Martingale debunked by Gary

Post Reply
User avatar
SteveHopwood
Owner
Posts: 9754
Joined: Tue Nov 15, 2011 8:43 am
Location: Misterton - an insignificant village in England. Very pleasant to live in.

Martingale debunked by Gary

Post by SteveHopwood »

I am going to place copies of this in several forums to make sure its message is widely understood.

At some stage, every trader will meet Martingale for the first time and become very excited about it. The idea is sooooooo seductive. At its simplest, M involves this:
  • - every time a trade is a loser, double the lot size for the next trade.
    - eventually, a trade will hit its take profit target and recoup all the losses from previous trades.
The problem is the 'Death Run'. Do a bit of mental arithmetic. Start with 0.01 lots and work out how many levels deep you have to go before your lot size is one that you would not even trade in your worst nightmares.

And then consider this; the next trade is double the size that was too hideous for your nightmare.

And then consider this unbreakable, unshakable, unavoidable law of Forex trading; if something can happen, and it is bad for you, then it will happen and probably sooner rather than later. If it happens later, then it was toying with you and you were ball-achingly lucky.

Garyfritz has emerged as this forum's National Treasure. Stunningly clever, mathematically adept at levels I cannot even begin to dream about and possessing an ability to analyse that makes me gasp, I asked him to do some M debunking.

Here is what he posted in one of our threads. As you read it, bear in mind Gary's conclusions which are these:
  • - In the vast majority of cases he studied, and in virtually ALL realistic cases, a Martingale system will not turn a losing system into a winner.
    - M applied to a successful trading system will increase the profits, but only at the cost of occasional heart-stopping drawdown. You can gain the same result by increasing the starting lot size of your trade and dropping M.
"I agree with Steve. If Martingales actually worked, every trader in the world would use them, and every trader would be a scadzillionaire.

I haven't tried proving this but I think it may be possible for a M to *improve* a system's results. E.g. it may be possible to turn a slight loser into a slight winner -- but at the cost of dramatically increased risk and a terrifying equity curve.

If you have infinite funds, so you NEVER EVER blow out of the M, then you could produce a closed-trade equity curve that looks like a straight line. But your open equity -- the actual day-to-day current value of your account -- is going to have frequent huge drawdowns. Since you have infinite funds, so you can survive any possible sequence of recovery trades, you don't care. You accept a tiny but guaranteed return on your infinite account.

But for those of us with finite funds -- which I suspect is most of us!! -- you can't guarantee you will NEVER blow out of the M. So you have to take enormous risks for small gains, and that's a very dangerous way to trade.

The exact progression of recovery levels -- 1.2.4.8.16, or 1.1.2.3.5.8.13, or whatever -- will change the results, but it will not change the M from "disaster waiting to happen" into "guaranteed money machine." It might reduce the chances of an inevitable blowup, but the blowup is still inevitable. The only question is whether you can survive the blowup, and whether the resulting profits (if any) after the blowup are worth the risk. I strongly suspect they aren't.

Martingales are "perpetual motion machines" that claim to create something from nothing. Unless you're Rumplestiltskin and can spin gold from straw, that doesn't work in the real world. Steve is right -- you're better off to learn how to trade, instead of relying on hocus-pocus to do the work for you.
"

Then his next post in the same thread:
"I decided I needed some numbers to back up my intuition.

I threw together a spreadsheet that does a crude simulation of a Martingale system. It simulates 1000 random trades, then trades it as you specify: win%, win size, loss size, M bailout point, M progression. It shows the results of trading those random trades without any Martingale, and trading them with your specified Martingale parameters.

BTW I should point out that while I was working with this, simulating a system with 50% wins, I saw one case where I got TWENTY-ONE CONSECUTIVE LOSSES. That's a freakishly uncommon result, a chance of 1 in 2 million, but it's the kind if thing that CAN happen. That's the kind of event that can kill you, no matter how well-capitalized you think you are for Martingale blowups.

So, what did I see in my simulations?

