Topic: random entries

I came up with an idea for a new strategy.

There are two assumptions behind it:

1. technical analysis doesn't work
2. fundamental analysis doesn't work

thus:

1. there is no reason to develop more technical indicators
2. there is no reason to waste time on fundamental analysis

consequently, it is worth trying different approach and in this case we are talking about random entries.
if the strategy wont be profitable it still might be useful since it could be used to evaluate the efficiency of the money management ratio.

random entries could be divided into three sub-categories:

- "totally" random entries: a program would open unspecified number of positions for any given time period (regardless of whether its 1min bar, 5min bar, 1h bar or any other time period) starting from the single position and keep adding/scaling until it reach the maximum level of risk allowed (for instance 1%, 10%, 1lot, 5lots etc.). since the strategy is based on pure randomness entries should be random regardless of whether the previous position/positions is winning or loosing. i know there are some libraries for C that deal with it. by implementing this concept we should be relatively independent of the time period, which put significant boundaries over randomness.

- "semi" random entries:  a program would open a limited number of random positions (1,2,3...) for any given time period (regardless of whether its 1min bar, 5min bar, 1h bar or any other time period) until it reach the maximum level of risk allowed (for instance 1%, 10%, 1lot, 10lots etc.) based on pure randomness. this method will allow us to test whether any particular time period is more efficient for this strategy than other periods.

- "pseudo" random entries: a program would open just one position for any given time period until it reach the maximum level of risk allowed. this might be quite useful for any money management test.

if anyone is willing to program the strategy please dont hesitate to post in this topic.

Re: random entries

That's an interesting idea. All that's really needed is a new opening logic condition.

Re: random entries

You can experiment with Random Filter.

http://i.imgur.com/kTDww4O.png

Re: random entries

I have some more ideas to play with randomness but at this stage we need to find out whether this basic concept can be easily implemented to be used on FSB.

Re: random entries

Popov wrote:

You can experiment with Random Filter.

Thank you, I will.
My initial impression is that this indicator is able to open only one position either on bar opening, bar close or on similar conditions. Since this feature reduces the entire 'randomness concept' it may not be exactly what i'm looking for.

Re: random entries

If you use daily or hourly charts, sticking to bar open may be a limit, but if you go to 1 min chart it may not be an issue. Especially if you set relatively small probability for entry 1-5%.

Re: random entries

Here I set 1% chance for entry (50% long/short) and 1 % chance for exit

I forced program to recalculate by pressing F5. It gives different result at every calculation.

http://i.imgur.com/jz8n9he.png

Re: random entries

you genius developers are getting way too esoteric for jgp to follow.  But I have been playing around with developing FST strategies wherein the simple entry logic is essentially over-ridden by a manual [buy/sell] entry based upon visual inspection of price action combined with visual inspection of values of multiple indicators on multiple time frames [like iPanel Indicators ].  So what matters thereafter is how that trade [particularly the exit] is managed by the FST attached to that chart.  Seems to me that there may some way to semi-randomly exit the trade at the end of the bar, based upon the price action during that bar or [2] at the end of the next bar, based upon the price action during that next bar or on those two bars, or [3] at the end of the third bar based upon the price action during that third bar or during those three bars usw, usw.  I don't understand what you doing here Master Popov with this random entry magic, but can a random exit logic condition be developed and somehow be optimized on previous data streams?  Once again I apologize for not understanding what you're doing, now, or perhaps have already done.   


http://forex-indicators.net/mt4-indicators/ipanel_indicators

9 (edited by automated 2013-02-28 11:34:50)

Re: random entries

Popov wrote:

If you use daily or hourly charts, sticking to bar open may be a limit, but if you go to 1 min chart it may not be an issue. Especially if you set relatively small probability for entry 1-5%.

