because the 1.01 is artificial, as I said, I added 1 to every cell(I was trying to graph a lot of periods, and needed all of them to start at 1 
So in other words, 1.01 is 1%, 0.98 is -2%
here is the raw data, without the +1
Right
Just a confirmation: the datas are the mean of the whole variations on weekly bases, using the latest week close as input and not the first week, right?
Thatās really important 
hehe, ofcourse, the mean is the mean of the ROC(+1) of each symbol, its just calculated out of the table I gave you
Iām sorry mate but there must be something wrong in your data. The changes are really small (0,0018 for eg) which is highly unlikely :S
Lemme know
Uhm or maybe is normal, given the fact that is average weekly change, when multiplied on a yearly average, the numbers looks more normal
Iām probably wrong on this but isnāt a diversified forex of 28 currency pairings an oxymoron?
Thats good, I was worried when I read your replyā¦ Iāll take a look at the numbers again, and come back to you tomorrow.
Satunya, yeah, it is. But it is so to minimize risk, not to maximize profitsā¦
Hey Satunya,
Thanks for your insight, could you provide me a link to it?
Sorry Richard for disappering from time to time, but gathering some answers I need from a statistics prof is harder than I thought 
Stay tuned
No worrys, I know enough that what you guys are doing is the hard part, so Im watching the thread and looking forward for your progress Let me know if there is anything I can do
Just saw this thread today and read through it, so Iām almost caught up. Didnāt think there were threads like this on BP.
Iāll try and help out when I can.
Hey guys,
A quick update on whatās going on in my mind, my concerns and my ideas in general. Iād love to hear your thoughts if you have knowledge in this field.
- Why am I waiting to build Markowitz CAL model? After all we got all the data!
Yes, but Iām not sure if they could give out a real predictive model. Iāve talked with a Finance friend, and he high-lighted that using past data for estimating future Er doesnāt mean that market is going to respect that mean: this is a too theoretical approach, in practice things are slightly different. In his view, I should find a way to analyze āscenariosā as he said and this brings me to the second point.
- In order to analyze each scenario to ādecideā expected return implies a few things:
-
We need an un-biased model (ndr: Markowitz tells you go long 20% this and short 5% that), which to be honest I donāt have in my knowledge book
. Maybe black-litterman? Not so sure tho, and I donāt really know it so wellā¦ -
Iām a Price action trader, so what makes sense to me is to analyze each scenario at a Support/Resistance etc etc level, which bring up 2 problems: Itās discrectional, not objective, and therefore not statistical appliable, and any econometrics professional would kill me for building a model like this based on discrectional trading. Other issue is that Price action analysis is not going to keep you in so many securities constantly to make up a enoughly diversified portfolio.
These issues brings me to the next idea: using an algo in order to rank pairs related to their ātrendā and momentum. An interesting model can be found in earlier pages, from a friendly user who suggested me to read about RON SCHELLING. Here I need someone to make it happen, as my programming skills are terrible!
So to recap, main objectives I have:
- programming Ronās model (or maybe use it as base to implent a better version)
- finding a proper un-biased model, cos equally weighted portfolio is OUT of my mind, as itās not optimal by definition.
Best,
Cry
Hey there, Iāve been reading through yours and MGās posts and think a combination of the both will work brilliantly. Iāve been trying to quantify the change in lot size based on a currencies relative strength. I had an idea that if you took an average of a basket of currencies (For EUR: (EURGBP+EURUSD+EURJPY+EURCAD+EURNZD+EURAUD)/6, AUD: (AUDCAD+AUDCHF+1/EURAUD+1/GBPAUD+AUDNZD+AUDUSD)/6) then an average of x bars ago and get a % based on that and compare it with other currencies, though I donāt know how to translate that into a lot modification. Iāve not tested it out but I thought Iād add it here to get your thoughts.
P.S. Just thought of a general formula to determine the multiplier, this is just a general formula for now.
Lots = Lots * ((pair strength/average strength)*(pair pip range/average pip range))
Where the average strength is absolute average ie. no negative values.
The reason why Iāve included the pip range is because Iāve seen an example of EURGBP heading up with EURJPY and GBPJPY heading up yet GBPJPY had a moved more pips because of itās higher value. It moved less %. Just another thing to consider.
