Menu

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

Dreamliners’ MPAS (Mechanical Price Action Strategy) System 486 replies

Percentage vs. Pips per Trade vs. Pips per Day/Hour! 7 replies

MACD_Colored Indicator for 4 Hour Strategy (300+ Pips per Month) 157 replies

my MT4 have a stroke? 2 replies

Attachments: Two-Stroke System (Mechanical Price Action +300 pips per month)
Exit Attachments

Two-Stroke System (Mechanical Price Action +300 pips per month)

  • Post #81
  • Quote
  • Edited 1:48pm Aug 5, 2007 12:30pm | Edited 1:48pm
http://img507.imageshack.us/img507/3...enshot5np6.png


I still am not clear as to why you find it so difficult to determine the trade side..... in the image above as you can see there are 2 MAs 1 is a 10 period the other is a 55 period.....
price action is very clear above 10 MA when it is going with the longer term trend, when price and MA turn counter to long term MA there are still good signals but as you can see the price wants to return to the main trend by making attempts to break back through the 10 MA.....this should be a clear bias as to price direction.

and yes I realize the right side of the screen is the unknown but momentum is a universal law that gives us an edge.

also in my post #74 only differance is the MA which is a 14 period, this same thing is apparant, once the price is moving momentum is on our side, it is only during the rollover or a strongly ranging trend that we will have the most losses and this is where MM comes in and observation...ie is there still a strong slope to the MA or is it flattening out.

just tryin to help
Same Whore .... Different Dress
don't drop the ball now.......you are going in the right direction.....
Same Whore .... Different Dress
@TS i am still confused with your system........can you make simple resume please..? tq and one question : you never buy USDCAD..?
@inflection so... you ALWAYS buy : eurusd, eurjpy, audusd, audjpy, gbpusd and ALWAYS sell : usdcad are you NEVER buy : usdcad are you NEVER sell : eurusd, eurjpy,audusd, audjpy,gbpusd thank you fro youe explaination...!!!
Hi, Inflection! Just finished to read your thread, the system is still dangerous to trade live IMO, but you are doing a great job, your approach is so energetic and robust, and I think you are on the right track. I think using MAs would definitely bring more power to the system, now you have 53% of right trades, but using just 1 MA would increase it to 60-65%, only that could make a huge difference. There are some other improvements could be made.
  • Post #86
  • Quote
  • Edited 6:30pm Aug 11, 2007 4:22pm | Edited 6:30pm
Quoting rogerha
Inflection, that was a great post, and you should be proud of your ethics and demonstrating your lack of greed.

I myself have been ditching most of my indicators and looking at basic price breakout strategies, but for your information, one indicator I have found invaluable and continues to be so is the Fractal Dimension Indicator (FDI). Since this is a trend following strategy, it may benefit your cause. I do not have the programming skills in Meta Trader that I do in E-Signal, but the difference it made to the results for all of my trend following strategies was very remarkable. Its a basic trending/ranging filter which may allow you to refine your entries. There is a version coded for it for MT4, but I am not sure where to find it unfortunately.

Anyway, I don't want to crash your thread with a ton of indicators, and I hope it doesn't open the floodgates for that, but I thought I might mention it to the forum to see if any others had experience of it, or could look at the validity of using it on this system.

Good Thread, Good Luck, I shall be following this with much interest.
Thank you rogerha, I think using fractals might be even more useful, than using MAs, that's my feeling it has to be tested. And I'm pretty sure in MT4, the indicator is called "Fractals", I tried it to test some ideas and indicator seems to work very good, and for that strategy could be very useful. It definitely has to to be tried.
G'day Alex,

Quoting alexfot
I think using MAs would definitely bring more power to the system, now you have 53% of right trades, but using just 1 MA would increase it to 60-65%, only that could make a huge difference. There are some other improvements could be made.
Can you explain how you came up with these numbers? 60%? 65%? A greater % of winners does not equate to a greater % of winnings. I believe the idea is to keep things as simple as possible. In this instance, the only purpose I see for MA's is to guide us in choosing a trade direction, ala JohnDoe's example.

