Indicators development for T101's not very simple method

It definitely should work Empori, schmerex had similar problem, but he's figured in the end how to do that. Use your mouse like 4-5 times to select the label. The EA just might be busy at that time, over several clicks you should be able to select the label. When it's selected - it' movable.

Here is how to achieve 99% of modeling quality for backtesting if somebody's interested. Painful, but possible

http://www.tradingsystemforex.com/ge...g-quality.html

Hey, Don when you have a chance.... on what indicators and what rules your future EA would be based on? Did you mean just classic T101 rules and using your backtest indicator for automated backtesting?

Thanks

Orest,

Since Aplari data is not available, and someone would have to buy data otherwise, I downloaded all the data from MetaQuotes and made it available, so that everyone has the same starttime if they are using non-IBFX pairs and MasterIA_Mode as true in my upcoming indicators. All user using IBFX pairs will have the same startime if they use MasterIA_Mode.

As far as backtest, my plan is:
To start at the starttime for the basket in MasterIA_Mode, and make rules so that when the indicator is placed on the chart, the indicator prints at what times in the past met the following conditions, as per my understanding of T101's "simple" rules, as stated in the PDF T101 made easy, as well as individual posts made my trader101, below in my words:
1. BuyCondition1, which is also only CloseSellCondition, is when All Sells are at top and all Buys are at bottom. If this happens, BuyCondition1 will be set to True, and BuyCondition2, SellCondition1 and SellCondition2 will be set to False. Also all Sells will be closed.
2. If BuyCondition1 is True, and if a Buy pair reaches top position or Sell pair reaches bottom position, BuyCondition2 becomes True and a BuyAll happens.
3. SellCondition1, which is also only CloseBuyCondition, is when All Buys are at top and all Sells are at bottom. If this happens, SellCondition1 will be set to True, and SellCondition2, BuyCondition1 and BuyCondition2 will be set to False.
4. If SellCondition1 is True, and if a Sell pair reaches top position or Buy pair reaches bottom position, SellCondition2 becomes True and a SellAll happens.

This is my understanding of the simple rules, disregarding the nuances that showed up in the thread mostly. But I need to rest before I code more. Was up til 3 downloading history data. Since we are not really backtesting in the MT4 sense, but doing it in another way, since it is a basket, if the code is right, then at most we will be off not even by spread amount (since I count spread, just now's spread vs first bar's spread, for each pair).

Edit: When I talk about how far we should be off, what I mean is that for EVERY bar, we look at the open time of the first bar (starttime), for each pair, and the last bar (most recent bar, aka endtime), for each pair. Each bar uses 2 prices only from each pair to calculate the gains. I'm not saying there aren't going to be gaps, I'm just saying what is most important are the two prices for each pair, because that is what NOW (at in point in time) is made of. This doesn't have to be perfect. Afterall, we are using MasterIA_Mode to get an overall view of the market over time, to know where we are now. Then with more recent IA, can make more specific starttime closer to now to make IA to make entry and exit trading decisions by. I want to know if there is a pattern somehow where the conditions are actually met that trader101 talked about, which I've described above, as far as my understanding. If I can get those occurences, I can print what date and time it was, and what every important variable's value was at that time (e.g. BuysValue,SellsValue,TotalForBuysAndSells, TotalSpreadCost, TotalBuysSpreadcost, TotalSellsSpreadcost, ValueIfBuyingAll, ValueIfSellingAll), to calculate what would have happened if we were buying then or selling then. More manual than backtest in a way, but not as manual in other ways. Most would be done by prints being made by indicator into experts tab in terminal when conditions were met in past, but showing up and printed in an instance without having to actually having to do a "backtest". But one would be able to take that data, go back in time with that enddate (manual backtesting mode), and see what everything was at the time by objects drawn on chart.

Regards,

Don

Don,
is it on weekly time frame (rather starting every week) and resetting itself on the start of the next week? Yes, would be interesting to see the results. Take a rest.

The Sentinal indicator also follow Fibs, etc. from what I've been able to show in the past. But again, have conflicting rules, but can plainly see on chart as you have that price obeys that area, whether looking at one pair or multiple pairs gains together. If you look at the basket as a BuyAll, it doesn't matter what pairs you Buy. You don't have to Buy all of them or Sell all of them. The indicator should show this, but unfortunately I'm having trouble making a simple user input doing that, where choosing UseIBFX pairs makes it one simple boolean.

