[yaniv_av]

Hi All!

I have just finished developing a new system, base on martingales.
Now, I know, some of you would say immediately, without even hearing the concept of the system, “you’ll finish up dead”.
But what would you say if I tell you that I tested the system on 1 minute data, since December 1999 (I simply don’t have older data…)?
That means almost 10 years of 1 minute bars - more than 3.5 bars in the test!!
In the worst case, the system was needed to double up 8 times.
No 9 doubles since 1999 (and probably never…)
Now, with initial account of 1000$, and first trading amount of 0.02 lot (mini account), this system survives since 1999 until now with a profit of 23,000$, which is 2200%!!! (See first chart). All the profits ware reinvested to the account.
Of cource I can start with account of 2000$ with first trading amount of 0.02 lot, which means that I can survive 1 more doubling - up to 9 doubles, and the profit will be “only” 1300%...

I just want to hear your opinion about this, because indeed – it looks too good…

[Elovv]

It seems fantastic!!

[PrymeTyme]

hmmm


yea looks interresting ..
but waht if it doubles up 8 times at my very first trade ?

yea looks like it wont happen .. but you never know

i simply dont trust mister martingale...

just my 2 cents

[Cobra]

Yaniv

So now that you have our attention would you mind telling us when do you start trading on the 1 min and when and where do you double up. Which currency, indicators ( if any) etc. ?

A chart would also help....

thx.

[Cobra]

Originally Posted by PrymeTyme

hmmm

Originally Posted by PrymeTyme

yea looks interresting ..
but waht if it doubles up 8 times at my very first trade ?

yea looks like it wont happen .. but you never know

i simply dont trust mister martingale...

just my 2 cents

Why do you say it looks like it wont happen? Based on what will it not happen? How many pips apart did you double up?

[Davidee]

Originally Posted by yaniv_av

Hi All!

I have just finished developing a new system, base on martingales.
Now, I know, some of you would say immediately, without even hearing the concept of the system, “you’ll finish up dead”.
But what would you say if I tell you that I tested the system on 1 minute data, since December 1999 (I simply don’t have older data…)?
That means almost 10 years of 1 minute bars - more than 3.5 bars in the test!!
[font=Arial]In the worst...

If it's using 1 minute charts I wonder how profitable it will be after the cost of trading, sounds like there will be an awful lot of trades with it for not many pips profit? Can you give me more details about the system please, then perhaps I can comment further. I also instinctively dislike the idea of doubling up on a losing streak.

[Knacky]

Originally Posted by yaniv_av

Hi All!

I have just finished developing a new system, base on martingales.
Now, I know, some of you would say immediately, without even hearing the concept of the system, “you’ll finish up dead”.
But what would you say if I tell you that I tested the system on 1 minute data, since December 1999 (I simply don’t have older data…)?
That means almost 10 years of 1 minute bars - more than 3.5 bars in the test!!
[font=Arial]In the worst...

Very important is to watch for the modeling quality , which in this case is 25% . This is very low. Only backtests with modeling quality 90% worth .

[hanover]

Originally Posted by yaniv_av

I just want to hear your opinion about this, because indeed – it looks too good…

ALL martingales look 'too good' because they guarantee a 100% winning sequence...... until the 'death trade' double-up wipes out your account.

No disrespect, but there's no point in asking for others' opinions, because nobody can tell you when a 1-in-a-million event will occur. It might not be until the year 2050, or it might be tomorrow; the probabilities of each are equal. However, the more often a martingale-based system trades, the more frequent the gains, but the higher the probability of wipeout: a 1-in-a-million event is more likely to occur in a million trades than it is in 100 trades.

There's nothing wrong with trading any system, as long as you're aware of the return-versus-risk trade-offs involved, and are willing to live with the consequences. In the case of martingale, it's a 99.9999% win rate, for a 0.0001% risk of total wipeout. There is no shortcut to success; no return without some kind of risk.

I posted my thoughts on Martingale in some more detail recently here. The whole thread is a good read, IMHO.

[EDIT] 2700% in 10 years (120 months) equates to ~ 2.8% compounded per month, and without the risk of Martingale. So an alternative question might be: is it possible to average 2.8% per month, using 'conventional' MM?

[kaas]

The same trick can be used for roulette (in theory because normal roulette has a table limit).

The problem is indeed the risk. I am not a math guru, but i will try to explain.

The roulette system says put 1$ on black, if it is read , put 2$ on black, if it is read again, put 4$ on black. The chances that it is four times red are smaller than 2 times etc etc.

OK, in theory the chance that it is black (withour zero and double zero) is 50%

so 2 times is : 0.5 = 50%
so 3 times is : 0.5*0,5 = 25%
so 4 times is : 0.5*0.5*0.5 = 12.5%

So the occurence in the events are declining, however beacuse you are doubling up, the losses are doubling up, therefoure you can never win.

In theory you can go on forever if there is no (zero and double zero) if you have deep anough pockets, however if you can only double up 8 or 9 times you will go broke.

The same for doubling up on Forex, the chances are low that it can occure, however the losses are very high. Forex is a zero sum game, and especially with spreads, slippage and/or commission youre risk increases more than 50% when you double up and the rewards are less than 50% .

The 'house' always win with this kind of games.

[SamTa]

Trust me, if you forward test this system, it will not give you the same results.
Especially with 1M TF, I would probably won't say this if it was on bigger timeframes.

Good luck !

