Megamuel's Money Making Journal

[tiger2008]

this is an interesting thread.

I have been trying to learn trade for a while and have experimented with numours strategies. still can't find the one and i don't think it exists. I guess, it is what you get used too. I am still using demo account, currently just using 5ema crossing 13ema as buying or selling signal, this has provided a better result than any other thing i have tried, but there are false signals. i would love to know if there was an indicator that can spot these signals.

T

 

[jonboy123]

this is an interesting thread.

I have been trying to learn trade for a while and have experimented with numours strategies. still can't find the one and i don't think it exists. I guess, it is what you get used too. I am still using demo account, currently just using 5ema crossing 13ema as buying or selling signal, this has provided a better result than any other thing i have tried, but there are false signals. i would love to know if there was an indicator that can spot these signals.

T

Alexander Elder says to not use MA below 8....as its a trend follower.... low MA wont give reliable trend indicators. ... and more prone to whipsaws/false noise....

But if thats working, stick with it..... ill have a test later.

The best indicator to spot those signals? ...Your Judgement....

 

[tenbobtrader]

What is the market

Hello all,

So, I've decided to start a journal for 2 reasons:

1. For me to record my progress, and,
2. To post the many questions I have along the way!

Some of you may have seen me posting here and there but you probably have no idea who I am. Here is some background:

My name is Sam, I am 24 and I live in Chester, UK. I have been interested in trading now since July last year. I started out thinking I'd make easy money buying shares, waiting a few months and then selling them! I soon realised it wasn't that easy, and after the commission it would take me years to even break even on the small number of shares I was buying with my tiny amount of savings! Then came along this bear market and now recession to render those shares close to worthless! I like to think I've come a long way since then... I've read a few books, spend a lot of time on here reading and posting, and I have also taken up spreadbetting.

Over the last few months I've been testing out various strategies/ideas/techniques but mainly gambling as I have no 'Edge' as yet! As can be imagined I've had very limited success so far, although I'm pleased to say that I haven't lost anything yet - I'm actually up on what I started with which at least shows I have some concept of risk management I think! I've read lots about newbie traders blowing their accounts!

Anyway, I'm ready to take my trading to the next level now. I know I still have so much to learn but I am ready, and I am determined to succeed in trading. My reason for wanting this so much is quite a strong one. Since the age of 16 I have suffered with panic attacks. They caused me to drop out of uni and have been the cause of many problems in my social life and also when trying to pursue a career. I now find myself at 24 working behind the bar of my local pub, with no confidence and a fear of pushing myself into anything new. I am reasonably intelligent and I am physically and intellectually capable of much more, I just really struggle with my anxiety and panic attacks. I really believe that trading is for me and if I can succeed (when I succeed) as a trader it will help me massively with my confidence in all aspects of my life. Anyway, enough of the sad stuff!

Where am I now? I've saved £2000 capital - Don't laugh! It took me ages to save that as I have a very crap job! I've tried a number of strategies such as Elder's triple screen strategy, trading in line with companies releasing their results, and various day trading strategies using many combinations of useless indicators, however, I've not found anything I am comfortable with yet!

I hear a lot on here about there being thousands of profitable trading strategies out there as long as you stick to the rules... Well, where are those strategies??? I've not found one yet! Maybe you guys can point me in the right direction? Any suggestions are welcome!

I'm thinking about trying Mr. Charts' simple trading strategy and I have contemplated buying MrSpreadBettings system, but I am not willing to spend any more money on trading yet as I have been neglecting my girlfriend - LOL! I just want to go for it and start trading something. I don't mind loosing money, I think just going for it is the best way to learn. You can read a million books but it won't make you a good trader!

My goal for this year is to become a profitable trader and I would like to earn more than I do in my pub job (which isn't much) so that I can leave it and be a full time trader. Ok its a pretty tough goal but I think you have to aim high right?

So there it is:

Me,
£2000,
My Computer,
A whole lot of determination, and,
No strategy (Yet!).

Hope you will all enjoy following my progress!

Sam.

Sam

keep your 2K in your pocket and your A whole lot of determination in tact !

decide what the market is 1st

then choose a strategy

http://www.cisco-futures.com/trading_model_development.html

"A balanced Market begins with the end of a trend.

Each succeeding day adds to the balance until a new trend begins.

The trend beginning is observed as a breakout from the balanced distribution.

A single day session may or may not be a valid locator of value. Value determined from a three day Overlay is far superior.

The trader must be prepared for either a trending or congesting market.

The market generated data gives the support price, the resistance price, and breakout points.

Markets tend to run and pause in all time frames.

Starting from a balance, a market will test the balance limit, then breakout starting a trend.

Then the trend slows, testing the propensity to continue the movement.

Finally, the next balance forms. And so it goes as long as the market exists.

