To use chat, please login.
Back to contacts

Portfolio Companion Thread

bizWiz
Sep 24 2009 at 16:24
397 posts

fxanalytics posted:
You can trade for a living; I've done it. You either need a very large porfolio before you begin or you need to manage funds for others.


so which way did you try? both maybe?

reading your posts, i clearly see you're an experienced trader, and maybe even one of the old-school traders, making a low yet stable return of 4%-5% per year. I agree that if you're able to produce such returns over and over again, it's much better than those 5% a day traders.

Sleep is for the weak.
fxanalytics
Sep 24 2009 at 16:49
17 posts
I was lucky enough to get my start in finance very young and I worked with a number of successful traders early on. I was also exposed to a lot (A LOT) of unsuccessful traders. It was harder to define in the 80's but you could see trade activity and exposure (leverage and liquidity) seemed to be linked; that was really driven home when LTCM went under. Unfortunately, even in the professional world, the guys who make 100% for a year or two get all the attention, and when they blow up, there is always some other new exciting approach or person who quickly takes their place. People don't get excited about the slow and steady guys until they are moving enormous amounts of money like Buffet or Soros.

Me, I was just a small time operator that managed a small portfolio for a group of investors - I've always been the low and slow type. Unfortunately some risks are hard to quantify and can't be predicted. We were part of the group that got wiped out when Refco went under. That significantly changed my risk profile and I have traded since... I'm still active in the industry though, been around it too long to quit. This past summer I blew the dust off the old model and with some improved computing began to rework the old approach. Slowly putting small amounts of cash aside to get back to live trading again. We'll see where it goes from there; if all goes well I'll eventually be managing funds again. It can be a long road to raise enough money to make it worthwhile, but if you do, a few good years in the market and your set. Then you can be much more restrictive about who you work with etc. Once you've made your money, you can trade just for yourself if that's your wish.

bizWiz
Sep 24 2009 at 17:28
397 posts
i must admit it's a pleasure reading your posts!

appreciate the your honesty.

so you didn't get anything back from the refco scandal? and why would you let it affect you as a trader? if you were successful, why didn't you just start over?

if you had the skills and the connections, all you're left to do is trade.

Sleep is for the weak.
fxanalytics
Sep 25 2009 at 13:49
17 posts
It's all about the money... After Refco there was very little left. I had spent three years supporting myself, a lot of capital on R&D and on the business itself to turn the model into something that would handle the kind of funding it takes to make it worth while. I also found myself without a way to generate livable income after the collapse. I don't live in one of the financial hubs so running out and landing a Series 3 job wasn't a possibility.

Just starting the trading over wasn't an option and to be honest the emotional impact of it took some time to get over as well. Decisions I had made about the broker cost people their hard earned money. I don't take my clients capital lightly.

Once I did get back to work, the demands of those jobs made it difficult to trade and monitor the market, it would also have been impossible to have taken on clients as this adds an entire layer of operational requirements and difficulties. I did spend quite a lot of time trying to leverage the success I had as CTA to get refunded with a hedge fund, but my professional track-record was just too short and I didn't have enough AUM.

It goes without saying I learned a lot through that experience, things that I think would make me a better 'manager' of a CTA - a lot of hindsight if you will. Now I just need to get the pieces back in place. It's patchwork right now; I work on things when I can, work when I can and keep grinding forward. This past summer I fired up the grid of computers (8 dual processor workstations) and began a large scale regression. Spent several weeks on it, running and rerunning the data-mining process. I have to admit that the first time around my testing may not have been as thorough as it should. You think you are being statistically significant when you run hundreds of simulations when in reality it takes thousands to really determine if what you are looking at has any real potential.

When I was done with it, I had ran 5k simulations over 1mil datapoints... things looked okay but I'm not convinced the testing methodology is right. I can see issue problems with they way the monte-carlo is setup. I think it 'biases' the outcome to a positive expectation.