Hey Heel,
Thanks for your contribution. What you say should require a bit of empirical testing
What I donāt like about it, is its simplicity: considering only the average gives quite a lot of problems from a statistical point of view. Thatās why indicators like variance/covariance are implied, in order to āremoveā the noise of the mean value
Keep me posted with your ideas and tests on it,
Best,
Cry
Hi, I like your idea to, I choose a different route thoug, using some of the information posted in a earlier link, using 2 moving averages on RSI, and using the ratio beetween them to messure strength, if both are rallying then buy, if both are falling out of bed sell, I used the 26 symbols, and bought the 2 strongest, and sold the 2 weakest on the days close. And it provided wery interesting results:
The bottom line is the Equity on this strategy, and the rest is just a benchmark(GSPC)
Unfortunatly, I just got lucky, and more testing just prooved that it was heavily curve fitted:P Just chaning the time period for the test slightly would make you loose everything, and if you flipped the buy/sell signal you would get rich again, so I couldnt trust it;)
This whole idea of trading a āportfolioā of pairs, in the way mastergunner presents it, is entirely flawedā¦ In order to make consistent profits doing this you need to have a statistically profitable entry and exit method that is consistently profitable FOR EACH AND EVERY pair you are tradingā¦
In essence you would already need a winning system in order for this to work, and the ONLY benefit trading so many tiny positions over many pairs, instead of a larger position over one or two pairs that the system works best on, is that it theoretically may smooth out an equity curve a bitā¦ Doesnāt mean it will make performance any better, just a little smootherā¦ In practice the performance will be less as you are trading efficiency for a smoother performance curveā¦
And this is only a theoretical benefit because there is always going to be correlation among pairs, and in practice this will more often then not cause wins and losses to cluster together if the same entry system is being usedā¦ Thereby negating the ONLY benefit trading a āportfolioā of pairs this way would theoretically have.
Also most of the non-major pairs have spreads that are larger then the majorsā¦ So your losing efficiency thereā¦ If you have a technical system that is consistently profitable on 28 pairs individually (doesnāt exist) then choose the several that it is the most profitable on and focus on thoseā¦
Iām not saying to limit the pairs you tradeā¦ Take as many trades as opportunity arisesā¦ But a shotgun approach to opening a plethora of positions just for the sake of spreading around the money isnāt going to workā¦
A decent argument. I think one benefit though is the mindset. This is a very patient approach and might get the impatient trader to step back and get out of the habit of over trading. This is black and whiteā¦hold for large pips. Also, many people donāt have time to trade intradayā¦so it might make sense for them. If Iām not mistaken, basic technical analysis is being applied to varying degrees, so there is an edge between that and the good RR with the trend.
You do have valid points though, I am not about to trade it either. When price is ranging I can get trades, contrary to this type of trading where you just sit and wait.
One thing Iāve seen around here over the yearsā¦the pendulum will swing. Coming off ICTās overload of tools, MG was at the right place at the right time.
Yepā¦ I am noticing the pattern.
Remember that English boob with the yellow Porscheā¦??? Canāt remember his nameā¦
Hey Christian,
I could start boring you with empirical proofs that what you are saying is partly wrong, but Iāll go on with a text explantion (if you desire a mathematical approach, which is surely more objective, just ask):
-
I donāt want you all to get confused: this thread has NOTHING to do with MGās one. It can be easily proven that how MG is trading, is not efficient and it has not an optimal portfolio (ie. higher standard deviation for the same Expected return). But it sounds like Iām repeating myself, as I explained this all already in past pages. What we are trying to do here, is exactly this: finding the best way to build an optimal risky portfolio.
-
Thatās absolutly wrong. Trading a portfolio has another benefit: reducing risk (ie. less standard deviation). Risk sharing/pooling and all other principles that are used by Insurance companies, are the same that drive modern Portfolio and financial economics theories. (see attachement to get an example of how more traded securities emphasize this effect)
-
If you go get a look at the correlation table we built, there are many pairs having a very low correlation ( some have even 0.0n among them) making diversification works. Note also that, once again, this is NOT MG thread: I donāt intend to trade 28 pairs, I intend to build a portfolio off the pairing universe, picking the ones that gives higher Sharpe ratio. Security analysis will tell us WHICH and HOW weighted pairs to be traded (and therefore how many).
-
Yes thatās a good point, but keep in mind that if you consider EMH (Efficient Market Hypothesis), Forex is a highly efficient market in comparison to Equity or even more to Fixed-income market. This implies that the fees paid in FX are way lower than the ones paid in those market.
So yes: non-majors have higher spreads, but this is absolutly a relative concept, and taking it as absolute, not putting it in comparison with other markets, is simply myopic. -
This is surely not a shotgun approach: this implies long lasting trades (also in order to reduce commissions of point 4).
Once again, stop refering to this thread as it was a branch of MGās one, cos is not 
Thanks for your interest,
Cry
-
yes you reduce risk somewhatā¦ Thatās what I meant by āsmoothing the equity curveā ā¦ This does not increase performanceā¦ Actually it lowers it as one has to be sacrificed for the other.
-
correlation tables only correlate directional movementā¦ Pairs that arenāt correlated by the correlation standards are still correlated to each other in the sense that market reversals and market sentiment are market generalā¦ A shift in market sentiment effects all pairs at the same time, and will throw a wrench in the respective setups, causing it to swerve from the initial prediction of the setupā¦ Even though they may not go in the same direction, all of the pairs paths are altered simultaneously.
The fact is the US economy affects all forex pairsā¦ Even ones that donāt have USD in the pairā¦ So there is commonality in the forex market that makes non-correlation impossible.
If you have a valid and consistent entry and exit methodā¦ There is no need for a āportfolioā ā¦ If you donāt have a valid and consistent entry and exit method then a āportfolioā wonāt help you anyways.