Martingale revisited

[Betwrong]

~

I have no explanation how it is ever possible, but winning chances below 37% bring about an exponential surge in variance, and martingale no longer remains safe. I can't come up with any plausible reason for this phenomenon but I've read about it before and can confirm it (call it broscience or a gambling street wisdom of sorts, or whatever)

I think it's simply a question of proper adjustment. If you lowered your "increase on loss" exactly in accordance with how the win chance was lowered, your strategy would work just fine

Without any additional info, proper adjustment assumes linear change. This is not what I observed. If variance changes abruptly (like spikes exponentially), and you don't know by how much exactly, there is no way to lower the increase on loss that would match this spike. Finding it out empirically would probably mean busting in the process

I think, you can bust in the process anyway , but what I meant was that with a proper adjustment you could create a betting strategy that would work exactly like your previous one in regards to the probability of busting.

For example, if you were betting with 40% win chance and increasing 160% on loss, you'd have to switch to 80% increase on loss when betting with 20% win chance; to 40% increase on loss when betting with 10% win chance, and so on.

[1714736117]

1714736117

[1714736117]

1714736117

[1714736117]

1714736117

[deisik]

With such strategy martingale is useless. Any martingale, with any "improvements"

Then you are on your own

Without any additional info, proper adjustment assumes linear change. This is not what I observed. If variance changes abruptly (like spikes exponentially), and you don't know by how much exactly, there is no way to lower the increase on loss that would match this spike. Finding it out empirically would probably mean busting in the process

I think, you can bust in the process anyway

Okay, let's now ignore traffic lights cuz we are still going to kick the bucket somewhere down the road, right?

For example, if you were betting with 40% win chance and increasing 160% on loss, you'd have to switch to 80% increase on loss when betting with 20% win chance; to 40% increase on loss when betting with 10% win chance, and so on

You are now suggesting what I already went through in my previous post

You are implicitly assuming a linear change. But if there is a linear change, then increasing 160% on loss with a 40% win chance should be the same in the long run as increasing 80% with a 20% win chance, which is otherwise known as six of one and half a dozen of the other. However, the change is non-linear. It basically means you can bust sooner with an 80% increase instead of 160% with a win chance only a few percentage points less. But if you want to check it for yourself, go for it since who am I to stop you?

[BITCOIN4X]

Yes, but as i wrote above, if you for example spinning dice with 50 chance to win, you will not get in average "one win/one lose/ one win / one lose"
but rather you can get 10 loses in a row and 3 wins after and then again 2 loses. With such strategy martingale is useless. Any martingale, with any "improvements".

5 times in a row applying this technique, but I failed and lost. No matter how good the technique is, we also have to have a lucky day in gambling. Technique will only be a medium that helps us get luck.

I believe that there is no perfect gambling technique, even though someone can prove that it works but that doesn't apply to everyone because we have to have luck.

[deisik]

5 times in a row applying this technique, but I failed and lost. No matter how good the technique is, we also have to have a lucky day in gambling. Technique will only be a medium that helps us get luck

You should probably stay away from using martingale

By and large, it all comes down to properly balancing your chances of winning versus losing so that on a specific timeframe (a month, a year, a decade) your chances of coming through and making it would be higher than busting and losing your balance (actually, way higher, for a mighty safety margin). If you can't do that, martingale will be a losing strategy for you. Long story short, blindly applying it and hoping for the best will likely end in a disaster. The good news is that it is the same as with any other complicated strategy out there that requires knowledge, expertise, and genuine understanding of how things hang in practice

I believe that there is no perfect gambling technique, even though someone can prove that it works but that doesn't apply to everyone because we have to have luck

With a thoroughly thought-out martingale strategy you must be not so much lucky to win as unlucky to bust

[BITCOIN4X]

The good news is that it is the same as with any other complicated strategy out there that requires knowledge, expertise, and genuine understanding of how things hang in practice

That must be true, it all depends on the knowledge and experience practiced in the game. Maybe for you this strategy is useful, but not for me because of lack of knowledge when applying it.

