Synthetic hedges, cointegration, mean reversion and similar stuff | Page 7 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Simple Mean Reversion 621 replies

Pairs Trading: Reversion to the Mean 22 replies

Forward test of my new mean reversion strategy 11 replies

Question for synthetic and non-synthetic currency pairs 0 replies

Cointegration, Synthetic hedges, mean reversion in R, Tech Thread 0 replies

Attachments: Synthetic hedges, cointegration, mean reversion and similar stuff
Exit Attachments

Synthetic hedges, cointegration, mean reversion and similar stuff

  • #121
  • Nov 18, 2010 7:56am Nov 18, 2010 7:56am
hi 7bit,

my lot sizes are not changing in realtime, should they be changing?

regards,
vip
  • #122
  • Nov 18, 2010 7:58am Nov 18, 2010 7:58am
I've taken 2 trades on a demo account yesterday to see how it behaves. First trade was on the 15M chart and the second was on the 1H. I've closed the 15M trade this morning as it was back to the mean. The other on the 1H chart is still open.

I'll post the beginning and closing plot for the 15M trade later and add some thoughts about it so maybe we can improve this baby.
  • #123
  • Nov 18, 2010 8:04am Nov 18, 2010 8:04am
Quoting guess121
hi 7bit,

my lot sizes are not changing in realtime, should they be changing?

regards,
vip
If you are on a 5M chart, every 5 minutes you add a tick to the 2000 values of the plot and remove the first so it most likely won't have any effect on the basket composition. Come back in two hours and you'll have 24 new ticks and then you will see a shift in the composition.
  • #124
  • Nov 18, 2010 9:27am Nov 18, 2010 9:27am
Quoting luiboy
I want to believe this is a holy grail. Can I?
Its only a tool that has to be used very carefully. Its not a complete trading system and it can not guarantee anything. But it might be possible to use it in a profitable way, just don't believe it blindly.

For example don't fit a trend with data from before an event of world-wide longterm impact and expect the whole market to behave like it did before. Just look at how they play ping-pong with the EURUSD (or the market as a whole) from G20 to G20 (plot all G20 summits into an EURUSD chart) to see what I mean (somebody did this in the EURUSD thread a while back but I can't find this posting now).
  • #125
  • Nov 18, 2010 10:56am Nov 18, 2010 10:56am
Quoting 7bit
Its only a tool that has to be used very carefully. Its not a complete trading system and it can not guarantee anything. But it might be possible to use it in a profitable way, just don't believe it blindly.

For example don't fit a trend with data from before an event of world-wide longterm impact and expect the whole market to behave like it did before. Just look at how they play ping-pong with the EURUSD (or the market as a whole) from G20 to G20 (plot all G20 summits into an EURUSD chart) to see what I mean (somebody did this in the EURUSD...
Thanks for the advice. But I actually do not know what's going on here. I just know how to follow instructions. Like Old Dog said "a little knowledge is very dangerous!" But I'll do my best to learn and understand what's behind all this.

I had setup a new demo account with reasonable opening balance and lot size. I want to start with H4 to be safe.
Attached Image (click to enlarge)
Click to Enlarge Name: H4.jpeg Size: 152 KB
  • #126
  • Edited 12:12pm Nov 18, 2010 12:00pm | Edited 12:12pm
So here is the trade on the 15M chart.

I got in at 1 stdev as you can see on the first picture and got out at the mean as per the second picture. It happened about 60-62 ticks later (15 minutes periods) and as you can see, both plots are not starting the same way.

What strikes me is that the mean moved a lot (which is normal since we're adding values and recent values have more weight). The mean moved quite a bit in fact. On the second plot, the point where I entered the trade is ABOVE the 1 stdev line while it was right on it at the moment I got in. This would mean that scaling out and then doubling is the equivalent of a martingale since you are not aiming for the same mean anymore.

As for the trade itself, I went with the unit in the header and scored the following pips:

Pair ////////// Lot ///// Pips @ close
GBP/USD /// 1000 /////// 104.5
NZD/USD // -628 /////// -81.1
USD/CAD /// 657 /////// -75.1
USD/CHF // -212 //////// 51.1

More with the trade on 1H when it closes.
Attached Image(s) (click to enlarge)
Click to Enlarge Name: Test Arbomat.jpg Size: 82 KB
Click to Enlarge Name: Test Arbomat 2.jpg Size: 81 KB
  • #127
  • Nov 18, 2010 1:27pm Nov 18, 2010 1:27pm
Hi. This is because de Begin is not the same in both case IMHO. 7bit is it posible to leave the begin position in the same time all the time to observe the fluctuation of the hedge over time?

Regard
Sorry my poor english, it's not my first language
  • #128
  • Nov 18, 2010 1:34pm Nov 18, 2010 1:34pm
Quoting abroco
Hi. This is because de Begin is not the same in both case IMHO. 7bit is it posible to leave the begin position in the same time all the time to observe the fluctuation of the hedge over time?
Yes. Just draw the back line into your chart (draw a vertical line and call it back). I have mentioned this a dozen times already (in almost every third posting I mention this).

The next version will also have a line that you call now and it will only use the bars before now to do the regression. I will post this version later, I just have to clean up the code a bit.
  • #129
  • Edited 2:25pm Nov 18, 2010 1:39pm | Edited 2:25pm
The regression was fitted to the bars before the red line (2010.02.10 0:00) and the same coefficients were used for everything after it.
Attached Image

Attached Image
Attached File(s)
File Type: mq4 trendomat.mq4   13 KB | 1,140 downloads
  • #130
  • Nov 18, 2010 2:42pm Nov 18, 2010 2:42pm
Attached Image (click to enlarge)
Click to Enlarge Name: 2010-11-18-213722_1023x514_scrot.png Size: 57 KB

Attached Image (click to enlarge)
Click to Enlarge Name: 2010-11-18-213854_1023x589_scrot.png Size: 50 KB

Attached Image (click to enlarge)
Click to Enlarge Name: 2010-11-18-213815_1023x590_scrot.png Size: 49 KB
  • #131
  • Nov 18, 2010 2:48pm Nov 18, 2010 2:48pm
Quoting luiboy
I just know how to follow instructions.
This is because there are no instructions yet. We are all here to find them. We have a tool and now we need to find out what we can do with it or whether it is of any use at all.

