Two Systems Journal | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Top 100 best Trading Systems: post good Trading Systems here 62 replies

Are the best systems also the systems that fail most spectacularly? 55 replies

ROTrade Systems Journal - Free trading talk 10 replies

Systems, Systems and Systems... 1 reply

Attachments: Two Systems Journal
Exit Attachments

Two Systems Journal

  • #21
  • May 24, 2011 10:27am May 24, 2011 10:27am
Quoting zznbrm
stevefx,

Your "Three Fours" strategy seems interesting. But before creating an EA to trade it live, I wanted to backtest it to see if May 2011 was a "typical" month for the strategy. EURUSD Results for the past 17 months are below (in pips). I did not spend much time verifying all data. I did run some of the tests in visual mode to verify the backtester was working per the strategy...I think it is.

I'll do some more verifying of data and try some other pairs.

Send me a PM if you are interested in the full excel spreadsheet of results for...
Thanks very much for your effort zz. PM sent.
  • #22
  • May 24, 2011 2:13pm May 24, 2011 2:13pm
Quoting zznbrm
stevefx,

Your "Three Fours" strategy seems interesting...

Edit: Added second set of results for GBPUSD over same period
I just wanted to follow up a little on this report. I checked the first 12 reported losses on the EU using the spreadsheet provided. About 6 out of 12 of them were incorrect.

Among the various errors: counting near set pendings, or near triggers as real set pendings and real triggers. Taking trades after 1PM. Counted a +95 point trade as -95 point trade. Extra non-existant trade. Failed to count a big winner. I've only been through 10% of it and its already a couple hundreds points of discrepancy.

I don't mean to bash your efforts at all-I'm just saying that its not nearly an even roughly true reflection of the systems performance as I've designed it.

One very positive thing that I did get out of it, was that on three occasions price missed its profit target by a pip or less, and then went back to hit full stop. This is because of the nature of the TP's, being at 25% areas, are strong S/R's, and will sometimes deflect by a pip. So adjusting that should be a very positive thing in the long run.

I can provide the chart I used to check the spreadsheet. I only marked it up on days where there were mistakes, and put X's on or near the problemed area, and checks where it got it right.

Thanks Again for your input-should prove very helpful.
Attached File(s)
File Type: tpl statscheckforeu.tpl   4 KB | 257 downloads
  • #23
  • May 24, 2011 3:17pm May 24, 2011 3:17pm
Ok, so on with the show.

EU: not triggered
GC: triggered long, 1/2 taken 1 box, 1/2 close BE+2, +10
GNZD: triggered long, hit full stop, -76, 2nd round triggered long, 1/2 taken 1 box, 1/2 hit trailing stop, +49, total -27
AU: not triggered
AJ: not triggered
NZU: triggered short, closed EOD -9
EJ: triggered short, closed EOD @BE
AUDCAD: not triggered
GU: not triggered
NZJPY: triggered short, hit full stop, triggered short 2nd round closed EOD, total -49

Daily Total: -75
Running Total: +1077

Observations: first decent sized losing day of record. Nothing much more to say about it. Tommorow is another day.
  • #24
  • Edited 3:42pm May 24, 2011 3:21pm | Edited 3:42pm
Quoting zznbrm
LOL! Now that's funny!

1) Are you using a broker with US ET server time? My timestamps are in US ET. In your strategy description, you quote times in EST. So, that's what I used. Well, actually I used ET (Eastern Time) because technically we are current on EDT( Eastern Daylight Time).

2) Yes, the +95/-95 is a bug when End Of Day is reached for a short trade that is in stage 3 (trailing stop) mode.
I use IBFX-which uses GMT time. When its 8 o'clock PM for me its 12AM on the chart-ie I get a period separator at 8pm but the chart says 00. I was referring to EST cause thats what I think in. (If that helps clear up the funniness). Here are the first two opening day bars on the EU for me, if this helps clear it up any further.

