Hedge and Correlation Strategy

1 indicator of news !

Yes, option 1 is renentry. Any body wants entry at 90 or 95, they can set with present EA. But What we all really expect is a recovery for bad trades.

something to thing about:

now we check every combination for correlation we take 50 bars in the daily chart. it is the same if we would calculate the stochastic-difference in the daily. the pro is we are not bound to set a referencepoint (50 bars). this is the multitimeframing-trading i spoke at some time.

and what about this idea:

you know if we get a tradesignal for Pair1 and Pair2, in this case Pair1 sell and Pair 2 buy. this trade could ending in 4 different scenarios like you'll see in the picture below.
we could filter out the first two examples (a,b) when we check if both pairs cross some Moving average or breakout of a defined channel.

one of the example could be traded if Pair1 is volatiler than Pair2 and cross of MA or breakout of a channel(c.).

so for this example:

a. never trade because no crossing or breakout occurs.
b. never trade because Pair2 is volatiler than pair1.
c. trade this because Pair1 cross MA or Breakout-Channel AND is volatiler than Pair2.
d. trade this because both Pairs cross MA or breakout-channel.

lets discuss this if you are interested in.

Hi ... it's me again.. i at this moment trying to do my own analysis and ways of improvement on my trade using dreamliner strategy.

ok here goes...
earlier i've been opening trade on stochs that crosses above 80 or 90% and take profit on closing 20% which all of you were doing...

well for the past few weeks, i've been trying this method below and its giving me an overwhelming results of few hundred pips/ week.

i look for the same entry point rule for example of the stochs show gbpusd above 80% while eurusd is way below and the gap is somewhere 60-80%, where i would open a sell on gbpusd and buy on eurusd... this method works provided if you have the holding power.. normally i had a few screnarios where i had to wait sometimes few hours - 20 days to take profit.. meaning at the open of this trade most of time there's a retracement to the opppsite direction and obviosuly if combine both sell and buy trade its easily negative in huge drawdown ( now why not take advantage of this drawdown while maintaning the objective)

Now why not still sell at gbpusd and buy at eurusd and make sure we wait for the target of 20% to be reach to either profit or loss and at the same time open opposite trade buy gbpusd and sell eurusd, if this trade is in positive, lets take some decent profit and close the trade while we wait for the original trade to close to 20% stocks level at profit or loss.

if the entry rule matches.
1st objective must be executed

2nd objective make opposite trade.

you might have a question, what if the 1st trade went on profit, then close all four trade to breakeven.

We are interested to open 1st and 2nd trade hoping to take profit on the 2nd trade while mainting the 1st trade objective to be met.

the result is outstanding.

I've posted ealier of my trades without this method above, hopefully next week i will do so and to let you know what i mean.


adious.

The EA already waits between the 2 trades for the "trade context" to be available;
I think the "trade context" state is handled by the broker, so when it says "ready" then it means the trade can be sent and the broker should accept it.
What i've seen happening sometimes is trade rejected due to "invalid price", which means the slippage may be too small, or bid/ask has changed during trade handling: it may take several 100s milliseconds before the order gets to the broker due to transmission times, and that can be enough for price quotes to have moved away from requested opening price, even though the EA always refreshes the rates before sending the trade.
Same for closing.

I can add a loop to repeat until trades are executed.

Sq

The attached one is the overlay chart of gbpcad/eurchf. It is hourly chart and EA place trades gbp buy at 1.5888. Sell eurchf at 1.3153.

This was the bad tade i was talking about.

Thanks sq. I have now adjusted the slippage to 5.

sq, can you help to develop a version with pip target take profit? i think it should be very easy to code, but I'm not programmer. Many thanks.

Squlaou :

I attach more overlay charts with stoc diff indi and basket profit indi. Both are 5 min chart.

a) Euraud/euusd
b) gbpcad/eurchf.

I would consider following for entry with overlay chart and indicators:

Option 1 : The curve in basket profit indi is between "0" line and bottom/top. The stoc diff must be > 50. The buy and sell trades shall be placed as per stoc indi. I mean if blue line is at bottom then buy. Red line on top then sell.

