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:
- 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.
- 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;
- 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.