Question for experienced trade system creators and theorists | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Question for experienced Fibonacci traders 3 replies

Question for experienced traders: what am I doing wrong? 19 replies

Question for experienced MT4 Traders. 3 replies

Expectancy question for experienced traders 2 replies

Question for experienced traders 0 replies

Attachments: Question for experienced trade system creators and theorists
Exit Attachments

Question for experienced trade system creators and theorists

  • #41
  • Apr 30, 2013 12:15pm Apr 30, 2013 12:15pm
Conclusion:
I need at least WIN %: 58 to make it worthwhile.

Worst expectation: to End after 1000 Trades: -22.5% ( if I'm unlucky and End it on the expected Worst MAX DRAW DOWN)
Money is important, Love is essential.
  • #42
  • May 1, 2013 5:34pm May 1, 2013 5:34pm
Quoting Ati Baia
Conclusion: I need at least WIN %: 58 to make it worthwhile. Worst expectation: to End after 1000 Trades: -22.5% ( if I'm unlucky and End it on the expected Worst MAX DRAW DOWN)
Found with the 17 pairs and over the 5 years backtest data a couple of strategy options to generate higher Win% by RR 1:1.

But the number of resulting trades is in comparison low - so I'm considering to lower me MIN required WIN% to 57 %.

the other thing is: in the end I will not trade all 17 pairs (just wanted to test it on a broader selection for WORST CASE scenario)

I will now check:
- how each pair on all 4 TF combined works out (are there any pairs which basically do not work good for my strategy)

- how each TF for all pairs combined works out (e.g doe my strategies not work well on M30)
Money is important, Love is essential.
  • #43
  • May 1, 2013 6:01pm May 1, 2013 6:01pm
I believe that if you have a your system profitable and have an understanding on what you are doing sometimes risk to reward varies especially if you are using a variable stop....Drawdown and having a losing trade is part of the business after all,,,
"I made money every time you place your Stop loss too close to ME"
  • #44
  • May 1, 2013 7:48pm May 1, 2013 7:48pm
Quoting Masterm1nd
I believe that if you have a your system profitable and have an understanding on what you are doing sometimes risk to reward varies especially if you are using a variable stop....Drawdown and having a losing trade is part of the business after all,,,
In general I think Risk:Reward can vary: for manual trading it did vary a lot (but probably partially because of human reaction (for better or worse) but for my automated system I choose to keep it at a fixed ration of R:R 1:1.

You are absolutely right about losing trades being part and as a result Draw-Down:
I believe especially for a fully automatized system which hopefully runs 24/7 it is necessary to try to get a good balance between
- best Win% and acceptable trade frequency
- and knowing my expected worst Case scenario as good as possible
Money is important, Love is essential.
  • #45
  • May 2, 2013 4:33am May 2, 2013 4:33am
Quoting Ati Baia
- how each pair on all 4 TF combined works out (are there any pairs which basically do not work good for my strategy) - how each TF for all pairs combined works out (e.g doe my strategies not work well on M30)

That was interesting: on all systems and variations the WIN % increased with each higher TF.
Maybe it's the spread in relation to SL or the general greater SL for higher TF not sure - will find out.


With the different pairs it was not as clear cut - even though there are some common outcomes.


All together:I'm mostly satisfied that the systems react quite predictable to key input parameter changes.
basically: within a certain range Number of trades and WIN % can be quite predictable adjusted (more trades lower Win% ect)

I think this is a very positive sign.
Money is important, Love is essential.
  • #46
  • May 2, 2013 8:34am May 2, 2013 8:34am
the last thing for now before coding the final EA: will be: running tests on different Trading hours.

Mainly: I will use a selection of my current findings and run them with a filter which will open new trades only
between
- 08:00 - 17:00 for Sydney / Tokio / London /NewYork timezone independently
Maybe my systems work best within a particular trading time.


Hope to be starting forward tests next week.
Money is important, Love is essential.
  • #47
  • Edited May 3, 2013 3:29pm May 2, 2013 6:20pm | Edited May 3, 2013 3:29pm
The different trading hour tests did not reveal any major things.


I did make an unexpected decision: for the start to skip TF M5 completely - never had expected that
reason:
Quoting Ati Baia
That was interesting: on all systems and variations the WIN % increased with each higher TF..
Money is important, Love is essential.
  • #48
  • May 4, 2013 7:08am May 4, 2013 7:08am
OANDA SPREAD WEEKEND
Attached Image (click to enlarge)
Click to Enlarge Name: weekend.png Size: 87 KB


Manually I did not keep trades over weekend: I traded mainly M1 but with the automated system my trades duration is on average between 1-8 hours (depending on pair and TF)

I have already implemented New Order TimeFilters on a daily hour bases (e.g. only open new trades between 10:00 - 15:00.