Option2 : When red line and blue line cross each other in stoc doff indi.
If blue line was previously at bottom and crossing over red line then i would enter buy for blue line currency pair. If red line was previously on top and crossing over blue line then i would enter sell for that particular currency pair. I would not consider this as best entry as after stoc cross, we will not be sure which way the pairs move.

Exit : Cash profit.

Hello all, I wanted to congratulate you all on an excellent thread here! I think this is a great start, and this concept and it's tools have real potential. However, I wanted to address some of the issues that I, and others have seen, and provide potential solutions to them.

First off, although some of my solutions are mathematically based...I am NO mathematician! I will be as descriptive as possible, but I simply lack the math vocabulary to illustrate these ideas in proper formulas. Bear with me, and feel free to ask any questions for clarification. I will be happy to meet on skype for anyone who can actually code this, and needs clarification.

second, i am using two pair trading to illustrate my examples, but they would be equally valid on trips or quads trades.

ISSUE 1: Using stochastics to determine when correlation is "out of sync" - Someone mentioned that stochastics could be this systems "achillies heel". While I think there is a bigger issue...this one is simple to address, and improve upon.

Solution: determine the average fluctuation of two correlated currencies in terms of percent, over a specified period of time. Then, determine the standard deviation of this average. Then, set the system to enter a pair trade when a specified standard deviation of this average occurs.

For example: Over the last 24 hours, EUR/USD fluctuates an average of 0.1% in value from EUR/CAD. We determine we want to enter a trade if EUR/USD pulls away more than 0.3% in value from EUR/CAD...which would be 1 standard deviation from the 0.1% average.

Why this is better: First, rather than entering on an arbitrary number, you can enter on a specified distance from AVERAGE fluctuation... key word is average. Stochastics only show directional big price movements relative to other price movements in the period setting, but they do not measure how much a "NORMAL" fluctuation for a particular correlated market is, nor how it may change from 1 day to the next. By also taking into consideration how fast and how much volatility rises and falls, this concept will provide a larger edge.

ISSUE 2: Reward:Risk ratio - i.e. DRAWDOWN. Some drawdown is natural, but it seems that the amount of drawdown incurred from many who have posted can be destabilizing to both accounts and confidence. I have a suggestion that may be able to reduce this, and help determine expected profit.

Solution: In our example above, by determining average fluctuation (0.1%), and our 1 standard deviation of this average (0.3%) as our entry, we decide to enter a trade, and keep a 1:1 Reward:Risk ratio...so since we are making a potential 0.2% average move as our two pairs pull together, we decide to stop out if our pairs diverge by more than 0.5%, or 0.2% more than our entry price. These percentages can be converted with some approximation into pips...so you have a much better idea of much you will make, and how much you are risking.

Why this is better: Even a system that has 99% winners can lose money if it doesn't manage risk properly. Black swan events can and do occur, and the market can stay irrational longer than anyone can stay solvent. By using a 1:1 Reward:Risk, never again will a single losing trade erase even 2 winning trades. It's hard to make a profit if 1 loser can destroy 3 or 4 winners...and likewise, it's easy to make a profit when it only takes a single winner to make up for a loser (or more!).
Besides... who here is psychologically immune to drawdown? I'm not! and unless you can say you have never closed a trade early when it was pulling down...or you didn't second guess your system when you had a couple big losers, then you really can't afford that much drawdown either.

ISSUE 3: Poor entries due to reliance solely on indicators for entries. Since this is a system that profits by trading when price (not stochastics) returns to the mean... it would be best to let price have the final say on when it's about to head back towards the mean.

Solution: Include a price action confirmation before entering. Once our "indicator" recognizes a large price deviation in an otherwise highly corrolated pair set, it should wait for a single candle in any pair to close above the high (if going long), or below the low (if going short) of the previous candle before entering the trade. If a single candle will push too far back to the mean, zoom in on a smaller time frame, and wait for a single candle in that smaller TF to close above the high, or below the low of the previous candle.

