Does the martingale system really work? | Page 18 | Metals Mine

Options

Go to
Permalinks
Print Thread

Similar Threads

Martingale, Anti-martingale, and Compounding 40 replies

Martingale, Reverse Martingale, Modified Martingale, Maths 1 reply

How to change this Martingale to Reverse Martingale ? 3 replies

Martingale vs. Non Martingale (Simplified RoR vs Profit) 1 reply

Does this simple system by Richard Donchian really work? 134 replies

Attachments: Does the martingale system really work?
Exit Attachments

Does the martingale system really work?

  • #341
  • Nov 28, 2013 2:04pm Nov 28, 2013 2:04pm
Quoting FX_James
master kiwa's account is around 50% loss now, his pivot martingale trading Technics not working as well
Does that appear to be a martingale system to you?
Attached Image (click to enlarge)
Click to Enlarge Name: where.PNG Size: 37 KB
  • #342
  • Nov 28, 2013 2:06pm Nov 28, 2013 2:06pm
Quoting PipMeUp
{quote} If p is your winrate, NOT getting a single string of k losers (you say k=4) out of n trades (say n=200, approx one trade per session in a year) is given by {image} Certainly not the most sexy formula I agree...
Thank you for sharing PipmeUp. It's just the big time losers in forex seem to be pessimistic about everything in forex. Which is why they only focus on the negative, so IF the negative occurs they can FINALLY be correct about something. To bad they get no financial gain from their post.
  • #343
  • Nov 28, 2013 2:11pm Nov 28, 2013 2:11pm
Quoting hanover
{quote} Proof that mart...
Ok I undestand your point , i know that it wont increase your expected profit, but it does help you avoid losses more effectively than the naked betting since you'd have only a
0.01406408618241 % of losing ^7x times the sum, which is relatively low, but it will change the distribution of it in a short time span.What i mean is, that if we use martingale as a gambling method.We can do the following:

Ignore leverage and lagging of markets for now.Lets just simplify the method.
For example a 1 shot to 1 million dollars attempt.Start with X dollars and hedge yourself up to 1 million then stop.With the 0.01% chance of losing you actually have better chances than winning a prize in lotto 6/49
So you have a better chance of winning a lottery prize (not the jackpot) than losing with the system.

So it would be a 1 shot system, repeat martingale until you reach 1m$ or blow up your account, but eventually you would hit 1m$ quicker than blow your account up.But the key is to never repeat it or start with a small sum you can afford to lose.Of course you can't make sustainable profit with it, because its called gambling, but you can avoid losses more efficiently with martingale.

So martingale is a gambling strategy, no wonder since it originated from roulette, and secondly, its a 1 shot system,you cant make durable profit, but you can get much money with it if your luck is on your side, more efficiently than with any other gambling system on earth?

Why? Simple, since none of the casino games have higher odds than 49%, but in forex you can have higher odds (theoretically), although i didnt experienced it, i`m sure the best traders out there did.
"There's a sucker born every minute" - P.T. Barnum
  • #344
  • Nov 28, 2013 2:17pm Nov 28, 2013 2:17pm
Here’s an exception: where a capped martingale can give you an advantage

My previous post (#338) holds true when the trials are independent, i.e. each trial (trade) has no bearing on the outcome of other trials. Or in other words, when the probability of each trial is assumed to be equal.

However, in a forex setting, it’s perfectly possible that some trade setups have a higher probability of success than others, and this justifies a higher bet size. (Analogy: That’s how blackjack pros beat casinos: higher bets when there are more high cards remaining in the deck, and the probability is in their favor). Another example might be where markets deliver a period that favors your system, causing wins to cluster, or losses to cluster when conditions are unfavorable (e.g. a trend following system prospers in a trending market, and loses money in a sideways market). Statisticians call this clustering phenomenon ‘serial correlation’.

