A quantitative approach to trading Forex and Forex Derivatives | Page 6 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Quantitative and Algorithmic Trading 188 replies

Quantitative Approach to Backtesting 24 replies

My approach to forex trading 44 replies

Gill Blake's approach to forex trading 5 replies

The Shark's Trading Book - Quantitative Trading 2 replies

Attachments: A quantitative approach to trading Forex and Forex Derivatives
Exit Attachments

A quantitative approach to trading Forex and Forex Derivatives

  • #101
  • Edited 4:49am Sep 23, 2017 4:38am | Edited 4:49am
Quoting GuyBaker
{quote} In the context of your posts those responses I made are valid. I 100% agree with keeping the site clean, where we differ are the methods used. Can you honestly say that your posts here are in the spirit of keeping the site clean and honest? I don't think anyone would read your posts that way.
Well , you just can not understand what I say , or perhaps " there is none so deaf as he who will not hear "...
Ok , no problem , have a nice weekend.
PS : I believe that you and me will never get married , lol.
  • #102
  • Sep 23, 2017 8:10am Sep 23, 2017 8:10am
Quoting GuyBaker
{quote} i agree with that, as above, my suggestion is not to use them as a hedge at all, but as the main instrument traded, in VEEFX's case. Nadex being an exchange is more cost effective though, from my brief analysis typical payback is in the region of 90% depending on your timing, and it is possible to exit mid contract, it is essentially the same deal as what IG Markets call Digital 100
Is there a way to automate IG's BO ?
  • #103
  • Edited 1:26pm Sep 23, 2017 9:10am | Edited 1:26pm
Quote
This is the essence of quantitative analysis, we are simply looking for a correlation between given data and market movement. Get creative!

@GuyBaker - You mentioned this in the first page of this topic, could you please elaborate or point me to some more in-depth explanation/examples?
I want to develop a system that analyses the market based on this correlation; What do you think about statistics?
  • #104
  • Edited Sep 24, 2017 4:13am Sep 23, 2017 9:19am | Edited Sep 24, 2017 4:13am
Quoting tradista
{quote} Is there a way to automate IG's BO ?

They have a REST API but it is not available for demo accounts
I live entirely from the growth of my personal trading account
  • #105
  • Sep 25, 2017 12:55am Sep 25, 2017 12:55am
Quoting MrCash
{quote} @GuyBaker - You mentioned this in the first page of this topic, could you please elaborate or point me to some more in-depth explanation/examples? I want to develop a system that analyses the market based on this correlation; What do you think about statistics?

A correlation is anything that gives you an insight into probably future market direction. For example, interest rate rises generally lead to a currency gaining value, there is a correlation between interest rate changes and the value of a currency. Or it could be a technical correlation, for example, price tends to rise after crossing a certain moving average (just an example, I am not suggesting that necessarily is true), the idea is to find correlations in whatever data that can help us predict price movement such that we can overcome the cost of trading and produce a profit.

Investopedia has a detailed explanation herehttp://www.investopedia.com/terms/c/correlation.asp
I live entirely from the growth of my personal trading account
  • #106
  • Sep 25, 2017 4:43pm Sep 25, 2017 4:43pm
Lets speak about computing power obstacle.. How can it be solved for optimization problems with low budget
????
Take it easy.
  • #107
  • Sep 25, 2017 9:19pm Sep 25, 2017 9:19pm
Quoting Quants
Lets speak about computing power obstacle.. How can it be solved for optimization problems with low budget ????
By forgetting trading hopes and buying lottery ticket?
Joining to a quant firm as an employee is a lot budget alternative to lottery.

Joke aside, starting with a low budget, what will that leave for trading capital?
there is always, always another trade!!
  • #108
  • Sep 25, 2017 9:27pm Sep 25, 2017 9:27pm
Quoting profitfarmer
{quote} By forgetting trading hopes and buying lottery ticket? Joining to a quant firm as an employee is a lot budget alternative to lottery. Joke aside, starting with a low budget, what will that leave for trading capital?
Need an optimization technique to check 1 billion variables in short period of time. How can be this done efficiently without enourmous money spending.
Take it easy.
  • #109
  • Sep 25, 2017 10:35pm Sep 25, 2017 10:35pm
Quoting Quants
{quote} Need an optimization technique to check 1 billion variables in short period of time. How can be this done efficiently without enourmous money spending.
said above:
join to or bring to your ideas to a quant firm with existing computing power.
or
start to work on something simplier that suit your reality better ( ie no money for hardware)

may not be your preferred answer, but that doesnt mean there is no truth in it and all a joke...
there is always, always another trade!!
  • #110
  • Sep 26, 2017 2:48am Sep 26, 2017 2:48am
Quoting Quants
Lets speak about computing power obstacle.. How can it be solved for optimization problems with low budget ????

Yeah I know the problem well, many times my laptop would over heat or grind to a halt during my massive Excel sessions, the solution I used was to break the data into smaller chucks, it may be possible to do this using a cloud servers, Amazon Web Services et al. A billion variable though... hmm..
I live entirely from the growth of my personal trading account
  • #111
  • Edited Sep 27, 2017 1:06am Sep 26, 2017 2:29pm | Edited Sep 27, 2017 1:06am
Quoting Quants
Lets speak about computing power obstacle.. How can it be solved for optimization problems with low budget ????
I replied to a thread on backtesting in last 12 hours.... recurrsive backtesting. check it out before I delete it

here it is...