Cheers,
Art...
  • Post #88
  • Quote
  • Edited 11:28am Aug 12, 2007 11:15am | Edited 11:28am
Hi, Artamon! 60-65% is just my estimation, if you apply, let's say 13SMA for daily chart, it looks like about 60% days will be bullish, when price is above SMA, and vice versa with bearish, and it does not complicate the system too much, that was just my idea and it has to be tested of course.
Just instead W%R the SMA should be applied, I think W%R almost useless in this case, as I see things. And it has to be tried with different MAs, maybe 21 or 8 would work better.
Greetings all.

Inflection, i hope all is well with you and you are enjoying your family and friends. I see you've been away for a couple weeks and maybe that is a good thing to come back with a fresh look at things?

I commend you on your method of research. This thread (posts 1 to 88 above) has been a very enjoyable read for me. Your style of research and discovery has shown me additional methods to incorporate during the build process. I especialy appreciated, your candor in your post #56. I had gained much respect for you up to that point in the thread, and that particular post won me over as a listener and supporter.

Best wishes to each who has contributed so much to this thread. I will continue to give serrious thought to what you've presented.

Cheers,
Thom
Inflection
I too hope all is well and hope to see you back soon.
Same Whore .... Different Dress
Hi Friends -

My apologies for being absent so long! I have been really, really busy the last couple of months with a business plan for a new startup and was out of the country talking with potential investors. It has absolutely nothing to do with trading, but I'm really excited about the business and I hope to raise the money, because I think it's a cool idea and could be very helpful to a lot of people. It's an online language learning application, and I believe I have a good chance of securing a patent on the approach I am using.

My stock trading this year has been just terrible. Not catastrophic (thanks money management), but not paying the bills either. If "necessity is the mother of invention," then I figured I had better start inventing!

Anyway, can you tell that I'm trying really hard NOT to get a real job :-)

Besides that, I've been working on a proposal for my doctoral dissertation, still doing volunteer work, and generally trying to keep up with two little kids in the house. So, I hope you all will forgive my silence.

I will post again shortly with responses to several of the issues that have been raised in recent replies. I'll also talk about the little bit of work I have been able to do on this system in the meantime. Suffice it to say for now, I am even more optimistic that this can be a workable system.

More later.
Hiya Inflection,

boy, have you put a LOT of work into this. VERY commendable and if only everyone did.

Quoting Inflection
The reality that I have to keep in mind (coming from a history of stock trading), is that currencies are not like stocks. They don't just keep going up and up, or down and down. Currencies are a ratio. Some premium or discount from a 1:1 ratio. My 1 Euro gets me 1.3800 Dollars. Sure they fluctuate a lot, but over time they move around in really big ranges, as you indicated.

Having said that, this does not mean that there are not definable trends, sometimes very big, very long trends within these big macro ranges.

In fact, this is the part of the system I myself am most uncomfortable with. The problem with this is something no one has yet pointed out.
I was about to, then scrolled down and saw that you were doing it yourself.

Quote
Hindsight.

I chose the trade side by a simple frequency probability spanning the past time period. Simple count and divide. Such and such a pair had xx% up days and xx% down days. Therefore, I'm better off going long. Let's see what that would have done.

The problem is that I'm looking back ... from today.

But if I were in 2004 or 2001, and making the same calculation on the prior 5 years, it would look different. I haven't actually done that study, but it might be a good idea.

The simple truth is that in order for a system to have a really good chance of working in the future, it must have also worked in the past, over a long period of time, through various market changes.
Yea, but the problem is, you used the PAST TREND DIRECTION as a PREMISE for your main trade DIRECTION. You have no means of knowin when this will give signs that it's changing in the future, because your trend bias lies in the PAST. Now, we all don't know the future, but what I mean is all your assumptions and calculations rely on the sole factor that the USD is going to die off forever if you choose long EURUSD / short USDCAD.
The second both cycles peak, you're basically busted, as you won't find anymore trends to stick around into / develop, since the direction will now be opposite to your original assumption.

Hence this is more of a "what would have worked the last years, had one known the trend in advance"(or been able to recognize it in the early stages and then latch on, like jacko et al. Trendlines spring to mind) thing than something tradeable in the future UNLESS you have some way of forecasting USD direction for the next few years.