I cound provide a far more meticulously marked up template showing exactly where and how our results differ, if that would help.
Attached Image (click to enlarge)
Click to Enlarge Name: TimestampEU.png Size: 251 KB
  • #25
  • May 24, 2011 4:46pm May 24, 2011 4:46pm
Quoting zznbrm
But then you are not thinking in EST (GMT-5), you are thinking in EDT (GMT-4).

It seems you want the "day" to always start at 00:00 GMT. That is 20:00 EDT (during the summer months) and 19:00 EST (during the winter months).

But in your strategy description you say the "day" starts at 8 PM (20:00) EST.

20:00 EST is always 01:00 GMT.



That would be a good idea. Although a simple screenshot would be better than a template since our brokers are on different times, don't you think?

Here is a screenshot of EU for today and yesterday....
Yes, sorry about the time discrepancy. Its a bit confusing to me. This is correct, the day should always start at 00 GMT, which would change from summer to winter according to Eastern time.

What you showed on your chart matches my chart perfectly. That should be the set-up everyday. My fault for not correctly expressing the starting times.

(I've been up for about 25 hours.) Anyway, perhaps all that makes my showing the previous differences a bit of a mute point.

-thanks for bearing with me
  • #26
  • Edited 5:20pm May 24, 2011 4:54pm | Edited 5:20pm
Quoting stevefx
Yes, sorry about the time discrepancy. Its a bit confusing to me. This is correct, the day should always start at 00 GMT, which would change from summer to winter according to Eastern time.

What you showed on your chart matches my chart perfectly. That should be the set-up everyday. My fault for not correctly expressing the starting times.

(I've been up for about 25 hours.) Anyway, perhaps all that makes my showing the previous differences a bit of a mute point.

-thanks for bearing with me
I will adjust the code to always start the day at 00:00 GMT and re-run the tests.

Is this correct:

Day starts at 00:00 GMT
Open no new trades starting at 17:00 GMT
Close any open trades at 19:00 GMT
  • #27
  • May 24, 2011 5:29pm May 24, 2011 5:29pm
Quoting zznbrm
I will adjust the code to always start the day at 00:00 GMT and re-run the tests.

Is this correct:

Day starts at 00:00 GMT
Open no new trades starting at 17:00 GMT
Close any open trades at 19:00 GMT
Yes, thats it. Thanks very much.

-Steve
  • #28
  • May 25, 2011 8:36am May 25, 2011 8:36am
Quoting zznbrm
Day starts at 00:00 GMT
Open no new trades starting at 17:00 GMT
Close any open trades at 19:00 GMT
I deleted previous posts with invalid data.

Attaching EURUSD results for past 17 months. Used M1 for test data. Keep in mind my timestamps are in New York time (GMT-5 during winter months, GMT-4 during summer months).

Also attaching video of simulator for a recent day that shows how simulator works. Please verify against your system rules.

Thanks!

Inserted Video
Attached File(s)
File Type: xls SfqTester-EURUSD.xls   210 KB | 270 downloads
  • #29
  • May 25, 2011 4:59pm May 25, 2011 4:59pm
Hello.
Thanks zznbrm for results.
This clearly show that need something to improve in this starategy.
  • #30
  • Edited 6:17pm May 25, 2011 5:41pm | Edited 6:17pm
Quoting zznbrm
I deleted previous posts with invalid data.

Attaching EURUSD results for past 17 months. Used M1 for test data. Keep in mind my timestamps are in New York time (GMT-5 during winter months, GMT-4 during summer months).

Also attaching video of simulator for a recent day that shows how simulator works. Please verify against your system rules.

Thanks!
Hi ZZ, it looks right to me. I can't zoom in to make sure all the counts are right for the stops, but from a distance it looks correct. The way its setting up and closing trades looks correct. Nice simulator. Thanks for redoing the tests, and making the video.

Correcting the starting times has already swung the results into profitability, which is good to see.

Looking at the results you gave me yesterday has also already allowed me to pinpoint weaknesses in the trading rules which I can correct. I just want to take some time to try and verify these new results manually, and see what else I can learn, and then I'll say something more, suggest changes, etc.

