Does martingale really works?

[dooglus]

Python is a great beginner language for example, it's like today's BASIC except that it's suitable for a lot more advanced programming.

I too recommend Python as a first language to learn.  You might not end up using it, but it's a great way to get started.

Turns out it's not for everyone however.  Just this morning:

11:07:59 (760085) <x Sunset x> → (1) <@dooglus> Hey:) had a look at python, it's extremely hard to master it hehe

[1714061721]

1714061721

[1714061721]

1714061721

[1714061721]

1714061721

[1714061721]

1714061721

[1714061721]

1714061721

[kuverty]

If the numbers are off on either side, you can predict either a period of a higher amount of wins or losses, as the statistics HAVE TO balance out.

This is your fundamental misunderstanding.

The statistics don't have to balance out in the way you think they do.  They balance out, but in a way that gives you no predictive power.

Here's why they really balance out:

Suppose you flip a coin 10 times and it comes up heads only twice.  20% when it's meant to be 50% is way off.  So you would say "keep playing and it has to balance out".  Which is kind of true, but not for the reason you think.

If I play a thousand more times, I expect to get about 500 heads and 500 tails.  That brings the totals to 502 heads and 508 tails.

Now 49.7% of the flips have been heads.  Yay, it has "balanced out" from 20% heads to 49.7% heads even though we saw exactly 500 of both heads and tails.

And that's why it "balances out".  Not because you can predict an excess of tails in the coming rolls to balance it, but because you can predict that the 2 vs 8 weirdness you saw at the start will become insignificant over time as you flip more and more times.

Dooglus put it very well!

I like the following example myself: think about flipping a coin.

You flip the coin 20 times. Every time, you get heads. Now, do you expect that the next one will be tails more probably than the first one, or a few of the next flips?
If you think so, that it has to balance out and hence the next flip is more probably tails than heads, then think about this situation:
Imagine I had flipped the coin 20 times just before, without you knowing, and got 20 heads in a row. Then you come in and see the coin and decide you want to try some coin flipping.
Do you still believe that the 20 previous flips will have any significance as to your odds? Then how so? Surely there is no cosmic memory that knows the coin has been flipped, nor does the coin remember it. Every flip is a similar event, exactly, and it has no relation to the previous ones whatsoever.

[kuverty]

Dooglus, could you do the bitcoin community a favor and start a new gambling site based on the Monty Hall problem? I would really love to see the threads discussing that if martingale gets people confused.

[boumalo]

Another in the anecdotal series of "Does martingale really works?"

This time I present evidence for the "no, it bloody doesn't!" side of the argument:

It would be cool to be able to see the biggest martingales lost for the day or the week either on jd first page or in the blog

They sold expedia for 100m$

I think they sold their travel company *to* expedia.

To know the site profit without the Nawaka winnings you just add the Nakawa winnings to how much has been won, his final winnings are believed to be in an order of magnitude of 3000BTC if I remember correctly

I don't have exact numbers, because I don't know for sure which accounts were his.  But I estimate that he risked 2.5 million BTC and won 10k BTC.  The 2.5 million he risked should have netted the house 25k BTC.

So the current profit is 35k BTC short of where it should be, due to the BTC he won (10k) and the BTC he should have lost and didn't (25k).

That's probably where you're getting the '30000' number from.

Oh ok he supposably won 10k!

If he didn't have 25k to lose, he wouldn't have lost 25k though so it only represent a very fictive expected gain

[nimda]

First, I love how martingale charts look. A perfect martingale will be a straight line with periodic drops of varying magnitude, where deeper drops are exponentially more rare.

Second, is there any way to calculate the probability of a bust on a martingale procession while taking into account the ever-increasing bankroll it provides?

I think it wouldn't be exactly straight, even connecting the peaks, because a:
  -1 -2 +4
and a
  -1 -2 -4 -8 -16 +32
both add 1 to the total bankroll, but take different numbers of steps to do so.