I have been playing around with this strategy and it gives some interesting results when combined with moving averages (I'm considering the MA more like a filter than the indicator). I believe this concept might be further improved if we add some statistics indicators, such as:

-enter when the average bid/ask spread (X period) is < than (Y)
-enter after (X bars) in the same direction
-exit after (% of previous day volatility measured in PIPS)
-exit when average volatility (X period) is > than median volatility (Y period)

more advanced version of this concept would calculate the average successful combination of methods used and simply keep opening positions until it either hit the maximum level of risk allowed or the average win/loss ratio becomes negative (in such case it could reverse the strategy).

the goal is to avoid using technical indicators and rely more on probability and statistics.

btw. the platform is sometimes crashing when i refresh or change the indicator

10

Re: random entries

automated wrote:

... I believe this concept might be further improved if we add some statistics indicators, such as ...

If you are interested in some reading on this subject, some of the work published by Bill Williams on using chaos theory may offer some assistance in finding a suitable 'filter' for your random entry/exit ideas.  Try the reference at http://www.alpari-us.com/en/alpari_acad … chaos.html as a starting point.

There is some really interesting research into working with random techniques if you search around, and don't mind reading through a lot of detail on math theory.  The best example of what may be possible that I have been able to find is by Guy R. Fleury, with the following links making good starting points:
  http://alphapowertrading.com/index.php/ … g-the-game
  http://alphapowertrading.com/index.php/ … he-game-ii
  http://alphapowertrading.com/index.php/ … e-game-iii

There is an excel file on the last of these pages that has simulator for a randomized model using stock data, that may form a decent starting point for further study.  I don't believe that it would be a very difficult task to convert the excel math functions into a set of custom FSB indicators for further study if the idea was something that interested a few of the users.

NOTE: I have no relationship with the owners of the links above are please don't take the above as an endorsement of a product or service, these are links from my personal stash that I have read in the past, and yes ... I do read too much smile

Re: random entries

Ab, one never reads too much wink And thanks for the links!

I've read about similar approaches in other, bigger fx forums. The somewhat gurus of random entry techniques have coined it a coin toss system, so by default you have 50/50 chance of winning or losing. Interesting about it is that they claim to trade a coin toss system profitably, but at the same time they have a clear entry, which, to my understanding, defines the system as not a coin toss system.

Automated, you can pretty much use standard FSB indis as additional statistic indicators.

Re: random entries

Ab, thanks for the links.

It seems that Mr. Fleury is trying to apply the concept of randomness to the buy&hold strategy. It is interesting indeed but it will be even more interesting to see how does this strategy perform if we would choose 10 random stocks/currency pairs in a strong downtrend. I may understand that in case of stock market it has more chance to succeed as "stocks always rise" if you diversify them well, but unfortunately its not always the case in forex.

What might be worth the effort is to test Mr. Fleury's concept of 'random' diversification. We would need a platforms that is able to simulate a portfolio of at least 10 different instruments. As far as I'm concerned its not possible in FSB at the moment..

13 (edited by automated 2013-02-28 15:03:27)

Re: random entries

footon wrote:

Ab, one never reads too much wink And thanks for the links!

I've read about similar approaches in other, bigger fx forums. The somewhat gurus of random entry techniques have coined it a coin toss system, so by default you have 50/50 chance of winning or losing. Interesting about it is that they claim to trade a coin toss system profitably, but at the same time they have a clear entry, which, to my understanding, defines the system as not a coin toss system.

Automated, you can pretty much use standard FSB indis as additional statistic indicators.

footon, could you please provide a guidance on how may I implement the following:

-enter when the average bid/ask spread (X period) is < than (Y)
-enter after (X bars) in the same direction
-exit after (% of previous day volatility measured in PIPS)
-exit when average volatility (X period) is > than median volatility (Y period)

using standard FSB indicators ?

Re: random entries

-enter when the average bid/ask spread (X period) is < than (Y)

As you most probably know, there is no bid/ask in FSB, but it can be doable in FST.

-enter after (X bars) in the same direction

You can look the thread in the link to get ideas. http://forexsb.com/forum/topic/19/three … -in-a-row/

-exit after (% of previous day volatility measured in PIPS)
-exit when average volatility (X period) is > than median volatility (Y period)

Define volatility first, then we can look at the implementation.