With a random coin-flip system -- 50% wins, win size = loss size -- the Martingale is also a coin-flip. Sometimes it helps, sometimes it hurts. No benefit that I can see. If the system result is random, the M results are random. This was true for all progressions I tried.

With a winning system, it gets more interesting. With 50% wins, a 2:1 W/L ratio, and bailout at level 5, the Martingale consistently made about 2.0-2.5x more profit. HOWEVER you could get very similar results without a Martingale, just by increasing your fixed position size by about 2.5x.

There are a zillion ways you could play the numbers, and I've attached the spreadsheet for your own experimentation. Let us know if you find a miracle Martingale. From my experiments, my conclusions are:

  • A Martingale will not turn a losing system into a winner, unless you NEVER hit the blow-up case. In a normal situation where you occasionally hit the bail-out level, the Martingale usually loses significantly worse than the basic non-M system.
  • If the system is a random coin-flip, the Martingale results will also be random. Sometimes better, sometimes worse.
  • If the system is a winner, a Martingale will increase the profits, but you have all the Martingale problems. You can achieve similar returns just by increasing your position size, and you will have fixed known risks, a more consistent and sane equity curve, and no worry of a Martingale blowup.
  • If you have a high-win% system, your chances of long strings of losses are reduced (but NOT eliminated). E.g. for a system that wins 70% of its trades, the chance of N consecutive losses is (1-70%)^N; the chance of 7 consecutive losses is 0.30^7 = 0.022%. If you can survive 6 consecutive losses, you have only 1 chance in 4500 of blowing up with 7 losses. But remember those 21 consecutive losses with 50% wins... If you never blow up, your closed-equity curve is a beautiful straight line. But once again with that 70% system you could get about the same results just by increasing your position size by about 1.7x, and you have none of the Martingale risks.
  • Bailing after a few consecutive losses results in smaller blowup losses, but the blowups happen more often. Bailing after more consecutive losses causes fewer blowup losses, but they're bigger. The end result tends to be similar even though the equity curves may look very different.
  • If you're lucky, you won't hit many blowups, and you'll make more money than without the Martingale. If you're NOT lucky, you may hit a lot of blowups, and you'll lose far more than you would have without the Martingale. If you're like me, the Trading Gods will guarantee you see the latter result, and the Martingale will kill you.

Therefore: I conclude that Martingale Mania is a will-o-the-wisp, a fantasy. Martingales hugely increase your risks without providing any benefit. Martingales won't help you unless your system is already a winner, and if your system is a winner, a Martingale is a bad way to increase its returns.

Just Say No to Martingales.
"

EDIT: Added the Martingale simulator spreadsheet -- Gary 6/20/12
EDIT: Major update, converted to fixed risk% instead of fixed position size -- Gary 7/2/13
You do not have the required permissions to view the files attached to this post.
Read the effing manual, ok?

Afterprime is the official SHF broker. Read about them at https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?p=175790#p175790.

I still suffer from OCCD. Good thing, really.

Anyone here feeling generous? My paypal account is always in the market for a tiny donation. [email protected] is the account.

To see The Weekly Roundup of stuff you guys might have missed Click here

My special thanks to Thomas (tomele) for all the incredible work he does here.
rob2360
Trader
Posts: 10
Joined: Thu Feb 09, 2012 10:59 am

Re: Martingale debunked by Gary

Post by rob2360 »

Great topic. Martingaling to create profits on a losing strategy - is exactly like what alchemists in the 1600s spent a lot of time trying to doing with lead, namely turn base metal into gold. They failed and in the long run you will too. As you use "M" you gotta ask yourself - 'do you feel lucky, punk, well do ya?'

The expected return on a Martingale version of a strategy is exactly that of the strategy - the key difference is do you want the losses spaced probabilistically in-line with the strategy (PL) or loss weighted power N, with a probability PL/N? all that means is you can have the losses as you go or play dodge-the-bullet and get the "big one" now and again. Overall expectancy should be equal. Having said that, the expectancy of any lottery is negative but that doesn't stop them lining up in droves to play...