Meanwhile, dice is my favorite game besides sports gambling. I can only win a few bets or more if the winning percentage is set at 60% or more. That is the best opportunity for me so far on dice. At least I can produce something even though its not much.

[shoreno]

The good news is that it is the same as with any other complicated strategy out there that requires knowledge, expertise, and genuine understanding of how things hang in practice

That must be true, it all depends on the knowledge and experience practiced in the game. Maybe for you this strategy is useful, but not for me because of lack of knowledge when applying it.

Meanwhile, dice is my favorite game besides sports gambling. I can only win a few bets or more if the winning percentage is set at 60% or more. That is the best opportunity for me so far on dice. At least I can produce something even though its not much.

60 percent is or more is more winnable compare to lower percents but you can only earn less with it and if you want to earn big you need to adjust and up your bets but be careful because higher percent chance are also deadly  . you can loose imdiately upon upping your bets   .  so many newbies i see play that strat with bigger bets and sadly they end up easily  .   this is why i play only with low chance win rate but with martingale strat because this makes me last long and still be able to hit my targeted multi with a nice profit

[Betwrong]

With such strategy martingale is useless. Any martingale, with any "improvements"

Then you are on your own

Without any additional info, proper adjustment assumes linear change. This is not what I observed. If variance changes abruptly (like spikes exponentially), and you don't know by how much exactly, there is no way to lower the increase on loss that would match this spike. Finding it out empirically would probably mean busting in the process

I think, you can bust in the process anyway

Okay, let's now ignore traffic lights cuz we are still going to kick the bucket somewhere down the road, right?

For example, if you were betting with 40% win chance and increasing 160% on loss, you'd have to switch to 80% increase on loss when betting with 20% win chance; to 40% increase on loss when betting with 10% win chance, and so on

You are now suggesting what I already went through in my previous post

You are implicitly assuming a linear change. But if there is a linear change, then increasing 160% on loss with a 40% win chance should be the same in the long run as increasing 80% with a 20% win chance, which is otherwise known as six of one and half a dozen of the other. However, the change is non-linear. It basically means you can bust sooner with an 80% increase instead of 160% with a win chance only a few percentage points less. But if you want to check it for yourself, go for it since who am I to stop you?

In fact I tried it quite some times already, and the result was always a win. Below are some latest examples:







But I wouldn't try this strategy with BTC though.

I understand that 10% increase on loss is a bit too much for betting with 1% win chance, and that I was just lucky that I didn't bust. With 2% increase on loss it looks more like your strategy



with which you can catch a nice outlier before getting busted.

[deisik]

with which you can catch a nice outlier before getting busted

You are thinking in correct direction

However, you only made so many rolls. As I suspect, to reveal a particular pattern of variance at so low win chances, you would have to literally make billions of bets and not bust in the process. If you bust, all your effort will be spent in vain. With win chances close to 50% we can be damn sure that the variance of variance (let's call it second-order variance) is less random that the first-order variance (this also confronts us with the question as to how truly random is random)

To cut to the chase, with lower multipliers you can safely assume that you will see fewer longer losing streaks than shorter ones. However, with small chances second-order variance may get out of hand, and you will see an entirely different picture. In practice, it basically means that you can't use the approach described here with very small odds as you are set to hit a very long losing streak kind of "all of a sudden" that would likely wipe you out

[Betwrong]

with which you can catch a nice outlier before getting busted

You are thinking in correct direction

However, you only made so many rolls. As I suspect, to reveal a particular pattern of variance at so low win chances, you would have to literally make billions of bets and not bust in the process. If you bust, all your effort will be spent in vain. With win chances close to 50% we can be damn sure that the variance of variance (let's call it second-order variance) is less random that the first-order variance (this also confronts us with the question as to how truly random is random)

To cut to the chase, with lower multipliers you can safely assume that you will see fewer longer losing streaks than shorter ones. However, with small chances second-order variance may get out of hand, and you will see an entirely different picture. In practice, it basically means that you can't use the approach described here with very small odds as you are set to hit a very long losing streak kind of "all of a sudden" that would likely wipe you out

It's true that with higher multipliers we can expect longer losing streaks, but don't you think that we can estimate how much longer they can be?