A day trader is very interested in the run-pause picture within the day, There can be several run and pause phases within a day.

Identify key reference points for trading action.

The day activity is taking place within the context of the longer term [multi-day] auction"


take it slow if your in a rush

 

[FX4Newbies]

Hello all,

I am new to this forum and was reading some of the posts. Not much activity for a while though. Lots of good ideas and advice was shared, but there are some basics to trade successfully. First off, it does not start with knowing lots of fancy indicators and systems, though they come in handy sometimes. You must be able to look at a naked chart and see what price has been doing...where it's coming from and where it's going. Then the trader applies whatever else to that foundation.

I have been posting some charts in a thread I called Goldmine! Take a look and tell me what you think. I think I can help you make some solid progress Sam if you are still looking.

All the best.

 

[megamuel]

So, 2 years and 8 months after starting this thread I have decided to fire it up again as I believe I am now getting very close to 'making it' as a trader. I have come a long way in the last 3 years. Tried a lot and dismissed a lot. Gone from trying to find a discretionary method to systems and automation. I have had many ideas coded up. Most were terrible! This year I have mainly been working on Pocrel's pin bar and pivot point strategy ( http://www.trade2win.com/boards/trading-journals/120484-pin-bar-system.html ) I had numerous EA's coded up based on his setup and I tested it extensively. At first I thought I was getting somewhere when I had lots of profitable back tests and optimizations. However, it wasn't until I had a chat with Travis ( http://www.trade2win.com/boards/trading-journals/85510-my-journal-2-a.html ) that I realised I was testing it all wrong and what I had really done was to curve fit the settings to the sample data. Once I started testing using in and out samples I discovered that the setup doesn't have an edge at all. The edge was purely in Pocrel's skill as a trader. I was pretty disheartened and close, very close to giving up trading.

I decided to have one more go at designing a system and I had a very basic idea coded up. Very basic, but it just made/makes sense. Possibly it was one of those light bulb moments that are often spoke of! Straight away I was getting better results in back tests and it is looking very promising so far.

As I have a lot going on with testing and such, instead of filling out notepads and scraps of paper everywhere, I have decided to start posting again on this journal to record my thoughts and progress and to welcome any comments and suggestions people may have. I also hope to help anyone who may have any questions about system development, optimization and back testing as I have a fair bit of experience now.

Currently I am testing the system on EUR/USD and I will post some results up soon. After that I will also test it on 6 other pairs (USD/CHF, USD/JPY, GBP/USD, AUD/USD, EUR/GBP, and EUR/CHF. All of which is going to take a fair bit of time but that's fine. I need to wait for some of my longer term investments to (hopefully!) bounce back anyway so that I have some capital to trade with once I am ready.

Hope that some of the guys that supported me when I first started this journal are still around!

Updates coming soon.

Sam.

 

[forker]

Good luck.

 

[pocrel]

Best of luck Sam I shall be keeping my eye on you, I am sure you will be a great success

 

[VendorExterminator]

Best of luck Sam I shall be keeping my eye on you, I am sure you will be a great success

So you're a vendor as well then Porcel....?? Your journal, as useful or not as it was, had the distinct air of a "set-up". Did anyone else feel that?

 

[Jason101]

Good luck Sam. I hope all that hard work pays off.

 

[wackypete2]

Good luck and don't give up!

Peter

 

[pocrel]

So you're a vendor as well then Porcel....?? Your journal, as useful or not as it was, had the distinct air of a "set-up". Did anyone else feel that?

Not sure what thats got to do with Sam journal if you want to have a go at me please use my journal and will be happy to answer any questions on my live trade calls (which i have seem to have faked aswell according to you in my elaborate set-up)

 

[megamuel]

Funny how things go sometimes isn't it... So only the other day I kick start this journal after 2 years and 8 months and say that I believe I am close to 'Making it'... Now I've had a big set back (again) and I am back to square one!!! I've been pretty beaten up about it and had a break from 'trading' for the last few days. Oh well, it's not the first set back and it won't be the last so I guess I just have to take it and carry on. Anyway, I guess I will document it here and welcome opinions on what to do next.

So basically I have been testing my latest system. I have been using an in sample and an out sample to reduce the possibility of curve fitting. For anyone not familiar with in/out samples, the general idea is that, rather than optimise your system over your whole set of data, you split the data set into two. You then optimise on the first sample (in sample), then when you have profitable settings, you then test them on the out sample. You will find that most combinations will fail on the out sample because they have been curve to the in sample. What you want is something that works on both the in sample and the out sample.

