Does the martingale system really work?

Martingale always work, you only need infinity money, when you don't have it - the game is more difficulty.

In principle the more money you have and the lower your starting lot size, the more succesful it will be, it also depends on your trading systems win/loss. If your max losing streak for your system is 10 straight loses and you have enough money to double up 20 or 30 times, then I would say that is a pretty minimal risk. Then again, sooner or later you will run into that 1 long streak that could wipe ur account, be it 1 month or 100 years.

You don't let your losses get that far. You stop out at a smaller profit, break even or a smaller loss. Every system has its mechanical limits.

you should always average you profit,not loss.

If you are placing only one trade, it will work. The more you'll trade, the less it will work. Eventually, sooner or later, it will stop working and you'll lose everything. It doesn't matter how you would, if you would, modify your martingale strategy... the odds will always be against you in the long-run.

Martingale only works on certain pairs, definitely not for EUR/USD since it can travel up down 1000-2000 pips at any crazy period within 1-3 months and may take 1 whole year to return. If EUR/USD was at 1.4 and u buy it until it drop 1.2 will martingale work? obviously no, martingale is a form of money management betting.

A 1st position is placed in, if trader get lucky he's right, if hes wrong he plan the next entry on a good position. Any grid of less than 150 pips is a sucide imho, some trader use martingale is 10-90 pips difference with x 2 each time and sooner or later they are treading in even deeper under water shit.

Martingale must be accompanied with a clear rules of understanding how support resistance works, meaning to say.

If Price 1.0200 sell 1.0 lot aud/usd did not work out, it moves up to 1.0350, negative in $1500 under water, a 2nd trade was done at 2.0 at 1.0350 of sell 2.0 lot.

Martingale is not for trader to make sure all his positions get green and to keep out stacking, it's mainly to move price back to former breakout area mainly. Martingale is also meant for longer period trading, sometimes a trader may need to wait up to 1 week - 2month for positions to be closed, if negative swaps would be even worst.

For example, if 1.0290 was the breakout point of resistance, we will look it at a higher probability of resistance turns support.

If trader get lucky the breakout is not huge/fundamental, 1.0350 once touches 1.0290.

2.0x60pips = $1200 green

whereas earlier 1.0 lot from 1.0200 would be -$900 on 1.0290.

Trader should dump all positions at this point $1200 -$900 = $300 profits in the end, if trader choose to close all/partial but his earlier 1.02 must be close to respect the martingale system to make sure initial balance and equity is higher and not under-water anymore, if failing to do so, may end up in collapsed trading once price keep breaking higher.

a simple review of martingale failure would be in pairs like eur/usd, eur/aud, gbp/aud or even jpy pairs which can move from north to south. Accumulation of negative pips will end up in more damage and emotions trading.

That last statement basically removes most of your currency pairs, as all pairs have their tsunami moments. It largely depends on what range you want the grid to survive.

Prices don't move straight up or straight down. They go in "waves" up and down. Often times with bar charts, we forget all the movement within the bars. That is what counts the most.

Don't see why you could not use martingale on other pairs. But it is best to use smaller lot sizes and restrict how many levels you use.

Martingale works but only if you take very low risk and do not use a x2 factor. The EA i'm using allready made 26% in less than 3 months with a limited risk (live account) and on demo it made an accout go from 20k to 85k in 18 months with only one occasion where a potential -30% loss could have occured (which did not) With 80k and low risk it just cannot burn

And that right there is the problem with Martingale: The longer it works, the more distant your fear of the account blowing up becomes.. then when it finally does, you never would have expected it.

True blow out; losing more than you ever expected to lose.

Anyone can turn a profit doubling down, but it's what's lost when things don't work out that matter most.

In your demo, you saw it grow from 20k to 85k, but that doesn't do anything but give you 2 more levels of martingale to bet before the account vanishes.. only two. The nature of the betting system is to double up each level, doubling the risk each level reached. That means an account that doubled in size would only be able to stand one additional level of betting before it too would blow. Yikes.

