Playing Russian Roulette with Martingales

I thought I would open up a discussion regarding that dreaded word "The Martingale Progression".

While many fear the concept given it's ultimate demise in account blowups, there are some things going for it when compared to more traditional trading methods. I am simply raising the topic as a discussion piece. It does not mean that I am an advocate of the Martingale approach so let's don't let this discussion get heated. It is for any that are simply interested in discussing these ideas.

I actually am neutral in my opinion until it can be demonstrated to me with sufficient voracity that it simply can never be made to work. Yes, I previously extensively traded Martingales and faced unnerving situations and recognise the addictive qualities of these progressions.....but the retail traders game is hard enough with such a high failure rate (though most of us won't admit it), so Martingales in my opinion are just a way that a trader perhaps can perhaps with a bit of lady luck stay in the game for a time before going the way of the dodo. You are simply using a gamblers method of seeing whether luck remains on your side. I am happy to play that game with a very small slice of my trading capital provided it is play money only.

For example, if you are lucky enough to survive an account blow-up for a period of time, there is a significant likelihood that you can at least return a 100% return on your initial trading capital whereby you withdraw your bounty (wiping the brow) and can commence trading with profits only. There are many examples of where this is quite regularly achieved however the blind faith advocated by many Martingale enthusiasts usually means that they never harvest the Martingale and ultimately end up in traders heaven.

So let's assume for this discussion:
1) that we are prepared to lose a small bounty of play money say $5K.
2) that we regularly harvest the Martingale; and
3) that the cumulative impact of the Martingale is only allowed to get to say a 15% max drawdown on account capital. If we get hit, we take the 15% hit and then resume but adjust our risk as a percentage of the reduced capital. That way if we are unlucky we will continue to lose our capital but in sequentially lower $ hits. Each loss however represents 15% of the trade capital at the commencement of that progression.

Now the reason we accept that Martingales are a gamblers fallacy is that overall expectancy using probability can be demonstrated ASSUMING a totally random market to be less that even odds (considering frictional costs) using the Law of Large Numbers but I would like to challenge this assumption for the purposes of this discussion. What this expectancy outcome means is that you will never be able to exceed your 15% hit with profits in the long term. What Martingaler's therefore do with this knowledge is regularly harvest when profits build with the aim of just being lucky. There is no problem with this idea at all. The problem arises when greed get's in the way and you never withdraw profits.

While statistically it can be demonstrated that the Law of Large numbers will ultimately catch up with you in a random market, let's assume the following:

1. That our loss point for a single progression requires a particular market pattern that can be demonstrated through backtesting may only ever occur say 3 times in a 5-10 year back-test. The odds of this happenning therefore on the commencement of the progression (assuming a 5 year trading life) may be extremely small.....yet over an infinite lifetime..it will happen. Ok so random luck may allow us to survive an undefined but finite period. You harvest when you can to hopefully reach the paradise tha means you are now only playing with profits.
2. That there is a slight element of non-randomness in the market that gives the gambler a better than even chance in relation to the expectancy outcome of a random market. If there is a small non-random element to the market then probability will work in your favour that particular patterns may not repeat in the same frequency as they would in a totally random market;
3. Let's also asume that we scale our annual return to be approximately 30% per annum. I understand this expectation may be regarded as too miserely by many retail traders out there but the reason we choose this low return criterion is because we want to design Martingales that we can demonstrate through backtest a long term survival rate through backtest trade history. You will get hit, but at what frequency. If you can achieve a 30% return for say 5 years you have probably been hit a few times but the return is certainly better than that achieved by a traditional retail trader over the long term. For example a simple Martingale long may be able to survive unless you have an instance where the instrument plummets (almost linearly) over a range of say 20%. For this example, anything other than almost a linear decent may mean that you can reach your average profit target and extract yourself from this unnerving situation at no loss due to the progression saving you with the progressively improvement in average cost. So what are the chances of this market pattern in a backtest history of say 5 to 10 years. In a random market you would expect the Maringale fate to get you but the reality is that the number of actual times this is achieved is far lower. This may just be foolish luck.....but what if the market is subtly non-random?​

So let's assume you can design a virtually unlimited number of Martingale progressions using some exotic market patterns that through extensive backtesting does not appear to happen all that often. For example, we can develop a Martingale that is long, only, short only, hedged, sideways only or combinations thereof...etc......it is the progression itself that represents the Martingale technique and not the trade technique.