Note:
Sentinal/Sentinel indicator is also seen by me as "make-your-own" USDIndex, except you put what pairs you trade by putting a 1.0 if you are trading a pair, or 0.0 if not trading it, in user inputs.

That indicator, like my BuyVsSells indicator, is based on buying and selling the way I have described doing it in code - buying is close-open, selling is open-close. Buying USD would involve selling EURUSD, GBPUSD, AUDUSD, and maybe NZDUSD if you wanted, and buying USDJPY, USDCHF, and USDCAD all simultaneously. If buy line goes up, doing this would be profitable. If line goes down, doing the opposite would be profitable. Just like we have in this situation. Also one starttime, and the indicator shows now at the end and everything in between start and now(end price/time, added together to make total gain, except spread must be taken into account - where depends on if open buying of buy pairs: yes, open selling of sell pairs: no, open buying of sell pairs: yes, opening sell of buy pairs: no, close buying of buy pairs: no, close selling of sell pairs: yes; close buying of sell pairs: no; close sell of buy pairs: yes). BTW, the buysvssells indicator has to have all this done to get the lines made to be correct.

I posted a version of Sentinal/Sentinel that I corrected of Kevin's last version at the end of that thread, and comparing USD to this basket trade (which has been described basically as trading EUR against JPY, which EURJPY looking close to our T101 index seems to show is correct observation, as well as me crossing out the basket currencies to see how Trader101 came up with this, and it shows that the basket is balanced well, but I observed this really would only be true if you trade the same "percentage amount" of each pair, instead of the same number of lots for each pair, and I made a money management equation that automates choosing number of lots to trade that can do this if you want it implemented.

I know this is a lot of information, but let's say I've been into baskets for a long time. The combined pairs are much smoother than individual pairs, so easier to trade. And they say one should diversify. Can't do better diversifying in forex than this method IMHO. I track equity because that's the bottom line: Am I making money or losing it? My BuyAll or BuysVsSells and Sentinel indis tells me that, depending on if I am trading that strategy or not, and they're great for comparison. Taking USD into account seems to always be a good idea, you know? Because if you think about it, the currency market is basically made like the Sentinel, where USD is against everything else. When USD goes down, everything else goes up, and when USD goes up, everything else goes down. See what I'm saying?

Regards,

Don

Orest,

I put the code into init() for now for start (since the starttime I usually don't want to change), but after a couple of ya'll have said something about it, there are multiple ways to do it.

Now as I'm thinking about the ways:
1. We could have StartWeeksBack, which always starts at the beginning of the week, but it would be put in start(), since I would want it to start over at the beginning of the new week seeing that the week start time has changed and using the first bar of the week (MT4 bar using start time, -1).
2. We could have StartWeeksBack, which I put in, which always starts at the beginning of the week when you put the indicator on chart, but it would be put in init(), since I would not want it to start over at the beginning of the new week, but keep rolling on from original start time, using the first bar of the week (MT4 bar using start time, -1).
3. Have user inputs for Starttime and endtime, so you can look at a snapshot in time. If the starttime is not the beginning of the week, or the endtime is the last bar of the week, numbers could be wrong/off by one bar.
4. Use MasterIA_Mode that would be put into init(), and would use the pair with the least history, would use it's first bar, get the time, use that as basket start time, and then use that time to get the bar on each pair with that open time - 1 bar, to get correct bar for each pair, so all will have that same start time and correct price that goes with that open time. Look in history center to check correctness and comments on chart. If any of the previous three methods had start times that are earlier than this method's start time, this method's start time would be used, and the user would be alerted that we must use a later start time because of not enough history to reliably show accurate backtest.

For the middle of the code, I can't say I know how to do it although I think I have actually done it before, but it would have where it checks for last bar and next bar along the path during i--, and if the i bar is missing for a pair, the indicator would fill in an average of last bar and next bar (if next bar is not missing) for the missing bar in history. This would fill in gaps as best as possible for gaps in history, and also would smooth out equity/profit lines, too. Edit: I remember how I did it kind of: I took the gap, and all bars that would be in between, I filled with "gap in price between two bars", divided by number of bars between the two times of those bars), and price of last bar before gap starts + 1/n of total of gap (in price).

