds than heads which is incorrect. So Phildo is not exactly wrong, but in this particular part, leex is 100% correct. :)

tl'dr There is no such thing called evening out in practical cases/finite rolls.

Let us check. :P

Server seed hash
3aaac305ec911f3dc3965c731dfe6db16902eb23822487a593114e68922c9e32
Server seed
a151b24df86e5dea2ef3--snip---
Client seed
ndnhc


What is the 89th nonce roll?



Edit: PrimeDice. I can give thousands, just doesn't want it too long, lol.

I'll agree with if you show me your betting simulations and state which dice game algorithm you used.

All I can tell you is I have made a dice simulator for JD and PD and if you group the results by 10s you do reach a point where you approach equal distribution of numbers.

Using this method I can within a good % know that a bet will win in a worst case situation.

To clarify, if you make infinite rolls (10^100000000000000000000000000000000000000000000000000000000000000000000000000...)
the heads and tails tend to even out. But when you take 10 or a trillion rolls, there is no such concept.

Yeah, it is mostly correct, but if it was to even out, as the rolls go by, the tails would have more odds than heads which is incorrect. So Phildo is not exactly wrong, but in this particular part, leex is 100% correct. :)

tl'dr There is no such thing called evening out in practical cases/finite rolls.

But the local probability always converge to the basic probability.

I just rolled 20000000 times with the virtual coin and turned out:  10000630 heads and 9999370 tails, which is very close to eachother.

Maximum consecutive is 29

[IMG.]http://i62.tinypic.com/35k34nc.png[/img]

But the local probability always converge to the basic probability.

I just rolled 20000000 times with the virtual coin and turned out:  10000630 heads and 9999370 tails, which is very close to eachother.

Maximum consecutive is 29

http://i62.tinypic.com/35k34nc.png

Of course. That is expected. But still the difference is ~1250. Try making it 0-10.

In dice by evening out I mean tending towards an equal number of 0.0000, 0.0001, 0.0002 to 99.9999

Was there ever a time they where equal?

Was there ever a time they where equal?