Topic: FSB v4.2.0.3 unstable - beta tests

FSB v2.4.0.3 - 1 Unstable, released on 2007-04-15


Changes with Forex Strategy Builder 2.4.0.3 - 1 Unstable

   * Donchian Channel indicator - added
   * Take Profit - added (2000 pips limit)
   * Stop Loss  - added (-2000 pips stop)
   * The range of Stop Limit increased to +/- 2000 pips


Edit:

FSB v2.4.0.3 - 2 Unstable, released on 2007-04-18

Changes with Forex Strategy Builder 2.4.0.3 - 2 Unstable

  * More symbols - added
  * Take Profit - logic improved
  * Stop Loss - logic improved
  * Stop Limit - logic improved
  * Trailing Stop -  logic improved
  * Close logic conditions - only when the "Closing point of the position" is "Bar Closing"


Edit:
    The download link is closed!

Re: FSB v4.2.0.3 unstable - beta tests

A possible usage of Stop Loss:

Forex Strategy Builder V 2.4.0.3-1 Unstable
Strategy name: MACD crossover and Stop Loss
Exported on: 4/15/2007 7:00:31 PM

Market: EURUSD 1 Day
Spread: 4 pips
Swap Number: 1 pip

Balance: 4390 pips
Minimum account: -245 pips
Maximum drawdown: 1293 pips
Time in position: 93 %

The same direction signal - does Nothing
The opposite direction signal - Reverses the position

[Opening point of the position]
Bar Opening
     Enter the market at the beginning of the bar
     Base price  -  Open

[Opening logic condition]
MACD
     The MACD line crosses the zero line upward
     Smoothing method  -  Exponential
     Base price  -  Close
     Slow MA period  -  21
     Fast MA period  -  14
     Signal line period  -  9
     Use previous bar value  -  Yes

[Closing point of the position]
Stop Loss
     Stop Loss  -  -200

This strategy is included in the distribution.


In the shown example the main logic is crossover of the MACD.
It is determined here: The opposite direction signal - Reverses the position

The Stop Loss level limits the maximum loss to 200 pips (plus the spread and the rollover fee)

Re: FSB v4.2.0.3 unstable - beta tests

Open bug# 1  - The Stop Loss conditional orders have to be canceled if the target order has been canceled. Now they present in the journal as Confirmed.
Edit: Ops. When the target entry order is canceled the exit order is canceled too. The problem is when the entry order is executed and then the exit order becomes confirmd. It has to close the position at the SL level, but if the position has been reversed earlier this exit order is not canceled.
If (like in the shown example) we have two exit points (SL and the revrsing) FSB sends two exit orders. When the first is executed, the other has to be canceled.

4

Re: FSB v4.2.0.3 unstable - beta tests

Now that I have tried my hand on FSB, and realizing that its just as easy as click and select then the parameters. This present a good opportunities for using genetic algorithm to find profitable strategy faster than human could.

If you don't have the time to implement genetic algoritjm, may I suggest make FSB to randomly chose combinations of technical indicators and it parameters and write down the profitable combination in a file for latter review.

It may even discover logics that no man have even attempted and it just as easy as leaving the computer running. I believe this feature could be easily implemented, may I ask that it be included in the earlierst next version release.

Re: FSB v4.2.0.3 unstable - beta tests

I've worked on a genetic algorithm 5 years ago. It was a huge program written on Visual Basic and representing the results in Excel.
This program develops 6 independent models that vote a signal every bar. The models (depending from the results) can evolve, "marry" each other, bear a child model (whit a slight genetic  mutation). The end result of this was huge over optimization.

I will include  something like this in FSB. The function is called "Strategy Generator". It will be released in next 6 month.

For now I what to finish the website. Also the online help is a mess yet.

Re: FSB v4.2.0.3 unstable - beta tests

Any updates on the Strategy Generator? I can't see it in the unstable version and it does sound very interesting. I've been trying a few strategies out in the FSB. Can I ask which post/strategy appears the hot-pick at the moment please? It is a little tricky to find specific strategies by poll/seniority.

Thanks.

7 (edited by andonov 2007-04-18 10:03:50)

Re: FSB v4.2.0.3 unstable - beta tests

re:kabuse
i'm a new member, but as i see the purpose of the FSB program is everybody  h i m s e l f  to test strategies, before using them
in real market. and (the purpose is) not to rate specific strategies publicly best. so everyone could have his favourite strategies,
test them and then use them. arrividerci!

8

Re: FSB v4.2.0.3 unstable - beta tests

In the future version can, we have a feature do download the stock and index data directly from Yahoo  and save them in a seperate folder for stock.

I would also like to see live streaming forex data and BUY/SELL signal trigger by the strategy we build , to access if it "work" in the real world. It would be good if FSB also keep track on it buy and sell and track it profit and loss and give some statistic how the system is performing.

Before being linked to forex broker for online trading, a live fx data would make FSB useful as an advisor to people who still trade fx manually. This will help guide them on the BUY and SELL strategy they developed.

For a start FSB may use the data acquired by Meta Trader ( all the data are stored in a folder inside the Meta Trader { allow user to specify where to look for this folder} ) in that way we could make FSB "live" with borrowed streaming data.

Re: FSB v4.2.0.3 unstable - beta tests

This with the Meta Trader saved data files is very interesting. I'll check this now!  Tanks!

Edit:
Yes, you are right. I found data in .hst format. Do you know something about this format and de period of updating?

10

Re: FSB v4.2.0.3 unstable - beta tests

I open it with MS Excel and it look like a csv format that have a different extension name.

For export, I manually open it with MS Excel and resave it into the FSB folder under csv extension and delete the "m" xxxxm60 to make it comply with the FSB naming convention.

Re: FSB v4.2.0.3 unstable - beta tests

Well! I have to integrate a converter in FSB.

12

Re: FSB v4.2.0.3 unstable - beta tests

Opps......it was exported using Meta Trader export function  and not via MS Excel.

Re: FSB v4.2.0.3 unstable - beta tests

I'm not sure about the rules of trading of the indexes, shares and metals quoted by MetaTrader. It is not difficult to include it in FSB in the next versions.

I cannot find info how to decode the .hst format :-(

14

Re: FSB v4.2.0.3 unstable - beta tests

he optimizing function in the unstable did not perform like that in the stable function, it fail to find the optinum value the stable function found.

When I tranfer the strategy file from the stable version to the unstable version, the unstable version gave the same result. Hence I have to conclude that the is some issue with the coding of the optimizing function in the unstable function.

Re: FSB v4.2.0.3 unstable - beta tests

The optimizer uses no liner, random algorithm to find the extreme value. You have to try several times. This algorithm can reach maximum optimization for minimum time, but it may not succeed in the first try.

The best usage of optimizer is to start it 2-3 times. After that if the values are near to the limits adjust it and try again 2-3 times.

If the result is not so good, you have to start from the initial values (cancel the optimization and again).

You can test the strategy for curve fitting by changing the parameters one by one a little bit up and down. If there is a significant change of the performance the strategy is over-optimised.

16

Re: FSB v4.2.0.3 unstable - beta tests

Now I know why the optimizer behave this way, for the stable version, it do charge any cost for Spread & Swap whereas in the unstable version it is 4 pips for Spread and 1 for Swap Numbers. [ Maybe the user have to be warn to include cost of transaction in the attempt to come up with wining strategy, else it don't work in real trading)

No wonder it was so easy to make money using the stable version. I was "making" in excess of 70K pips for EURUSD M1 in 20 days

But still FSB is the easiest way for a novice like me who don't fully under what spread & swap number are, to come up with a wining forex strategy.

For USDCHF Daily it make 34,670 pips over ~30 yrs averaging  1,100 pips per year and the profit line is a straight line up. Using risk managing technique ( a feature I had earlier requested) I am sure it could grow to a much higher number than this.


Forex Strategy Builder V 2.4.0.3-1 Unstable
Strategy name: Daily_30K_AllCurrency
Exported on: 4/18/2007 2:00:48 AM

Market: USDCHF 1 Day
Spread: 4 pips
Swap Number: 1 pip

Balance: 34670 pips
Minimum account: -17 pips
Maximum drawdown: 589 pips
Time in position: 44 %

17

Re: FSB v4.2.0.3 unstable - beta tests

What is this 0 pips spread in the Stable and 4 in Unstable?!?
The spread and the swap values are of your choice.  When you save your strategy with 0/0 spread/swap it will be loaded next time with the same values.
I think the default values are 4/1. But of course you can override your default strategy (New.xml) with other numbers.

Be careful!!

Re: FSB v4.2.0.3 unstable - beta tests

Melvin is right.

We need some kind of warning when are doing something unusual (unsafe).
-- zero swap
-- zero spread
-- Manual operating with "Use prev. bar value"
-- some non-logical indicators (if any)!
...

Re: FSB v4.2.0.3 unstable - beta tests

FSB v2.4.0.3 - 2 Unstable, released on 2007-04-18

Changes with Forex Strategy Builder 2.4.0.3 - 2 Unstable

  * More symbols - added
  * Take Profit - logic improved
  * Stop Loss - logic improved
  * Stop Limit - logic improved
  * Trailing Stop -  logic improved
  * Close logic conditions - only when the "Closing point of the position" is "Bar Closing"

Re: FSB v4.2.0.3 unstable - beta tests

Can the optimization function have an option to optimize over a certain period, aim is to get a strategy with a as straight line upward as possible.

Re: FSB v4.2.0.3 unstable - beta tests

Popov - can you specify a little more about "logic improved" for the stop and limit?
This is one of the biggest achilles heels in FSB, and I believe this is causing the euphoria in this forum that suddenly everybody is a successful system builder.

Re: FSB v4.2.0.3 unstable - beta tests

The problem of Stop Limit is coming from the interpolating algorithm's inability to find the right order of exit prices (in the same bar).
This problem can be fixed with an in-deep scanner (lower periods searching).

Do you know other solution?
Or, to mark the indicator as unsafe! (Remove it from FSB!!)

The same logic is used in the derivative indicators: Stop Loss and Take Profit

The Stop Limit in older versions before 2.4.0.3-2 send unlimited exit order. In most cases this is not problem, but is possible the position to be closed/reversed earlier by the opposite signal.

Example:
  buy at: 1.0100
  Stop Limit: +-50

if the position has been closed at 1.0140 (by an opposite entry signal), the exit orders are not canceled. FSB will cancel it when the market touch the exit price (FSB cancels an exit order when there is no open position).

If the position has been reversed at 1.0110  FSB sends buy exit orders at 1.0160/1.0060
The problem is coming  when the market reaches the nearest exit 1.0150 (from the initial position) and we are short, but we have a sell exit order?!? I think FSB calculates this situation correctly, but I'm not so sure.

The new version sends valid to the end of the period orders. And if there is reversing before the exit FSB simply sends the new exit orders.

We need more tests to be sure that the new logic works well!

This does not cover the closing in the same bar problem!

edit:
The exit logic does not average the exit levels (if the position's amount has been changed).
FSB is very strict to the exit signals. It cannot increase, reduce or reverse the existing position. An exit signal is "exit" - close all the open lots.

Re: FSB v4.2.0.3 unstable - beta tests

You are correct. The best solution is to look into a deeper time frame, to figure out the execution order.
But, the current implementation, not only that it is not safe, it is completely misleading.

I know, it is nice to have many beginner strategy writers come to this forum and share their 5-minute get rich strategy, and get all excited about something that will never work in real life, but I believe there should be a different approach here.

Accuracy of backtesting should come first. Nice GUI, fancy indicators and other bells, are nice but are only a secondary priority. Always.

So, to answer your question - I think that stop/limit exits cannot be removed from FSB - these are the basic exit rules and must stay there.

One solution that is possible (without deep scan) is to do the following:
1. Assume the worst possible execution order
2. Apply stops at next bar only.

This means that once you enter the trade, it stays "naked" until the next bar, then, the next bar has a logic of:
    if not SellAtStop then SellAtLimit

Which means, if the next bar hits both limit and stop orders, it is considered a loss since you have no way of knowing which one hit first.

Of course, if you plan on doing a deep scan sometimes soon, I wouldnt bother working on this logic, since it has its own inaccuracies - only in this inaccuracy case, it is a pessimistic inaccuracy rather than an optimistic one, so if a user will be able to make a working strategy, there is more chance it will actually work in real life.

Re: FSB v4.2.0.3 unstable - beta tests

Stop at the next bar is a solution except in the case when the levels have been passed.
- If the exit (stop or limit) is not reached in the current bar - the problem is transfered to the next bar (if both levels are in both side of the open price)
- if a level has been passed - where to close?  It seems - the opening price is the best place.

And lot to do if the user set Stop Limit +-5 pips? The deep scan cannot help us here, except a tick scan!

This situation has to be changed as soon as possible! The fastest is to mark the "Stop Limit" and "Take Profit" as "For demo only!!!" like Fibonacci.

Re: FSB v4.2.0.3 unstable - beta tests

Yes, I think you should mark these for demo. People seem to be counting on them now, so its best if they know the status and not be overly optimistic about their systems.

- If the exit has passed (open price of next bar is higher than the PT or lower than the SL) then you exit at open (overshoot)
- If both levels reached at the next bar, you assume the worst and mark this as a loss, since you have no way of knowing for sure which one hit first.
- If no level was reached on the next bar, you continue the logic to the following bar