For example, on a dice site with 1% house edge:

Odds of losing 42 bets in a row at 40% win chance: Once every 2,078,009,285.66 bets.

I don't know about you, but I would consider a setup when you are immune to 42 losses in a row a pretty safe one in this case.

Now we should find out for how many losses in a row we must be immune when betting with, say, 1% win chance to have the same expectancy of busting(once in 2 billion bets), and that's it, don't you think?

[deisik]

Now we should find out for how many losses in a row we must be immune when betting with, say, 1% win chance to have the same expectancy of busting(once in 2 billion bets), and that's it, don't you think?

That's the problem that I'm trying to address

You again assume linear relationship but you miss the larger picture. It is not about averages at all (in this case) as the whole thing with the approach practiced and discussed here is about outliers. However, outliers are the opposite of averages. Put differently, you can't apply stats to outliers as stats is just a substitute word or placeholder for averages. In practice, you bust specifically because you encounter an outlier that you can't cope with, and averages, whatever they may say to the contrary, won't help you

It just happens that with chances less than 37% (give or take) outliers become extreme even if averages still remain linear on a long enough timeframe. I don't know why it happens but it does. To sum it up, the entire story is about variance and how you deal with it

[Betwrong]

~
It just happens that with chances less than 37% (give or take) outliers become extreme even if averages still remain linear on a long enough timeframe. I don't know why it happens but it does. To sum it up, the entire story is about variance and how you deal with it

Are you sure you've done enough tests with chances less than 37% to make such conclusions? As far as I know, you've done millions of rolls with 37% win chance. But how many rolls with, say, 1% win chance were performed?

Maybe the variance from standard deviation can be higher with lower win chance, but I still feel it can be tamed with a large enough bankroll. Or, do you think that even 1 billion DOGE wouldn't be enough for that?

[deisik]

~
It just happens that with chances less than 37% (give or take) outliers become extreme even if averages still remain linear on a long enough timeframe. I don't know why it happens but it does. To sum it up, the entire story is about variance and how you deal with it

Are you sure you've done enough tests with chances less than 37% to make such conclusions? As far as I know, you've done millions of rolls with 37% win chance. But how many rolls with, say, 1% win chance were performed?

Maybe the variance from standard deviation can be higher with lower win chance, but I still feel it can be tamed with a large enough bankroll. Or, do you think that even 1 billion DOGE wouldn't be enough for that?

I think one billion bets will be enough to see the pattern

Unless you bust in the process, of course. Indeed, you can set the increase on loss to 0, so that to avoid busting altogether, but it may take too long. Anyway, we don't necessarily need to run so many bets to draw some conclusions, and it is probably worth a try after all (I may run some tests)

I started out at exactly 37% win chance, then I decreased it to 36% (to win more, obviously), and I almost busted. After that I only increased the odds as my profits grew (I finished with a 42% win chance, to be exact). Ironically, I had first read about that magic number (37%) in the freebitco.in thread somewhere around here, and I discarded it as a mere superstition

[Betwrong]

~
It just happens that with chances less than 37% (give or take) outliers become extreme even if averages still remain linear on a long enough timeframe. I don't know why it happens but it does. To sum it up, the entire story is about variance and how you deal with it

Are you sure you've done enough tests with chances less than 37% to make such conclusions? As far as I know, you've done millions of rolls with 37% win chance. But how many rolls with, say, 1% win chance were performed?

Maybe the variance from standard deviation can be higher with lower win chance, but I still feel it can be tamed with a large enough bankroll. Or, do you think that even 1 billion DOGE wouldn't be enough for that?