In summary, then, the idea is to have higher position sizes for higher probability setups; but of course you need to know which setups are superior, before placing the trades. Where many traders go wrong with martingale is that they use it to try to recover recent losses. There is no mathematical basis for this, because the market takes no cognizance of the trader's personal P/L. It is simply gambling. Moreover, recent losses have no greater effect on eventual overall P/L than any historical losses.

The best exploitation of serial correlation that I’ve ever seen is Forexhard’s SSBO system. As each fakeout occurs from a consolidation zone, the probability of a good size breakout increases, justifying a significant increase in position size. The concept of breakouts from consolidation is among the most robust of trading methods, assuming that one can identify the highest probability situations, and SSBO employs a mechanism that does exactly that.

However, all trading systems ultimately are a balance between maximizing income and minimizing risk of ruin (intolerable or irretrievable drawdown). Increasing position sizes causes risk of ruin to increase exponentially. Hence, while it’s justifiable to increase position size in favorable situations, the martingale progression (1,2,4,8,16,….) is probably too steep for most situations where serial correlation exists. The betting progression should remain approximately commensurate with the increase in probability. In other words, to justify a bet that’s 16x the original bet size, the probability of a successful trade should be ~ 16x as great as it was originally, which (for most systems) is very unlikely to be the case.

The attached XLS allows you to test serial correlation. You can change the probability of a winning trade, following a win or a loss, over a maximum of 6 trials, and then test different MM systems accordingly. Feel welcome to extend it to more than 6 trials, if you wish.

David
Attached File(s)
File Type: xls Martingale comparison testing.xls   115 KB | 612 downloads
  • #345
  • Nov 28, 2013 2:17pm Nov 28, 2013 2:17pm
Quoting Master_kiwa
{quote} Martingale isn't used to change expectancy. It is only used to recover your string of losses much quicker. Which is why talking about win %, or anything regarding % is pointless. As it is mearly an avg. It NEVER translates into real life situations. I could have a 86 % win rate, but the amount of trades actually placed will determine if my account margin calls or not. If I place 1000 trades a 86% win rate means I closed 860 trades for a win. Yet what if my first 140 trades were all losers. That would mean martingale won't work. So what should...

You wont have situations where your first 140 trades will be losers and the other 860 winners in the first place(even if you are willingly losing trades), thats why i drawed the probability distribution curve, the probabilities are spread out evenly in a gaussian curve shape, where the chances that 140 losing trades will be in a row in an 50-50 system are 0.5^140 = 7.1746481373430634031294954664444e-43 which is probably way past the 10th sigma zone, so the probability of the Earth blowing up has a better chance than that scenario so your argument is invalid
"There's a sucker born every minute" - P.T. Barnum
  • #346
  • Nov 28, 2013 2:30pm Nov 28, 2013 2:30pm
Quoting hanover
Here’s an exception: where a capped martingale can give you an advantage My previous post (#338) holds true when the trials are independent, i.e. each trial (trade) has no bearing on the outcome of other trials. Or in other words, when the probability of each trial is assumed to be equal. However, in a forex setting, it’s perfectly possible that some trade setups have a higher probability of success than others, and this justifies a higher bet size. (Analogy: That’s how blackjack pros beat casinos: higher bets when there are more high cards remaining...
Yes i understand you man but do you understand me? This martingale would work in a 1 shot system, so in certain cases it would end up in failure, which is fine,it's just gambling with, but if we set a goal of lets say 1m $ and hedge towards that goal, we still avoid the losing strikes more efficiently with martingale, since if we'd use naked system with 51%, the probability distribution would still suggest that unexpected things can happen, a win rate is not a constant, its a variable.

We must work with the distribution curve, and if we'd use the normal betting on a 51% system, after 1000 trades we still have a higher chance of ending negative than with martingale, since we hedge and avoid losses way better that effectively.

