Does martingale really works?

[dooglus]

175,833,708 had positive profit
179,380,878 had negative profit
2,722,802 had zero profit (and so I can't immediately tell if they won or lost, being 0 BTC bets)

>>> 175833708 * 100.00 / (175833708 + 179380878) = 49.5007

49.5007% of the non-zero bets at 49.5% won!

Wow , that seems really accurate to the odds.

Mathematically , what would the average error % of the odds be ? That seems too close to the given odds to be true.

The law of large numbers says that the more bets we have, the closer the actual percentage win rate should get to the expected percentage win rate.  We've had over 300 million bets at 49.5% so it's not surprising that it's close.

Of the 355214586 non-zero bets at 49.5%, we would expect to see 175831220.07 wins and 179383365.93 losses, but we saw 175833708 wins, which is 2487.93 too many.  So in a sense we're not close to expectation - we saw over 2 thousand too many wins.  But in another sense we're very close - because that's only 0.0007% too many.

As time goes on, we don't expect that 2487 number to get any smaller.  We just expect the total number of bets to get bigger, which causes the 2487 number to become a smaller percentage of the total.

I know that doesn't answer your question, but that's because I don't know the answer.    A cheap answer would be "the average would be 0% error, because all the positive errors cancel out all the negative errors".  But you're probably asking for the average of the absolute value of the error, which I don't know how to calculate.

[cp1]

It should just go as the square root of N, so you'd expect to be within about .01 %

[dooglus]

So I contacted a mathematician guy to ask for help in calculating my chance of winning the bet I made here recently.

He pointed me at this page, and in particular this part of it:



But I have no idea how to evaluate that for the bet I made.  He concluded that I was very likely to win the bet, but didn't show his working.

Is there anyone who can explain how to use the above formula to calculate the odds of seeing at least a 28-streak of losing bets in a sequence of 357,937,388 bets at 49.5%?

[BayAreaCoins]

So I contacted a mathematician guy to ask for help in calculating my chance of winning the bet I made here recently.

He pointed me at this page, and in particular this part of it:



But I have no idea how to evaluate that for the bet I made.  He concluded that I was very likely to win the bet, but didn't show his working.

Is there anyone who can explain how to use the above formula to calculate the odds of seeing at least a 28-streak of losing bets in a sequence of 357,937,388 bets at 49.5%?

Did he take in to account that the bets where spread over who knows how many accounts?  Speaking of that... how many different accounts had made a 49.5% bet? Jw

[dooglus]

Did he take in to account that the bets where spread over who knows how many accounts?  Speaking of that... how many different accounts had made a 49.5% bet? Jw

Oh, and only 22,807 different accounts made bets at 49.5%.

I think that since the number of bets is massive compared to the number of accounts, we can assume it's all just one account without much loss of accuracy.

[TheJuice]

I actually use a reverse-martigale system when I bet in vegas on the craps table (only real betting I do besides poker). It makes it more fun because when you bet bit if you win it menas you are up a lot. As opposed to the normal martingale where a big bet is just to be even.

For instance when I'm in a casino and have a budget of $1k.

Each round is say 5 minutes (in craps)

$10 min bet (so 100 rounds) for 500 minutes of 'fun' or ~8 hours.

My modified start in craps is basically

bet $10 to open -> win 2x bet
after 7 wins I take 25% off the table (after 100 plays there is a 52% change of this happening) - which is $1280 * .25 = 320 pocket
if the 8th is a win 960 * 2 = 1920 (i take 25% off again) = 480 pocket (or 800 total)
if a 9th win 1440 * 2 = 2880 (take 25% off) = 720 pocket (1520 total). [this happens ~18% of the time with a 100 plays]

etc until I lose or hit max bet at table. My personal best is a streak of 11 at which time I had 4.5k to open (may bet for the table was 10k I think, so another 2 wins and I would have walked)... [and lost] but had already pocketed $3k.

Either way I get to play for hours and once you get to ~5 wins people start gathering around the table looking at the crazy guy betting 1k upfront (at a 10 min table).

