GBP/USD Portfolio #1
21 replies
Matusiak Adrian
9 years ago #113044
Hello forum members,
I have created my first (seems robust) GBPUSD portfolio.
Most of strategies are on H1 , but few are on M30.
Mostly, robustness test gives result of 2/3 to 1/2 net profit as shown on raport and around 1-3% bigger drawdown than shown.
Every strategy tested on Real Tick method on Alpari,Aplari UK, Alpari US , Dukascopy, FXDD and Forex.com data.
Whats strange,
Alpari (all) + Dukas are doing great results on Real Tick + Robustness.
FXDD + Forex.com makes losses on Real Tick Method, but when I run robustness test, then result of RT seems to be almost like RT on Alpari, Dukas.
Anyway, here’s a raport.
Strategies are tested on MM with details of:
Start deposit: $1000
Risk 2%
Lot if no MM: 0.01
Pls leave any comment about it. I don’t know much about Sharpe ratio and any other “clever” stats yet.
——————–
edit:
Demo account @ Alpari UK with forward test
http://www.myfxbook.com/members/adrian8891/portfolio-gbpusd/1123911
^ Account closed due to Alpari Insolvency.
———————-
Mark Fric
9 years ago #128129
Hello,
it looks really nice, I’d be interested to see how it performs on real or demo account.
There’s just one problem with many EAs trading on the same currency – if your broker doesn’t support hedging the strategies will influence each other.
They are tested with assumption that there’s no other open position in GBPUSD, but when you’ll be trading it for real then multiple EAs could try to open positions to opposite directions.
Non-hedging broker wouldn’t allow this.
Mark
StrategyQuant architect
Matusiak Adrian
9 years ago #128131
Indeed Mark, that’s true. Currently I use brokers that allow hedging. Soon I will start some demo account to see real performance.
Patrick
9 years ago #128151
i am interested in your portfolio performance since 2003 for example…
also the PF should be higher…
I am also interested in your demo and real results.
Best regards
Patrick
by the way, very nice business card 😉
Matusiak Adrian
9 years ago #128152
Soon I will start demo account. – I will connect it at myfxbook as soon as possible and post here.
About loger period – most of my data is available from 01.04.2010 , only Dukascopy has larger database. I will try to make it at longer period, but personally I think that market is changing so far that I don’t look for holy grail.
tnickel
9 years ago #128190
Hi Adrian,
very interested portfolio.
We will see if the portfolio runs in the future with the same profit.
thomas
https://monitortool.jimdofree.com/
Matusiak Adrian
9 years ago #128195
We will see if the portfolio runs in the future with the same profit.
I am also interested in your demo and real results.
First post modified – I have added demo account link for forward test
Matusiak Adrian
9 years ago #128211
Good day to all,
I see that there has been quite mismatch between MT test and SQ test.
It may be lead by data – tick used for SQ test are from REAL servers of Alpari UK, and account has been on tests at DEMO servers.
Anyway, I attach screenshot of one of strategies that mismatches. Please take a look, maybe You have some idea about those mismatches. I also attach source file of this strategy so “codes” might look at rules of strategy.
What is important:
1. time shift between test and SQ
2. I started forward test at 12.01.2015 around 8pm (20:00) so first and 2nd trade in SQ should be skipped
3. Every of trades on SQ and FT has finished with Stoploss (not by exit rule)
4. Test on SQ has been performed on “Real Tick” method
Any ideas?
Mark Fric
9 years ago #128307
hm, I think it is a difference in data, DEMO servers usually have much smaller spreads and they don’t change that much as for REAL accounts.
Also, there is virtually no slippage in DEMO.
It looks that your strategy goes for very little profit or stop, it is difficult to test this reliably.
Do you somehow record real tick data from your account that you then use in SQ?
Mark
StrategyQuant architect
Matusiak Adrian
9 years ago #128754
Hello Mark,
I have updated 1st post with screen of 1st period test time.
I don’t think that strategy goes for very little profit:
Average Win:
28.12 pips / $3.80
Average Loss:
-25.74 pips / -$4.97
Mark Fric
9 years ago #128802
ok, I got a very good comment from another user who created his own successful portfolio of strategies (trades for over half year with all months in profit so far).
it is important to understand that backtest is always only an approximation, strategies that work in backtest and even pass robustness tests can stil fail in live tradingfor many reasons – curve fitting or simply the strategy rules cannot be properly backtested for some internal reasons.
SQ can help you find potentially good strategies, but whether they really work or not can be revealed only by running them on demo or small real account.
So when you generate portfolio in SQ it is only the first step, the second step is to verify that these strategies perform as expected also during live trading.
It is called incubation period in some books – it verifies if strategy is profitable and if its results in live trading match the results of backtest.
Only after strategy passes this test for 3 or more months it can be considered to be traded with more money.
Mark
StrategyQuant architect
Matusiak Adrian
9 years ago #128977
clonex / Ivan Hudec
9 years ago #129021
Omg, big DD. What is the reason?
Matusiak Adrian
9 years ago #129032
It would be good to know the reason. Seems that strategies are generated bad. I dont know if its because of SQ or problem is on my side.
Now I have generated new strategies on expensive data (€800 per year) so I will try to generate something on such data.
Mark Fric
9 years ago #129056
SQ is a tool, the problem shouldn’t be in SQ 🙂
There are several possibilities:
– your strategies were curve fitted to historical data and don’t have real edge – it is difficult to determine if this is the case,do they wor on EURUSD or other currencies?
– they come through normal drawdown period – strategies could have drawdown periods that last for weeks or even months.
Have you verified that your portfolio consists of non-correlated strategies? If they are correlated then when one goes to loss the others will likely be losing too.
– they may work on data that you generated them on, but they will not work with your broker – your broker could have different timezone, he could include Sunday or doesn’t include this, all these things affect how the strategy will behave in real trading.
In general you should focus on finding strategies that work on your broker as same as they work in StrategyQuant with the data you use for generation Then you can trust your generation process.
There will always be strategies that simply don’t work when traded live, instead of wasting time with them try to look for strategies that work in both broker and SQ.
Mark
StrategyQuant architect
Matusiak Adrian
9 years ago #129079
– your strategies were curve fitted to historical data and don’t have real edge – it is difficult to determine if this is the case,do they wor on EURUSD or other currencies?
EURUSD only
– they come through normal drawdown period – strategies could have drawdown periods that last for weeks or even months.
Have you verified that your portfolio consists of non-correlated strategies? If they are correlated then when one goes to loss the others will likely be losing too.
Drawdown of 90% ? C’mon, please see at report. There never was such hudge DD and of course it has to happen just after start of testing period? 😉
– they may work on data that you generated them on, but they will not work with your broker – your broker could have different timezone, he could include Sunday or doesn’t include this, all these things affect how the strategy will behave in real trading.
I know, thats why I generate strategy that works 24/7 , to not look after timezone and trade day.