​The key point to this discussion is this. Like a coin toss, each progression can be considered a single event. However if the market is not totally random then use a particular pattern till it fails and then move on to a different Martingale pattern and so on and so forth so you just never repeat a partocular pattern that has been hit on your trading lifetime.

Is there any merit to this approach?

Obviously there will be those that view the market as totally random. What if there is a kernel of non-randomness about it. Does this influence mathematically the ultimate fate of the Martingale?

PS With regards to the proftable side of the Martingale let's also assume that you trail your stops behind a profitable cluster with no defined profit target to take advantage of an anti-martingale on the other side of the trade.

I had made this 100% in two weeks, playing martingale and high leverage. I Know I'm gona be hit soon, so I try to MM the situation: I know I can't keep this 90% succs rate and sooner than later them will catch me... and them will take 20% or 30% of the original stake. So, I will run still them catch me and then I will sit ana reanalyse the leveraging-up progress. I challenging my self to made at least 500% in 6 months.

That's great pipburner. So given you have made profits to equal your initial capital, do you go again? I am assuming that after say 2 successful runs with 2x your initial stake you could then apply risk management and reduce your exposure by half so you then have four lives to reduce your return to your beginning stake.

As opposed to conventional approaches of re-investing the entire stake, how do you play it?

This looks to be a fun thread ,something to take ones mind away from the rules and strictness of trading our regular accounts. certainly for me i have a side of my personality which im having to manage everyday of my trading which is quite opposite to my overall personality .its one of risk and excitement and chance not generally good traits for trading,so i think setting aside a small amount of capital that one can afford to loose would keep those reckless traits satisfied .

I have a system that i have worked on for many years and have yet to trade it on a live account for the very reason that it would require a martingale approach.the problem being is that it requires you remain in the market .you are opening and closing multiple trades in profits with the extremes being the drawdown .the thing with martingale is there are many variations of martingale.infact this system ive been working on would be traded as a supa martingale which actually is the only form of martingale that has a positive expectancy,the double edged sword of this is there is no stop point.this form of martingale does not blow you out of the water it just has no point to stop.as with all martingales the stop point does not exist so because of this fact you have to cheat .in the fair game theory its not possible to win .so the only way you can win is to cheat.as you have already mentioned.the more you employ a trading strat that is more supa martingale the better and more controlled will be the risk. although this reduces the reward factor.rince and repeat with smaller gains is the way i will try to implement this.

lets have some fun

Hi dicemann555 :-)

I am hoping we get lot's of contributions from those who like a little flutter on the side. I was having a look at supermartingales and submartingales and found this in wiki:.

1. Every martingale is also a submartingale and a supermartingale. Conversely, any stochastic process that is both a submartingale and a supermartingale is a martingale.
2. Consider again the gambler who wins $1 when a coin comes up heads and loses $1 when the coin comes up tails. Suppose now that the coin may be biased, so that it comes up heads with probability p.
   1. If p is equal to 1/2, the gambler on average neither wins nor loses money, and the gambler's fortune over time is a martingale.
   2. If p is less than 1/2, the gambler loses money on average, and the gambler's fortune over time is a supermartingale.
   3. If p is greater than 1/2, the gambler wins money on average, and the gambler's fortune over time is a submartingale.

   
   

3. A convex function of a martingale is a submartingale, by Jensen's inequality. For example, the square of the gambler's fortune in the fair coin game is a submartingale (which also follows from the fact that Xn2 − n is a martingale). Similarly, a concave function of a martingale is a supermartingale.