If you do, PLEASE tell me, I sure wouldn't mind knowing


But really - you're effectively suggesting trend trading, no? Surely there are many ways of doing trend following. Yours could work, too. I haven't gone into the deeper mechanics since after I read "going over xxxx years, I assume direction is" I stopped and skipped down, looking for the part where you realize that this is the crucial hook...

Now this doesn't mean it can't work. But you have to realize you're relying on the trends to sustain, and they've been going on for quite a while already now.

Just saying - be aware of what it is you're doing, why it works or not and whether you can sustain your premise into the future safely, too

But man, a LOT of work in this, a lot. Respect.

Take care,

SeekingLight
Trust price. Know yourself.
Okay, let me answer a few questions and issues that have been raised in recent replies.

Oh, but before I do that, I should first mention that I would prefer people to post replies on the forum rather than sending PMs. I get tons of email already and I just cannot reply to all the PMs. Unless what you're sending really is something "personal" please post on the thread for the benefit of everyone else. Thanks!

All right, the first thing to address are the numerous posts suggesting that we use various MAs for determining the trade side. I've already mentioned that I don't think this is a good idea for several reasons.

1. I've already tested it with short and long MAs from 30days to 200days. It does not improve results. If someone could take the spreadsheets and prove me wrong, that would be great. I don't have a lot of patience for suggestions that a person should go back and redo many days of painstaking, detailed, trade by trade analysis, because "if you just look at the chart, you can see that 60% of the time such and such and so and so." It is notoriously difficult to estimate 60% of anything by looking at a chart, especially moving averages. That's why moving averages are so deceptive, and why almost every newcomer to trading at some point believes that the simple cross of one moving average and another is the magic bullet to certain success. It's deceptive because all markets in all timeframes alternate between trending and ranging. I don't want to be narrowminded, but I am not convinced that prices above/below a moving average are a good way to determine the trade side. The reason is that by definition, prices tend to fluctuate closely around their moving averages. This is why it appears to me that most moving average cross systems require a lot of extra rules and filters to work. IMHO, there are better ways to trade with MAs. The first is contrarian, meaning that prices have rapidly extended far away from their moving averages into extremely oversold/overbought conditions, presenting a good reversal entry. The second is to use MAs to identify pullback entries. There are two excellent methods here at FF which use these approaches: Cornflower and Phillip Nel. I'll come back to this later, because one of the ways to smooth the equity curve is to diversify by using multiple systems/approaches on the same pairs, and I think that could be beneficial to this system as well.

2. The second reason I'm not in favor of using MAs to determine the trade side is that I ALREADY HAVE A METHOD for determining the trade side. I also already have a way of identifying the trend. There's no need for moving averages. The method we are using to determine the trade side is a statistical calculation of the %age of up days and down days over some recent period. I think in theory this gives you a basic idea of recent sentiment by showing quite simply how much of the time the pair traded up and down. This is a simple measure of open-close price action. This also correlates conceptually with the method we are using to identify an emerging trend - the Wm%R oscillator - which is measuring consistently higher/lower closes. It's just a way of aggregating all the bullish and bearish action over some period of time. The only challenge so far has been determining what period of time to use, e.g. 30 days, 60 days, etc.

3. Related to this, there seems to be some confusion in thinking that this system says one should ALWAYS trade on the long side or the short side of some particular pair. This is incorrect. The system will switch between long and short. The only question is how frequently this is done. For example, if we base our calculation of %age up/down days on the past year or two or three, then the system will change sides very slowly. There could be big fluctuations and retracements during that period. Alternatively, we could use a shorter timeframe, such as 30 days. A few people have suggested even shorter timeframes, such as weekly. I believe this is the motivation behind the suggestions for using short MAs. However, I will say it again, that in my opinion more frequent switching does not pay off for this approach. I'm sure that some of the difference in thinking has to do with the dominance of hyperactive trading approaches on these forums. What I mean is this: for every thread discussing "slow go" approaches involving trading on daily timeframes or even longer, there are a dozen others discussing 1hour, 30min, 15min, and even 5min approaches. My goal is to slow down. I am trying to find a simple, mechanical system that allows me to constantly participate in the market so that I can engage longer term trends as soon as they emerge. There are many successful hyperactive traders out there, but that's not for me.