[dooglus]

I actually use a reverse-martigale system when I bet in vegas on the craps table

We saw a good example of reverse martingale working on JD yesterday:

Martingale loses:

    2014-06-01 08:10:58 *** angel (781321) [#1220009235] bet 0.032 BTC at 49.5% and lost ***
    2014-06-01 08:10:59 *** angel (781321) [#1220009276] bet 0.064 BTC at 49.5% and lost ***
    2014-06-01 08:11:00 *** angel (781321) [#1220009323] bet 0.128 BTC at 49.5% and lost ***
    2014-06-01 08:11:02 *** angel (781321) [#1220009367] bet 0.256 BTC at 49.5% and lost ***
    2014-06-01 08:11:03 *** angel (781321) [#1220009413] bet 0.512 BTC at 49.5% and lost ***
    2014-06-01 08:11:04 *** angel (781321) [#1220009463] bet 1.024 BTC at 49.5% and lost ***
    2014-06-01 08:11:06 *** angel (781321) [#1220009500] bet 2.048 BTC at 49.5% and lost ***
    2014-06-01 08:11:07 *** angel (781321) [#1220009560] bet 4.096 BTC at 49.5% and lost ***
    2014-06-01 08:11:10 *** angel (781321) [#1220009649] bet 3.7319836 BTC at 49.5% and lost ***

Reverse martingale wins:

    2014-06-01 08:24:11 *** angel (781321) [#1220035199] bet 0.91221859 BTC at 49.5% and won 0.91221859 BTC ***
    2014-06-01 08:24:13 *** angel (781321) [#1220035298] bet 0.91221859 BTC at 49.5% and won 0.91221859 BTC ***
    2014-06-01 08:24:46 *** angel (781321) [#1220036343] bet 2.73666192 BTC at 49.5% and won 2.73666192 BTC ***
    2014-06-01 08:25:54 *** angel (781321) [#1220038473] bet 2.73931514 BTC at 49.5% and won 2.73931514 BTC ***
    2014-06-01 08:26:38 chat: <angel> ok
    2014-06-01 08:26:50 chat: <angel> go big or go home
    2014-06-01 08:26:55 chat: <angel>
    2014-06-01 08:27:18 *** angel (781321) [#1220041208] bet 4.1089753 BTC at 49.5% and won 4.10897530 BTC ***
    2014-06-01 08:27:25 chat: <angel> DONE
    2014-06-01 08:27:37 chat: <angel> #degenlife

[Jbanna]

dandiilion here,
doog man, it was an interesting bet. I calculated that the statistically potential streak loss was 27.64 by taking sample data from occurring bets that were happening live. After getting your data of actual % of 49.5% bets the number came out to be 27.82. To get these numbers I used log(# of losses @ 49.5%) / -log (0.505). So basically, it seemed to me that there was a 50% chance that a 27.64 had occurred, and we'd seen the 27 already! Anything higher than 27.64 was more of a favorable advantage in me and BAC's favor. I ran more numbers and came out with the probability that we had a 70% chance to win.
Was never the best with stats.. Can anybody verify that this was the correct formula to use? Would be a shame that I lost 30k due to a stupid math error along the way. Are there any mathematicians looking in here that can figure this doozy out? I can't quite seem to get a good night's sleep until this is resolved...

oh, and double or nothin?

[dooglus]

oh, and double or nothin?

I never intended to bet that much.  I just realised I had to put some kind of a cap on things, and picked 50 BTC pretty much at random.  I didn't expect anyone to go that high.

I don't think it looks good to JD investors to have me looking like an out-of-control gambler.  I imagine it is concerning for them.  So I'm going to have to pass, and keep my gambling in check.

[Jbanna]

lol was just kidding. I'd trust doog with the credentials to my bank account. I myself am the crazy gambler.

[boraf]

Trend trading works similar to reverse martingale with condition of liquidating all trades after certain range drop in price.

It has proven to work well on strong directional market.

 

[NakedPony]

Math is not very helpfull sometimes.

I had ~85 000 bets with Martingale, made it from 0.2 to 7.5 BTC.
At this period I never had more than 12 losses in a row.
Which is only 1/2^13 = 1/4096 chance.
So, I should have at least 10 of these deadly strikes.