What you calculated there works if we take infinite amount of trades, where yes the percent will converge to the mean , but we dont have unlimited trades we may take 10000 trades in total in a lifetime, so we also have to calculate the odds of being wrong in a limited set, and that in my understanding is more effective if we hedge our position with a martingale or kelly or whatever.We wont gain an edge, but we avoid losses more effectively in a limited number of trades than on naked.Not the martingale gives the edge, its the system on which we use the martingale, martingale just lets our goal to be earned quicker than without,and if its a 1$ goal then its certainly more effective than without.And its also more effective than on any casino game.
"There's a sucker born every minute" - P.T. Barnum
  • #347
  • Nov 28, 2013 2:30pm Nov 28, 2013 2:30pm
Quoting Proximus
{quote} Ok I undestand your point ,...... But the key is to never repeat it or start with a small sum you can afford to lose.Of course you can't make sustainable profit with it, because its called gambling, .......
Agreed, it's gambling. I already suggested something like this in post #333 (start with a small amount that you can afford to lose). But surely the goal in trading is to exploit a proven edge, and minimize risk of ruin? That's exactly what casinos do, and they profit handsomely: small edge, massive capital backing relative to table maximums (equates to conservative MM), croupiers must follow fixed rules (equates to consistency and discipline in trading).

@everybody: Here's yet another XLS that I've posted many times before. It allows you to calculate the probability of X consecutive losses over a sequence of Y trades, for a given win rate. It uses a similar formula to the one quoted by PipMeUp in post #340.
Attached File(s)
File Type: xls Probability of X consecutive losing trades.xls   25 KB | 636 downloads
  • #348
  • Nov 28, 2013 2:41pm Nov 28, 2013 2:41pm
Quoting Proximus
{quote} Yes i understand you man but do you understand me? This martingale would work in a 1 shot system, so in certain cases it would end up in failure, which is fine, but if we set a goal of lets say 1m $ and hedge towards that goal, we still avoid the losing strikes more efficiently with martingale, since if we'd use naked system with 51%, the probability distribution would still suggest that unexpected things can happen, a win rate is not a constant, its a variable. We must work with the distribution curve, and if we'd use the normal betting...
I still stand by what I said in post #338, i.e. I agree that martingale improves win rate for any +EV system, but the losses (if they do occur) will be commensurately bigger. The higher the number of martingale levels, the greater this effect. For example, you could have a system that has a > 99.9% probability of generating a winning sequence (due to martingale's recovery mechanism) but a < 0.01% probability of a massive loss.

Re calculating the odds of being wrong:

IMO no amount of testing can determine (without a prohibitively high amount of statistical error) the likelihood of whether an extremely low probability event will occur sooner, rather than later. Hence I don't believe that an uncapped martingale can be tested accurately. You can have 12 months without disaster, 2 years, 5 years, whatever. But the 'death sequence' needs to occur only once. I guess it's a bit like saying "I've been driving for 20 years without accident, therefore I'll never have one". Or "I'm only half as likely to have one as a person who's been accident free for 10 years". The flaws in such reasoning are obvious.
  • #349
  • Edited Nov 29, 2013 10:42am Nov 28, 2013 2:46pm | Edited Nov 29, 2013 10:42am
Quoting Master_kiwa
{quote} Martingale isn't used to change expectancy. It is only used to recover your string of losses much quicker.
Agreed, but it's at the risk of making the losses exponentially bigger. I'm not saying that "martingale doesn't work", I'm just saying that it's a trade off between recovery rate and risk that comes down to personal preference. But since martingale increases risk or ruin without any commensurate gain in expectancy: it's not for me. I'm happier to let a positive expectancy system gradually recover losses over time, and with significantly less risk.

Now it's getting to the point where I'm merely typing the same material repeatedly, so I'll end it here. Thanks for the lively discussion.
  • #350
  • Nov 28, 2013 2:47pm Nov 28, 2013 2:47pm
Quoting hanover
{quote} Agreed, it's gambling. I already suggested something like this in post #333 (start with a small amount that you can afford to lose). But surely the goal in trading is to exploit a proven edge, and minimize risk of ruin? That's exactly what casinos do, and they profit handsomely: small edge, massive capital backing relative to table maximums (equates to conservative MM), croupiers must follow fixed rules (equates to consistency and discipline in trading). @everybody: Here's yet another XLS that I've posted many times before. It allows you...