For ending code, it depends on how I close. If automatic backtesting, I use the startbar/first bar's close, and then progress one bar for each tick that comes in, updating all objects on chart. This is my simulation of a visual backtest. Pause by moving the blue play arrow below the red stop square, play by moving red stop square above blue play arrow. If manual mode is chosen, user double-clicks the vertical green (start) line and moves it to the right, realizing they have actually double-clicked and moved the red (end) line, and when a new tick comes in, the indicator will update objects and values on chart by using start and end times and prices, always including spread. If the market is not open, or the user is impatient, they may manually Refresh chart by right-clicking on chart and choosing Refresh. If using MasterIA_mode, I have thought that both automatic and manual may need to be set to false, but haven't thought it through yet. I have thought about the MasterIA_Mode as being used for looking through all pairs from a starttime, and printing when conditions were met between starttime and endtime, so that printed data could be used for analysis.

I figure multiple IA, if started a long time ago, although their numbers wouldn't match up over time (because of them having different starttime), would have the same time in which they met these conditions, and this would show the move of the market, times of peaks and valleys for the pairs, if indeed trader101 was correct. The indicators I've fixed show this to be the case from the little I've seen. In last 9-10 years (I removed CADJPY since it had bars only back to 2007, while the rest of the non-IBFX basket had bars back to 1999), the currency basket's profit has fluctuated in a range of thousands of pips. But further research is needed. But indicators must be fixed first (actually the only one that may be wrong is one I "fixed" that orest posted, and another one I've done but haven't posted that zznbrm's used orest's indicator to make, which trader101 said zznbrm's indicator was the closest indicator that had been made that mimics his excel document he used to track pair movements.

As anyone can see, that's a lot, even before making sorting pairs from most to least profit, and making objects on chart. A little overwhelmed, but trying to bite off each piece and chew it before I move on to implementing another piece. Rome wasn't built in a day. I may have posted my latest indicators in a day, but they took time to make.

My latest indicators may be right for all I know, I can't remember but I think they are since I used long code and "spelled it out", like using Pair1,Pair2,Pair3,...Pair13,Pair14, spelling out same code for all pairs instead of for(int j=0;j<14;j++), and then simply using Pair[j] in one line instead of 14 lines where only the pair has changed, and I'm trying to understand sorting using for loops, and if I can get that down, then I should be close to wrapping it all up and implementing and double-checking math. After all, we have the indicators, we just must make sure they are right, but want I guess all the options above). Whew!

The indicators I posted of mine should work fine, just not have the MasterIA_Mode yet, and should work fine with StartWeeksBack set as long as you have enough history (because history checking using iBars is not there yet). If the comments on backtest when using the StartWeeksBack shows the same start times for all pairs, and the open price for a pair in comments matches history center's open for bar with that time, then you should be good to go. If not, download the history (through MT4 or save hours and use my downloaded history, it all comes from the same place and is same history - MetaQuotes), and recalculate al timeframes (you'll have to do this step no matter what).

As rules are made by testers of the new indicators, etc., rules can be implemented in versions of the EA, right? I haven't seen the original rules work for me, so I've got to either prove or disprove the theory first, before I am going to be able to move forward. And these indicators are the key to doing that. If I prove the theory to myself, then in doing so I also provide timing of trends, see where I'm at in the bigger trend, etc. That's why you getting your EA right on opening and closing all these pairs would be great code to implement in my EA or my stuff I've shared you can implement in yours. If you understand what I've written, I'm sure you could make the code faster than I can, using my existing code I've given here, zznbrm's indicator, and Sentinel indicator, shortening it or changing it slightly, to get exactly what we all need. If I make the EA, I will probably implement code from 'CloseEverything' script to make sure everything is done fast and it's got checking in it to make sure. I'd like reliability, and have used OrderReliable but haven't gotten it to work, where if code from CloseEveything could be used for opening and closing positions, it would be fast and have better reliability that orders are put through or closed, and done fast.

