Swing momentum trading - the robots are taking over

Hi Steve,
Last night I was sailing along through the front end of the Tokyo Session on a live account with the SS bot and Scalper....
I managed to shoot myself in the foot with a really stupid math error sending an order. Then as senior citizens are prone to do I reloaded the gun and shot myself in the other foot trying to fix it on the fly.
This may not be the time or the place to bring this up since I'm also running Simone on demo and I know how busy you are but any chance of a future upgrade to MPTM such as a pending order function might save me from myself.
I'm really surprised MPTM does not have a pending order function now since all your robots do it so smoothly.
If you get a chance it would be great and as always...... Thanks for the fun

whip

Not a hope. mptm is an existing, open trade manager; besides, it is a monster already.

Tell me what you want in terms of lot size, tp and sl, and I will knock together personally tailored pending long and short scripts instead.

Thanks Steve,
.4 lots...60 sl......25tp are good round numbers for what I am doing and I owe you big-time.......

whip

Here you go.

When you need to change the lot size, or tp/sl for a 5 digit crim:

 

1. Right click the script and select 'Modify'
2. When the code editing window appears, make the changes and hit the F5 key to recompile the scripts
3. Close the editor

One Full Truck load of thanks for this Steve. The Shih Tzu and I are in the truck and working....I'll load them up and try them when we get back to the house......Again I owe you be-time......Have a Great Day...

whip

Thanks for this Steve......scripts work like a charm and MPTM honors them....Great Stuff

whip

Still trading the indicator manually but have moved up to the 1hr chart as I an working on this whilst in work and checking for alerts every 15 mins was just interupting my work too much. A good day today using 0.04 lots profit was £107. My mptm setting are TP and SL 1000, be after 200 with half close.

Will keep trying this for a while and may even go live although everytime I try manual trading live I manage to screw it up

Paul

[although everytime I try manual trading live I manage to screw it up

Paul[/quote]
Welcome to the club. I'm the current President.

Same problem. That is why I write trading robots.

hi guys,

i've been backtesting it on GU and EU lately, and came up with quite interesting results, without using MPTM (don't know how to run 2 EAs together in backtesting...), just hard TP/SL settings.

And i got quite interesting results on EU over the last 2 months (from Apr 1ts till today):



- I changed the RSI trend filter to H1 TF instead of D1 (just a personal feeling that H1 RSI gives more sensitive trend change signals than D1...)

- the settings:
- EU pair;
- M15 chart; (don't know if there is any recommended TF for this strat -- Steve, snap my fingers if i missed this information )
- 24h trading
Lot=1; (i like round numbers...)
CandleShift=4;
StopLoss=800; //points, equivalent to 80 pips
TakeProfit=500;
RsiPeriod=20; //H1 RSI period
dist2=20;
dist1=10;

It gives $7500+ profit in past 2 months; (with 1 lot)

GU gave only about 3900 profit max after optimization;

Attached are the backtest results on EU -- note that i did the backtest during night time, so spread is about 25 points (2.5 pips) regardless of trades time in the day (which is typically only 0.6~1.0 pips on EU).
So this is a pessimistic result.


Initial deposit 10000.00
Total net profit 7587.90
Gross profit 13990.70
Gross loss -6402.80
Profit factor 2.19
Expected payoff 210.78
Absolute drawdown 638.00
Maximal drawdown 1799.00 (9.76%)
Relative drawdown 11.12% (1175.80)
Total trades 36
Short positions (won %) 27 (85.19%)
Long positions (won %) 9 (55.56%)
Profit trades (% of total) 28 (77.78%)
Loss trades (% of total) 8 (22.22%)
Largest profit trade 500.00 loss trade -802.10
Average profit trade 499.67 loss trade -800.35
Maximum consecutive wins (profit in money) 6 (3000.00)
consecutive losses (loss in money) 1 (-802.10)
Maximal consecutive profit (count of wins) 3000.00 (6)
consecutive loss (count of losses) -802.10 (1)



What i find interesting in the result, appart from the absolute profit, is in red above.