Well finally something we can agree upon, and i like to emphasize again i never said martingale will improve winrate, all i said is that it is more effective statistically visioned than naked trading, of course viewing the risk its bad, but when using martingale you have to know that you can lose all your cash.So its gambling true, but it is the ultimate gambling if you apply to forex, whereas in forex you can tip the balance of the 50-50 where in any other game you cant.I`m not really into blackjack, i played it once but i`m not familiar with its probabilities but i heard that its better than the casino average.

But i would not recommend martingale to anyone, i just wrote all my posts as a though experiment that ,yes its possible to a certain case to hedge your losses effectively, but i`m personally more risk aware to do such thing.And i`m mostly unlucky so i would always be on the tails of the distribution curve

Also i dont think that spreadsheet is correct, when you calculate winrate and distributions in forex you want to view it as a dependent system not an independent one since trend/reversal, the previous candle affects the nexts one's probability so you cant view it as an independent one, so the formula p^n is better where p is the probability and n is the lose streak.
"There's a sucker born every minute" - P.T. Barnum
  • #351
  • Edited Nov 29, 2013 10:43am Nov 28, 2013 3:07pm | Edited Nov 29, 2013 10:43am
Quoting Proximus
i like to emphasize again i never said martingale will improve winrate
My bad. I meant that winrate is improved if we view each sequence of trades as a single entity (the "percentage of positive outcomes" in the screenshots ion post #338). I should have said that martingale improves recovery rate rather than win rate.

Quoting Proximus
i dont think that spreadsheet is correct, when you calculate winrate and distributions in forex you want to view it as a dependent system not an independent one since trend/reversal, the previous candle affects the nexts one's probability so you cant view it as an independent one, so the formula p^n is better where p is the probability and n is the lose streak.
IMO degree of dependency depends on how the system chooses its setups. Most of the EAs I've written treat each 'fresh' setup as being independent of the others.

Thanks for the discussion.
  • #352
  • Nov 28, 2013 3:43pm Nov 28, 2013 3:43pm
Quoting hanover
{quote} My bad. I meant that winrate is improved if we view the sequence as a single entity (the "percentage of positive outcomes" in the screenshots ion post #338). I should have said that martingale improves recovery rate rather than win rate. {quote} IMO degree of dependency depends on how the system chooses its setups. Most of the EAs I've written treat each 'fresh' setup as being independent of the others. Thanks for the discussion.

OF course, no money management improves the winrate, that is impossible, but it will improve the total profit, by distributing the win/lose trades differently than the normal distribution, so at a small sample you indeed have a relative larger winrate, but on a large sample it would converge to the mean , and will be distributed just like the normal distribution curve so it would not change its overall winrate.Winrate can only be improved if you choose better subsets ,where the entire set is the entire market which has a 100% probability, the better subset of trades you take the better will be your probability, if such thing exists in the market, and apparrently it does since so many people improve their tehniques just be selecting different trades, and it does improve their winrate too.

And i dont really like to use the word winrate since that is asumed if we know the entire population, but we dont, it is an infinite population, and we can only aproximate the winrate or the mean (given the 0,1 coded series), and not even static,so to say that oh yes the limit of the series will be a fixed number, since the series will always change cos the market is changing, and you cant really calculate any fixed stuff on it.The most you can do is estimate the population's properties with a randomly selected sample.And thats why i kinda dont agree with PipMeUp's calculations because he did 2 errors, 1 he calculated with to the limit of infinity,where all distributions are mostly equal,so you cant distinguish them and of course then you can say that martingale is bad because it gives the same statistics, where if you'd do the tests on limited numbers and relatively small sample it would be very different, and 2nd error was that he disregarded the fact that the market is a constant static series, which is not true, its dinamic and always changing, you can only estimate it you cant draw definite conclusions.
And even if we say that this system is 50 or 51%, that is only true on that specified set, but we need an estimate on the entire population.