I've seen people play like:
 -1 -3 -7 -15 -31 +63
before, which *does* lead to a perfect straight line, since the net gain is equal to the number of steps.  The 6 numbers in the previous line sum to 6.  It's equivalent to starting a new martingale of size 1 each bet:

First bet, bet 1 (and lose)
Second bet, bet 2 for the first martingale, and 1 for the new one you're starting: so 3 in total
Third bet, bet 4 for the first, 2 for the 2nd and 1 for the new one: 7 in total
...
Sixth bet: bet 32 for the 1st, 16 for the 2nd, ..., and 1 for the new one and this time we win, finishing off all 6 concurrent martingales at once.

I think the first person I noticed playing this way was the owner of the BTC Guild poll, Eleuthria.  It didn't work out too well for him in the end, and unfortunately I can't remember his account number to show the chart.

Oh, you're right. Normal martingale is only a straight line if you ignore the losses

That's an interesting variation on the martingale, too. It busts much quicker, but it's also a faster earner. Maybe something to try with tiny amounts.

[kuverty]

First, I love how martingale charts look. A perfect martingale will be a straight line with periodic drops of varying magnitude, where deeper drops are exponentially more rare.

Second, is there any way to calculate the probability of a bust on a martingale procession while taking into account the ever-increasing bankroll it provides?

I think it wouldn't be exactly straight, even connecting the peaks, because a:
  -1 -2 +4
and a
  -1 -2 -4 -8 -16 +32
both add 1 to the total bankroll, but take different numbers of steps to do so.

I've seen people play like:
 -1 -3 -7 -15 -31 +63
before, which *does* lead to a perfect straight line, since the net gain is equal to the number of steps.  The 6 numbers in the previous line sum to 6.  It's equivalent to starting a new martingale of size 1 each bet:

First bet, bet 1 (and lose)
Second bet, bet 2 for the first martingale, and 1 for the new one you're starting: so 3 in total
Third bet, bet 4 for the first, 2 for the 2nd and 1 for the new one: 7 in total
...
Sixth bet: bet 32 for the 1st, 16 for the 2nd, ..., and 1 for the new one and this time we win, finishing off all 6 concurrent martingales at once.

I think the first person I noticed playing this way was the owner of the BTC Guild poll, Eleuthria.  It didn't work out too well for him in the end, and unfortunately I can't remember his account number to show the chart.

Oh, you're right. Normal martingale is only a straight line if you ignore the losses

That's an interesting variation on the martingale, too. It busts much quicker, but it's also a faster earner. Maybe something to try with tiny amounts.

Well, nothing too special. Really just like a "one step quicker to lose" version of martingale.

With regular martingale the bets are 2ⁿ (starting at n=0): 1, 2, 4, 8, 16, 32...
The other is 2ⁿ⁺¹-1, so it's 1, 3, 7, 15, 31...

Sure it's faster because you the bets are risen faster. Why stop there, you could choose to bet 1, 10, 100, 1000, 10000... an even faster earner and faster loser. Or bet n factorial; 1, 2, 6, 24, 120, 720... (exponential growth is for the basic martingaler, boooriiing)

I myself think it's too slow losing all my money with these noob strategies, betting strategies based on sequences given by primitive recursive functions are super boring anyway. The Ackermann function can give nice sequences to follow, use it to your needs, that's a real pro tip. You make good dough if you keep winning!

[boumalo]

First, I love how martingale charts look. A perfect martingale will be a straight line with periodic drops of varying magnitude, where deeper drops are exponentially more rare.

Second, is there any way to calculate the probability of a bust on a martingale procession while taking into account the ever-increasing bankroll it provides?

I think it wouldn't be exactly straight, even connecting the peaks, because a:
  -1 -2 +4
and a
  -1 -2 -4 -8 -16 +32
both add 1 to the total bankroll, but take different numbers of steps to do so.

I've seen people play like:
 -1 -3 -7 -15 -31 +63
before, which *does* lead to a perfect straight line, since the net gain is equal to the number of steps.  The 6 numbers in the previous line sum to 6.  It's equivalent to starting a new martingale of size 1 each bet:

First bet, bet 1 (and lose)
Second bet, bet 2 for the first martingale, and 1 for the new one you're starting: so 3 in total
Third bet, bet 4 for the first, 2 for the 2nd and 1 for the new one: 7 in total
...
Sixth bet: bet 32 for the 1st, 16 for the 2nd, ..., and 1 for the new one and this time we win, finishing off all 6 concurrent martingales at once.

