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

Multi 10:4 Trader EA
https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?t=1619		 Page 32 of 160
Author:  snailbeard [ Sat Mar 23, 2013 11:30 am ]
Post subject:  Re: Multi 10:4 Trader EA
Backtesting

Although it is hard to do this on Multi NB version it could be a useful exercise. I have made some changes to my copy of the code so that it can run on one pair.

I'll have to make more performance changes to reduce the time it takes to run even over small periods (two months) if I intend to use every tick. When I understand the code better I might be able let certain blocks of code run on a periodic basis.

Things to watch out for:
a) can't be sure that external indicators produce the correct results in BT mode.
b) no external EAs running to help with trailing stops in BT mode.

This is very preliminary but
a) there was only one pending order (GPDUSD) between 2012.01.02 to 2012.03.02
b) the order could have been a winner with good stop management
c) the price moved away from the pending order over several days so the pending order could have been re-entered at a better position or an extra pending order added at a better position
d) the price moved so far away from the pending order that the price was beyond the pending stoploss
e) the price came back several days later and triggered the original order
f) the price crossed major pivots at W-S1 and M-S1 (chance to take partial profits and/or move stop loss)
g) the trend was over at M-S1 and price started to move back towards the original entry point.
h) 3 to 4 days later price crossed back over the entry point
i) and a few more days later price triggered the stop loss (-£192)
Interestingly even though priced moved over several days so that P>DP>WP (and sometimes DO>DP) there was no pending buy order.

Anyway, I don't know if this is valid back test and unfortunately I am tied up most of the weekend but I'll do more investigation & testing as soon as I can.

Regards

Brian
Author:  MrLong [ Sat Mar 23, 2013 11:46 am ]
Post subject:  Re: Multi 10:4 Trader EA
Hi Traders,

To backtest, you need to do the following.

string Pairs[] = { "EURUSD" }; <== enter the pair to backtest

//CreateSymbolList(); <== Highlight out, like //
//CreateADRList(); <== Highlight out, like //

extern bool useBoundryFilter =false; <== make sure it's set to false.

extern bool useOneDailyTrade = true;

There is a few bugs in 1.08 and 1.07, I've fixed them, I just need to check when the markets are open, then I'll release version 1.08b.

That's it.

Andy
Author:  zentauro67 [ Sat Mar 23, 2013 2:39 pm ]
Post subject:  Re: Multi 10:4 Trader EA
I am very optimistic about the future of this ea but for now the results are a bit disappointing. Maybe there is something important that should be added. EA closely follows 10.4 but needs a human touch to filter low-probability trades. Perhaps the key is to consider the angle of the 240MA,avoiding trades when 240 is too flat and setting minimum levels of slope for 240.
Author:  McNish [ Sat Mar 23, 2013 3:48 pm ]
Post subject:  Re: Multi 10:4 Trader EA
slipshod wrote: This particular robot helps with entries, .... For people who may be struggling with having learned the rules of a method but can't bring themselves to stick to it, a robot like this is extremely useful.
I'll re-post some part from mine on 10.4 thread. This I believe can be a nice addition to the EA incorporating 10.5 along, like the BBBE is ( 10.2 + 10.3). 10.5 can be easily hard coded with it and can work like "Bob's proven profitable strategy of banking profits two to three hours into each trading session ( Read levels here), rinse and repeat. This strategy is particularly suited to those who cannot be in front of the PC all the time."

Repeat.
Observed that 10.4 have minimal stuff, 240 - Trend, RSI - Pullback in trend, Pivots - Trading Trend Pullbacks. As I also learn couple more things, 50*6 = 300*1 but better as faster game, and more probability. So I go Long @

1. WMS1 - TP two levels i.e WMR1, SL Two levels - WMS2, lot 1.
2. WP - Same, TP two lvl, sl 2 lvl, lot 1
3. WMR1 - TP 1 lvl, SL 1 lvl, lot 2.

Price will almost always range in or pass through WMS1 to WMR1 and this is the min pip distance for the week and this will be achieved if the 3 trades above get filled and TP, even covering costs.

Taking this over to 10.5 too and banking all pips and let 10.4 do the reserves.

Quite a few of us can't handle SL at 240 if it is very far away. I for one can't. So for us, I think the SL at two levels behind is a good area.

Regards,
Author:  Jonathan2FI [ Sat Mar 23, 2013 6:52 pm ]
Post subject:  Re: Multi 10:4 Trader EA
snailbeard wrote:Backtesting

Although it is hard to do this on Multi NB version it could be a useful exercise. I have made some changes to my copy of the code so that it can run on one pair.

I'll have to make more performance changes to reduce the time it takes to run even over small periods (two months) if I intend to use every tick. When I understand the code better I might be able let certain blocks of code run on a periodic basis.

Things to watch out for:
a) can't be sure that external indicators produce the correct results in BT mode.
b) no external EAs running to help with trailing stops in BT mode.