I did my testing on 10 years of data. Optimised over 2001-2006 and back tested over 2006-2011. So from about 10,000 optimisation results, I then back tested everything on the out sample that had over £2000 profit (trading 0.1 lots), lower than 15% drawdown and over 100 trades. Probably around 300 different combinations. Out of the 300 I was left with 35 candidates that did well in both in and out sample. Hey presto I though, here are my systems to start trading with. I just had to backtest them over the whole data set and pick the best performers to begin forward testing with. So I began backtesting but I decided to include a bit of extra data I had, so from Jan 1999 - Sept 2011. Well, this is when I went from being jubilant to being on the verge of tears. In every one of the 35 backtests, the systems performed badly in the 1999-2001 period. In short, I had picked systems that had also been curve fit to the out sample. Here is an image to display what I mean:

As you can see the system performs well in the optimisation/in sample (Red). It performs well in the out sample (Green). But in the extra bit of data that I included (Blue) it performs badly. This was the same for all 34 other settings. They are all curve fit.

Quite what I do now I am not sure... Carry on developing this method/add new filters etc. Or scrap it and make a return to my well used drawing board?! Once again I am feeling a bit lost. Thanks for the messages of support. Pete, your "Don't give up" message came EXACTLY at the right time. Thanks.

Any suggestions are most welcome. Thanks for reading.

Sam.

 

[piggybridges]

Hi,
out of interest, When you said you were now getting very close to making it as a trader, were you just going from the results of the various backtests and stuff, or on actual results from your live trading over a decent amount of time?
Cheers.

 

[Adamus]

Hi Sam,
since you asked, I think I've been in this place in system development too. I'd say that you are either trying to hard with optimisation - firstly, it's best not to bother optimising anything to do with entries, that quickly causes curve-fitting, and secondly, what are the optimisation results like as a whole? do you get nice result vs optimised parameter values?

Or you are trying to make a system trade all market types when really the system's only suited to certain types - trends, ranging, volatile, non-volatile etc.

You'd do well having a look at a whole load of the system's trades. Were the entries any good? What was the market doing during the bad drawdowns? And during the good times? Were exits ruining the results? Were you trying to trail a stop when the market was choppy? Or did you use profit targets when the market was offering you big trends? Etc etc.

Sorry it's a bit rushed, bit pushed for time right now.

 

[megamuel]

Hi,
out of interest, When you said you were now getting very close to making it as a trader, were you just going from the results of the various backtests and stuff, or on actual results from your live trading over a decent amount of time?
Cheers.

Hi Piggy,

Yes I was saying it just based on the results of my testing. At the time I believed I had found something that was not curve fit and had worked over the past 10 years and I was hopeful that it would continue to work in forward testing. Well now I know (by accident) that it was actually curve fit and that it won't work in forward testing.

I know that back testing is not the be all and end all if that's what you are getting at. I know it has limitations. Even more so now! So I would always forward test something on a small live account before properly funding my account.

Oh well, we live and learn. I'll get over it!

Sam.

 

[megamuel]

Hi Sam,
since you asked, I think I've been in this place in system development too. I'd say that you are either trying to hard with optimisation - firstly, it's best not to bother optimising anything to do with entries, that quickly causes curve-fitting, and secondly, what are the optimisation results like as a whole? do you get nice result vs optimised parameter values?

Or you are trying to make a system trade all market types when really the system's only suited to certain types - trends, ranging, volatile, non-volatile etc.

You'd do well having a look at a whole load of the system's trades. Were the entries any good? What was the market doing during the bad drawdowns? And during the good times? Were exits ruining the results? Were you trying to trail a stop when the market was choppy? Or did you use profit targets when the market was offering you big trends? Etc etc.

Sorry it's a bit rushed, bit pushed for time right now.

Thanks for the response Adam.

Yes I think you are right. Effectively what I did was to optimize over the 10 years, by accident. Or at least without knowing. I read this article which touches on it:

Out of Sample Testing: The Power and Correct Execution of This Simulation Technique | Mechanical Forex

The problem is that if you start “picking” results to get those that perform well in out of sample testing you are effectively introducing a strong selection bias which is equivalent to running an optimization of the full testing period.

Could you expand on this - "do you get nice result vs optimised parameter values?". I am not the smartest cookie, I am already out of my depth!

I guess you are right about looking over the trades. I tried mainly fixed stops and only toyed a little bit with trailing stops. For some reason trailing stops don't appeal to me at all but I can't quite put a finger on why. Maybe I should look into them a bit more.

I think I have been optimising too many parameters. I mean I had over 10,000 combinations. Of course there are going to be many profitable combinations in there.

Hmm... I'll keep plugging away. Better sleep now though

 

[Adamus]

First of all, don't optimise on more than one parameter. Why? Because that is the road to curve-fitting. If you run 10,000 optimisations, of course you are going to find one that is profitable! If the other 9999 are unprofitable, that tells you that your basic system isn't profitable and that your one profitable parameter combination is just a fluke and has no predicitive ability.