I think the first person I noticed playing this way was the owner of the BTC Guild poll, Eleuthria.  It didn't work out too well for him in the end, and unfortunately I can't remember his account number to show the chart.

Oh, you're right. Normal martingale is only a straight line if you ignore the losses

That's an interesting variation on the martingale, too. It busts much quicker, but it's also a faster earner. Maybe something to try with tiny amounts.

Well, nothing too special. Really just like a "one step quicker to lose" version of martingale.

With regular martingale the bets are 2ⁿ (starting at n=0): 1, 2, 4, 8, 16, 32...
The other is 2ⁿ⁺¹-1, so it's 1, 3, 7, 15, 31...

Sure it's faster because you the bets are risen faster. Why stop there, you could choose to bet 1, 10, 100, 1000, 10000... an even faster earner and faster loser. Or bet n factorial; 1, 2, 6, 24, 120, 720... (exponential growth is for the basic martingaler, boooriiing)

I myself think it's too slow losing all my money with these noob strategies, betting strategies based on sequences given by primitive recursive functions are super boring anyway. The Ackermann function can give nice sequences to follow, use it to your needs, that's a real pro tip. You make good dough if you keep winning!

You make good dough if you win 

I think martingale can be funny because you have a false sense of winning each time : +1 +1 +1 +1 ect. until you bust which will happen sooner than late

[kuverty]

Probability question for coin flippers
How many times, on average, would you be expected to flip a coin to get THREE HEADS IN A ROW? How about this sequence; heads, tails, heads.

I suggest it is a good exercise for any gambler to work it out. Simple but interesting if you've not thought about it before.

[dooglus]

Here's a list of the bets in the losing streak of length 16 from earlier today.  He bet L L H H L L H H all the way through.  Notice he would have won the 65.536 bet if he had made it and followed the same pattern:

2014-06-13 19:20:24 #1246171032:  0.001 BTC @ 45% lo:03.6353 profit:  0.0012

2014-06-13 19:20:26 #1246171083:  0.001 BTC @ 45% lo:79.3823 profit: -0.001
2014-06-13 19:20:27 #1246171139:  0.002 BTC @ 45% lo:85.5822 profit: -0.002
2014-06-13 19:20:29 #1246171195:  0.004 BTC @ 45% hi:06.3283 profit: -0.004
2014-06-13 19:20:30 #1246171244:  0.008 BTC @ 45% hi:32.1576 profit: -0.008
2014-06-13 19:20:32 #1246171297:  0.016 BTC @ 45% lo:45.8491 profit: -0.016
2014-06-13 19:20:33 #1246171348:  0.032 BTC @ 45% lo:79.6849 profit: -0.032
2014-06-13 19:20:35 #1246171395:  0.064 BTC @ 45% hi:10.5289 profit: -0.064
2014-06-13 19:20:36 #1246171448:  0.128 BTC @ 45% hi:40.4183 profit: -0.128
2014-06-13 19:20:38 #1246171495:  0.256 BTC @ 45% lo:70.8842 profit: -0.256
2014-06-13 19:20:40 #1246171547:  0.512 BTC @ 45% lo:71.0415 profit: -0.512
2014-06-13 19:20:41 #1246171593:  1.024 BTC @ 45% hi:38.4590 profit: -1.024
2014-06-13 19:20:43 #1246171638:  2.048 BTC @ 45% hi:26.5505 profit: -2.048
2014-06-13 19:20:44 #1246171691:  4.096 BTC @ 45% lo:91.5917 profit: -4.096
2014-06-13 19:20:46 #1246171745:  8.192 BTC @ 45% lo:83.1348 profit: -8.192
2014-06-13 19:20:47 #1246171791: 16.384 BTC @ 45% hi:35.0831 profit:-16.384
2014-06-13 19:20:49 #1246171843: 32.768 BTC @ 45% hi:07.9822 profit:-32.768

2014-06-13 19:50:03 #1246226691: 0.000001 BTC @ 45% lo:18.3393 profit:0.0000012

[dooglus]

Imagine I had flipped the coin 20 times just before, without you knowing, and got 20 heads in a row.