This topic is a great area of interest for me. The only way I could find myself avoiding all the different biases and pitfalls that stems from backtesting is to do what I call recursive backtest". What I mean by this is to slice and dice your backtest duration into subjective and objective building blocks. If your strategy is intraday shooting for a daily % target, pick a duration and then recursively test the strategy of each day within the duration of the backtest. Then do the same on a random distribution to ensure your strategy will hold merit through thick and thin periods of all market cycles. Do the same for weekly targets, monthly targets, quarterly targets, annual targets or whatever fits your trading plan. Doing this way is the most foolproof way to confirm whether your strategy will hold long term merit to cover all market conditions. Survival and sustainability should be the focus measuring all the necessary parameters (consecutive loses, win/loss rate, average win/loss, duration, peak/tough equity drawdown etc

I have a automated script (which took me 2 yrs to develop) that runs over 85 million combinations of my signals which does recursive loops 8 levels deep for each chosen pair, for each chosen timeframe and for each chosen "subset window" within the backtest duration. It would take years to manually conduct this type of detailed and robust backtesting via strategy tester!

There is no such thing as a coin flip or a 50:50 probability distribution in my opinion. The sampling size has to be huge to rule out any probability of guess or luck within your model. Either you get a positive expectancy from your backtest or you don't! All my backtests also, by default, factors a 5-8 pips towards trading costs and slippage. What this means is any trade has to be more than 8 pips in profit in order to be considered a valid profitable signal! Think about the rigor and hard work and patience strategy development requires. Any other route would simply result in failure and disappointment and lost opportunity.
I feel most intraday strategies end up in blown accounts because of lack of historical data available in MT4 to do a robust backtest. I backtested a strategy for a year by forward and backtesting every month to get robust stats. Are you willing to put in the extra effort before you throw a dime into your live trading? Most take the easy route of depositing small funds and then complain about blown accounts and about their broker. I was one of them during my journey!

Hope this helps.

Edit -typos and more detail

Edit: I have over 250GB of what you see in the screenshot :-)
Attached Image
Staying in my lane...
  • #112
  • Sep 27, 2017 1:10am Sep 27, 2017 1:10am
Quoting VEEFX
{quote} I replied to a thread on backtesting in last 12 hours.... recurrsive backtesting. check it out before I delete it here it is... This topic is a great area of interest for me. The only way I could find myself avoiding all the different biases and pitfalls that stems from backtesting is to do what I call recursive backtest". What I mean by this is to slice and dice your backtest duration into subjective and objective building blocks. If your strategy is intraday shooting for a daily % target, pick a duration
Which language do you use for it ? We have similar optimization algorithms and strong computers but i speak about much more advanced stuff..
Hitting billion variables or hundreds of million in short period of time.
Take it easy.
  • #113
  • Sep 27, 2017 1:39am Sep 27, 2017 1:39am
Quoting Quants
{quote} Which language do you use for it ? We have similar optimization algorithms and strong computers but i speak about much more advanced stuff.. Hitting billion variables or hundreds of million in short period of time.
good old mql on CrapT4. It's all local on my laptop but plan to go API route perhaps with Oanda using python. I wouldn't trust any server unless I administer it in my home office :-)
Staying in my lane...
  • #114
  • Sep 27, 2017 2:03am Sep 27, 2017 2:03am
Quoting Quants
{quote} Which language do you use for it ? We have similar optimization algorithms and strong computers but i speak about much more advanced stuff.. Hitting billion variables or hundreds of million in short period of time.

Do a search for Bitcoin mining servers, they have the same problem, IE processing power, maybe you will find some solutions there.
I live entirely from the growth of my personal trading account
  • #115
  • Sep 27, 2017 3:08am Sep 27, 2017 3:08am
Quoting VEEFX
Possibilities with computational finance is endless and a never ending endeavor. it's very easy to go nuts with trying out everything and get scattered all over the place (atleast that was my experience with my system development project). I had to constantly remind myself to stick to what I know about the markets. By nature, scalability is in the front and center of all my system development so need to watch every step to ensure staying on the path. The more we discuss, the more distraction takes us away from our objectives.

To an extent I think keep it simple is a good rule.
I live entirely from the growth of my personal trading account
  • #116
  • Edited 12:59pm Oct 1, 2017 5:44am | Edited 12:59pm
As promised I will post my results here at the end of each month, so here is the score for September; a disappointing -3.14%, first losing month in 7, I noticed a discrepancy between the theoretical results and the actual of some 5% during September also the same execution issue in August, it is unusual to get exactly the theoretical results in live trading, but generally the two numbers are within a percent or two per quarter, the upshot, I found an execution system bug which meant some trades were being missed, possibly I missed as much as 10% return in the last two months, which is very gloomy as for me that is in the region of £21K. Bug fixed now, but this issue highlights the importance of keeping detailed records and auditing in every way you can think of, particularly comparing actual trading to theoretical results, and monitoring your edge to see if it is performing as your analysis suggests it should.
I live entirely from the growth of my personal trading account
  • #117
  • Last Post: Oct 17, 2017 12:58am Oct 17, 2017 12:58am
Quoting GuyBaker
As promised I will post my results here at the end of each month, so here is the score for September; a disappointing -3.14%, first losing month in 7, I noticed a discrepancy between the theoretical results and the actual of some 5% during September also the same execution issue in August, it is unusual to get exactly the theoretical results in live trading, but generally the two numbers are within a percent or two per quarter, the upshot, I found an execution system bug which meant some trades were being missed, possibly I missed as much as 10%...

thanks for the thread, it's a good read. Sorry to hear there is a bug, what caused it? And may I ask, is your approach based on candle stick pattern or something else? And do you think machine learning could help to generate better signals?
0 traders viewing now