[yaniv_av]

Thanks for your comments.
I’ll try to answer some of your questions here.
First, just want to emphasize – I am dealing with martingale systems a lot of times. Indeed, when I back tested any one of them with huge past data (since 1999), eventually, none of them has survived - except this one.
That’s why I’m optimistic about it.
Of course, a “black swan” can happen anytime. The real question, as some of you mentioned, is the risk via reward. Now, given these numbers I wrote, I think that 2700% if a fare reward.
Consider the following conservative approach - You can wait until the account will double itself (~2 years in average), then withdraw the profit and start again. If a “black swan” will happen 1 in the next 10 years (again – in the past 10 years it never happened) you still will end up with a profit of 300% for 10 years, which is 2.5% per month – not compounded (without capitalization – since the profits aren’t reinvested)
With aggressive approach, where all the profits are reinvested, the profit is 2.8% compounded per month for 10 years!! I don’t know any trader who can achieve consistently these numbers for this long time… maybe in a very good year. But 2.8% per month for 10 years?? Not realistic in my opinion.

Now for your comments:

“So now that you have our attention would you mind telling us when do you start trading on the 1 min and when and where do you double up. Which currency, indicators ( if any) etc. ?”

This system is trading the 1M chart, on the EURUSD only, with 2 pips spread.
I’m sorry, but for now I’m not going to elaborate this system internals.

“Why do you say it looks like it wont happen? Based on what will it not happen”

Well, It’s not so simple. The market has a movement’s patterns. I assume you’ll agree that for example it can’t move in one direction 1000 pips, without any correction of 10 pips on the way, right? I exaggerated with the example of course but this is the heart of the system. It consists on the thesis that the market must make some corrections on the way.
So you can’t say that the past isn’t relevant at all. Statistically speaking, you can’t say that these are independent variables. The past is very relevant here. Of course, as I say, the past doesn’t guarantee that this “black swan” won’t happens in the future, but statistically, the odds are very small.

“If it's using 1 minute charts I wonder how profitable it will be after the cost of trading, sounds like there will be an awful lot of trades with it for not many pips profit? Can you give me more details about the system please, then perhaps I can comment further. I also instinctively dislike the idea of doubling up on a losing streak.”

Well, the system works on 1M chart, but it trades with TP/SL of about 100 pips. So
in 10 years, this system opens 7081 trades including the doubling. This means 60 trades per month (again – including the doubling!) So this is not too much…

“Very important is to watch for the modeling quality , which in this case is 25% . This is very low. Only backtests with modeling quality 90% worth .”

As far as I know, the modeling quality of 1M data is always 25%, so it’s not relevant. If I’ll produce 1H data from these 1M data, I’ll get 90% modeling quality.
The data is from Alpari data bank.

Hanover, I this I answered most of your comment in my first section.
And again, “an alternative question might be: is it possible to average 2.8% per month, using 'conventional' MM” – for my opinion it’s not realistic to achieve 2.8% per month consistently for 10 years…

“The same trick can be used for roulette (in theory because normal roulette has a table limit).”

Kass, as I mention above, this is totally different from roulette. In roulette, the events are completely statistically independent. Here, this is not the case since the market has patterns, so the past is related to the future. (Like my previous example – it can’t move 10000 pips without a correction of 10 pips on the way…)

“Trust me, if you forward test this system, it will not give you the same results.
Especially with 1M TF, I would probably won't say this if it was on bigger timeframes.”

I forward test this system for 3 months, and it trades exactly as when I back tested it on the same data…
Since it works on market movements, I think the backtests + real trading will be almost the same.

[kaas]

Kass, as I mention above, this is totally different from roulette. In roulette, the events are completely statistically independent. Here, this is not the case since the market has patterns, so the past is related to the future. (Like my previous example – it can’t move 10000 pips without a correction of 10 pips on the way…)

I agree that the market has patterns en certain behavior, otherwise it was pure luck/gamble, however my comment was regarding the double-up method.

[Sauron]

Nothing is certain, only the change so there's no way we could assure you that the system is going to work. You seek for an answer which doesn't exist.
If it looks good on paper just try it to see what will happen. If the black swan hits at least you tried. If it never hits you'll be rich.

Make a test with other pairs and tell us the results.

[yaniv_av]

Originally Posted by kaas

Kass, as I mention above, this is totally different from roulette. In roulette, the events are completely statistically independent. Here, this is not the case since the market has patterns, so the past is related to the future. (Like my previous example – it can’t move 10000 pips without a correction of 10 pips on the way…)

I agree that the market has patterns en certain behavior, otherwise it was pure luck/gamble, however my comment was regarding the double-up method.

Well, as I wrote before:

Consider the following conservative approach - You can wait until the account will double itself (~2 years in average), then withdraw the profit and start again. If a “black swan” will happen 1 in the next 10 years (again – in the past 10 years it never happened) you still will end up with a profit of 300% for 10 years, which is 2.5% per month – not compounded (without capitalization – since the profits aren’t reinvested)

This risk reward ratio seems very fare to me...
Of course with roulette you'll end up dead. But here, you can blow your account 2 times in the next 10 years (while until now even 1 event like this didn't happened), ans still end up with a profit of 200%... (according to the conservative approach I described)
And BTW - the numbers I gave are after spreads...
Oh, and the "house" in forex (the broker) only interested in the commissions (spreads), not like in the casino where it want all your money (I'm speaking about brokers who covers against themselves...)

[Ucglobal]

so suystem does not have a loss obviously, but what is the max drawdown in the 10 years tested?