What if someone pre-flipped a coin until getting 20 heads in a row before a soccer match.

Then they could easily win the toss that decides who gets to kick-off by calling tails! 

[Campsis]

Here's a list of the bets in the losing streak of length 16 from earlier today.  He bet L L H H L L H H all the way through.  Notice he would have won the 65.536 bet if he had made it and followed the same pattern:

2014-06-13 19:20:24 #1246171032:  0.001 BTC @ 45% lo:03.6353 profit:  0.0012

2014-06-13 19:20:26 #1246171083:  0.001 BTC @ 45% lo:79.3823 profit: -0.001
2014-06-13 19:20:27 #1246171139:  0.002 BTC @ 45% lo:85.5822 profit: -0.002
2014-06-13 19:20:29 #1246171195:  0.004 BTC @ 45% hi:06.3283 profit: -0.004
2014-06-13 19:20:30 #1246171244:  0.008 BTC @ 45% hi:32.1576 profit: -0.008
2014-06-13 19:20:32 #1246171297:  0.016 BTC @ 45% lo:45.8491 profit: -0.016
2014-06-13 19:20:33 #1246171348:  0.032 BTC @ 45% lo:79.6849 profit: -0.032
2014-06-13 19:20:35 #1246171395:  0.064 BTC @ 45% hi:10.5289 profit: -0.064
2014-06-13 19:20:36 #1246171448:  0.128 BTC @ 45% hi:40.4183 profit: -0.128
2014-06-13 19:20:38 #1246171495:  0.256 BTC @ 45% lo:70.8842 profit: -0.256
2014-06-13 19:20:40 #1246171547:  0.512 BTC @ 45% lo:71.0415 profit: -0.512
2014-06-13 19:20:41 #1246171593:  1.024 BTC @ 45% hi:38.4590 profit: -1.024
2014-06-13 19:20:43 #1246171638:  2.048 BTC @ 45% hi:26.5505 profit: -2.048
2014-06-13 19:20:44 #1246171691:  4.096 BTC @ 45% lo:91.5917 profit: -4.096
2014-06-13 19:20:46 #1246171745:  8.192 BTC @ 45% lo:83.1348 profit: -8.192
2014-06-13 19:20:47 #1246171791: 16.384 BTC @ 45% hi:35.0831 profit:-16.384
2014-06-13 19:20:49 #1246171843: 32.768 BTC @ 45% hi:07.9822 profit:-32.768

2014-06-13 19:50:03 #1246226691: 0.000001 BTC @ 45% lo:18.3393 profit:0.0000012

He must of been really pissed off when he realised he would of won the last bet (x Sunset x)

[dooglus]

If he didn't have 25k to lose, he wouldn't have lost 25k though so it only represent a very fictive expected gain

Exactly.

People sometimes look at the sequence of bets he made, and use a monte-carlo simulation to calculate the chance of him winning as much as he did, given the bets that he made.

But I think that's kind of flawed - because if he hadn't won more than he lost early on, he wouldn't have been able to afford to even make that many bets.

When we look at the actual site profit (19k) and compare it to 1% of the total amount wagered (53k) we see that actual profits fall 34k short of expectation.  And so the natural question is "where is the missing 34k?  who won 34k?".  The answer is (roughly) "he won 10k, and failed to lose 25k (and we made up the extra 1k by being lucky)".

[PacoMartin]

You make good dough if you keep winning!

It's also a good idea to buy stocks when they are low, and sell them when they are high.

Betting strategies should be viewed as for your own entertainment. The underlying probability of the game never gets better or worse because of a betting strategy.

But viewed that way, Just Dice gives you the chance at playing lottery style bets without the lottery commission taking half your money. Max payout is 990,000x in JD. If you want to buy 1000 million dollar lottery tickets for $1, then the overhead is about $500 at the lottery. On JD the overhead is $10, and all of it goes at the jackpot instead of wasting away on mini payouts.

Factory workers who pool their money every week and purchase a box of Powerball tickets would be better off if someone would place that money into JD. They could produce a paper and show where every bet went, not like those cases where one ticket is mysteriously replaced, only to be claimed 6 months later.