Could you please explain (without giving away your technique) what you mean by supermartingale. On reading the wiki article I am not sure how a submartingale can deliver positive expectancy?

To add to the last post....

The power of progressions only has a net benefit through overlaying the progression over a trading strategy with a positive bias. Without that positive bias, the long term less than even expectancy will getchya!!!

One of the terms we often hear on this forum is positive expectancy.....but do we really know when we achieve this aim? In a nutshell a positive expectancy can only truly be verified statistically using large timeframes and sample sets and unfortunately in this retail world and the length of time we trade for, it is very difficult to statistically verify whether a trading strategy has any real positive bias or whether the reported positive result was just a lucky winning streak during particularly favourable trading conditions.

But in the investment time horizon, positive expectancy can be statistically validated to exist. For proof of the pudding, just refer to this interesting article.

So why don't we use this info in our retail trading activities and then utilise concepts such as risk management armed with a portfolio of trading instruments and powerful progression systems such as Martingales (sub) or safer anti-martingales to drive the dollar further while strictly managing risk?

Copernicus
thanks for the detailed answer and yes quite correct sub martingale its some time since i read the papers on this .i have a very interesting paper on this topic on my old laptop which i will try and locate and post here.

going to read your article you posted thanks

No probs diceman. I gathered that you were using a sub martingale technique from your response. :-)

I have spent some time looking into that guys blog and when studying it, you can clearly see where a Martingale progression can come into it's own when you ensure a positive bias is applied to your trading technique. Given the stigma attached to the Martingale progression, it is easy to dismiss it....but when applied with a robust trading strategy that applies positive expectancy, it is hard to dispute it's power.

Hi Copernicus,

you are too much into statistics, so pardon me if this post is not what you were looking for. I disagree with randomness of the market. Yes, you can see some randomness during intraday noise, whether it is caused by bots running on stats rather than fundamentals, or by broker settlement, or by famous stop hunts. But imo on the long run market always acts on underlying fundamentals. This beeing the case, I consider martingale very healthy approach . I should point out here my last account is in loss - and partially is due to martingaling, as when your tradeplan screwes up for whatever reason (time/personal issues in my case) martingale speeds up the burning process. Also another danger of martingale is your psychic health, as if you chew one loss, then bigger loss, when running for even bigger third loss - you are all set for heartattack.

But the point is, if your FA is right (and I am always convinced my FA is right) the fact you loose on one trade also means price is much better than at you original entry. When price is better and you know where market is heading, investing more is well justified.

Also, am not sure if this qualifies as martingale, but suppose you have a trend and waiting for reversal. Then market starts forming consolidation range, you see it run out of steam and decide to enter. But hardly ever you manage to hit precise bottom and so you enter @ mid of consolidation range with stops just below - beeing your long term entry. Should market short term chop range and test its boundaries, I usually "kinda scalp" and martingale up my position @ low of range, while martingale TPs are set from mid to top of range - and I ofc hope some of my entries on low of range will survive before break and therefore I will be loaded with bigger position on this range level.

Anyway, fun thread, looking forward for your posts.

T

Hi Tom

Thanks for your input. I can see where you are coming from. Using FA in my opinion is a very healthy way to look for positive bias in a trading plan. :-)

Unfortunately the lack of timely information makes it exceedingly difficult for me to competently apply FA. The alternative for me therefore is to look at the very long timeframe to observe where fundamentals should be playing a role in price directionality to keep away from the inevitable random noise that occurs on a short term basis.

That is why I am a stastics nerd.

I think that there are many ways we can do this and statistics can help here.

For example on a monthly chart on say EURUSD you can see the direction fundamentals seem to be playing. Without scrutinising the fundamentals, you can therefore use TA as a trade entry tool to take advantage of this fundamental bias but in a very selective manner where windows of opportunity lie on the longer term timeframe.