For Example: Say you are watching the 30 minute TF, and you have your indicator/ea/whatever setup to wait for a 2 standard deviation move away from the mean price fluctuation in the eur/usd, and the eur/cad. They have been pulling apart for 3 candles, and as the 4th 30 min candle closes, it is over a distance of 2 standard deviations away from average movement. Rather than buy/sell then... the EA would then "zoom in" on the 5 min T/F. It would wait for a 5 min candle in EITHER pair to close above the high of the previous candle (if going long), or below the low of the previous candle (if going short)... it would also only take the trade if AFTER this conformation occurred, and if it still provided a minimum 2 standard deviation move to equilibrium. If for some reason the 5 min candle shoots real fast, and when it closes you have less than 2 standard deviations away from the average change, then it would SKIP THE TRADE.

Why this is better: We only make money when price...not indicators, closes the gap. And although drawdown is a necessary evil, it can be mentally defeating. Therefore, we should only take trades that demonstrate reasons why they will drawdown less frequently, and by less pips, than other trades. In other words...just take the A+ trades. By waiting for price to have some say in it's future direction, we will take less trades, BUT they will (almost by definition), be higher probability trades.

ISSUE 4: overall risk management via trade management. Not only is it emotionally difficult to have a decent profit turn into a loss...it is bad trading. these causes losses of capital and confidence. Once Green, Never Red. It is necessary to insure yourself against loss, and guarantee yourself some profits via proper trade management.

Solution: Never let any trade fall below a 1:1 reward:risk ratio. By simply trailing a stop, you insure you are cutting your risk almost immedietly, and you are locking in profits shortly after, so winners stay winners, even if they would pull back and become losers shortly afterward.

For example: you enter a trade that two pairs have pulled away 2.3 standard deviations from the average price variation. You wait until one of the pairs shows smaller Time Frame price action confirmation, and when that smaller TF candle completes, you are 2.1 standard deviations, or 0.5% from a reversion to the mean. You enter a trade, one long EUR/USD, one short EUR/CAD, and your initial risk you decide should be 0.3%. Lets say your pairs pull back together 0.1%. You are now 0.4% from your profit target, and 0.4% from your stop loss. Then, your trades pull back together 0.2%. Your are now 0.2% from your profit target, SO, you slide your stop up to only risk 0.2%. At this point, you have now guaranteed a profit, eliminated the risk of loss, while maintaining your chance to make your full profit target.

Why it works: There is ONLY 1 THING YOU CAN CONTROL ONCE YOU ENTER A TRADE... HOW MUCH YOU LOSE! You never have control over how much, or even whether you win. Also, it would be foolish to come 3 or 5 or 7 pips from a target on your pairs, with a potential 35 pip loss, and then have it fall all the way back and lose money. By focusing on trade management, you quickly eliminate the possibility of losing, while maintaining the possibility of winning. The secret to successful trading is putting yourself into a "no lose" situation as soon as you reasonably can. You will have some potential big winners stop out at break even, or small profits...but they will be few, because most big winners don't give back profits. Just ask anyone on the wrong side of a strong trend how many opportunities they had to get out before the trade went against them...generally none! Eliminate your losses...and your winners will take care of themselves.

I have more (believe it or not), but I'll be shocked if anyone reads all of this, so i'll save more for later, if anyone is interested. Also, if there are flaws in my logic, please point them out.

I believe with the proper measurement tools, entry criteria, stops, and trade management, this can pump out profits, day after day, week after week.

Greg

An edit to the post above, regarding issue 2 and 4:

Although the system as is does have a stop loss...when the stoch diff is less than 50, close trade. However, a more aggressive capital protection system would provide better capital preservation in those situations when the stoch just rides above 80 and below 20 for quite a while (it does happen!)

Remember...for this particular system...it seems the people struggling are doing so because 1 or 2 big losers are wiping out a week of winners. by incorporating an absolute stop that is determined by statistical movement and current volatility, the big losers will be completely eliminated....

Hi Greg,

Very interesting ideas. I think we can´t require Stochastics more that what they can do, I mean, Stochastics have their limitations and we have to accept that.

I think we have to look for another way to take advantage of the gaps between correlated pairs.

I have to read again your post......interesting.