Excellent Work-Thank You!
  • #31
  • May 25, 2011 7:18pm May 25, 2011 7:18pm
EU: not triggered, according to my data missed setting its pending for an otherwise successful trade by 2/10's of pip. This would be a discretionary trade.

GC: triggered long 2x, hit full stop, -88, triggered long 3rd time, 1/2 taken 1 box, 2nd closed BE+2, +11, total -77

GZ: tiggered short, 1/2 taken 1 box, 1/2 close BE+2, +25, triggered short 2nd time, closed EOD -13, total +12

AU: triggered long twice, both time profit taken 1 box, 2nd 1/2 close BE+2, +22

AJ: triggered long, 1/2 taken 1 box, 2nd closed EOD, +36

NZU: not triggered

EJ: triggered long, 1/2 taken 1 box, 2nd closed BE+2, +16

AUDCAD: triggered long 2x, both time 1/2 taken 1 box, 2nd close BE+2, total +14

GU: not triggered

NZJPY: not triggered

Daily Total: +23

Running Total: +1100
  • #32
  • May 27, 2011 9:01pm May 27, 2011 9:01pm
EU: triggered short, hit full stop, (-40), triggered short 2nd time, 1/2 taken 1 box, 1/2 close EOD (+43), total +3
GC: triggered long, 1/2 taken 1 box, 1/2 BE+2, (+12), triggered long 2nd time, close EOD (-3), total +9
GZ: triggered long, 1/2 closed 1 box, 2nd close BE+2, total +18
AU: triggered long 2x, hit full stop 2x, total -82
AJ: triggered short, 1/2 taken 1 box, 2nd closed EOD, total +23
NZU: triggered short, hit full stop -28
EJ: triggered short, hit full stop, -39, triggered short 2nd time, 1/2 taken 1 box, 2nd close BE+2, +11, triggered short 3rd time, 1/2 taken 1 box, 2nd hit trailing stop, +56, triggered long close EOD BE, total +28
AUCAD: not triggered
GU: triggered short, hit full stop, -75
NZDJPY: triggered short, 1/2 taken 1 box, 2nd close EOD, +17

Daily Total: -87
Running Total: +1023


5/27/11

EU: triggered short, 1/2 taken 1 box, 2nd closed BE+2, (+19), triggered short 2nd time, hit full stop (-52), total -33
GC: triggered long, hit full stop (-22), triggered long 2nd time, 1/2 taken 1 box, 2nd close BE+2, (+10), triggered short 2x more, hit full stop (-44) total -56
GZ: triggered long, close EOD -5
EJ: triggered long, 1/2 taken 1 box, 2nd closed BE+2, +16
AJ: triggered short, 1/2 taken 1 box, 2nd close EOD, +25
NZU: not triggered
EJ: not triggered
AUDCAD: not triggered
GU: triggered short, hit full stop -49

Daily Total: -102
Running Total: +921

Two tough Days, More Comments Later
  • #33
  • May 27, 2011 9:18pm May 27, 2011 9:18pm
The biggest weakness in the trading rules, which the Sim results made me aware of, and after I started seeing very frequently in the live results, has to do with trades that get very close to hitting their profit targets, and then fall back to hit a full stop.

There is no provision for this, and its responsible for a large part of the overall losses. Its not something you'd allow if you were actually watching your trade, so its not something that should be allowed systematically.

So the revision is just to set trades which are in profit at least 50% 1 box to BE+1 as soon as the hit 50% level. This should reduce drawdown dramatically.

I also have six charts which show where I found about 1100 points worth of discrepancy and the SIM results for the EU, provided by ZZ. These don't represent all possible discrepancies, since I only checked about 50 to 60% of the losers, and none of the winners.

If there was a discrepancy which had to do with a data point that was less than 1 pip, I overlooked it. If it was 1 pip or more, then I pointed it out.