4. There were some other suggestions for using different indicators, such as fractal dimensional index or other pattern based fractals. I find fractal dimension analysis to be very interesting, but from what I've read from the people who developed that indicator, it still appears to me to be something of an artform to interpret and trade. In fact, they suggest using other combined indicators for confirmation and specific entries. It would be great if someone could show me how one of these kinds of indicators performs better with this system, but I don't have access to FDI.

Okay, I'll be back later with updates.

All the best!
Let me try to explain a little bit more about the way this idea works by means of a couple of charts.

I have attached two charts of EURUSD showing a period of some major volatility around 2006, during which the recent long term bullish trend began.

The first chart is a basic candlestick chart, green candles for up days, red candles for down days. The second chart is the same period with the Wm%R paintbar study applied. Green bars are uptrend (i.e. higher closes), red bars are downtrend, yellow bars are indecisive.

A number of people have posted replies here and sent PMs to me with the strong caution that you cannot use long periods of history to project the future trend. This is entirely correct. You can't say, "it's been going up for a year and it will keep going up."

But that's not what we're doing here.

I am NOT projecting a trend. I am calculating the statistical frequency of up and down days over a recent period of time (e.g. the past 6 months, 1 year, 2 years, etc.). It's not so much an indicator of direction or trend, as it is a measure of bullish/bearish sentiment. It's really just an abstract measurement, as if we took the entire last year as a single bar and looked at what %age of that bar was constituted by upward direction and what %age by downward direction. We're just breaking that period of time into 24 hour chunks, and counting how many those were up and how many were down.

The theory is that if we trade on the side of the statistical frequency of up or down days (over some "recent" period of time), then we will have a slight edge on the probability of winning. The idea behind using a period of 6 months, or a year, or even more, is that a larger set of data makes a %age more relevant. If you only have 10 days of data and you say x% of days were up/down, it's kind of meaningless because you have so few numbers.

As we move forward, we are recalculating this measure every day, so it will change over time, and we would accordingly switch trade sides. But it changes slowly. I may not have been as rigorous in testing as I could be, from what I can see, switching sides more frequently does not pay off.

Now let's take a look at the charts and hopefully I can explain how the system can work while being agnostic to the shorter term trend.

First, look at the simple candlestick chart. What I want you to notice is that whether the trend is up or down, whether prices are above or below the various moving averages, whether MACD is rising or falling, whether the moon is waxing or waning ... there are ALWAYS up days and down days, green candles and red candles. In fact, that number is relatively close to 50%, plus or minus a few percentage points. It's the plus or minus part that matters. But what you have to realize is that even if the overall trend is down and I am trading long every day, I will still win 47% ... 48% ... 49% of the time. Of course, I will also lose, but you get the idea.

Now, look at the paintbar chart. Notice that even in the major downtrend leading to the March 2006 reversal, there were multi-month retracement periods (indicated by the green trend bars). What this means is that even if I win on the daily price action less than 50% of the time, but I am positioned to enter an emerging trend (whether a major trend or retracement), then I can still make money.

That does not mean that this particular period would have been profitable for this system (in fact, it probably would not have been). But it shows how this kind of system can flow with the larger cycles. The system is always trading on one side or the other, changing sides as the frequency of up/down days changes. Thus, it is always positioned to engage either a major trend, or a retracement. What matters from the hard right edge to the future is how long the trend will run, or how steep and extended the retracement will be, or how tight the range will close. Unfortunately, no one can really predict those things very well, and even if they could, unforeseen events can throw a real wrench in the works. As I mentioned in one of the very first posts, I believe that the solution to the paradox is participation, not prediction. Find a mechanical way to be always present in the market, booking small profits and waiting for a trend to engage.

I hope that helps clarify how I'm looking at this thing. And as always, I reserve the right to disagree with myself :-)

Your thoughts?
Attached Image(s) (click to enlarge)
Click to Enlarge Name: TREND1.JPG Size: 158 KB
Click to Enlarge Name: TREND2.JPG Size: 159 KB
  • Post #95
  • Quote
  • Edited Sep 7, 2007 5:44am Sep 6, 2007 5:40pm | Edited Sep 7, 2007 5:44am
Hi Friends -

I really need some help.

I am currently working on several aspects of testing this system:

1. backtesting over many years (from 1998) on all major pairs
2. optimizing the tradeside calculation factor
3. testing the model of portfolio management that I discussed previously
4. supercharging the system with amplitude trading (something to be discussed later)

I stink at programming. I use Tradestation, but never got the hang of their so-called "Easy"Language. I've also used Metatrader, but don't know how to develop strategies in that language either.

The only way I know how to develop these models is with spreadsheets and historical data. Unfortunately, I'm not a very good VBA programmer either.

If there is anyone here who is a rock-and-roll VBA coder or can develop Tradestation/Metatrader strategies in your sleep ... then could you please PM me? It would be really great to see if there is any faster way to do this testing. I try to be totally thorough and so I end up spending of lot of time checking my work and validating trades manually, but we're talking about tens of thousands of trades!

I'll keep working on it. I promise. But if there is anyone else who leans toward the quantitative/analytic/scientific side of trading and you find the direction this system is going to be interesting ... please please please contact me.

Thanks!

- Lucas
Is there another name for this oscillator? I have been trying to look it up, but not having muhc luck finding anything on it.
Hi, Inflection. I just discovered your thread and read through the whole thing. Thanks for your effort!
I really like your morale and your professional approach.

About your system, there is one thing that bugs me;
you say in post 6 (point 3) that you should "Exit at the CLOSE". This is impossible, since knowledge about the close is a priori, that is, something you know after it has happened. This might not make a whole lot of difference, but it's worth taking into account. In effect, you cannot trade on a bar that is finalized.

Secondly, I am able to program very basic things, so I want to try out your system mechanically in an automated setting both backtesting and also forward testing. For this, however, I need some clear "computerizable" logic. There's one thing that I would like interpreted:
Quote
EURUSD

1. Enter LONG at the open of the day with 2 positions.
2. If %R indicates an UP trend just before the close, hold 1 position.
3. If the day will close lower than the open or prev close, then exit one or both positions.
4. If the day will close higher than the open and prev close, then hold one or both positions.
5. If the prev bar was an EXIT, re-enter at the open of the next bar with 1 or 2 positions.
2 questions:
How do I know where it will close (in computer terms)?
How do I determine if I should exit one OR two positions?


@maninjapan
I think you should look for Williams %R
origato

With all respect,
Cronos
The entries look pretty straight forward, but I dont think I quite get the exits. if the %R is up then hold one, if close is lower than open then exit..
but what about if close is HIGHER than open but %R is flat?
and does the close is lower than open exit only apply to bar of entry?
And what would be an exit signal with %R? If it just comes goes to flat? or does it need to go to Down?

The answers are probably quite simple, Im jsut a little confused....
Hi,

I am a quant developer from London and got the exact same idea a couple of weeks ago. The similarity is frightening. If you are interested we can work together to polish this thing but certainly extremely promising stuff... I think we can easily smooth the equity curve and figure out a money management principle that could shoot the profitability well over 50%.

regards
bullrock
.
Quoting CronosFX
Hi, Inflection. I just discovered your thread and read through the whole thing. Thanks for your effort!
I really like your morale and your professional approach.

About your system, there is one thing that bugs me;
you say in post 6 (point 3) that you should "Exit at the CLOSE". This is impossible, since knowledge about the close is a priori, that is, something you know after it has happened. This might not make a whole lot of difference, but it's worth taking into account. In effect, you cannot trade on a bar that is finalized.

Secondly, I am able to program very basic things, so I want to try out your system mechanically in an automated setting both backtesting and also forward testing. For this, however, I need some clear "computerizable" logic. There's one thing that I would like interpreted:
2 questions:
How do I know where it will close (in computer terms)?
How do I determine if I should exit one OR two positions?


@maninjapan
I think you should look for Williams %R
origato

With all respect,
Cronos
Cronos. The close is relative. The close in fact never happens but when you say that I will start to draw the next bar from 5am GMT then your close will be at 4:59:99:99 GMT in computer terms. Because from there you will start to draw the next bar. It just a question of agreement. What is the timespan of one bar and when that time span starts. From there you can construct the chart and the close is effectively happens when you stop drawing a bar and start the next.
.
0 traders viewing now