Regards,

Don

Why do I want to watch the pairs I don't trade?
Because thats the way this method works. When the basket gives me the signal I'm waiting for, I trade it under it's abbreviated version.
With the 7Buys and 7Sells basket, you get:
3USD Buy, 3USD Sell; 2EUR Buy, 2EUR Sell... and so on, each 1 buy balance a 1 sell.
When you buy or sell the whole basket, you get on one side: 2GBP, 2AUD, 2NZD, 4EUR / on the other side: 2USD, 2CHF, 6 JPY.
I tried a couple of combinations and found that this one match perfectly the behaviour of the whole basket: EURJPY, GBPJPY, AUDJPY, EURCHF, NZDUSD.
I tested it with the BuySell14wBasket indi you can perfectly superpose the 2 images. I guess Don would have the same result with his buy vs sell...

5 pairs are way easier to handle than 14 !
Anyway if I was to trade the whole basket I would still stay away from USDCHF and GBPCHF which, in this basket are renegade pairs and consistently move against their group.
It's a liability to your trades.
The more I think to this EA the more I believe that it should be distinct from the "dashboard".
Now that your dashboard has reached a kind of perfection, a slick EA with all the options might be easier to handle and modify as needed? What do you think?
Schmurex


Hi guys, Have been out of T101 for awhile, but you have rekindled my interest.

I kind of disagree, but I think on the right track.

If you list all the pairs and think of them as BUY them all, and look at what is weak and strong, I think you will see that you are actually trading the European Currencies (EUR,CHF, and GBP) Stronger than the YEN, and USD

IF you list all the pairs and think of them as SELLS, you see you are trading the JPY and USD to be STRONGER than the European Currencies EUR, CHF, and GBP.

So what you are really trading, buy or sell, no matter what is actually based on the Strength of the YEN. This is because the Yen is never the reference currency of any pair.

So I think throwing out the EJ would be a mistake. IN fact, the entire basket FOLLOWS the EJ and the GJ since you are basically trading European against Yen.

I agree with with your 5 pairs, shmurex, but I would replace the EURCHF with the GU. This is because the EURCHF is all european.

I believe that the EJ and the GJ are the leaders. I made some money with the original Orest, then quit it because I was uncomfortable with the fact that a more solid method was needed for stoploss and profit target.

Here are the rules I came up with and maybe you might see if you think it has merit, and refine it:

Watch the GJ EJ AJ GU NU. When 4 of the 5 have jumped, up, with the GJ or EJ as anchor, BUY basket.

SELL basket is opposite.

I had a pretty good success rate with this.

Alternate... method, but maybe more risky, is to BUY Basket when the 3 YEN pairs have jumped with GJ or EJ in anchor. (Only did this one a couple of times).

Another way is to trade the yen basket only. Do as above, but only open the GJ EJ AJ UJ CJ and NJ are optional.

Anyway, that's my observatons, hope it makes sense.

Shakey

Don, sounds like a plan. You do realize you can't test multi-currency EA in MT4, don't you? Just another annoying limitation.

Hey Shakey, glad to see you here. I agree with many of your points.
Unfortunately we've seen so many variations of the system and tools - I'm frankly tired of all of them. That's why I quit trading this method some time ago, it was too much and most importantly I had no consistent results (it's not like I have them now ). So, let's see... I kind a agree with some of seasoned trader, who say:
- trading must be boring and simple.
- if you are overloaded with tools (indis, ea, etc) you most likely doing something wrong
- if you are not following the system you most likely doing something wrong

I'll try something different this week. Very simple. I'll try to define the basket direction for the week (note: not predict, but just observe where the price is moving to) by watching Weekly column. I'll give it a day (Monday) to tell me that info. My main criteria would be gained pips - must be more than 200 pips combined and pairs positions - must be clearly visible trend direction. Tuesday morning (GMT) I'll most likely make just one basket trade for a week, let's how it goes.

the only tool i ever really liked for this method were the T101dashboard or the profitgraphs. that's all you really need. but yes, this can get way too complicated IF you start adding indicators other than those. these are just observations. if someone want to fix zznbrm's dashboard, i'll tell you what to fix and what to put into it. it is not right on the math so it's giving false signals.

Orest,

You don't understand my backtest indicator. I know you can't backtest in MT4 a multi-currency strategy. That's why I built the indicator.

What is a backtest?

A backtest says what the probability that your strategy will be profitable, given historical data.

If I put in Buy and Sell Conditions (which I left off the most recent indicator, the one I posted here), and use i-- (meaning the indicator is going from left to right - like time), and tell the indicator to print pertinent information when a condition was met, then wouldn't I have the information I needed at the first tick when the indicator is first loaded?

(note: do you think with that information I could subtract basket total at one point in time from basket total at another point in time, to see if buying when buy conditions were all met, closing the buy when that condition is met would have produced a profit or loss?)

Plus I have manual and automatic visual backtesting. The user can move the vertical green and red lines in manual mode, refresh the chart, and see what the pertinent values and order of pairs visually at any endtime point in time given that the vertical green line is starttime, red line is endtime, and refreshing chart updates the obects and values to show the gain or loss from start to end. Automatic just does this automatically, stepping forward in time on any chart, with one starttime, but endtime moves forward one bar with each incoming tick.

What else does backtesting in terminal give you that you can't get from this (even if this needs changes, and takes coding to input the conditions to print when those conditions are met)? And like you said, you can't backtest in MT4 a multi-pair strategy. Does the backtest indicator I described above not do what a backtest would do? It could be improved to be closer to the exact information a backtest spits out at the end, except being printed to experts tab or to a file via MQL4.

Regards,

Don

[quote=Orest;3463101]Hey Shakey, glad to see you here. I agree with many of your points.
Unfortunately we've seen so many variations of the system and tools - I'm frankly tired of all of them. That's why I quit trading this method some time ago, it was too much and most importantly I had no consistent results (it's not like I have them now ). So, let's see... I kind a agree with some of seasoned trader, who say:
- trading must be boring and simple.
- if you are overloaded with tools (indis, ea, etc) you most likely doing something wrong
- if you are not following . . .

Thank you Orest, Pretty much my sentiments as well.

Love the concept, but found it un-nerving working without more concrete Stops and Targets. Pretty much why I moved on. That fibo indy looks interesting, so might give it a look, along with the dashboard, for targets, Stops.

Your dashboard made me a few bucks, and I greatly appreciate your work.

Shakey

The indicators I posted here are correct:
http://www.forexfactory.com/showpost...postcount=1041

I did forget to change the default Lots in code for the backtest indicator to 0.1 instead of 0.3. But of it is changed to 0.1 like the BuysVsSells indicator, it reads the same values as the BuysVsSells indicator. Both indicators use the long math, and although I used i-- where it's recommended to use i++, since I do the calculations like I do, I get the same result using either.

The one that Orest got from finimej and changed, and then I changed and posted, the changes I made were correct, and the indicator works correctly. I posted it here:
http://www.forexfactory.com/showpost...postcount=1044

So no indicators I've posted on this thread have been wrong.

One indicator I changed for hazelj80 but didn't post, that one is not correct, but I have trouble reading the other code, so I don't know if I'll be able to change it or not. So I am posting it now because I know the first column is the same as my other two indicators I posted, and since I added spread and currency value correctly to this indicator, I will post it. I think hazelj80 said the indicator I'm posting now has problems in individual currency calculations, but I don't know, since I'm trying to understand the rest of the code (I looked for and changed the areas I was asked to at the time, which was be able to not use ibfx pairs, add spread, and add currency value). So the first column, numbers 1-14 I know are correct, except their rounded to whole dollars and don't have cents since I'm using currency value I'd like to have cents too (unless it's too much trouble for the coder who picks this up and changes it for hazelj80 if I'm not able to further).

Note: Because I thought I had posted something wrong, and the reality was I found something, and was too tired to realize it was not wrong, but thought it was. And then trying to add MasterIA_Mode to multiple indicators had me a little overwhelmed, which all combined showed me I needed rest. I came back a little more rested, and found the only problems I have are in indicators I have not posted yet because I am trying to understand some styles of coding better and I don't want to post something until I know that it works, or at least what I've done works, in the case of the indicator on this post.