I've played for 2 weeks,  dreaming about new car   (lol)
And then .. 17 losses in a row. At least I was clever enough to stop with big bets.
And - 20 losses in a row in next 5 mins..
All the next week I had only losses.
So - its really BIG NUMBERS, and your luck on the first place.

[boumalo]

Math is not very helpfull sometimes.

I had ~85 000 bets with Martingale, made it from 0.2 to 7.5 BTC.
At this period I never had more than 12 losses in a row.
Which is only 1/2^13 = 1/4096 chance.
So, I should have at least 10 of these deadly strikes.

I've played for 2 weeks,  dreaming about new car   (lol)
And then .. 17 losses in a row. At least I was clever enough to stop with big bets.
And - 20 losses in a row in next 5 mins..
All the next week I had only losses.
So - its really BIG NUMBERS, and your luck on the first place.

It must be fun to win at martingale and it seems like the system works when you win, but as you said it just mean you are not unlucky at first and then that you are lucky

oh, and double or nothin?

I never intended to bet that much.  I just realised I had to put some kind of a cap on things, and picked 50 BTC pretty much at random.  I didn't expect anyone to go that high.

I don't think it looks good to JD investors to have me looking like an out-of-control gambler.  I imagine it is concerning for them.  So I'm going to have to pass, and keep my gambling in check.

You must be interested in gambling since you created a gambling site; as long as your bets are not out of control and fair the JD investors probably don't mind since they are open minded

[BayAreaCoins]

Dandiilion's $5000 has been paid out.

[dooglus]

Martingale should never be used in the long run
Few bets, maybe a hundred ok...anything after that it's a no
You'll find success when you first start gambling with martingale...it's a matter of cashing out at the right time

It really depends how many times you can afford to double.

If you have 7 and start at 1, you can only afford two doubles (1 + 2 + 4 = 7) before you bust.
If you have 7 and start at 0.001, you can afford 12 or 13 doubles, and so will probably last longer before you bust.

However you play, the odds aren't in your favour.  Martingale doesn't change that fact.

[dooglus]

However you play, the odds aren't in your favour.  Martingale doesn't change that fact.

Of course, that doesn't mean you can't win...

We recently had a guy playing at Just-Dice.  He played on 4 different accounts, calling himself steven (user 522621), march (718719) and charlie (755590 and 764137).  He made around 74,000 bets, and ended up something like 90 BTC in profit!

Here's a chart of his play.  It took him about 21,000 bets to get into his stride and start using martingale, but once he did, it went really well.  This is typical of how martingale player charts look, only the drops are usually bigger than the rises:



Then one day he stopped playing and we've not seen him since...  Good for charlie/march/steven! 

[Farmer17]

Then one day he stopped playing and we've not seen him since...  Good for charlie/march/steven! 

Interesting. The player lost ~90 btc before using martingale, and in the end got 90 btc profits, and he knew when to stop.

Just curious, how did you link the accounts?
IP? deposit address? or withdrawal address?

[dooglus]

Interesting. The player lost ~90 btc before using martingale, and in the end got 90 btc profits, and he knew when to stop.

Just curious, how did you link the accounts?
IP? deposit address? or withdrawal address?

He linked the accounts for us after I spotted that the players were playing very similarly and asked him about it.

I suspect he didn't know when to stop, kept playing, lost, and didn't keep talking to us, so we didn't get to link the losing accounts.

It's often the way that players like to talk when they're winning but go quiet when they lose, making it harder to know who is really still 'up'.

[dooglus]

extremely interesting...when was this?
90BTC is a heck of an amount to win with today's prices, let alone being 90BTC down and continuing to play!

We see quite a few players with lots of coins to throw around.  I guess there are lots of early adoptors still around.

steven played from mid Feb to the end of Apr.
march player the first half of May.
the 1st charlie played in the middle of May and the 2nd at the end of May.

We've seen bigger martingale wins.  A guy called something like Balotelli took us for hundreds of coins.  Diablo too.

I'll dig up their charts if I can find their userids.

[dooglus]

We've seen bigger martingale wins.  A guy called something like Balotelli took us for hundreds of coins.  Diablo too.

I'll dig up their charts if I can find their userids.

Here's Balotelli.  I was forgetting - he switched from using martingale to using what he called "nakowa style" and lost everything very quickly: