stevehopwoodforex.com
https://www.stevehopwoodforex.com/phpBB3/
Print view

Multi 10:4 Trader EA
https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?t=1619		 Page 93 of 160
Author:  MrLong [ Thu Apr 18, 2013 2:36 pm ]
Post subject:  Re: Multi 10:4 Trader EA
On that drawdown we just got, using my method described here: http://www.stevehopwoodforex.com/phpBB3 ... &start=910

I just scalped 3%.
Author:  fxYogi [ Thu Apr 18, 2013 3:00 pm ]
Post subject:  Re: Multi 10:4 Trader EA
MrLong wrote:
Hi,

We have that built into the EA, just needs switching on:
extern bool useFilterCurrencyNumber = false;//Allow a currency, like JPY to trade NumberToTrade as below
extern int NumberToTrade = 2; <== number allowed to trade at the same time.

Andy
Thanks Andy, this is brilliant!
I am sure this will curb the gambling side of the EA. In fact in the user manual (which is still in the making ;) ) I would strongly recommend this option enabled to all users (unless they are gamblers). Or perhaps even better make this enabled by default.

If I understand your present project with reducing drawdown, it is based on the cyclic nature of profitable and drawdown periods. So you expect that the profitable and DD periods will follow each other in a predictable manner. Then why not wait for the end of a DD period on a separate demo account used as a "Luck indicator", and then start running the EA on a live account or using its signals to place manual live trades when the Lucky Streak starts?

It makes sense, in fact I have used a similar strategy in an online casino when playing blcakjack for money. If you play perfect strategy then you must have an edge against the casino, but as we know the casinos are crims, they see your cards, but you don't see theirs. And as an additional handicap, they deal you any card they want to make you lose, and you can't detect cheating. Except if you are running a statistical analysis spreadsheet while playing, that shows you the pattern of losing and winning hands clustered in winning and losing streaks.

To make the long story short, from the win/loss ratio over many dozens of bets I have seen that they are cheating, the odds were grotesquely squed in the casino's favor. My only option to get out of their trap was to bet high in the winning streaks and bet minimum during the losing streaks. I was able to do this only because the algorithm of the casino that prescribes the win-loss distribution did contain such streaks to model reality (and prevent martingale), which they didn't do very well (reality is not so predictable ;) )

There is good potential here, the major challenge is fixing the criteria of the end of losing and winning cycles. I would draw the variation of the account growth on a graph and see if there are clearly recognizable periods of Gain/DD, and what marks their ends.
Author:  glabbeek [ Thu Apr 18, 2013 3:19 pm ]
Post subject:  Re: Multi 10:4 Trader EA
As most of you know, I'm not a fan of baskets. A basket only has a function if it is based on a correlation.

This can be used to minimize the drawdowns as we experienced with the JPY. You could make baskets based only on the same currency in a pair. So if you have 8 trades open and 4 are based on the JPY then those 4 make up a basket.
You could set an stoploss so that this basket is only allowed to drop a certain amount into the red.

This way you don't restrict profitability by not accepting certain pairs, but you do limit possible drawdown.
Author:  fxYogi [ Thu Apr 18, 2013 3:35 pm ]
Post subject:  Re: Multi 10:4 Trader EA
slipshod wrote:Excellent post Yogi, you've described the cause of the excessive DD in a nutshell.

I imagine it might be quite difficult to implement what you're suggesting - and indeed if you choose to enter, say, 2 JPY trades instead of all of them then which ones do you go with & which do you leave out? Its getting into the realm of needing to use the CSS indicator.
As you have seen Andy has already provided the simplest solution to the problem, by enabling the useFilterCurrencyNumber = true switch. I just didn't know what it is for, and this is still true for most of the switches in my case. But hopefully someone who knows them all will publish a list with explanations.

About how to choose between the pairs based on the same currency, that is not a simple matter to implement in an EA because many factors need to be considered. But perhaps the best criteria is the relative size of the pullback compared to the large moves in trend direction. The closer the price retraces to the 240MA the better the risk/reward ratio of the trade (since stop loss is at the MA). Of course there are other things to consider as well, but lets keep things simple one step at a time.
The solution Andy has provided is good enough for now.

I am more interested in implementing your suggestions and code around page 62 right now. I understand the point of Andy that he wants to be consistent and stick with the written rules of NB. However, the system evolved somewhat since those rules were written down, and nobody cared to update the fact that NB doesn't uses the AO for entry, because it more often than not enters a trade while the price still moves counter-trend. This is dangerous and makes no sense. The right entry is when the counter-trend has ended and price resumes the main trend. The entries based on RSI only do this better.

The best solution would be if the user could chose between the two different entry strategies via a switch. But since Andy is busy with other important problems to solve, and does not want to change entry logic, the user will have to do this in the code, which I will try now. Thanks for the alternative trendDirect(). :)
Author:  mjws00 [ Thu Apr 18, 2013 4:17 pm ]
Post subject:  Re: Multi 10:4 Trader EA
fxYogi wrote: The best solution would be if the user could chose between the two different entry strategies via a switch.
Use slipshod's code. It is pretty and lets you pick methods. It is somewhere around p62. Or just Delete the references to AO in trendDirect(). Both fix the entry issue nicely. If you get stuck someone will give you a hand.

If you limit the pairs a little, the EA is not nearly the gambler. 6 JPY pairs running completely overwhelms all other inputs to the system. I'm not sure how the EA handles USD if you limit pairs to 2.

We may want to trade, USDCAD, AUDUSD, USDJPY, EURUSD as non correlated ideas. If it disallows that it will need a tweak.

Mike
Author:  MrLong [ Thu Apr 18, 2013 4:18 pm ]
Post subject:  Re: Multi 10:4 Trader EA
aagarcia wrote:I am seeing these errors though in Experts. Andy, are these errors insignificant?

2013.04.18 08:00:04 Multi_10_4_EAv2.01a GBPJPY,H4: object name passed to ObjectFind function cannot be an uninitialized or empty string

2013.04.18 08:00:04 Multi_10_4_EAv2.01a GBPJPY,H4: symbol name for MarketInfo function must be a string

2013.04.18 16:16:11 '2137049': Signal - failed parse signals

Give me 5 minutes and I'll upload the latest update, let me know if it fixes the problem.
Author:  MrLong [ Thu Apr 18, 2013 4:33 pm ]
Post subject:  Re: Multi 10:4 Trader EA
mjws00 wrote:
fxYogi wrote: The best solution would be if the user could chose between the two different entry strategies via a switch.
Use slipshod's code. It is pretty and lets you pick methods. It is somewhere around p62. Or just Delete the references to AO in trendDirect(). Both fix the entry issue nicely. If you get stuck someone will give you a hand.

If you limit the pairs a little, the EA is not nearly the gambler. 6 JPY pairs running completely overwhelms all other inputs to the system. I'm not sure how the EA handles USD if you limit pairs to 2.

We may want to trade, USDCAD, AUDUSD, USDJPY, EURUSD as non correlated ideas. If it disallows that it will need a tweak.

Mike
Hi Mike,

It works on a first come basis, if you select 3 then the EA will only allow 3 USD trades to run at the same time, but if the EA is adding to on another pivot it will allow to add to a trade.

Andy
Author:  sponn [ Thu Apr 18, 2013 4:57 pm ]
Post subject:  Re: Multi 10:4 Trader EA
Guys is somebody testing 10.4 ea with other than GMT based crims ? I want to test it live on armada account and they are GMT +3 so I Wonder if it is better to use tchem or use GP (or IFBX) and personal Empty4 copier. Any thoughts ?
Author:  fxYogi [ Thu Apr 18, 2013 8:04 pm ]
Post subject:  Re: Multi 10:4 Trader EA
mjws00 wrote: Use slipshod's code. It is pretty and lets you pick methods. It is somewhere around p62. Or just Delete the references to AO in trendDirect(). Both fix the entry issue nicely. If you get stuck someone will give you a hand.

If you limit the pairs a little, the EA is not nearly the gambler. 6 JPY pairs running completely overwhelms all other inputs to the system. I'm not sure how the EA handles USD if you limit pairs to 2.

We may want to trade, USDCAD, AUDUSD, USDJPY, EURUSD as non correlated ideas. If it disallows that it will need a tweak.

Mike
Thanks Mike, I have edited and compiled the code with the second variant of slipshod's trendDirect() from page 64. It works so far, but I will see the results only after observing how it opens several trades.

The selection of simultaneously traded pairs with minimum correlation is indeed a problem. My understanding is that in a pair there is usually one currency which is the dominant (its value fluctuates more than the other). Like JPY in most of the pairs. The influence of the non-dominant currency on the price movement is much less, and in some cases it can even be neglected. So some currencies like CHF which move slowly may be used as reference currencies (instead of USD). Then if you trade pairs like CHFJPY, USDCHF, GBPCHF etc. their moves (and the profit/loss of our trades) will be determined by the dominant currencies in these pairs ie. JPY, USD, and GBP. Thus the correlation effect of the slow CHF will not pose a big danger of excessive DD.

Sometimes the roles of dominant/passive currencies may change in a pair, so the mentioned principle is not fixed forever and not perfect, but I think it is worth exploring this line of thought. What is your opinion, what is the ranking order of currencies on the list of dominance (which moves faster than the other)?
Author:  dietcoke [ Thu Apr 18, 2013 9:07 pm ]
Post subject:  Re: Multi 10:4 Trader EA
Andy,

Just wanted to congratulate you an excellent piece of work and point out a little bug in the EA.

The PairIndex variable gets set to the value of the current index (i) in "start"

However, in the places it is used in the grid create and update code, you actually should be using ArraySize[Pairs]. I don't know if you've noticed, the grid tends to misbehave a bit when you load the EA.

try changing

Code: Select all

PairIndex = i;
to

Code: Select all

PairIndex = ArraySize(Pairs);
I'm finding the grid loads more reliably with this change
All times are UTC Page 93 of 160

Powered by phpBB® Forum Software © phpBB Limited