Ah yes...It seems I wasn't very clear there. your right...Stoch has its limitations...so what I propose is for someone who can code an indy should get rid of the relatively arbitrary stochastic measurement alltogether, and create a new indy that would show a typically correlated pair set that is temporarily out of correlation by measuring a standard deviation based distance away from the "average change." it wouldn't be stochastics...

i hate to compare because it is nothing like a bollinger band, but the similarities would be that bollinger bands take into consideration changes in volatility, as well as standard deviation away from the average price over a period.

in effect, rather than "price on this one is over 80, and under 20 on this one, so i should sell/buy" it would be like

"over the last 72 hours, price on this pair set has fluctuated by an average of 14 pips up or down, and that corrisponds to an 80% overall price correlation. Right now, they are 53 pips away from their average, which is 2 standard deviations away. 2 standard deviations means prices only gets that far from one another 5% of the time. So, there is a 95% chance that if volatility does not increase, our pairs will start to revert to the mean...giving us a high probability of profit. Of course, if volatility continues to increase, there is still a good chance that we will have a mean reversion move. "

The bottom line: Rather than using a relatively arbitrary way to determine when correlated pairs are out of sync...use a precise statistical measurement to more thoroughly evaluate the opportunity. This will allow us to seperate the wheat from the chaff...

In other words, not all 80/20 stoch differentials are equally good. This would tell you exactly how good they potentially are.

Very well thought out, I think. Removes the indicator based issues, leaving the pure logic and price action. Impressive, and should be possible to mod the existing indicators to show deviation from average.

Over to the experts.

thanks greg,

this is what i tried to told in my post also, but you pin it on a point. this wisdom could applied to most every system and every trading method.

some thoughts to issue 2: there is a way to cut losses short, because this is the only thing we could control BEFORE we enter a trade. Let winners run as long as you can is what we do not know in future. because moneymanagament is not in itself which makes a system profitable. the system has to be a positiv expectancy than we could push it with proper MM.
i wish i could express my thoughts like you Greg, but it is not my native language.

Mediator

i think this could mathematical expressed with the TMA indicator. it is a Triangulated Moving Average with ATR Bands like the bollingers are.

Very nice thoughts Greg, I would love to hear more of them.
I think dealing with it in a statistical way would be a big improvement in comparison to the stochastics.

Mediator, if you have any problems with expressing your more complexthoughts, feel free to PM them to me and I'll give my best to translate them for you.

aw shucks, thanks! it seems i have an unfair advantage over you mediator since i am american, though if it wasn't for spell check, i couldn't express my thoughts like me either!

As far as letting winners run, I do agree with you, to an extent. Since this is a correlated pair strategy...you're not likely to get any more profits than statistical variance will allow (volatility plays a huge role...but even that has its limits).... So rather than have a fixed target....how about a fixed point where stops get moved to break even. If one were to enter a trade that was 2 Stand. Dev away from mean price change, one could reasonably expect to return to mean price change...AND, probably move 2 stand dev. in the other direction (like bollinger bands, price moves from band to band)... but after that, a further move is unlikley, as you are now hoping for a directional trade, not a mean reversion trade. it would invalidate the entire premise of this system

So, how about trailing 1 Std dev below price as it moves in your direction. Effectively locking in profits as they come, and eliminating the risk of a winner turning into a loser...? and if u do happen to get a HUGE variance in your favor...well, u trailed it the whole way...Just a thought...

And actually, money management alone CAN make some trading methods profitable. Trend trading systems in particular can be pure random entry (heads go long, tails go short), and with aggressive stop trailing (trailing stop a certain ATR, always moving it as price moves in your direction, or as volatility decreases ATR, ATR moves by that decreased amount)

In fact, over a 10 year period, using an ATR based stop trailing method, and trading the daily TF, one could be profitable in the futures market of their choice by simply flipping a coin when stopped out...ignoring all trend, and going long on heads, short on tails. (except possibly stock index futures...)

A study was conducted by Van Tharpe, a pro trader, author, and trading mentor, as well as Larry Hite, the market wizard (if i remember correctly)

Both took over 30 randomly selected futures markets... flipped a coin to determine entry direction...and lo and behold! every one was profitable over a 10 year period... in fact, i want to say profit was made each and every every year on each market....but with some large drawdowns.... though don't quote me on that.

Point is... it's a game of risk management, risk management, risk management. I'd guess almost all trades are positive at some points.... so do your best to find efficient ways to eliminate all risk after that positive point is hit, and you can make a lot of money.

ANyway... lets see if someone will code this up for us... cuz like I said, it would be really nice to have a better way of seperating the A+ setups from the rest.