Btw, I notice that your lot size numbers seem too small to be useful. You should increase base_units to get higher numbers to have more precision after they are rounded to the nearest integer.
  • #132
  • Nov 18, 2010 2:55pm Nov 18, 2010 2:55pm
thank you 7bit for a new version to play with, but what does 'use_diff' exactly do?
  • #133
  • Edited 3:16pm Nov 18, 2010 3:01pm | Edited 3:16pm
Quoting ultfrisbee
What strikes me is that the mean moved a lot (which is normal since we're adding values and recent values have more weight).
They have all the same weight but the coefficients change and so not only there are added more points on one or the other side of the mean, also the shape of the curve changes slightly.

I have currently a trade running where it already moved beyond 4 or 5 standard deviations, the whole curve is already changing its shape and trying to bring it back into something more horizontally looking and I keep adjusting the positions (not martingaling, only moving around the same money from A to B). I have anchored the beginning of the regression to a fixed time with the "back" line, so the total number of bars is increasing. I have put a profit target at the other side of the mean at where -2*stddev is now because I expect it to do the same into the opposite direction soon. as the number of bars increases stddev and the total size of the swings keep increasing too. If it eventually swings back it has room for an equally big move to the other side.

I intentionally don't close this because I want to know how this will end.
  • #134
  • Edited 3:41pm Nov 18, 2010 3:15pm | Edited 3:41pm
Quoting Mediator
thank you 7bit for a new version to play with, but what does 'use_diff' exactly do?
it runs the regression on the price differences (first difference), If I have understood the few articles I read about it correctly then this is how it is usually done. The old version did it with the absolute price. This brutally enforced an absolute trend or a perfect range during the training period, no matter what the basket really wanted to do if left alone.

It is now trying to fit and hedge the relative moves of the individual candles as good as possible and the resulting overall trend is only a side effect of this. It puts much more emphasis on the hedging of the short-term correlations (which seem to stay constant) and this results in a smoother curve and also seems to give coefficients that don't fluctuate so much.

If I do the same with the arbomat.mqh (ranging mode) I cannot produce these artificial almost perfect horizontal ranges anymore. It results in something different that has extremely stable coefficients but cannot completely remove the overall trend anymore. But it still looks predictable and tradeable. I will post this modified version of arbomat soon.
  • #135
  • Nov 18, 2010 4:25pm Nov 18, 2010 4:25pm
Quoting 7bit
Yes. Just draw the back line into your chart (draw a vertical line and call it back). I have mentioned this a dozen times already (in almost every third posting I mention this).

The next version will also have a line that you call now and it will only use the bars before now to do the regression. I will post this version later, I just have to clean up the code a bit.
Back now. Lovely touches. Thank you, I will be using them alot.

JC
  • #136
  • Nov 18, 2010 7:20pm Nov 18, 2010 7:20pm
[quote=7bit;4183197]Yes. Just draw the back line into your chart (draw a vertical line and call it back). I have mentioned this a dozen times already (in almost every third posting I mention this).

Thank you very much 7bit. Finally I undestand perfectly.
Now I'm playing with the back vertical line

Regard
  • #137
  • Nov 19, 2010 11:27am Nov 19, 2010 11:27am
Hi 7bit

Do you need new common file or .dll for trend-o-mat?

JC
  • #138
  • Edited 1:01pm Nov 19, 2010 12:21pm | Edited 1:01pm
Quoting Waka
Hi 7bit

Do you need new common file or .dll for trend-o-mat?

JC
No, it should work as usual. Are you getting any errors? (Don't post the MT4 crash log! (If anything then post the DebugView output on level 2 and only after you have read (and followed) the instructions that are always written at the top of all my mq4 and mqh files). Please post only legitimate errors, errors in my code, errors that do not go away after following the installation instructions. I don't write the detailed instructions into the files only to repeat them again and again and again in the forum, I write them once and at a prominent place (at the very beginning) that can't be accidentally overlooked by anybody who looks into the file. And you don't install mql4 files from some random stranger who looks like Santa Claus without at least briefly looking at their contents!

If it crashes then most likely you forgot to set the path to Rterm. The downloaded file will always point to my installation or R which is almost guaranteed to be different from yours.) Here in Germany the folder is called "c:/Programme/..." and in an English version of Windows it is "c:/Program Files/..." for example. Also I still have R-2.11.1 and you might already have R-2.12.

You can modify the startR() function in mt4R.mqh to always ignore the path and hardcode it to always use the one of your installation if you don't want to set the path all the time with downloaded EAs. Then you have to do this only once when if you update mt4R itself.
  • #139
  • Nov 19, 2010 1:04pm Nov 19, 2010 1:04pm
Any (un)successful trades or interesting insights, anybody today?
  • #140
  • Nov 19, 2010 1:16pm Nov 19, 2010 1:16pm
I will be looking at this in more detail , once my exams are over.

It's frustrating, knowing this is here to drown in, but no time to look.

JC

Oh BTW, I removed arb-o-mat, put trend-o-mat on, then put arb-o-mat back on and all working fine. I assume no one else had, what seems to be, conflict between the two, but both work now.
0 traders viewing now