Also, do this, look back at your demo on that 30% drawdown. If your lot size starting didn't change as the account grew (which I assume it didn't since you said that the account size of 85k was a lot safer,) then note the account balance where the drawdown occurred. If that was 30% of $60k+, then you lucked out that your account grew to that size before the drawdown took place. If the same trading conditions happened when the account was only $20k, then the account would have been destroyed.

Worse yet, that 30% drawdown on demo you talk about would only pick up the pace as the market moved against it.

Let's say the system is basic and initially risks 1 unit to make 1 unit. So 1:1 at the first level. At level two, we gotta double the risk to make 1 unit back, plus the first unit of risk, putting total risk at 3 units to 1 reward. For example sake, let's make each unit 1% (I know you're probably using much less, this is just for illustrative purposes,) that means by the time we get to 30% drawdown, if that level failed and price continued moving against the basket, you'd see 60% drawdown by the next level's entry (and the one after that kisses the account goodbye.)

This example is a bit harsh, but it follows the basic strategy around martingale (doubling up each failed bet.) A lot of EA's and systems try to make it safer, doing things like spacing out the order levels based on market trends or ATR, to not quite doubling up each level (or having some other multiplier in use,) to using specific entries on where to start a basket, etc... there are some really sophisticated EA's out there all devoted to this. Sadly, at best they just delay the eventual loss, and do so by reducing the potential return.

The market doesn't have to respect your basket of trades, it doesn't know you and doesn't care about you. It will do whatever it wants and couldn't care less about your account.

I'm not picking on you here, this is more so people can assess the risks themselves as they read this thread. I just pointed out your post since you said "just cannot burn", and that absolute is very much so not the case.

Martingale is like picking up pennies in front of a moving steam roller.. you keep thinking you're a few bucks richer til you get run over.

If you're going to use a martingale system, you MUST be very aware of the risk involved and you must be prepared to lose the entire balance of your account.

(side note: besides having issues with the risk profile of a martingale system, I rather dislike how the system appeals to new traders who are the most common victims of their own greed in the forex market... they flock to systems like this and never ask themselves why they can't find an example of one that's lasted more than a few years at best before blowing up.)

Yet it's true, a blow up could happen but i'm managing my account and checking the different price level and resistance to choose to stop a trade or not. Untill now, the only occasion i had to do so was when the EURCHF was locked by the swiss governement. Leaving an account unattended when using martingale would be crazy

Past history are not an indication of the future results, i have done martingale before and i know its deadly point, being in negative drawdown is really unhealthy.

The biggest mistake of martingale will be not willing to be a loser early and tries to protect profits accumulation which later endangers the account.

Like i said it should be work on some pairs only, most major pairs or even exotic pairs are actually trend pairs, they do move one point to another, depends what timeframe u are looking at.

Like jack pointed out, martingale system actually do risk the entire of the account at some point of time, it's just a matter of time. You could have 10,000 or even 50,000 pips points before hitting margin call, but it may be a matter of time it don't work due to collapsed trading and accumulations of dd and negative pips, still depending on pairs you are trading.

It is true that the unknown variable is how far prices will go. We use the past to get ideas of the typical ranges. We can prepare with a best effort for tsunamis that occur once in a while, but we must still enter the market in order to have the opportunity to earn profits.

Martingale, position sized 'correctly' (within reasonable expected ranges) can produce massive returns in a short period of time.

I've seen that all strategies, particularly aggressive strategies usually you want to have real profit target and then use the betting strategy to hit the ultimate target. Say for example your goal is to get 500% or 1000% of your starting capital, and you have $10,000 to work with. You would start off with $2000 with nano lots (100 units of base currency) and see if you can reach the 500% or 1000%. Say 1000%. Say the first 3 trials you wipe out but the 4th one you reach your 1000% goal. You are still up 100% of the total starting capital ($10,000) for that trial. There are other variations like refilling after the drawdown reaches 50% or xx% vs waiting for a full blowout.