Also what you can do is get a huge chunk of historical data and perform a Z-test to test to see where will the most likely be the selected variable in the normal distribution graph, and use that in conjunction with placing martingale's.
"There's a sucker born every minute" - P.T. Barnum
  • #353
  • Nov 28, 2013 5:06pm Nov 28, 2013 5:06pm
Quote
Thank you for sharing PipmeUp. It's just the big time losers in forex seem to be pessimistic about everything in forex. Which is why they only focus on the negative, so IF the negative occurs they can FINALLY be correct about something. To bad they get no financial gain from their post.
Hmm except I'm part of the skepticals. If you plug numbers in the formula you'll see that the probability of not losing is vanishingly small when the number of trades increases even with a very high win rate...

Quote
we still avoid the losing strikes more efficiently with martingale
Not correct. Your probability of winning is independent of the betting size.

Quote
Yes i understand you man but do you understand me? This martingale would work in a 1 shot system, so in certain cases it would end up in failure, which is fine,it's just gambling with, but if we set a goal of lets say 1m $ and hedge towards that goal, we still avoid the losing strikes more efficiently with martingale, since if we'd use naked system with 51%, the probability distribution would still suggest that unexpected things can happen, a win rate is not a constant, its a variable.
A stopped martingale also has 0 expectancy. Therefore this is indeed playing pure luck lotery. Unfortunately your argument doesn't stand. There are two stopping bounds: the one million $$ goal and the backruptcy. The martingale betting makes the progression toward $1M linear while the risk is exponential. You can say that it is only a risk (until it is a realized loss) and that it isn't important how high is the risk as long as it isn't realized. But this is not true. The sequence of wins and losses is a markovian process (a random walk). Except that each winner makes one step up while a loser makes 2^k steps down. Because you start with X dollars which is "a very small amount you can easily afford to lose". You are way nearer of the backruptcy level than of the goal. X-2^k<=0 is way easier to reach than X+N >= 1000000. The probability of winning, compared to a national lotery and very much higher. But the FX martinagle lotery ticket is much more expensive.

So no it isn't ultimate gambling. If you want to gamble, you're better off using a method like "Rags to Riches in 10 or 20 Trades - Simplicity at its finest". At least here you "only" needs a 67% winrate. It doesn't even need to be consistent but only for 20 trades. While a martingale requires more 80-90% win rate to survive long enough. Mainly because it would look like as if it was autocorrelated.

Quote
OF course, no money management improves the winrate, that is impossible, but it will improve the total profit, by distributing the win/lose trades differently than the normal distribution
I don't think so. The convergence to the Normal distribution is amazingly fast: http://www.forexfactory.com/showthre...58#post7089758 (last picture and the post right after)

Quote
And i dont really like to use the word winrate since that is asumed if we know the entire population, but we dont, it is an infinite population, and we can only aproximate the winrate or the mean (given the 0,1 coded series), and not even static,so to say that oh yes the limit of the series will be a fixed number, since the series will always change cos the market is changing, and you cant really calculate any fixed stuff on it.
I do agree with that. It is a simplification. But again it converges really quickly: http://www.forexfactory.com/showthre...32#post7091332 and http://www.forexfactory.com/showthre...52#post6623952

Quote
And thats why i kinda dont agree with PipMeUp's calculations because he did 2 errors, 1 he calculated with to the limit of infinity
Well I must give credit to whom owns the credit. The formula is from De Moivre (1735). The sums are finite.

Quote
he disregarded the fact that the market is a constant static series, which is not true, its dinamic and always changing
No I didn't. The heteroskedasticity of the market distribution is irrelevant in this context. Either this is a random system in such a case it will unknowingly "play" anywhere on the chart. Or your system is so good it is able to select (with probability 1) the market conditions in which it operates. In the first case the marginale probability applies. In the second it is the PDF of the selected market. Inbetween a mixute of both.