For extra credit, are "recovery trades"by their very nature, Martingale?
garyfritz

Re: Martingale debunked by Gary

Post by garyfritz »

Yes, I would say so. I haven't seen any recovery setup that DIDN'T look like a Martingale.
User avatar
SteveHopwood
Owner
Posts: 9754
Joined: Tue Nov 15, 2011 8:43 am
Location: Misterton - an insignificant village in England. Very pleasant to live in.

Re: Martingale debunked by Gary

Post by SteveHopwood »

On t'other hand Recovery is self-limiting - something a lot of people forget. Scoobs robot in its various versions has this wrong.

Suppose Recovery is set to 4 levels i.e. 1.1.2.4, 1.2.3.6, 1.2.4.8 or whatever. If the market reaches what would be L5, then the original L1 trade is supposed to close, and L5 becomes the new L1 and matters proceed from there. The trader is supposed to try to manage the position back to an ultimate breakeven.

What is not supposed to happen is, the market hits max_recovery_levels + 1 price and the entire position closes - the 'bail out' option.

Mind, Gary might be able to debunk this as well. :lol:

:D
Read the effing manual, ok?

Afterprime is the official SHF broker. Read about them at https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?p=175790#p175790.

I still suffer from OCCD. Good thing, really.

Anyone here feeling generous? My paypal account is always in the market for a tiny donation. [email protected] is the account.

To see The Weekly Roundup of stuff you guys might have missed Click here

My special thanks to Thomas (tomele) for all the incredible work he does here.
garyfritz

Re: Martingale debunked by Gary

Post by garyfritz »

Not unless/until I understand it. :lol:

So if you're trading 1.1.2.4 and you have L1/L2/L3/L4 open and hit the next level... what do you mean by "L5 becomes the new L1" ? Close the 1.1.2.4 positions (a normal bailout exit) and start a new recovery with the "L5 = new L1" trade at a size of 1?
What is not supposed to happen is, the market hits max_recovery_levels + 1 price and the entire position closes - the 'bail out' option.
That's the trouble with Martingales. The market has a way of exceeding max_recovery_levels far more often than you'd ever dream. Kind of like those "100 year floods" that happen every 5-10 years...
rob2360
Trader
Posts: 10
Joined: Thu Feb 09, 2012 10:59 am

Re: Martingale debunked by Gary

Post by rob2360 »

Nothing here to report. Move along now..
Last edited by rob2360 on Mon May 07, 2012 9:42 pm, edited 1 time in total.
User avatar
SteveHopwood
Owner
Posts: 9754
Joined: Tue Nov 15, 2011 8:43 am
Location: Misterton - an insignificant village in England. Very pleasant to live in.

Re: Martingale debunked by Gary

Post by SteveHopwood »

garyfritz wrote:Not unless/until I understand it. :lol:

So if you're trading 1.1.2.4 and you have L1/L2/L3/L4 open and hit the next level... what do you mean by "L5 becomes the new L1" ? Close the 1.1.2.4 positions (a normal bailout exit) and start a new recovery with the "L5 = new L1" trade at a size of 1?
No.

Trade 1 is the initial trade - L1. RecoveryLevels represents the max number of levels, i.e. 3 on a 1.2.6 system, 4 in a 1.1.2.4 systerm. 5 om a 1.2.3.6.9 system etc

So, we have open an L1 trade - the initial trade. Subsequent Recovery trades go through to, say, L4.