If you use a structured format with goals/targets, then you can forward test it with repetition in demo and live accts to see how viable your strategy is. This is more objective than just looking at one blowout and saying 'this doesn't work' Your total market risk decreases as the equity increases.

i don't believe in demo trading, i am not a robot too, minimum any testing should be done on 0.01 lot live trades to keep emotions intact, that's from my perspectives only. i do know martingale do make money, but price entry have to be alot sharper and sniper.

but the biggest problem as mentioned is always the dd and risk involved which i don't recommend to keep trades for too long.

Martingale works but only if you can cut the losses！

Good post (as always).

Folks, ignore Jack's wisdom and experience at your peril.

I've posted on Martingale many times before (see myth #12 here, for example); no point in repeating myself.

There's a great video here on building a risk model. As the presenter says at one point "Martingale is beyond retarded". Harsh but true.

Bottom line is that you can't argue with the underlying math. If you have an edge, Martingale can't enhance it; if you don't, no kind of MM can provide one.

You can either waste time experimenting with different MMs, or spend that time studying charts and learning how to trade profitably.

David

Don't know why so many people insist that something is not possible. It's all how you apply it. I explained a few post earlier how you can take advantage of more aggressive risk models in a structured format. Then you can test different ranges and position sizing to see what is most likely to be successful instead of just taking someone's word for it.

Careless martingale-ing (or any risk management technique) will ultimately fail.

because when intelligent people speak they usually discount luck entirely from their arguments since it is transient and not a factor in overall profitability. when unintelligent people speak they often base their whole argument on the results of luck and / or statistically insignificant data and have a myopic view on a topic, as opposed to an intelligent person who has a more well-rounded and grander perspective on things.

having the argument that "martingale is fine as long as you stop before busting your account" isn't an argument.

continually re-adjusting make-believe profit / loss targets / streaks doesn't change anything in the end. you either bust-out fast, or you rig your game so that your risk of ruin is further away but at the same time your profits become worthless.

anyone with the slightest understanding of risk of ruin will understand why martingale is a joke. always has been. always will be. there is simply no argument in favor of martingale that holds-up. anyone who thinks there is does not understand something critically important.

risk of ruin in conjunction with risk to reward is critical in all areas of life.

losers average losers. threads like this help me fill my ignore list further, so thanks everyone.

About MM not being able to alter expectancy: it's not impossible because of people's insistence. It's impossible because of the underlying math. The XLS attached to this post demonstrates that no matter what MM is employed, overall expectancy remains unaffected.

A more aggressive risk model increases both trade return and risk in like proportion. It also increases risk of ruin exponentially. This can easily be demonstrated mathematically, just like 2+2=4.

There are a number of reasons why martingale is complete stupidity. If you want to trade more aggressively, then you can simply increase your risk per trade from (say) 1% to whatever%, consistently on every trade, and increase risk of ruin accordingly. But martingale progressively escalates risk randomly and arbitrarily, without any statistical justification for doing so.

Analogy: suppose I'm playing basic 5 card draw poker. I lose the first pot, so I double my bet to recover my loss. Then I lose again, so I double my bet, and so on. Do you see something that's missing? I'm not taking my hand strength into account. I can justify an increase in my bet size, if I hold something like a full house, because my probability of winning the pot is better than if I hold (say) a pair of 2s. But simply increasing bet size merely to recover recent losses makes no sense; it has no bearing on the pot odds. The same applies to any activity that involves uncertainty, including trading; with trading the probabilities (albeit fuzzy ones) are being provided by orderflow instead of card combinations.

Martingale is about accelerating recovery from recent losses. But why are your most recent trading losses any more significant than historical ones? They're not; to think that they are more significant is completely delusional. Everything else being equal, all losses contribute equally to overall bottom line.