You could also explain it another way: slice the market in periods of given, fixed, volatility and consider each slice separately. Each is now homoskedactic. The probability of the system will vary from slice to slice but for each one you have iid.

Quote
Also what you can do is get a huge chunk of historical data and perform a Z-test to test to see where will the most likely be the selected variable in the normal distribution graph, and use that in conjunction with placing martingale's.
This is a good idea.
Translation for people who are lost with the stats lingo: if you want to martingale, trade the ranges. (yes it is smarter than doing it in trends)
No greed. No fear. Just maths.
  • #354
  • Nov 28, 2013 5:40pm Nov 28, 2013 5:40pm
Quoting hanover
{quote} At the end of the day MM must be a personal decision, and I expect that each trader's perspective is influenced by his own financial situation. For example, in my own case, virtually all of my income comes from investments (both active and passive) in other markets. I don’t trade forex myself, but my friend (mim2005) and I are slowly building and forward testing an arsenal of EAs which will, if they prove profitable, become a small part of the portfolio (see this thread for more...

Hi Hanover,


Excellent post When it comes to putting 'important' money on the line, I share the exact same thoughts you've expressed. One can never be too cautions when it comes not to lose significant capital. Rule #1 Do everything possible to limit potential account downside, and all things being equal, the only likely possibleoutcomes to be left are, either break even (capital intact, time lost), or upside (rising equity - our natural preference​).

As a general rule of thumb, I consider things like martingales, grids, averaging down, etc, things to be avoided all together in serious trading. However, if a system that uses these elements proves to be reliable enough, it can trade a small account (funded with money which is given as lost; a gamble) on the condition that withdrawals will be made after when/if equity reaches a certain set % (usually 100%).



Quoting hanover
Here’s an exception: where a capped martingale can give you an advantage My previous post (#338) holds true when the trials are independent, i.e. each trial (trade) has no bearing on the outcome of other trials. Or in other words, when the probability of each trial is assumed to be equal. However, in a forex setting, it’s perfectly possible that some trade setups have a higher probability of success than others, and this justifies a higher bet size. (Analogy: That’s how blackjack pros beat casinos: higher bets when there are more high cards remaining...

My team has created a strategy somewhat similar to the one you described here. The strategy is a capped martingale (5 levels: 1, 2, 2, 4, 8). Entries are on dynamic S/R levels after strong direction bias has been determined. Each entry has a higher probability of being on the money than the previous one. Risk can be set from 4 to 18% account size per trade. Probability of getting a full stop at maximum risk is about 1:1000. Naturally, I have mixed feelings towards it. The strategy is very reliable, but comes with a 18% drawdown black swan event possibility attached to it, no matter how remote that possibility, it is not something I would ever like to see trading a big account, or mixed in a portfolio. It has been trading onlyon small accounts. ~1 year results are well within expectation.

I'm having a hard time even considering the possibility that MM could be an advantage (= edge?) under a particular exception to the rule. I have doubts it is the capped martingale per se that gives the advantage, as one could always ignore the first lesser probability entries, wait and take only the later higher probability ones (the number of trade opportunities would shrink significantly though). Dependent or independent outcome probability, would then be irrelevant. It makes no sense. I must have misunderstood something here. What have I missed?