[kuverty]

Have you ever seen a martingale player that does not care whether she is betting on head or tails, only that she is doubling her bet each time? I can understand that it is easier to bet on just the other and that bots do this, but that I've never seen a martingaler just bet on whichever suggests that there usually there are some magical beliefs involved.

Edit: well, there is just one quite like that a few messages back in this thread

[dooglus]

But viewed that way, Just Dice gives you the chance at playing lottery style bets without the lottery commission taking half your money. Max payout is 990,000x in JD. If you want to buy 1000 million dollar lottery tickets for $1, then the overhead is about $500 at the lottery. On JD the overhead is $10, and all of it goes at the jackpot instead of wasting away on mini payouts.

The only problem with that is that the "jackpot" (ie. maximum win) at Just-Dice is only 0.5% of the bankroll.  Currently the bankroll is around 50k BTC, and so the most you can win per roll is 250 BTC.  At a little under $600 per coin that is worth a little less than $150,000 which doesn't go far when shared by a large group of workers pooling their stakes.

[deadlyunknown]

It does work out to gain profit, but you will terribly in the end if you have a losing streak.

[PacoMartin]

To know the site profit without the Nawaka winnings you just add the Nakawa winnings to how much has been won, his final winnings are believed to be in an order of magnitude of 3000BTC if I remember correctly EDIT : he probably won around 10,000BTC from JD

When we look at the actual site profit (19k) and compare it to 1% of the total amount wagered (53k) we see that actual profits fall 34k short of expectation.  And so the natural question is "where is the missing 34k?  who won 34k?".  The answer is (roughly) "he won 10k, and failed to lose 25k (and we made up the extra 1k by being lucky)".

My point earlier was not that it was difficult to factor out Nawaka , it's just that there is no point in doing so while trying to make an investment decision. Management cut the max profit from 1% to 0.5% of the bank to reduce variability, but it is still a low house edge game.

You can't remove the bumps unless the you raise the house edge and greatly reduce the max bets. But then your profit is low because nobody plays.

[PacoMartin]

...which doesn't go far when shared by a large group of workers pooling their stakes.

You are probably correct. Given 1 BTC=$594.9
Betting 0.025 BTC ($14.87) at 1:10,000 odds yields a jackpot of 247.475 BTC ($147,223).

It would probably make more sense for ten guys to pick positions, and print out the paper and see which position wins. They are much more likely to see somebody win something over the course of a year or two, than just to waste their money on a tiny chance of splitting a megajackpot.

If they agree on a time when the bets will be placed, then the documentation will have the timestamp, and it won't be possible to cheat. Plus you can easily produce the entire history of bets under a name, and so you have a record of all bets. The person placing the bets can't hide a win, and simply do another bet at a later time.

I must be a very distrustful person, but I see stories about lawsuits all the time.

I just object to that 50% house edge.

[badpeanut]

i martingale sports betting!

[PacoMartin]

...there usually there are some magical beliefs involved.

Belief in divine intervention and magic in unlikely streaks is a pervasive part of all of human culture. You can show thousands of writings and stories. The Pythagoreans who invented the famous triangle theorem were a religious group.


==================
Did I ever tell you I've been struck by lightning seven times?
Once, when I was fixing a leak on the roof.
Once, when I was crossing the road to get the mail...

Did I ever tell you I was struck by lightning seven times?
Once, when I was in a field tending to my cows.

Did I ever tell you I've been struck by lightning seven times?
Once I was sitting in my truck listening to the radio.

Did I ever tell you I was struck by lightning seven different times?
Once, I was walking my dog along a country road.

I'm blind in the one eye...
I can't hardly hear...
I get twitches and shakes out of nowhere...
I lose my line of thought...
but you know what...
God, for some reason, keeps on sending me a wake-up...
It's His way of reminding me I'm lucky to be alive...
(sniffs the air) Storm's comin'.

"Along the way you bump into people who make a dent on your life...
Some people... get struck by lightning..."
"Some are born to sit by a river."
- THE CURIOUS CASE OF BENJAMIN BUTTON