To illustrate this, let's have a look at some hypothetical trade results, in pips: +10, -20, +40, -20, +50, -20, -10, +30, +10, +10, -20, -20. Now how can I increase my profit, by using only MM? Easy: by placing bigger positions on the winning trades, and smaller positions on the losing ones (or even better, not taking them at all). But how do I know, in advance, which trades are going to win, and which will lose? Unless I have some solid statistical evidence that some setups deliver better results than others, then the reality is that I haven't a clue, and hence it makes no difference which trades I put the bigger positions on, because I'm merely playing a lucky game of hit or miss. Over a long period of time, then due to the law of large numbers, the 'luck' will balance itself out, which means that I would have got much the same result by using the same (average) size all along the way.

Now if you can show statistically, in a given situation, that price is more likely to return to a mean, than not, then sure, you're justified in adding to a losing position. But that's a very different reason to simply trying to recover recent losses. Do you think that the market is going to change its behavior, out of sympathy to you, because you're in a losing streak? Of course not. Well, no change in behavior means that the probabilities have not improved, and hence you have no justification for increasing your position size.

Martingale exacerbates the dangers of averaging down, because the position sizes rapidly spiral out of control. Very soon they are no longer commensurate with the probabilities being offered by the market. For example, doubling up after 5 losses, you're risking 32x your original stake. To justify such an increase, is the probability of the current setup in the ballpark of being 32x more likely to succeed than the original one? No way!

Folk show their ignorance when they say "look, martingale works profitably, I've been running it for X months without any problem". Of course it friggin' well works! Why? Because the recovery algo will always being the account back to breakeven point. That's true with every martingale. Some martingales are lucky enough to encounter suitable market conditions for several years before they eventually and inevitably implode.

You can't backtest martingale reliably. Why? Because the probability of a very unlikely sequence of losses occurring would require a massive data sample, to give a meaningful result within reasonable confidence levels; the high statistical error renders the results worthless. You can have a clear run over 200 trades, 500 trades, 1000 trades in backtest, and it proves nothing, in terms of how soon the 'death sequence' is likely to occur, if you were to go live tomorrow.


I read over and over again that the risks involved in martingale can be reduced by applying certain 'mitigation' strategies (I assume that's what you mean by a "structured format"). Let's look at some of these:

1. Instead of doubling (1,2,4,8,16,...), use a less severe progression (e.g. 1,2,3,4,5,....). Problem: it now takes more than 2 or more consecutive winning trades to return to breakeven, whereas doubling requires only one. Hence you're crippling your recovery rate commensurately.

2. Put a cap on the number of consecutive losses. For example, lets say it's 5; after 5 losses we've lost 1+2+4+8+16 = 31 units. Now we have to win 31 trades @ 1 unit to recover from 5 losses. 31 wins to recover from only 5 losses? Not very smart MM. An even dumber way of crippling the recovery rate.

3. Start with tiny position sizes, to allow a larger number of progression steps. But tiny risk will decrease return in exactly like proportion. There are easier ways to make a few pennies each week; and no matter how tiny the starting position size, the possibility of ruin is always there.

4. Here's the dumbest one of all: reduce risk by periodically withdrawing money from the account. Firstly, this is counter-productive, as less money in the account means that you're decreasing the number of progression steps that the account can survive; hence in one sense you're actually increasing risk. Secondly, if you're using fixed frac sizing, a smaller account balance means commensurately smaller returns. And thirdly, there's no guarantee that implosion won't occur before you make the withdrawal. The need to withdraw money before the account gets completely scuttled, is surely an admission that the 'trader' is gambling, rather than trading.

It all comes back to the math in the XLS in the link at the beginning of the post. In all of these examples, varying the MM is merely re-balancing return and risk, and/or redistributing wins and losses; it can not, in itself, provide and edge.


Let's say a trading system has no real long-term edge. Then, due to spreads and commissions, and assuming it's taking 1:1 RR trades (taking higher RR trades will decrease the win rate even further), it will have a win rate of slightly less than 50%; let's say 45%. Now the probability of getting (say) 10 consecutive losses within any 50 trade sequence is around 10% (see table here), which is a lot higher than many folk might think. The irony then, is the longer that you trade, and the more frequently that you trade, then the more 50 trade sequences you'll have, and the greater the likelihood that the 'death sequence' will occur.