Cheers,
vox
"To hold, you must first open your hand. Let go." - Lao Tzu
  • #355
  • Edited 9:35pm Nov 28, 2013 7:03pm | Edited 9:35pm
@ Hanover

I've re-read the relevant posts, and think my doubt might come down to differences in semantic interpretation. A higher probability trade naturally deserves a bigger wage, whether it happens BEFORE or after, a lower probability trade. This is applying basic MM, and has nothing to with martingale. It only acquires martingale quality when this is done systematically as in AFTER every losing trade next trade size is to be increased, irrespective of the probability outcome of the later trade. That said, if I'm understanding correctly now, the strategy from my team I've mentioned above, would be a prime example of what you meant with a "capped martingale can give you an advantage", because each subsequent entry offers higher probability outcome, unlike pure martingale that completely disregards bet outcome probabilities. AND, it assumes limited funds unlike pure martingale. Anyway, if this is what you meant my last argument in last post is left standing somehow. The 'advantage' merit should belong to the higher probability of the trade itself, not to MM, and even less to martingale.

EDIT:

Finally it came to me. What you said can only mean martingale in Forex has an advantage over martingale in games of chance, because in Forex unlike games of chance, one can use probability (edge). Now that is the only thing that makes sense, and it makes perfect sense! Martingale's still a loser in forex though LOL

EDIT 2:

My apologies to you Hanover. I've misunderstood what you meant, and really should have known better. If you took the time to write something then it is founded on logic.


@ OP's question, btw:

My opinion is that using pure martingale in forex would be outright idiotic. Even when martingale is redone and implemented in a clever way that caps possible losses and takes subsequent trades higher probability as rationale for a higher wager, it is still a ticking timebomb of uncontrolled risk. Smart people trading serious money will never trade in such a dumb way.
"To hold, you must first open your hand. Let go." - Lao Tzu
  • #356
  • Edited 10:50pm Nov 28, 2013 10:08pm | Edited 10:50pm
Quoting vox dei
@ Hanover I've re-read the relevant posts, .....
Vox,

In hindsight I've been too lenient in my use of the term "martingale". Strictly speaking, it refers to doubling without limit (i.e. uncapped), and ignores system/game probabilities as it merely seeks to recover recently lost money. But I've seen others refer to less steep staking progressions, and progressions that are capped, as being martingale 'variants'.

However, none of this changes my core set of beliefs:

1. Higher (average) position sizes increase risk of ruin exponentially. Uncapped increases equate to potentially unlimited loss.

2. MM can never alter expectancy — it merely makes tradeoffs between return and risk, and/or redistributes wins and losses — unless you count the case where there's an advantage in placing bigger bets on higher probability outcomes. To justify larger position sizes, the probabilities need to be higher commensurately, and given the way that the probabilities with forex operate, it's is unlikely to be able to justify several doubling steps; the progression is too steep.

3. Increasing bet size to recover recent losses has no mathematical basis, because trader's/punter's P/L has no bearing on market/game probabilities. Recent losses have no greater effect on eventual bottom line than historical losses.

4. With martingale, the starting wager must be small enough to allow survival through several steps in the progression (preferably a very high number), but also large enough to deliver a worthwhile return.

5. No amount of testing can determine (without a prohibitively high amount of statistical error) the likelihood of whether an extremely low probability event will occur sooner, rather than later. Hence I don't believe that a backtest of uncapped martingale can deliver meaningful results.

Sorry for repeating myself, but that's a summary of the conclusions — whether right or wrong — that I've reached. Hope it clarifies.

David
  • #357
  • Nov 28, 2013 10:37pm Nov 28, 2013 10:37pm
Quoting hanover
{quote} Vox, In hindsight I've been too lenient in my use of the term "martingale". Strictly speaking, it refers to doubling without limit (i.e. uncapped), and ignores system/game probabilities as it merely seeks to recover lost money. But I've seen others refer to less steep staking progressions, and progessions that are capped, as being martingale 'variants'. None of this changes my core set of beliefs: 1. Higher (average) position sizes increase risk of ruin exponentially. Uncapped increases equate to potentially unlimited loss. 2. MM can never...
Hey David,

Basically, I agree with each and every point you've mentioned on the text above. For a moment I thought you were opening an exception to point 2 (MM can't alter expectancy), hence my state of confusion. The very thought of you stating such a nonsensical thing resulted in utter disbelief from my part. "No. Hanover would never say a such a thing, as it makes no sense whatsoever, hence I must be misunderstanding what he's saying" - I thought. Obviously what happened here is that I misunderstood you (the fact that English is not my native tongue is not excuse enough), and feel embarrassed for even thinking for a moment that you would actually say something like that.