This is very preliminary but
a) there was only one pending order (GPDUSD) between 2012.01.02 to 2012.03.02
b) the order could have been a winner with good stop management
c) the price moved away from the pending order over several days so the pending order could have been re-entered at a better position or an extra pending order added at a better position
d) the price moved so far away from the pending order that the price was beyond the pending stoploss
e) the price came back several days later and triggered the original order
f) the price crossed major pivots at W-S1 and M-S1 (chance to take partial profits and/or move stop loss)
g) the trend was over at M-S1 and price started to move back towards the original entry point.
h) 3 to 4 days later price crossed back over the entry point
i) and a few more days later price triggered the stop loss (-£192)
Interestingly even though priced moved over several days so that P>DP>WP (and sometimes DO>DP) there was no pending buy order.

Anyway, I don't know if this is valid back test and unfortunately I am tied up most of the weekend but I'll do more investigation & testing as soon as I can.

Regards

Brian
Brian - interesting to see someone also intent on BT and gaining statistical results to judge future performance. Have you looked at optimisation by pairs recognising the different volitility ? and do you use any of the FWD testing tools to avoid curve fitting?
I get updated axi clean tick data sent to me every month from a provider so I can backtest and I would be interested in sharing the cpu load on this exercise drop me a pm
Jonathan
Author:  Jonathan2FI [ Sat Mar 23, 2013 7:06 pm ]
Post subject:  Re: Multi 10:4 Trader EA
slipshod wrote:fx8000, mine went backwards a bit too over the last 2 days, but is well ahead for the week. Its a H4 strategy, will take time to determine how effective it is.

Andy, I'm not concerned right now about the lack of trades yesterday. After all, 1.07 was doing the same and it had been trading like a champion previously. If they continue not trading next week then maybe there's a problem, will have to wait and see :)

I've also noticed that different trades are launched on different brokers - I have no idea why that is!
Slipshod,
I can concure overall week Gross 7% equity curve retracement Thurs Friday -2% net 5% for the week- very pleased with that level of performance and amount of control needed........so far.
Where are you on the following on v1.07
usebothforpullback true
useaveragebe true
mabufferpips 25
maouterbufferpips 310
usesr1 true
Stealth true
No additional mptm currently ...although as I can only do end of day trade checking adj. I should use additional to lock in pips. I do use equity sentry as a belt and braces on live account otherwise as per described.Any comments greatly appreciated
Jonathan
Author:  glabbeek [ Sun Mar 24, 2013 6:51 am ]
Post subject:  Re: Multi 10:4 Trader EA
lisaireton wrote:some things just aren't meant to be automated. human intuition will trump an EA

And this is based on human intuition?

History has proven that human intuition has been the biggest bias among traders. True, the most famous traders aren't robot traders. But this is the reason that they are famous.

It is very difficult to translate an eye-ball system to a robot system, but it is for 99% possible.

Read the book: "the quants" and we will discuss human intuition.
Author:  slipshod [ Sun Mar 24, 2013 1:31 pm ]
Post subject:  Re: Multi 10:4 Trader EA
Indeed, I've heard that particular nugget of "wisdom" quoted as gospel in quite a few places, like something that must be true because everyone says it is. Problem is that if intuition is the only way to trade, then not only EA development is a waste of time, so is all the work done on any manually traded rule-based system. Reduce charts to candles and trade on what your gut's telling you is the only way to go!
Author:  jonrtaylor [ Sun Mar 24, 2013 6:42 pm ]
Post subject:  Re: Multi 10:4 Trader EA
Hi my broker uses a dot suffix after pair names (e.g. my broker would be EURUSD. vs EURUSD with other borkers.)

In this case, will the EA work as is? I know other brokers use various suffixes as well.

As to the question of intuition: Wouldn't a scientist develop a hypothesis intuitively, then go about confirming it empirically? And of course we should also consider the notion that a scientist doesn't develop his intuitions without lots of experience, observing various facts which heretofore haven't been organized into recognizable patterns or relationships.
Author:  bubo [ Sun Mar 24, 2013 8:23 pm ]
Post subject:  Re: Multi 10:4 Trader EA
MrLong wrote:Hi Traders,

To backtest, you need to do the following.

string Pairs[] = { "EURUSD" }; <== enter the pair to backtest

//CreateSymbolList(); <== Highlight out, like //
//CreateADRList(); <== Highlight out, like //

extern bool useBoundryFilter =false; <== make sure it's set to false.

extern bool useOneDailyTrade = true;
Andy - have you managed to backtest the way you suggest? I tried your solution before on v1.07 and it did not produce trades during a period when the EA set pending orders on demo.

I agree that an optimization orgy is hardly of much benefit, but it would be useful for debugging and learning about some behaviours of the EA via backtest over a few weeks time, even if backtests are never perfect.

Thank you also to Brian - looking forward to hearing more about your backtesting!
All times are UTC Page 32 of 160

Powered by phpBB® Forum Software © phpBB Limited