I've heard an argument that trading is risky anyway, i.e. that the risk of martingale's 'death sequence' is tiny, and therefore no worse than the risk of ruin by sizing consistently at (say) 1% per trade. Maybe so, but one important difference is that at 1% gives you time to stop and re-evaluate your trading system after a sequence of losses, with most of the account still intact. Not so with martingale.

Any professional trader will tell you that risk management is the cornerstone of long-term profitable trading. Martingale is the ultimate antithesis of prudent risk management. Suppose a country's government unexpectedly takes steps to artificially revalue its currency, or an earthquake, or a terrorist attack, or any other 'black swan' event occurs, causing a 1,000 pip spike; the trader who fixes his risk at 1% loses exactly 1%; the trader who indiscriminately uses martingale leaves his account wide open to devastation.

You cannot fix your risk at 1%. Why? Because a tsunami-like event will blow through your stop loss if you are on the 'wrong' side of the trade anyway. The idea behind having fixed stop losses (or any trading strategy) is that you will overall average larger profits than losses. How you do it really doesn't matter. Rather than using stops, you can average down and wait out a pullback, even 1000 pips if you setup the grid correctly. You have much more room to play with on the pullback, in terms of taking a smaller loss, break even, or profit before returning to the original entry price (plus spread/commissions). All mechanical systems will have their limits, depending on what the goals and targets are. If the grid is only setup for 100 pips, and prices move 120 pips, you are in trouble.

When you used fixed stop/trailing stop approach, instead you are betting that your predictions will be correct more times than incorrect; in essence a predictive model which is very subjective. Since you are betting so heavily on prediction, you would want to place all of your position sizes in one price (the initial trade) This usually doesn't work because by the time you realize your chosen direction is incorrect, you are at a hard loss. You must then re-enter on the next signal and bet that you are correct (again) Then when you are 'correct', how can you let your profits run? Your next prediction of when to exit better be even more accurate than the entry, as the goal is to have net profits exceed net losses. Even with a predictive model (for initial entry), you can still combine a grid entry for averaging and then employ net profit exit system to improve profitability.

I was actually referring to the "structured format as laid out here" http://www.forexfactory.com/showthre...31#post6118831 (#112) . With hard targets for a net take profit vs net drawdown or blow out, your betting strategy becomes cyclical, and you can run real scenarios to see how much profit/loss you can make over x number of repetitions. Can test different ranges of prices to see what the system can/cannot handle. In practice, how far do prices typically go (over 50% of the time) before a significant pullback? Again, with a grid approach, the only unknown variable is the range of prices from the initial grid trade that the market will trend. You can get an idea of how far prices will go by glancing at historical price data, combined with walk-forward testing or forward testing on tick data. When a tsunami comes (tsunami = price movement which falls well outside the limits of the grid), the best you can do is attempt to detect it, and get out of its way. The process is transparent, and it is easy to test, and limitations are well known; whereas purely predictive models for entry and exit usually rely on lucky streaks.

The poker example is not a good one, because each hand dealt to the player is completely randomized (the deck has no memory of previous hand outcome). A betting strategy isn't as strictly mathematical as trading where linear price movement is the sole determiner of an account's profit/loss. The edge in those types of betting strategies is more with your ability to "play the other players" (bluffing), not so much the cards. Blackjack would be a better comparison, where your edge can come solely from card counting to improve your odds above 50% that you can call a bust or hit correctly. A 2 card deck isn't so difficult to do this. But most dealers despise card counting and will thwart it: use 8 card decks, or will switch up dealers every 15-30 minutes, which requires a new card deck, or other strategies to take the edge away from the player.