I think one billion bets will be enough to see the pattern

Unless you bust in the process, of course. Indeed, you can set the increase on loss to 0, so that to avoid busting altogether, but it may take too long. Anyway, we don't necessarily need to run so many bets to draw some conclusions, and it is probably worth a try after all (I may run some tests)

I started out at exactly 37% win chance, then I decreased it to 36% (to win more, obviously), and I almost busted. After that I only increased the odds as my profits grew (I finished with a 42% win chance, to be exact). Ironically, I had first read about that magic number (37%) in the freebitco.in thread somewhere around here, and I discarded it as a mere superstition

Hm, interesting. And I thought there were no magic numbers in math!

But seriously, we need more those tests, I guess. For example, with my type of betting, my magic number is 10%. I mean, if I play dice with higher than 10% win chance, I always lose in the end. That's why I prefer something like 1%, 0.1% or even 0.01%. Like I won several day ago:



No big money, but surely big fun!

[deisik]

I started out at exactly 37% win chance, then I decreased it to 36% (to win more, obviously), and I almost busted. After that I only increased the odds as my profits grew (I finished with a 42% win chance, to be exact). Ironically, I had first read about that magic number (37%) in the freebitco.in thread somewhere around here, and I discarded it as a mere superstition

Hm, interesting. And I thought there were no magic numbers in math!

I'm already running certain tests

But they will take some time as the analysis will require a pretty impressive amount of bets (aka sample size) on the order of at least 100k rolls per each percentage. After I'm done, I will likely start off a separate thread about this magic number, as well as try to find the original post in the freebitco.in thread where 37% had been first mentioned in this context (I remember I replied to it). Methinks, it might be a game-changer or even an eye-opener of sorts for a lot of martingalers if these assumptions or superstitions prove true after all

[swogerino]

The good news is that it is the same as with any other complicated strategy out there that requires knowledge, expertise, and genuine understanding of how things hang in practice

That must be true, it all depends on the knowledge and experience practiced in the game. Maybe for you this strategy is useful, but not for me because of lack of knowledge when applying it.

Meanwhile, dice is my favorite game besides sports gambling. I can only win a few bets or more if the winning percentage is set at 60% or more. That is the best opportunity for me so far on dice. At least I can produce something even though its not much.

There is no need for any experience when trying martingale,it is all boiled down to doubling your bet after every lost bet.In theory it should work if you have an infinite bankroll then you can double your bet after every losing bet and you will eventually hit a winning bet.The problem is that this is only theory,for the practice there a lot of answers not only in this thread but also many others that martingale it will fail.

[deisik]

The good news is that it is the same as with any other complicated strategy out there that requires knowledge, expertise, and genuine understanding of how things hang in practice

That must be true, it all depends on the knowledge and experience practiced in the game. Maybe for you this strategy is useful, but not for me because of lack of knowledge when applying it.

Meanwhile, dice is my favorite game besides sports gambling. I can only win a few bets or more if the winning percentage is set at 60% or more. That is the best opportunity for me so far on dice. At least I can produce something even though its not much.

There is no need for any experience when trying martingale,it is all boiled down to doubling your bet after every lost bet.In theory it should work if you have an infinite bankroll

It is ironic how fast people are to point out that you need an infinite bankroll for martingale to work out

However, they forget that this holds true only for as infinite timeframes. As soon as we are talking about particular spans or periods of time (like a month or a year), which is always the case in real life, what it actually boils down to are the odds of surviving a losing streak of a certain number of rolls for this strategy to work out on average

It is not even so much interesting to explain the specifics of this approach (cause they are pretty simple, and you don't exactly need a master's degree to figure them out) as to get to understanding why people go on and on about martingale being a losing strategy. Any strategy in any domain will end up with a disaster if used mindlessly

[Betwrong]

I started out at exactly 37% win chance, then I decreased it to 36% (to win more, obviously), and I almost busted. After that I only increased the odds as my profits grew (I finished with a 42% win chance, to be exact). Ironically, I had first read about that magic number (37%) in the freebitco.in thread somewhere around here, and I discarded it as a mere superstition

Hm, interesting. And I thought there were no magic numbers in math!

