Hedge and Correlation Strategy

Exactly! Well...almost exactly! It does depend on how the news comes out.

Personally, I'd definately be out 15 min - 1 hour before news, but depending on how it came out, would define when I get back in.

If news came out as expected, and it was not an interest rate or NFP announcement... you could start looking as soon as 30 minutes after, but an hour to be safe.

If it was an interest rate (for any country) or NFP (for U.S.), if it was as expected, i'd stay out for at LEAST an hour...but probably the rest of that session.

If it was a surprise that isn't NFP or interest rate... but it caused a big move, i'd wait an hour or two at least.

Of course, if an interest rate announcement, and it was either surprising, or even just an expected change, i'd stay out for 24 hours.

when it comes to NFP....probably best not to trade at all that day. too much seemingly random volatility.

So, set the bot to stay out 1 hr before all major news, until 1 hr after, and simply turn it off on potential interest rate annoucement days, or NFP days... this is how i'd do it anyway.

There is a mathematical way to do it..but i'd rather not get into here...this should be good enough for now, unless a coder wants some input, then i'll get into a math based way.

Hi Josch,

Yes, i have two methods likely using Dreamliner startegy ..

1) as what i've stated

and another

2) i use 15tf and dailyTF, based on the rules of enter i will take additional step that is to also check if the blue on top for 15TF then the dailyTF also must be up. same goes for red. then i decide my entry point.

adious.

Really interesting and thought provoking posts you have put up Spinfoil...

Good start of the week for me. 38.7 pips from 5 double trades including one recovery as follows:

EURCHF-GBPCAD 27.9
EURJPY-GBPUSD -16.6
EURCHF-GBPAUD 3.3
USDCHF-USDJPY 6.5
EURCHF-GBPAUD 17.6

The last 4 I closed manually at the same time because of the DD on the EURJPY-GBPUSD. What I should have done is close 2 and 5 and let 3 and 4 run. After I closed them the EURCHF dropped like a stone and it would have a cracking trade. As it turns out I have probably left 50 to 70 pips on the table. But hey I was focusing on eliminating DD and I was not sure which way it was going.

I am still on M5, 85%, 10%.

Upon reviewing last week's performance I found that the GBPJPY sets have resulted in a loss or at least BE so I removed them. I am currently on (with results since start):


EURCADEURUSD (no trades since starting)
EURCHFGBPAUD (76 pips total)
EURCHFGBPCAD (114 pips)
EURCHFGBPCHF (-4.7 p)
EURJPYEURUSD (22.8)
EURJPYGBPUSD (70)
EURUSDGBPUSD (17.4)
GBPCADGBPUSD (no trades)
USDCHFUSDJPY (20)

I had other sets prior to this. Total pips since starting this strategy 388.7, 4.65% growth. Max DD ~4%.

Hope this helps.

This was the bad trade i was talking about last friday. The EA placed trdes buy gbpcad at 1.5888 and sell eurchf at 1.3153. Traded in hourly char. Had a max DD of 120 pips at one stage.

Today it closed with 11 pips loss. Not bad. My entry was at 80 % exit at 40.

My turnover in europeon session :

5 min TF : 65 Pips in one platform. dd 35 pips. Entry at 90 exit at 10

5min TF : 47 pipsin another platform. No dd. Entry at 90, exit at 10.

Hourly TF : No new trades, but the bad trade which was placed on last friday had max DD of 120 pips, but todat closed with marginal loss 11 pips. Entry at 80 exit at 40.

Hi Calillou

I am a new member I was reading your posts with stohastic EA you were using, I was having a bit of trouble putting it to work. I've copied and pasted it in EA directory but when I try to drag it to EUR/USD chart nothing happens, could you assist me please.

Still a rookie

Cheers

Navk...question for ya....that "bad trade" that drew down 120 pips...was it held over a news release involving one of the currencies?

Greg

Yes, it was canadian news release.