I actually think we are on the same page here....albeit using different methods. Where we agree is that a Martingale progression technique can actually be a very positive tool providing we use it correctly. For example let's say you want to scale into a trade and risk say a max of 2% on the trade. You can start the progression risking say 0.1% and use a Martingale to bring you to the 2% threshold and that way ensure a good average price on the bundle of trades acquired. If the trade goes against you (which many times it will, you only cop the 2% hit....but if you nail the entry you trail a generous stop and make hay while the sun shines with positive bias in your favour.

Yes, thats about it, and kinda my plan...though I have to admit I often screw up for many reasons. Lots of them are pretty silly, like - I win in poker - so I martingale market heavy and fack up..so..not like I was superpro, but paying rent so far . This system would work if you had set pending orders, but while entering manually, is very hard to stay on predefined SL percentage while using martingale.

But regarding system, I dont martingale that much my opened position, but rather shoot more once my first level of SL gets cut and market moves deeper - so am not staying on that higher leverage all the way market is going against me. Let me use my current GU trade as an example. As you can see, there are three SL, going from top to bottom those are Greed SL, Range SL and Regular long term SL. Greed SL does not count, as that is pure greed and will very possibly screw up. I hold my regular long term entry up to 1,522 - in fact is so far I dont care to move it few pips lower, as the little percentage change dont matter to me and depends on market conditions. Should Range SL break, I load up on first loading zone - and martingale - load more - and then it takes only retest of range for me to TP partial and pay for that broken Range SL....should it screw me over again, I load up martingale on zone 2. After long term SL gets hit, I usually eat rice and chicken to punish myself - is actually healthy.

I certainly agree on those TA entries following long term FA bias...but am a bit weak here, though working hard.

Gotta go sleep, bb later.

T

Nice stuff Tom. :-) I can see what you are doing and it makes sense to me.

The reason I think statistics plays out in your favour when looking at the longer timeframe in a nutshell is this.

1. The most fundamental unit of data on a forex pair is the tick. This comprises an indeterminate order quantity;
2. When you scale up to the M1 chart, the summation of those ticks produces an average price on M1 which comprises all those market orders during a period of a minute. There is not much sense to be gathered at that microlevel where coordinated price action is playing. At this level, market noise rules supreme due to the varying objevtives of the participant but produces patterns that we think are repeatable and represent non-random price movement. When I talk of 'random' here it really isn't random but a trader lacks perfect information of all the different objectives surrounding each particpant in making their trading decision during that minute. Some are rational market participants whereas others are irrational. Some are institutions that have risk management objectives at the back of the mind, a few are speculators like us etc........some are traders who are forced to liquidate due to bankruptcy :-(
3. As you step out in timeframe you get a better sense however of where collective agreement is occurring through market participants simply due to the summation of a large quantity of price data to gain an understanding of where price is collectively drifting towards. This is simply due to the fact that as we scale up we progressively accumulate more price action of the individual participants in an attempt to predict where price will move.
4. Obviously the highest timeframe therefore should portray where consensus is at in relation to price direction. That timeframe for practical purposes is the MN timeframe and theoretically this is where fundamental economics rules as opposed to the objectives of the individual market participants. Participants that want to play a game of death against the MN timeframe (the big shorts for example) have balls of steel. It would be a foolish gesture to attempt to go against this collective price directional bias. Of course you can use techniques of undersold/overbought to play a retracement game in this timeframe, but to not recognise the direction in the grand order of things is simply folly if you want to stay in the game a long time.

​If these assumptions have any merit, then within this general price direction of the MN timeframe in mind, there are better points than others to attempt an entry to take advantage of this collective price bias and typically this occurs at retracements in the higher order timeframe. If you revert to the H1 or D1 timeframe on entry you just have to accept that there is a fair bit of noise at that level but the balance of probability is in your favour that provided you take heed of the macro timeframe and trade in that direction, you should be right more times than wrong and this is where a Martingale entry system can be used to magnify returns provided you don't go crazy with it.