I' m sure Greg that coders will come.....

Aright Chris...just for you, since you asked

First off, it's already been addressed, but i'll mention it anyway. The E.A. needs to NOT trade around major news announcements. Anything in Red when it comes to the U.S., or anything in red if you are in that currency...and possibly correlated currencies (major EUR news...get out of GBP).

I'd stay clear for 15 min before, and 30 min after, tho this could easily be left to the individual trader to set parameters i'd imagine. Sudden strong emotional markets are poison to mean-revision strategies.

Now, when trying to determine an appropriate time period and length of time for a statistical (as opposed to arbitrary) indicator, it must be factored in that the forex market does not move at all times the same way. If one were to trade 1 hour bars, it would probably be best to have a minimum of 24 hours on the time period. Imagine having an 8 hour setting, just as london opens. Your indicator will show all sorts of crazy "A+" trades! The problem is... it isn't incorporating the average moves of london, so what seems to be a 3 standard deviation move (a move only 1 in 300 candles would reach) is really just a 1 standard deviation move in london (the type that 1 out of 3 candles experiences). In fact, i suspect those who trade 30 min bars or longer during the london open, for trades that trigger that first hour, i wouldn't be surprised if they don't do as well, or win as much... it's a lot easier to push a stoch to the 80/20 levels on pairs when a sleeping market gets suddenly kicked out of bed.... but that doesn't reflect true imbalance...some of that move is just volatility...which stochastic misses.

Obviously, if pair sets only push away from each other in price 1 in 300 hours...it's MUCH more likely you'll make good profits, quickly, with low (if any) drawdown, then if they do this every 3rd hour (read: little to no edge)

So, maybe once london opens, have a "staggered" time parameter setting... a "smart indicator" if you will. Come london open...it now calculates the average standard deviation starting with the hour before london, and the 5 hours after london but before U.S. opens... and it would take that same time period back for several trading days trading days....

So, if you have a 1 hour chart, and the "lookback" time parameter it uses to calculate your standard deviations is set at, say, 19... this would mean once london opens...it would calculate standard deviation using a completely different set of data than just the previous 19 hour candles.

It would look at the last hour (before london), and then would JUMP BACK A DAY, and calculate the hours that london is open, working backwards from the hour before the U.S. opens, until it hits the hour before london opens. That should cover 7 hours of the 19 period setting (1 hour for the previous hour before london...and then the 6 from the same time period... 2:00 am - 8:00 am Eastern Standard Time New York)...then, it would have 12 hours to go...so it would jump back one more day, and repeat, and then one last day, and repeat.

WHY SUCH A PARTICULAR INDICATOR? Isn't this nuts??? Maybe...but i don't think so. If your trading london... and you want to know when a pair set is greatly out of whack with it's average fluctuation...you want to compare apples to apples.

If you take a slow asian market, and the indicator uses the average volatility and corrilation and price swings from asia...then EVERYTHING during the london session will look great (it'll look like "amazingly big deviations from the average price change")

For the same reason public companies compare their earnings not to the previous quarter...but to the same quarter the previous year.... we should break down our statistical variation by times...because we want to see the TRUTH... the truth being:

"we are looking for an time when two or more highly correlated currency pairs have become temporarily highly uncorrelated, to the degree that we can sell one, buy the other, and have an extremely high probability of them closing the temporary price gap between them, with as little drawdown as possible."

So...the questions are...how high is "highly"?, what exactly is "the degree", and what makes us believe that THIS situation REALLY IS high probability? What makes this one better than the rest?

The only way to know, is to develop very specific, accurate tools.... and we can't get specific, accurate tools unless we input specific, accurate data.

Period.

So, if it's london open, and u wanna know if a price swing really is out of whack for a pair set...then you need to know how much it generally moves during previous london opens...not just the last 6 or 12 or 24 hours.

So much more... but i need a break from typing. will be back later to see others input.

Greg.

P.S. Don't confuse specific and accurate with "curve fitting" Specific and accurate is inputting the best data possible... curve fitting is throwing out the best data possible to only see a certain result.

So, you either take the whole london session...or none of it...but not the first hour today, the middle two hours from yesterday, the last hour from last thursday....etc.