Well, there ya go... To the coders out there, besides eliminating the stoch and replacing it with a standard deviation of fluctuation, I think staying out at least 15 - 30 min before major news releases, and staying out until an hour after the news would be a crucial (and somewhat simple?) component to add to the E.A.

Navk...am I correct in assuming you are trading using the E.A?

Sure. Just send me a pm with your problems.

Thank you Oc...I haven't even looked in the"trading systems" section of this site in months, primarily because I find nearly all utilize some perversion of price action, and i'm a pure price action trader.

However, this strategy holds real merit, and I am strongly considering incorporating it in my trading if certain modifications could be made... and I believe they would greatly improve the system.

It's really just a matter of understanding WHY something is working... stochastics in this case... it works because it shows when two or more otherwise correlated pairs become pushed further apart due to speculation than "actual" value would dictate...

right now, stochastics is used to measure this... but why does stochastics work? it works because it is a way of measuring when price is extremely out of kilter with "actual value"... but exactly how out of balance? and how much is enough to have a high probability trade?

This is where stoch breaks down... and statistical measurements rock.

I just like to ask why... and "how come"... and if the answer doesn't make sense...i keep looking.

But thank you for the compliment, I hope someone can make use of it for us all.

Greg

[quote=spinfoil;4394974]I'll chime in on this one, if you don't mind wallstreet...

So... when two countries have currencies with a high level of correlation, and one of those currencies rises a great deal in price compared to the other...but there has been no economic reason for this to happen... it doesn't "make sense", and the market will correct it...by buying the lagging currency, selling the leading one...or both.


Wha? No offense spinfoil, but I'm reading a lot of high-brow hyperbole here that really doesn't make a lot of sense. Lagging currency? Leading currency? As they used to say in the army, "sorry, but no such animal exists." Price is measure by weight, not whether it somehow "lags" or "leads." Accordingly, the pairs either correlate or they do not. Once correlation becomes either too negative or too positive then there is no correlation at all.

Oanda maintains the definitive currency correlation heatmap here:

http://fxtrade.oanda.com/analysis/currency-correlation


And as far as the Stochastic indicator goes, people lets forget about correlation altogether and just get back to the fundamentals:


"Developed by George C. Lane in the late 1950s, the Stochastic Oscillator is a momentum indicator that shows the location of the close relative to the high-low range over a set number of periods. According to an interview with Lane, the Stochastic Oscillator "doesn't follow price, it doesn't follow volume or anything like that. It follows the speed or the momentum of price. As a rule, the momentum changes direction before price."


As it just so happens, the stoch is pretty slow in even doing that much (remember this thing was developed in the '50's), what its best used for is measuring convergence and divergence-banks use sophisticated algorithms to quantify these differences and make big money doing it. How do I know? I used to work for one of these large players.

Just watch price move during the end of the NY session each day as each institution starts to square their positions-slowly creating the trends that move the market, then you'll understand whats really happening with "price action."


Thanks!

[quote=tradervic63;4398112]

Quoting spinfoil

Disliked

I'll chime in on this one, if you don't mind wallstreet...

Ignored

Vic, appreciate your comment on my comment... I may have not used the correct wording when I said "leading" and "lagging". I simply meant that if there is an established "average level of positive correlation" between two currenty pairs that has existed for X periods, then a short term (much shorter than X periods) divergence from this "average level of positive correlation" creates a potentially high probability profit situation by buying the currency pair that has dropped in value (this would be the "lagging" pair) relative to the other currency pair that has risen in value (the "leading" pair)

I didn't actually mean one currency can be used as an early indicator to dictate how the other will act...

And as far as "sophisticated algorithms" go... I don't know what banks use, having never worked for any trading institution, tho I will take your word for it... but I think a simple standard deviation of the average correlation level for X periods could be a "simple algorithm" that would be effective... and if one wanted a more complex system, albeit one that would probably work even better... then one could isolate particular time periods (London, US/london crossover, japan..etc) and find the average divergence in each session. For the purposes of calculating the "average divergence in each session" data, one would also want to eliminate periods right before and right after previous news announcements, as well as staying out of the market when news is being released.