I'm planning to do such on a Weekly day bases: main reason to skip new orders on fridays afternoon.

For existing trades: I might rethink to close all BreakEven/In Profit trades at a certain time on fridays.
Money is important, Love is essential.
  • #49
  • May 4, 2013 2:07pm May 4, 2013 2:07pm
Still need more retesting:
Yesterday night I came up to over Win% 81.25 still keep the RR by 1:1 (with some filters) over 6 years
- but of course too less trades.
So I guess there is still room to adjust some of the main parameters.

I coded an addition to the Daily TimeFilter to define also a Week start/end Time. So I will retest if I skip fridays NewYork session all together for new trade entry.
Money is important, Love is essential.
  • #50
  • Edited 4:26pm May 4, 2013 4:15pm | Edited 4:26pm
Quoting Ati Baia
Thanks for all your feedback so far: Yes I'm talking about 100% automatized trading the main question is: how to choose the best (what to consider) from successful systems based on the trades frequency and results. is it better( and why does one think so) to have - 220 trades with a Win of ca 65% - or 1050 trades with Win of 59% in both cases one could adjust the final result to be equal by adjusting the traded lot size accordingly: but what would you choose and why?
The answer is 1050 trades because by taking the same account risk every trade you compound after each high watermark. This will increase the return over a system with less trade opportunities, even if the strike rate is better

Quoting Ati Baia
Conclusion: I need at least WIN %: 58 to make it worthwhile. Worst expectation: to End after 1000 Trades: -22.5% ( if I'm unlucky and End it on the expected Worst MAX DRAW DOWN)
You underestimate the market my friend, your worst expectation is not the worst possible result. You are forgetting that your results are based on the past. The future is completely unknown. The risk cannot be pre determined. Most backtest optimised mechanical systems go straight to worst case scenario and worse when trading live on future data
BTC Return This Year: na
  • #51
  • May 5, 2013 1:28am May 5, 2013 1:28am
I think you have done great job with your backtesting.I think you should now do a foward test to compare.if 22% is your worst expected drawdown u could commit to pull the plug off the system if it reaches there...and observe if it will recover.
wo-yoy! wo-yoy! wo-yoy! wo-yoi! wo-yoy-yoy-yoy!
  • #52
  • May 5, 2013 12:25pm May 5, 2013 12:25pm
Quoting Ati Baia
Thanks for all your feedback so far: Yes I'm talking about 100% automatized trading the main question is: how to choose the best (what to consider) from successful systems based on the trades frequency and results. is it better( and why does one think so) to have - 220 trades with a Win of ca 65% - or 1050 trades with Win of 59% in both cases one could adjust the final result to be equal by adjusting the traded lot size accordingly: but what would you choose and why?
There are advanced algos called ensemble learning methods: one is random forest tree : Random Forests grows many classification trees. To classify a new object from an input vector, put the input vector down each of the trees in the forest. Each tree gives a classification, and we say the tree "votes" for that class http://www.stat.berkeley.edu/~breima...ts/cc_home.htm
  • #53
  • Edited 6:59pm May 7, 2013 1:31pm | Edited 6:59pm
Got a bit stuck with my backtest data - MT4 Historical Rates - Backtest

In the meanwhile a old post I saw: related to trading system development http://www.forexfactory.com/showthre...20#post2212020
Money is important, Love is essential.
  • #54
  • May 30, 2013 4:14am May 30, 2013 4:14am
And, how's it going? Forward trading definitely tells you a lot more than anything else. Then you can monitor your EA while the market is developing. Very rewarding. Of course you can still build in a variation of parameters. Like that you don't have to trade them all, but keep a log of the various responses and compare them. Good luck!
  • #55
  • May 30, 2013 6:22am May 30, 2013 6:22am
Quoting Ati Baia
Thanks for all your feedback so far: Yes I'm talking about 100% automatized trading the main question is: how to choose the best (what to consider) from successful systems based on the trades frequency and results. is it better( and why does one think so) to have - 220 trades with a Win of ca 65% - or 1050 trades with Win of 59% in both cases one could adjust the final result to be equal by adjusting the traded lot size accordingly: but what would you choose and why?
Hello Ati Baia,
didn't read all the thread, but the information you are giving is not enough.
The decision can be taken by calculating the math expectancy for your system.
I understand the winner win 1R (1 unit of risk) and the losers are -1R, right?