If you have a ton of code and lots of places where you could put different values, don't optimise them all - just check out a sensible value based on observing the trades in the different market types to see what the value's impact is. Maybe have an if-statement to use two values, one for volatile markets, one for non-volatile. Do that for all parameters. Your optimal system should have no optimisation requirements! In fact, you should also try to minimise the number of if-statements.

If it sounds insanely dogmatic, then you are still being suckered by the beauty of your optimised backtest results.

When you do decide that you must optimise a parameter, then chart the optimisation runs with the performance of the system, e.g. annual return over drawdown, on the y axis against the parameter value on the x. That's what I meant by "result vs optimised parameter value". If the system is robust and the optimisation process is not curve-fitting, then you will see a graph that has a peak value for the optimal, and nicely tailing off on either side for less optimal values. If it is not a suitable parameter to optimise, the graph will look like the seismograph of the ***ushima earthquake - one fraction wrong on the value, and you get a drastically different profit.

It makes sense to adhere to this protocol because the future is always different from the past, so the optimal value you choose now might not be optimal in the future. If you are taking your optimal value from the seismograph-like results, your system is doomed.

Hope that helps.

By the way, trends don't happen much, which is why trailing stops don't work much. Perhaps that's why you don't like them, a subconcious aversion from your testing experience. But when they do work, they really work. You can make ten times your target profit.

 

[megamuel]

First of all, don't optimise on more than one parameter. Why? Because that is the road to curve-fitting. If you run 10,000 optimisations, of course you are going to find one that is profitable! If the other 9999 are unprofitable, that tells you that your basic system isn't profitable and that your one profitable parameter combination is just a fluke and has no predicitive ability.

If you have a ton of code and lots of places where you could put different values, don't optimise them all - just check out a sensible value based on observing the trades in the different market types to see what the value's impact is. Maybe have an if-statement to use two values, one for volatile markets, one for non-volatile. Do that for all parameters. Your optimal system should have no optimisation requirements! In fact, you should also try to minimise the number of if-statements.

If it sounds insanely dogmatic, then you are still being suckered by the beauty of your optimised backtest results.

When you do decide that you must optimise a parameter, then chart the optimisation runs with the performance of the system, e.g. annual return over drawdown, on the y axis against the parameter value on the x. That's what I meant by "result vs optimised parameter value". If the system is robust and the optimisation process is not curve-fitting, then you will see a graph that has a peak value for the optimal, and nicely tailing off on either side for less optimal values. If it is not a suitable parameter to optimise, the graph will look like the seismograph of the ***ushima earthquake - one fraction wrong on the value, and you get a drastically different profit.

It makes sense to adhere to this protocol because the future is always different from the past, so the optimal value you choose now might not be optimal in the future. If you are taking your optimal value from the seismograph-like results, your system is doomed.

Hope that helps.

By the way, trends don't happen much, which is why trailing stops don't work much. Perhaps that's why you don't like them, a subconcious aversion from your testing experience. But when they do work, they really work. You can make ten times your target profit.

Yes I think I have committed a classic case of over optimisation here. The silly thing is, I bet in all of the 10,000 results that somewhere there will be a system that would actually work consistently on multiple time frames and pairs. But trying to find it would be like searching for a needle in a haystack. Guess I need to refine the idea a bit more. Or maybe scrap it. Or maybe add some filters. I think the concept is too simple. I know people say to keep it simple, but this is really simple. It's just open and close sequences of bars.

Damn, I wish I could just think of a good idea. All my ideas are rubbish. I need a good idea before I get it programmed next time. As I can't code systems myself I have to pay to have it done and I can't waste any more money on rubbish ideas. Not that it's expensive, it's just wasteful.

Thanks for the advice about back testing and optimisation. What platform do you use for testing? I don't think I can change my x and y axis in the way you described. So yes my results do look seismograph-like.

I think the things with trailing stops is that I don't like giving points back. Prefer to have fixed targets. But you are right, I think I should look into using them a bit more. But first, I need a decent idea. Back to the drawing board.

Sam.

 

[Hotch]

Sam,

Question: does the strategy "make sense"? As in, can you explain why every condition is in there bar "it makes money this way". One of my algos for example, all it does is, defines a range, defines a breakout, defines a pullback, if you breakout of a range and get a pullback enter. There's 1 variable involved (2 if you include the time frame), because that's as simple as I can get it, I could look at whether we're up or down on the day, that might even improve profitability, but as a trader I don't see why it SHOULD improve profitability, so I won't add it.

On the surface it might seem like I'm just missing out, but it means I can be far more comfortable with it. Your curve is fine, so what if you had a down year?

 

[Adamus]

Don't try to invent it all yourself! Plagiarise as many ideas as you can. Take them and adapt them. I have a list of over a hundred potential systems that other people have invented that I need to code up, check out and adapt to my nefarious purposes. We're not all Mozarts. And anyway, I bet the people with the great ideas are rubbish at implementing them.