Your technique above is applying those princiles, so I am totally on board with what you are trying to achieve here :-)

All the papers and study related to financial markets tells us the market is efficient.i believe this to be a truth.

When price moves in one direction with increased acceleration with out retracing it generally is not efficient.regardless of the t/frame it can be days or hours .price does not need to retrace to regain efficiency .the longer this state remains the more a corrective state is increasing in probability.the market will always regain efficiency.if you know the % of that efficiency for various price swings we can trade this using sub martingale.this corrective state is the ranges .if you can determine the return of price for collective price values,pips you can determine that price band whilst price regains its efficiency .price returns 68 % of the time within a range of 0 to 30 pips.think about this .its very efficient.

the thing that would make this positive for sub martingale is using the increased probability of prolonged range and its depth whilst reducing the number of extreme positions which black swan us.

just to add to my post.most trading systems eventually fail due to this reason.there are to things we need to look at which are totally separated.the system that generates signals .and the market mechanics its efficient truth.

hello,
Is your martingale capped to a certain level ? What would you do if you reach level 12 (1-2-4-8-16-32-64-128-256-512-10024-2048 ) ?what do you think about anti-martingale ? I have to admit that i love martingale tho.

When applying the martingale it is probably a wise strategy not to overdo it. Consider a Martingale as a form of retracement entry when you are sure you are trading in the direction of positive bias perhaps of 3 to 4 levels (limited by a stop condition you would naturally use for a single trade entry). What you therefore do is divide that typical stop distance into the number of levels you wish to apply and then apply the progression entry to it. What you are trying to achieve is a good average price on entry without overdoing it. You don't have to apply a strict doubling up methodology such as 1,2,4,8 but rather a progressive slower build to a maximum tolerance of 1,2,3,4 or something like that. Anti-Martingales are great when progressing with profits. The Snowball described in this forum is a worthy candidate to consider.

Great info thanks. Its funny that you mention Snowball since i discovered it just days ago and tbh i still dont uderstand the method/thread.

You need to play with it for a while and then you will get the jist. :-)

It is a particularly powerful progression technique to take advantage of strong momentum or progressive momentum accelerating trends. A real handy device to magnify returns provided you use it with discretion. I find the Snowball an excellent choice when attempting to apply mean reversion or breakout techniques. It basically uses a machine gun for entry (scaled entry into a trade) as opposed to a sniper single trade entry thereby over the Law of Large Numbers compounds returns from positive bias.

If you however are trading trends with a nominal momentum (a slow and steady trend) then you manage to breakeven using Snowball and that's about it. This is because your average cost of trades using the progression stays close to actual price over time. An exponentially growing trend however with a positive slope (long) or negative slope (short) is where Snowball comes into its own, or when a news event increases momentum considerably allowing the average cost to significantly depart from breakeven.

For example a robust trading strategy with positive bias that over say 20 years has a 9% return on a 10% drawdown can be accelerated to a 30-40% return with a similar drawdown using anti-martingale progressions such as Snowball that take advantage of volatile (strong momentum) periods.

Anti-Martingales (such as the Snowball) are the antithesis of Martingales in that the significant profit result is seldom but the bleed (death by a thousand cuts) is slow. For strong Martingales the profit build is very slow and the killer blow is significant. But both techniques have their uses when you are trading the non-random drift over long timespans provided you keep a lid on risk.

PS The Anti-Martingale respects the principle of letting profits run and cutting losses short whereas the Martingale is the exact opposite. As a principle, I far prefer the Anti-Martingale as it's application is wide but there are specific instances where (light) Martingale progressions can be used for good effect as entry techniques.

Martingale - losing strategy. If you want to earn good money, have stable and long trading - forget about the Martingale method and don't even remember. I should also note that using of hybrid Martingale in combination with a good trading strategy can make you a good profit. But, unfortunately, I know very few such cases.