15

Re: random entries

re: volatility - can FST count "ticks per unit time"? can FST count "downticks per unit time" and "upticks per unit time"? If so, would it be possible to develop a "spike" exit condition?

Re: random entries

jgp wrote:

re: volatility - can FST count "ticks per unit time"? can FST count "downticks per unit time" and "upticks per unit time"? If so, would it be possible to develop a "spike" exit condition?

Actually Volume in MT data is a tick counter in essence, so in a way you can already do it. Downticks and upticks needs a custom approach, but using standard volume you can trade higher TFs on M1 to have a similar effect, practicality of my suggestion is an unanswered question of course.

17

Re: random entries

to be useful an excess volatility [or spike] exit would have to occur as soon as possible after the spike is detected, not even waiting till the end of the bar, not even for a 1min bar.  just to be safe, if such a excess volatility can be developed, it would probably be best to close an open position no matter whether it is an excess up or down volatility.  the "excess"  ticks per sec should be adjustable and be some multiple of the average tps of previous bars.

Re: random entries

if such a excess volatility can be developed, it would probably be best to close an open position no matter whether it is an excess up or down volatility.

This is very easy to be added in the expert. It can be controlled from a starting option.

19

Re: random entries

I'm assuming, of course, that these sudden large spike or breakout price moves that seem to occur so frequently for whatever reason are positively correlated with detectable 'excess' tick-per-second activity. but maybe they aren't?

20

Re: random entries

I'll try again to submit this reply.  don't know where the previous one went.  I gather Master Popov said it would be easy to add an "excess tick activity" condition to the MT4FST Expert "input" conditions [such as SL and MinAcctBal] that would essentially close a trade as soon as that input condition was reached.  Is there any way through FSB we could get some idea of what that sudden excess tick activity setting should be??

Re: random entries

jgp wrote:

I'll try again to submit this reply.  don't know where the previous one went.  I gather Master Popov said it would be easy to add an "excess tick activity" condition to the MT4FST Expert "input" conditions [such as SL and MinAcctBal] that would essentially close a trade as soon as that input condition was reached.  Is there any way through FSB we could get some idea of what that sudden excess tick activity setting should be??

As I wrote earlier, try Volume indi.

22

Re: random entries

I don't understand.  I want something that will close a trade asap after an "excess' tick per second activity is reached, not at the end of bar, which might be  an hour or a day or a week later.  Something like the Trailing Stop Limit closing point, which is not predicated on a bar closing, but presumably closes a trade asap after a SL or TP is reached.  At least on my FSB there is no "Volume" closing point  indicator that can be used like "Trailing Stop Limit".  If an "excess" tick per second setting could be added to the Inputs tab of the MT4FST expert, that would provide the needed closing when reached [not at the end of some bar] but how could user-members determine what that setting should be? That's my question.

Re: random entries

I'm confused!

jgp wrote:

I want something that will close a trade asap after an "excess' tick per second activity is reached, not at the end of bar, which might be  an hour or a day or a week later.

jgp wrote:

Is there any way through FSB we could get some idea of what that sudden excess tick activity setting should be??

jgp wrote:

At least on my FSB there is no "Volume" closing point  indicator

Well, it is a close filter, you can get a rough idea.

24 (edited by jgp 2013-03-13 18:35:21)

Re: random entries

but I'm talking about god-only-knows-what-caused sudden "excess" tick/millisec spikes [which may or may not be correlated with price spikes] that can and do occur well within a 1min time bar. My uninformed guess would be that anytime [in an open trade] the #ticks/time increases a factor of 2 or 3 over the previous #ticks/time, I would want my EA to exit that trade asap.  But what do I know?

Re: random entries

Thanks for clearer post! I finally get it what you want to achieve. smile

Like the developer said, it must be added to the expert. Trading higher TF's and scanning through volumes measured in seconds/minutes interval is hard to achieve as an indicator. And even harder to test it in FSB. Only thing I can think of (which I thought previously, too) is setting up a level, a threshold defining a volume spike in other words, to close a position if that level is exceeded. This can be done on the lowest of timeframes in FSB, just to get a feel for it. Your approach is more dynamic one.