This way, you really have a "pure" average level of correlation, and can trade extreme divergences when they occur.

also, you mentioned "just watch price near the close of the N.Y session, and you'll really understand whats happening to price action".

My whole foray into this thread was to share insight, and hopefully have someone develop an improved E.A. based on some of my suggestions. I already watch price for approx 8-10 hours a day, and I would like to find automated ways to increase my profitability without me watching more screens than I already do.

That being said, I think we're on the same page actually... I just didn't explain myself well. Personally, I don't look at any market, currency or otherwise to find a "leading" one or "lagging" one.

Greg

A trader who trades using 5 min, 15 min or hourly chart has to see deviation in that particular day.

For a trader who trdes using daily chart, may want to see deviation in weekly.

Sell when deviation reaches 80 % to 85 % buy when deviation reaches 20 0r even 10 %.

Trading manually is easier as we do not have an EA to measure the deviation levels.

If the deviation is positive for that particular day, goto 5 min chart and wait for deviation to reach 20 % and trade.

is it possible add time for trade with this strategy?

between 13:00 - 15:00 not so good !

Hi All,

I just recently found this thread and it really caught my attention. The idea definitely has some potential in my opinion and worth spending time in research.

So, I steal Going Home EA from a parallel thread (I hope SteveHopewood is not going to be mad at me ) and used it as a base to create a quick and dirty mod trying to implement StDev idea that spinfoil suggested below.

EA calculates StDev for each session (3 columns abbreviated "A","E", and "U") for each pair. I use SMA in EA to find mean and I added appropriate parameters to change MA period and TF.
The second column is a compound one which is (1st pair fluctuation + 2nd pair fluctuation) / (1st pair stDev + 2nd pair stDev). I hope I got it right from spinfoil explanation. It did not take into account spread at all at this point. BTW spread is shown next to appropriate pairs.

I did not really have a chance to test it as I just put it together. I wanted to see if the implementation so far meets the requirements. And I am looking forward for your comments and suggestions.

Doing a quick test and looking at the second column I was wondering if it ever comes to 2 st.dev. to open a trade. For the last hour as I started the bot it has been always below 1 for all combinations. I will leave it working overnight and see how it goes.

This EA is just a proof of concept so please do not try to use it on real account.

P.S. I also use a template to blacken the screen for better view. It is optional.

Hi all,

just picking up with the thread after a few days break...

Spinfoil, you inputs are much appreciated and i believe we can code these into the EAs quite easily.
I have always believed in statistical analysis, and reenforced my opinion when i rencently found 7bit's thread about "R" port to MT4, and his baskets correlation optimization EA.
What we are doing here is considering only 2-pairs baskets and try to trade that with the same "mean reversion expectation".
7bit's uses R to find the best lot sizes for each pair of the basket so that its combined "price line" stays as much as possible within [-2..+2]*stddev, and expecting this situation to continue for long enough, so we can consider selling the basket when it reaches +2*stddev from its mean, (or better when it returns below +2*stddev), and buying the basket when it returns above -2*stddev. (that's just one possibility of trading this).

Your suggestions are close to those ideas, only we have not explored yet the lot sizing aspects.

I read you do not want to compare that with Bollinger Bands, but this is exactly what bollinger bands are: they are showing the +/-n*stddev zones around price line mean.
Only here the "price line" is our basket profit curve as shown by my "Basket Profit 2 Pairs" indicator.

However, the BB are calculating the stddev and the average lines over the same number of samples.
We would probably want a short-term MA of the price line (100 bars for instance), to stay close enough to the line, but the stddev being calculated over a much larger sample than just the last 100 bars, to better reflect the typical variation, and be less sensitive to short-term fluctuations.

Here is a picture showing the Basket Profit for +EURUSD/-GBPUSD basket on an M5 chart, with a Red BB(100,2stddev), and a green SMA(100) with a +/-15USD channel around it.
The gray lines are the profit levels in USD for a 0.1 lot basket (buy 0.1 EURUSD, sell 0.1 GBPUSD).