Thats about it-have a good weekend. Charts to come.
  • #34
  • May 27, 2011 9:24pm May 27, 2011 9:24pm
Here are the data checks.
Attached Image(s) (click to enlarge)
Click to Enlarge Name: datacheckeu1.png Size: 359 KB
Click to Enlarge Name: datacheckeu2.png Size: 485 KB
Click to Enlarge Name: datacheckeu3.png Size: 307 KB
Click to Enlarge Name: datacheckeu4.png Size: 317 KB
Click to Enlarge Name: datacheckeu5.png Size: 294 KB
Click to Enlarge Name: datacheckeu6.png Size: 214 KB
  • #35
  • May 27, 2011 10:54pm May 27, 2011 10:54pm
Quoting stevefx
Here are the data checks.
Comments below.

Chart1 - 1/19/2010
----------------------
These trades are there, see trades #27 and #28 in spreadsheet

Chart 2 - 6/21/2010
----------------------
My broker opened at 21:00 GMT, yours opened at 22:00 GMT.
My broker opened at a much higher price than yours, thus changing the
previous day's high price and the corresponding grid for the day.

Chart 3 - 5/6/2010
----------------------
I can't see your high/low from the previous day on the chart. But judging by the chart, looks like my prev high/low is different than yours. I assume a difference in data. Keep in mind, I was using M1 data for my tests. Sometimes there are gaps in data (especially M1 data). Sometimes it is spikes that one broker has that another doesn't. That happens with different brokers...not much to do about that. My 3 losses on my broker using my broker's M1 data looks correct.

Chart 4 - 5/14/2010
----------------------
On my chart, I missed the green set pending by 0.2 pips. Thus, I entered 1 grid section above you...making the outcome completely different from you. Again, that will happen with different data feeds.
Also, on this chart. The last trade was triggered at exactly 17:00 GMT. This is the exact same time we should not take any trades. In my code I was allowing trades on 17:00, but stopped allowing trades at 17:01. By bad here, you are correct, the last trade here should not have happened.

Chart 5 - 5/21/2010
----------------------
On my chart, I missed the green set pending by 0.7 pips. Again, looks like simple differences in data feed.

Chart 6 - 6/4/2010
----------------------
Same as the last trade on chart 4. Last trade triggered at 17:00 GMT. Again, bug in my code where this trade should not have been allowed.

As all forex traders can attest to, one minute (or second) or 0.1 pips here and there can make differences in individual trade outcomes.

Backtesting is not perfect. Though testing with tick data is very, very good, it is not perfect. All we can do is try to get a general idea of the profitability of a strategy. I used M1 data because it seemed like it would give me enough of an idea about the system performance. I have tick data, but chose not to use it for this testing.

We've found two bugs in my code. Which is actually good, that's why we test things out. The programming arena is really more about design and testing than it is about actual coding. Are there more bugs? Maybe, maybe not.

I can tell you that I am very pleased that my tests have helped you identify some potential adjustments to your strategy. That's was the point!!!!
  • #36
  • Last Post: May 27, 2011 11:23pm May 27, 2011 11:23pm
Quoting zznbrm
Comments below.

Chart1 - 1/19/2010
----------------------
These trades are there, see trades #27 and #28 in spreadsheet


Chart 6 - 6/4/2010
----------------------
Same as the last trade on chart 4. Last trade triggered at 17:00 GMT. Again, bug in my code where this trade should not have been allowed.

As all forex traders can attest to, one minute (or second) or 0.1 pips here and there can make differences in individual trade outcomes.

Backtesting is not perfect. Though testing with tick data is very, very good, it is not perfect. All we...
I found the trades for the 19th, which were marked on the 18th. I tried to check to make sure any reported problems were not time stamp related, but that one got by me.

I understand about the data issues-been around the block with this before. Maybe it all evens out in the end anyway, since none of the winners were checked.

Aside from that, your tests were very helpful indeed. I spotted a major problem. So thanks again!

The last thing I'm going to look into is to try and verify the wisdom of splitting the trades up into two parts, and whether or not the extra pips being made on the trailing stops justify the multiple losses on the BE+2 results. I'm not at all sure that they do, so your results will again help me to figure this out quickly.

-Best Regards
0 traders viewing now