If this is the case, the first system (64% reliability), the one that trades less is the more convenient. Why? it has an expectancy of 30cents: in average it makes 30cents for every $1 risked.

The other one (59% reliability), has an expectancy of only 18cents on every $1 risked.

There area also other important reasons why you should trade the first system: less trades means less trading costs, slippage and trade management.

Following are the calculations of the expectancy of the two systems:

SYSTEM 1 (64%)
E1 = [ (143 x $1 ) - (77 x $1) ] / 220 = 0.30

SYSTEM 2 (59%)
E2 = [ (619.5 x $1 ) - (430.5 x $1) ] / 1050 = 0.18

And your decision is now easy!

Ciao
-fibstalker
"markets traded by programs/HFT offer a significant trading edge..."
  • #56
  • May 30, 2013 8:08am May 30, 2013 8:08am
thanks flofri,

at the moment it is not going anywhere: as I just finished my MT4 Historical Rates - Backtest. I guess I need one or 2 days to finish all of that and hopefully resume next week my work here.

I definiately want to start soon some forward testing: one of my main concerns are the effect of multiple open correleting pairs and the spread issue at news time (using OANDA)

Quoting flofri
And, how's it going? Forward trading definitely tells you a lot more than anything else. Then you can monitor your EA while the market is developing. Very rewarding. Of course you can still build in a variation of parameters. Like that you don't have to trade them all, but keep a log of the various responses and compare them. Good luck!
Money is important, Love is essential.
  • #57
  • May 30, 2013 8:16am May 30, 2013 8:16am
Thanks fibstalker,

I did implement after my post the 'expectancy' part you mentioned.

Cheers

Ati

Quoting fibstalker
{quote} Hello Ati Baia,didn't read all the thread, but the information you are giving is not enough.The decision can be taken by calculating the math expectancy for your system.I understand the winner win 1R (1 unit of risk) and the losers are -1R, right? If this is the case, the first system (64% reliability), the one that trades less is the more convenient. Why? it has an expectancy of 30cents: in average it makes 30cents for every $1 risked. The other one (59% reliability), has an expectancy of only 18cents on every $1 risked. There area also other important reasons why you should trade the first system: less trades means less trading costs, slippage and trade management. Following are the calculations of the expectancy of the two systems: SYSTEM 1 (64%)E1 = [ (143 x $1 ) - (77 x $1) ] / 220 = 0.30 SYSTEM 2 (59%)E2 = [ (619.5 x $1 ) - (430.5 x $1) ] / 1050 = 0.18 And your decision is now easy! Ciao-fibstalker
Money is important, Love is essential.
  • #58
  • Jun 4, 2013 12:15pm Jun 4, 2013 12:15pm
nearly a month passed since I worked last on this project: because of my side-project which I recently finished: MT4 Historical Rates - Backtest.

Well I'm back at it. With the new historical data available I will resume some backtests and than start coding the EA. At the moment I'm quite positive about it: major concerns are mainly the huge spread increase at news time and how simultaneous trades will effect all.

But this I will have to see when forward testing.
Money is important, Love is essential.
  • #59
  • Jun 4, 2013 4:52pm Jun 4, 2013 4:52pm
1.
I skipped TF M5 and only concentrate an the remaining M15, M30 and H1

2.
Testing on 29 possible pairs where I have also data at least back to 2007 mostly back to 2003 nearly 9 years

Working on optimizing SL/TP in the hope to survive bigger Spreads and maybe adjust RiskReturn:


All pairs x 3 TF produce for the data I have about 10.000 trades:

Original results RiskReturn 1:1:
- using 0 spread: Win%: 64.59
- using 3 spread: Win%: 61.37
- using 6 spread: Win%: 57.41 - till here each of the 29 pairs (all 3 TF together) over the whole data are profitable
- using 8 spread: Win%: 54.53 - couple of pairs in negative over all data
- using 10 spread: Win%: 51.66 - not much left
Money is important, Love is essential.
  • #60
  • Jun 6, 2013 5:41pm Jun 6, 2013 5:41pm
Worked the last two days on something related: ALL Symbols: Print Infos to file (e.g. Avg.Move)
Money is important, Love is essential.
0 traders viewing now