Please accept my apologies,
Nélio (aka vox dei in FF)
"To hold, you must first open your hand. Let go." - Lao Tzu
  • #358
  • Nov 28, 2013 11:28pm Nov 28, 2013 11:28pm
Quoting vox dei
{quote} Hey David, Basically, I agree with each and every point you've mentioned on the text above. For a moment I thought you were opening an exception to point 2 (MM can't alter expectancy), hence my state of confusion. The very thought of you stating such a nonsensical thing resulted in utter disbelief from my part. "No. Hanover would never say a such a thing, as it makes no sense whatsoever, hence I must be misunderstanding what he's saying" - I thought. Obviously what happened here is that I misunderstood you (the fact that English is not my...
Vox,

Absolutely no need for any apology. I've been wrong many times before (especially in my market analysis, LOL). And there are plenty of folk who are more knowledgeable about math and statistics than I am; PipMeUp is one example.

Sorry for being so confusing in my earlier posts.

David
  • #359
  • Nov 29, 2013 2:01am Nov 29, 2013 2:01am
Quoting hanover
{quote} Vox, Absolutely no need for any apology. I've been wrong many times before (especially in my market analysis, LOL). And there are plenty of folk who are more knowledgeable about math and statistics than I am; PipMeUp is one example. Sorry for being so confusing in my earlier posts. David
Thanks David. You're a gentleman. I have much respect for you, hence the need to clear any misunderstanding.

Cheers,
vox
"To hold, you must first open your hand. Let go." - Lao Tzu
  • #360
  • Nov 29, 2013 3:32am Nov 29, 2013 3:32am
Quote
the strategy from my team I've mentioned above, would be a prime example of what you meant with a "capped martingale can give you an advantage", because each subsequent entry offers higher probability outcome, unlike pure martingale that completely disregards bet outcome probabilities.
The increase of probability is a required yet not sufficient condition. The optimal bet size of a system can be calculated from a choosen utility function, based on personal risk aversion. The most known example is the log return which leads to the Kelly bet size. If you don't know the characteristics of your system, i.e. its winrate and R:R and their associated confidence intervals, such a function still exist. You simply dont know it.

If you look at the Kelly curve you notice that it drops more sharply after the optimal point (too big size) than it does before (too small bet): too big dramatically increases the risk of ruin. If you double even only twice of three times you'll certainly get in this wrong side of the extremum. So you have to start near the leftmost side of the curve, that is a really too small size. In the example you give (1, 2, 2, 4, 8), you shall start close to 1/8 of the optimal size.

This is exactly Hanover's point 4.
Quote
4. With martingale, the starting wager must be small enough to allow survival through several steps in the progression (preferably a very high number), but also large enough to deliver a worthwhile return.

For serious enough trading the starting probability (the winrate of the first trade of the sequence) shall already provide an edge. Otherwise it is punting. So the idea is to study the system closely to make sure the last progression step of the martingale keeps on the correct side of the curve. The best place beeing exactly on the top.

Of course if your probability changes given your previous trades so does the Kelly curve. This means that if you can estimate the probability of a success given the recent history, you can always bet at Kelly size. And this would be optimal. If the probability increases after a loss it only looks like a martingale because the trade size increases after a loss (but certainly not double!). In reality it is always statistically optimal! BTW You can as well have a decreasing optimal bet size if the correlation is the opposite. In this case you optimize (limit) the DD.

Just for my curiosity, what kind of system is your team using? The only strategy that I found to seem providing this increase of proba after a loss is the counter trend line breakout in the direction of the higher TF trend. And I'm not even sure it really does.
No greed. No fear. Just maths.
0 traders viewing now