I'm already running certain tests

But they will take some time as the analysis will require a pretty impressive amount of bets (aka sample size) on the order of at least 100k rolls per each percentage. After I'm done, I will likely start off a separate thread about this magic number, as well as try to find the original post in the freebitco.in thread where 37% had been first mentioned in this context (I remember I replied to it). Methinks, it might be a game-changer or even an eye-opener of sorts for a lot of martingalers if these assumptions or superstitions prove true after all

Very interesting. Are you going to cover ranges below 1% win chance as well? Asking because those are my fav ranges. I can't say that I'm in positive profit overall, but I do win those bets pretty often, like this one, 3 days ago:



Regarding 37% win chance being the lowest one can run martingale successfully with, I thought the same thing when I was only starting with dice. At first I was betting with 49.50% many times, not even knowing that changing the win chance was possible. Then, after I learned how to change it, I was playing with various win chances and was like, "Man, what kind of fool would bet with lower than 20% win chance?!" It appears that I have become that fool eventually.

[deisik]

I started out at exactly 37% win chance, then I decreased it to 36% (to win more, obviously), and I almost busted. After that I only increased the odds as my profits grew (I finished with a 42% win chance, to be exact). Ironically, I had first read about that magic number (37%) in the freebitco.in thread somewhere around here, and I discarded it as a mere superstition

Hm, interesting. And I thought there were no magic numbers in math!

I'm already running certain tests

But they will take some time as the analysis will require a pretty impressive amount of bets (aka sample size) on the order of at least 100k rolls per each percentage. After I'm done, I will likely start off a separate thread about this magic number, as well as try to find the original post in the freebitco.in thread where 37% had been first mentioned in this context (I remember I replied to it). Methinks, it might be a game-changer or even an eye-opener of sorts for a lot of martingalers if these assumptions or superstitions prove true after all

Very interesting. Are you going to cover ranges below 1% win chance as well?

It didn't work out

Long story short, the results were totally inconclusive. I started with exactly 37% win chance and almost instantly hit a losing streak of 26 rolls:



Then I went for lower odds running series of 100k rolls (99999, to be exact), and I didn't encounter as long or longer streaks. It basically means that the sample size of 100k tests is just too small. It seems the size should be well over 1M rolls but it is impossible to test in a real casino. Indeed, we can still use an emulator but if the issue is with a hardware RNG, we won't see anything there, either

[Betwrong]

~
Then I went for lower odds running series of 100k rolls (99999, to be exact), and I didn't encounter as long or longer streaks. It basically means that the sample size of 100k tests is just too small. It seems the size should be well over 1M rolls but it is impossible to test in a real casino. Indeed, we can still use an emulator but if the issue is with a hardware RNG, we won't see anything there, either

Those were my thoughts too, when I was reading your post, but then I decided to not discourage you because imo any test is better than none. 1M can also be not enough, but definitely better than 100k.

There are sites showing probabilities of winning with particular pocket cards, based on millions real games played, not simulated ones, for Texas hold 'em poker. Dice sites where tens of billions bets were made, could, theoretically, provide similar stats too. But there is no such data on the Internet, as far as I can tell.

[STT]

Everyone should avoid using martingale, it lends no advantage in usage and amplifys risk in the hope of winning it all back.    Thats a normal mistaken strategy by most new bettors anyway, its better to have an allotted budget each day and not try to chase after losses just separate each day or attempt at gambling and hope to learn something along the way.
  Theres one good reason why martingale strategy always pops up and over decades has been repeated, its simple and it appeals to the base instinct of a beginner.   Any easy strategy is my preference but its inferior to some discipline and actual learning by experience.