Nice article btw on the myths. The spreadsheet is good to test a mathmatical model, where the goal is to profit the same TP for each repetition. It only proves my point that the only threat is the range of prices prior to blowout point or stoploss point. how often does prices go to 100 pips with zero retracement? 200? 500? 2000 tsunami? It gets better when you realize you can adjust the tp to get more than just a single TP target, as prices tend to retrace much more than just the initial TP level

My responses in blue typeface.

The idea behind having fixed stop losses (or any trading strategy) is that you will overall average larger profits than losses.
Agree, it is about ensuring that losses remain small, but also about fixing worst case damage, in the context of account survival. Also, if I get stopped out at (say) 20 pips, and the move subsequently moves another 100 pips against me, then that 80 pip saving has the same effect on my eventual bottom line as an 80 pip win.

How you do it really doesn't matter. Rather than using stops, you can average down and wait out a pullback, even 1000 pips if you setup the grid correctly. You have much more room to play with on the pullback, in terms of taking a smaller loss, break even, or profit before returning to the original entry price (plus spread/commissions). All mechanical systems will have their limits, depending on what the goals and targets are. If the grid is only setup for 100 pips, and prices move 120 pips, you are in trouble.
Agree that how one trades doesn't ultimately matter; and also about mechanical systems having their limits. Apparently we both have different approaches to trading; yours sounds more grid-oriented, while I tend to scale in around what I see as being pullbacks to key market levels, within the framework of the major trend.

When you used fixed stop/trailing stop approach, instead you are betting that your predictions will be correct more times than incorrect; in essence a predictive model which is very subjective. Since you are betting so heavily on prediction, you would want to place all of your position sizes in one price (the initial trade) This usually doesn't work because by the time you realize your chosen direction is incorrect, you are at a hard loss. You must then re-enter on the next signal and bet that you are correct (again) Then when you are 'correct', how can you let your profits run? Your next prediction of when to exit better be even more accurate than the entry, as the goal is to have net profits exceed net losses. Even with a predictive model (for initial entry), you can still combine a grid entry for averaging and then employ net profit exit system to improve profitability.
Can't comment as my approach differs. My comments in the previous post were pitched more at martingale MM, than the merits of different approaches to entries and exits. Agree that the more prediction-based the method is, then the greater the importance of getting the timing correct.

Can test different ranges of prices to see what the system can/cannot handle. In practice, how far do prices typically go (over 50% of the time) before a significant pullback?
The point about risk management is to ensure that you're covered in the non-typical situations, which do occur. I realize that's not what you're driving at here, but I believe that one must ask "if price behaves in a way that my trading model doesn't anticipate, what is the worst-case damage that can occur?"

........ The process is transparent, and it is easy to test, and limitations are well known; whereas purely predictive models for entry and exit usually rely on lucky streaks.
Again, our methods differ.

The poker example is not a good one, because each hand dealt to the player is completely randomized (the deck has no memory of previous hand outcome). A betting strategy isn't as strictly mathematical as trading where linear price movement is the sole determiner of an account's profit/loss. The edge in those types of betting strategies is more with your ability to "play the other players" (bluffing), not so much the cards. Blackjack would be a better comparison.......
I've played a lot more blackjack than I have poker, but more folk seem to understand poker than blackjack, hence I chose poker on this occasion. My point is not about gameplay mechanics, but just that one should be basing decisions on the probabilities being offered by the current hand, rather than the profit and loss outcomes of recent hands (or trades). Perhaps I could have been more clear about that.

Nice article btw on the myths. The spreadsheet is good to test a mathmatical model, where the goal is to profit the same TP for each repetition.
Agree that the XLS has limitations by covering 1:1 trades only, but the same math principles behind expectancy and MM hold good with other RRs, and also other techniques (trailing stop, scaling in/out etc), although the proofs would be significantly more complex.

All gambling/trading ultimately comes down to somehow finding a balance between (1) maximizing income, (2) managing risk to ensure that drawdown is kept within personally acceptable limits, (3) smoothing the consistency of income (if this is one's predominant income source), and (4) operating optimally within one's psychological and lifestyle constraints. MM plays a big part in the first 3 of these considerations.