Just drag and drop MT4's "Bollinger Bands" default indicator to the Basket Profit sub window, and select "Apply to" = "First indicator's data".

The green SMA and channel "simulates" the short-term 100SMA with a channel made of a fixed 2*stddev channel (i made-up the 15 for the "2*stddev" value, it would have to be calculated).
This shows how BB is affected by short-term fluctuations, and why we don't want to use BB.

The green channel would be our "trading zone": sell on high level cross-back, buy on low level cross back.
TP on dotted SMA for instance.

EDIT:
To better illustrate the idea of the Basket Profit line with SMA channel compared to the Stoch Diff performance, here is a screenshot of both indicators over the past 5 hours this morning:



I have marked 2 trades based on the SMA100+15USD-channel principle:
if you look at the Stoch Diff, there would be no trade, stoch diff is far away from the 80 mark, it is even in the OS area...
However, we would have made approx 2*12USD.
The third trade after those 2 would be in DD though (not marked), but i can see it heading back to the mean now...

What do you think ?

Sq

sq,

i have downloaded the Basket Profit indicator that plots the combined PL of the basket from a reference point. I have seen some patterns. Most of the times the PL goes out of home (away from ZERO line) and comes back home. Howmuch it goes varies by pairs and market condition. however if you observer last 1 week chart on M5 (scroll to left) you can easily identify the upper/lower boundaries. I sell the basket when PL goes above 100 for EURAUD/EURCHF 0.1 lot each spread 4, 6. this is with the hope it comes back to ZERO. i dont wait for ZERO line, i exit with fixed pips (15-20). My stoploss would be couple of points above week high.

i know the PL can keep going up and up, but for pos corelated pairs, it comes back eventually.

i am yet to find a ideal entry, exit criteria but i can tell this is working better than sto diff levels (entry 90, exit 10). when both agree thats a sure shot.

may be others can contribute and test this

thanks for the indi. after more tests we can create EA out of this

Yes Roundrock, trading correlated pairs like this shows better potential, but that would be on higher timeframes / longer term.

So I coded the SMA and StdDev lines into the Basket Profit indicator, to see if that would give an edge.
It plots the SMA(100) line on top of the Profit Basket line, and calcs the stddev of the Profit-line-to-SMA difference.
Similar as Bollinger Bands, except that we don't use the same amount of periods for the SMA and the StdDev:
we want a short-term SMA to stay close enough to the price line, but we want to use the stddev calculated over a much longer period (500 here) to keep the SD lines stable enough and avoid short-term "glitches".

Here are the mods:

- added "SMAPeriods" (100) and "StDevPeriods" (500) inputs:
this will plot the SMA of the Basket Profit line, and the +1,+2,-1,-2 StdDev lines around the SMA.
If the pairs are highly correlated, then, when the basket Profit line goes outside the -2/+2*SD zone, (=pairs are getting out of sync), it is "expected" to "come back home" = return to mean (=SMA line).

Default inputs of 500 bars for the SD periods may be too short, however, i find the result "stable enough".
Higher values would give even more stable SD ranges, at the cost of a very high cpu load when loading the indi on the chart (then cpu keeps quiet as we calc the SD lines only at loading, and only update the last values each tick).


Here is a sample from this morning EU/-GU (50*D1 correlation of 84%):




The 1*SD is about 16 at the moment, 2*SD = 32 (usd, for 0.1 lot each pair);
In my last post, i had quickly made up "15" for the 2*SD value, but this is more like 1*SD really.

As you can see, my "3rd trade" from my last post this morning would have turned into a winner also, only using the 1*SD lines...

With the present SD lines as in the picture above, we would have closed 2 winners for about 50usd win (0.1 lot) if trading it as crossback the 2*SD line and "going home" (to SMA).

And again, the StochDiff indicator would not have given us ANY of these trades.


Sq