Taking, say, 1.1.2.4 as our Recovery structure and starting at 1 lot, we have open:
  • - L1 1 lot.
    - L2 1 lot.
    - L3 2 lots
    - L4 4 lots
Then, the market continues to retrace against us and hits L5. L1 trade closes and L6 opens, but at the L1 lot size, so we have:
  • - L2 1 lot.
    - L3 2 lots
    - L4 4 lots
    - L5 1 lot.
Market continues to retrace against us and hits L6. L2 trade closes and L7 opens, but at the L2 lot size, so we have:
  • - L3 2 lots
    - L4 4 lots
    - L5 1 lot.
    - L6 1 lot
Market continues to retrace against us and hits L7. L3 trade closes and L7 opens, but at the L3 lot size, so we have:
  • - L4 4 lots
    - L5 1 lot.
    - L6 1 lot
    - L7 2 lots
And so on. There should never be more than 4 trades open in a 1.1.2.4 Recovery, or 3 in a 1.2.6, or 5 in a 1.1.2.4.8 etc Whenever the market reaches RecoveryLevels + 1, the oldest trade is closed and its lot size becomes the lot size for the new trade. Traders are then supposed to manage the situation manually, even if they were using an EA to send the first RecoveryLevels trades.

At the time I coded The Beast and all its variants, I could not see how to code the rolling stuff, but I probably could now. Easy, it aint - and that is when you actually grasp the concept in the first place, which most people fail to do. The variants of Scoobs FR miss the point here - they simply bail out at a predetermined level rather than managing the position as I describe.

Sing out if this is still not clear; it is not easy to explain.


:D
Read the effing manual, ok?

Afterprime is the official SHF broker. Read about them at https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?p=175790#p175790.

I still suffer from OCCD. Good thing, really.

Anyone here feeling generous? My paypal account is always in the market for a tiny donation. [email protected] is the account.

To see The Weekly Roundup of stuff you guys might have missed Click here

My special thanks to Thomas (tomele) for all the incredible work he does here.
User avatar
hiredwhip
Trader
Posts: 90
Joined: Tue Nov 15, 2011 7:36 pm
Location: Mountains, East Coast USA

Re: Martingale debunked by Gary

Post by hiredwhip »

Steve and I both gave this young man a vouch for this.....Have a read
You do not have the required permissions to view the files attached to this post.
A rainy day is of enormous value to both the farm and the farmer....Stay Hungry. Stay Foolish....
garyfritz

Re: Martingale debunked by Gary

Post by garyfritz »

SteveHopwood wrote:Then, the market continues to retrace against us and hits L5. L1 trade closes and L6 opens, but at the L1 lot size, ...
Market continues to retrace against us and hits L6. L2 trade closes and L7 opens, but at the L2 lot size, ...
I see. At first glance I don't see how this could be much/any better than a basic L1/L2/L3/L4/bail Martingale. You never have a "full" bail-out loss, but you get roughly the same effect one loss at a time. In fact it seems like it would be worse, since e.g. a standard Martingale only has one "band" worth of loss on the large L4 position, whereas this approach would let the market move 4 "bands" against it before you closed out the L4 position -- for a much larger loss.

Seems to me that it would be a "death by a thousand cuts" instead of "off with his head." Maybe less gruesome, but equally fatal.

In my copious free time I'll see if I can modify my simulation to support that kind of "recycling" Martingale. I'm not sure the crude model I used will support it, but I'll look.
User avatar
SteveHopwood
Owner
Posts: 9754
Joined: Tue Nov 15, 2011 8:43 am
Location: Misterton - an insignificant village in England. Very pleasant to live in.

Re: Martingale debunked by Gary

Post by SteveHopwood »

I am not in the best position to argue for it either, as I have never properly managed Recovery to its conclusion - this is why my account is hedged to buggery and should come out when I finally manage to learn to trade properly. About 2020. Bob could probably make a better case, but he is a tad busy right now. :lol:

:D
Read the effing manual, ok?

Afterprime is the official SHF broker. Read about them at https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?p=175790#p175790.

I still suffer from OCCD. Good thing, really.

Anyone here feeling generous? My paypal account is always in the market for a tiny donation. [email protected] is the account.

To see The Weekly Roundup of stuff you guys might have missed Click here

My special thanks to Thomas (tomele) for all the incredible work he does here.
Post Reply

Return to “Automated trading systems”