Regards,

Don

Orest,

The indicator I posted in my last post, if first column was last column, other columns were going from left to right like time does, and the first column showed the initial position of the pairs (on open of starttime's bar, minus spread since we do minus spread as soon as we put on the trade), then it would be a logical progression, just like time, from left to right, from time we opened basket to now. I just wanted you to understand what I was saying there.

As it stands, like I said, that was just a preference. What is in the last indicator I posted in my last post, is basically the same thing from right to left instead and not showing the initial position (buys at top, sells at bottom).

If you put it on a 30 minute chart right now, using 9 or 10 columns, you can see the first 30 minutes of the market (per FXDD on my screen, so not using ibfx pairs), the sells had overtaken the top (where initial position is buys at top).

But if using my BuysVsSells indicator you'd see that after the first hour, the Sells profit started falling, and if one would have bought all then, they'd be in profit right now. Not saying there was any rule to say to do so. I'm just saying, the BuysVsSells indicator shows clearly that if you were selling, your profit started dropping (or if not in profit, your loss would be increasing). But if buying, it would be opposite.

Regards,

Don

If you had the last indicator I posted put on M30 chart like I said, you would see the backtest in one way: The position and values for the last so many 30 minute bars.

Here is an example of using backtest indicator in manual mode to show the same thing, just a different viewpoint.

(DoBackTestManual user input set to true, using Lots set to 0.1 pips)
(Note: I have pictures numbered just like numbers below):
1. This is the state of indicator on chart after choosing user inputs I just said and clicking OK. The green line and red line are on top of each other. The green line uses the start of the bar. The red line uses the close of the bar. This is why we can see that sells took control immediately, because we know the default position is buys on top. At end of first 30 minutes, we can see the total P/L if selling all (if you would have sold the second the market opened), was 87.77 (that's profit, since spread has been paid on the close of the 1st half half hour, if we closed the position at that time).
2. Shows after I double-click what looks like green line.
3. Shows after I move it one bar to the right and manually refresh. I then see I have really moved the red line, or endtime/endbar.
4. This is after I move it another bar and manually refresh. Make sure correct bar is chosen after you update.
Repeat step 4 until you get to last (current) bar.

Now as long as you have enough history so that all pair's Open Bar times are the same, and close bar times are the same (use comments to see this), then you can place green line anywhere and red line anywhere, and manually refresh to see position of pairs, and values from start to end (green to red).

Regards,

Don

Actually as I look at it now, the indicator I last uploaded, it's columns are wrong it looks like. The pictures I just uploaded, when compared to the T101 Dashboard (v1.1 Don) I have on my chart right now, I see that where mine says 1:30 (meaning bar opened at 1:30, picture 5), dashboard shows 2:00. Where mine says 2:00 (picture 6), dashboard shows 2:30, etc.

I can't say I know how to fix the dashboard. I fixed the areas I know pretty well. If anyone can fix the rest, it would be appreciated.

Edit: After thinking about it closer, this would make sense what I'm seeing. The vertical line in MT4 shows the open time of the bar. But gains listed would be difference between open bar's open price and close bar's CLOSE price, so using the closetime of the bar would be appropriate in a column setting with multiple columns, so I don't actually see this as a problem, since the gains were what is listed at the CLOSE of the bar (where I have no control over time MT4 prints at bottom of vertical line).

Thanks,

Don

thank you, orest.

Re. lock profit functionality:
right, the best thing is to have it in pips. that automatically scales profit in money terms with the size of trade.

Re. the few variables I asked about:

I looked for them but could not find. please point me to the page no. / post #

thanks again.

hope everybody did a buy all around 07:30 GMT . you should be up around 600-650 pips. no more trade from what I see as far as a basket trade goes.

I did. I'm going to hold this buy for a week actually. I'm testing my new strategy that as soon as weekly trend is established it's going pretty much in that direction till Friday. Let's see...

That may work out but i have seen it do a complete flip flop. but the profitlock EA would have had you out anyways and that was the 1st week of February when i saw that happen. that week was good for about oh around 3000 pips probably. I'm going to wait till the buyall line pulls back to +100 before i enter anything else. then see if it rises slightly again and check the position of the pairs .