What is interesting is the consistency of the profitable trades, and consecutive lost trades is limited to 1 only.

I am using the EA-builtin hard TP/SL (the only one available in the EA as is, no MPTM while backtesting);
And looking at the resulting chart, we can clearly see that there should be a lot of room for MPTM to boost the results, with BE and JS or trailingstop.
Meaning also that the trades are usually taken in the right direction (SL not hit very often, even if drawdown can go near SL sometimes).

I have several other sets of parameters that give similar results on EU, i just picked this set as a typical "good" result for this 2 months period.

I did not try other RSI periods so far, just the H1/20 yet.

I will consider run the optimizer on this parameter as well, but it takes quite a few hours to get the results...
Even after having optimized the EA so much that it now takes only 3 seconds on my laptop to backtest a 2-months period, where the last posted EA takes 4 minutes to cover it...


So, i now have a few questions to improve this stuff:

- trend detection:

EA uses RSI on D1;
i find it too slow -- i might be wrong!
so i changed it to H1 RSI to make it more sensitive to trend changes.
Is that a good direction ?
D1 RSI merely gives only down-trend signals in the last few months, which means that the EA will take only short trades.
With H1 RSI, the results show 9 long trades and only 55% are winners (5 wins, 4 losses),
vs 27 short trades and 85% are winners (23 wins, 4 losses only).
So, we have more winners than loosers on long trades, but with TP=50 and SL=80, that gives 5*50-4*80=-70 pips on long trades...
We might consider NOT taking long trades after all...
So, my RSI settings are questionnable... or there is room to optimize here too.

- i tried a bunch of the optimizer "best results" sets (for Apr1-June8) on Jan1-March12 to see how they would have performed...
well, bad news: they are ALL loosers.
The worst period is feb05-march08: this period is a real killer!
I wonder if we could come up with better settings if we optimize for just this period, and see is those sets would still perform well in other periods...
i'll try that tomorrow...


Sq

With the trend detection RSI, another option is to keep in on the daily, but to change it to a low parameter like 2 or 3 to make it more reactive.

After another optimization round, but with RSI (H1) in the loop, (10 to 40 by steps of 5), the best results are given with smaller values of RSI periods: 10.
Giving +$9000 profit from jan1 till today on EU.
That includes the bad february period, so it is looking good.
It even doesn't drawdown during February, while going up fast after that.

Settings are: jan1-jun8 rsi(H1,10) shift2-sl60-tp40-dist2-20-dist1-16.zip



I tried an optimization round without the RSI filter (removing it in the EA with a simple test "RSIPeriod>0"), and it gave very bad results...

So it seems that not using any trend filter is not the way to go.
I think the "super signal" indicator gives bad signals when it is not trading into the trend.
I can clearly see this with my change of using H1 RSI instead of D1 RSI, where the bad trades are mainly the long trades, which are opposite of the longer-term trend.

Another fact I can see is that the Super Signal indicator gives premature entry signals.
If you look at a chart with the indi loaded, you can clearly see that it looks like magic for finding the swing points.
But in fact this is because it repaints all the time, giving many consecutive "strong signals" before finally converging to the real swing point that will stay displayed on the chart.
This means that the first "strong" signals are most of the time not the good ones, leading to whipsaws and poor performance.

We should try to improve the performance of the SS indi signals.
One good way that is in the EA is the "CandleShift" parameter.
The optimization results show that the best results are given with CandleShift = 4 (for RSI(H1,15), on M15, I haven't tried other TF yet), meaning that if we wait for 4 (M15) bars before taking the trade, and the swing point has not changed, then we have a better chance to catch a real swing, not just a 1- or 2-bar small retrace which turns around very quickly.
This produces a 4-bars lag to the signal, but with the benefit of a much stronger entry.
I can't think yet of a way to do it any better, though...

Note that the example result i gave here shows CandleShift=2 for RSI(H1,10) which gives better restult than 4.
But with RSI(15), 4 is much better...


Any thoughts, anyone ?



You will also notice that the backtest shows a strange glitch in the equity curve in march...
this is because my tick data lacks values from 03.12 till 03.26...
I can't download data for this period for some reason...
I'm on Alpari, but i'm not sure the tick data are being loaded from Alpari or from Metaquotes database.
Does anyone experience the same issue ?
Where can i find reliable tick data for MT4 ?
Thanks...


Sq

Here is sq's updated code. I will make it the post 1 version later - got to go out for most of today.

sq, sorry to have appeared to ignore your email. Thunderbird has been claiming to send emails that in reality have not been going out - I only found that out late yesterday.

Feel free to post updated versions whenever you want; you are most welcome to do so.

Hi Squalou,
I have a version of this bot with a couple of MPTM functions included (BEP, JSL) that I have been using for my back testing.
If you think this would be useful I can post a copy when I get home from work.

PeeJay

Hey Steve,

no prob', i know how it is to be busyyyyyyyyyyyyyyyyyyyyyyyyyyy

I wanted to have your impression on the changes i made to integrate the heart part of the Indi into the EA directly.

For all people: the changes i did were specifically targetted on cpu performance improvements, in order to boost backtesting.
I achieved something like x100 by not externally calling the SS Indicator (with iCustom()), which takes ages in MT4 because it loads it and unloads it at every call, and there are 4 calls to it per bar...

Now it only takes like 5 seconds to backtest over a 6 months period, where it used to take... 9 minutes ?

So now you can pull your tester, and turn the optimization ON without fear...
It takes about 10 minutes only to optimize over 6 months with 5 parameters to optimize (CandleShift 1-7, SL 10-100, TP 10-100, dist2 15-40 and dist1 5-30).


Another change i did was to use pips intead of Points in the SL/TP settings, so these are identical for any broker now.

Good optimization to all!



Sq.

What's funny is that i just spent the past hour adding that to my EA...
oh well... i will take yours if it has more features than mine!

I see Hiredwhip is using MPTM all the time, both for manual and EA trading, so i decided it would be good to add it.
It's probably better to have it builtin rather than sticking to the hard SL/TP.

I also switched to my "reliable" versions of the OrderXxxx() functions -- i do that for all EAs now... it gives much less errors.

We can combine both into a single nice SS EA for sure!

Email me your EA, i will merge it.
Thanks for that.


@Hiredwhip: what MPTM features do you use ?

Maybe we can just integrate those and not everything from MPTM
(burpp... Oops, sorry, i have some digesting problems with MPTM.... )

Sq.

something i just discovered while backtesting:

we do NOT CARE about "dist1" parameter: it is completely USELESS in the EA, as it was originally used for decioding to CLOSE trades, not open them...
but the EA does not close trades anymore -- they until SL/TP.

So, in fact, don't bother including dist1 in your optimization set, it will just double the optimization time with absolutely no added value!



Sq

Hi Sq,
Here is the set file I used last night......this does vary alittle every now and then. I'm also sending pending orders with Steve's bullet proof scripts.
Hope this helps...

whip

I have been manually trading M5 with some success (using MPTM to close trades).

I have noted the signals are often early relative to the swing formation. However, I have used the signals as they come using D1 RSI and daily/weekly/monthly fib levels for S/R. As a result, I may have some -ve UPL to bear until trades close, usually in profit.

However, I seem to have noticed that the H1 RSI may have been a more reliable trend filter when taking the signals as they come in. This seems to be consistent with your backtesting data. I have been using RSI 14 but not because I found that to be a better period, just because it is the "norm."

It makes some sense to me that the lower RSI period (10) on H1 would work better with candle shift 2 just from my observations in manual trading, particularly if trading from a lower timeframe.

I'm not recommending what settings to use, just reinforcing your findings from my observations manually trading.

Whip,

I see you have added trailing stop and adjusted your hedge settings from the previous set you posted. Was there a particular observation you had on your prior set that triggered this change? Or are you just testing different settings?