The issue now is, I've got to work two jobs to survive and time for testing is limited but I'm getting closer to being able to actually pull the trigger. I have 23 years to retirement, pretty sure I'll be able to give it another go in that window!

fxanalytics
Sep 25 2009 at 13:52
17 posts
On a different note, I'm going to use this first model simply as a test of the application here. My trade model can be quite slow to signal, on average it kicks out about a trade a week but I've had periods where it has left me on my thumbs for three or four weeks. I want to see how the analytics work etc so I'm just going to throw some technical trades in there across the few weeks and see how the application works and what kind of information it gives me.

fxanalytics
Sep 25 2009 at 16:21
17 posts
The only instruction I can find on the site are for adding an Oanda or an MT4 account. How do we add an FXCM Demo account? I've set one up, just can't figure or find out how to link it...

bizWiz
Sep 25 2009 at 16:28
397 posts

fxanalytics posted:
    It's all about the money... After Refco there was very little left. I had spent three years supporting myself, a lot of capital on R&D and on the business itself to turn the model into something that would handle the kind of funding it takes to make it worth while. I also found myself without a way to generate livable income after the collapse. I don't live in one of the financial hubs so running out and landing a Series 3 job wasn't a possibility.

Just starting the trading over wasn't an option and to be honest the emotional impact of it took some time to get over as well. Decisions I had made about the broker cost people their hard earned money. I don't take my clients capital lightly.

Once I did get back to work, the demands of those jobs made it difficult to trade and monitor the market, it would also have been impossible to have taken on clients as this adds an entire layer of operational requirements and difficulties. I did spend quite a lot of time trying to leverage the success I had as CTA to get refunded with a hedge fund, but my professional track-record was just too short and I didn't have enough AUM.

It goes without saying I learned a lot through that experience, things that I think would make me a better 'manager' of a CTA - a lot of hindsight if you will. Now I just need to get the pieces back in place. It's patchwork right now; I work on things when I can, work when I can and keep grinding forward. This past summer I fired up the grid of computers (8 dual processor workstations) and began a large scale regression. Spent several weeks on it, running and rerunning the data-mining process. I have to admit that the first time around my testing may not have been as thorough as it should. You think you are being statistically significant when you run hundreds of simulations when in reality it takes thousands to really determine if what you are looking at has any real potential.

When I was done with it, I had ran 5k simulations over 1mil datapoints... things looked okay but I'm not convinced the testing methodology is right. I can see issue problems with they way the monte-carlo is setup. I think it 'biases' the outcome to a positive expectation.

The issue now is, I've got to work two jobs to survive and time for testing is limited but I'm getting closer to being able to actually pull the trigger. I have 23 years to retirement, pretty sure I'll be able to give it another go in that window!



very interesting.

are you a programmer by any chance? as it does sound like you are.

have you had any experience with neural networks?

Sleep is for the weak.
bizWiz
Sep 25 2009 at 16:29
397 posts

fxanalytics posted:
    The only instruction I can find on the site are for adding an Oanda or an MT4 account. How do we add an FXCM Demo account? I've set one up, just can't figure or find out how to link it...


don't think you can, looks like only oanda and mt4 accounts are supported for now.

but why would you open a demo fxcm account? just open another mt4.. much better platform i would say.

Sleep is for the weak.
fxanalytics
Sep 25 2009 at 17:17
17 posts

biz0101 posted:
    
fxanalytics posted:
    The only instruction I can find on the site are for adding an Oanda or an MT4 account. How do we add an FXCM Demo account? I've set one up, just can't figure or find out how to link it...


don't think you can, looks like only oanda and mt4 accounts are supported for now.

but why would you open a demo fxcm account? just open another mt4.. much better platform i would say.


Thanks for the input...

To answer the other question, I'm not a programmer. More of a power user LOL. I tested a lot of different platforms including neural net and finally settled on the one that I felt gave me the least amount of bias. Most people on our level don't really consider the 'quirks' that are inherent to that type of computing and forecasting; all these applications have their own internal biases that can skew your forcast and trade entry. This is where all the research comes in; it takes a lot of time and a lot of testing to identify those anomalies.

When it was all said and done, I had the code I needed developed and then tested alongside a commercial application that I was paying nearly 20k a year for. The code cost more than that but the application is mine and should save money down the road. I also looked at converting the grid over to Lynux and Python but the guy I was working with flaked and disappeared into other projects. So I'll just keep what we have; it works (I think haha).

Unfortunately MT4 isn't robust enough for the code I run: The forecasting component requires parallel processing and something more robust (and compact) than C based math. There is also another component to the model, this might be easier to get into MT4 but I've never been comfortable with MT4 as a professional tool to handle funds. Things like Currenex with their large liquidity pools always seemed more appropriate to me; plus they allow for spreadsheet trading. In that scenario I can key in the model forecast, run my own prop analytics and when a trigger is generated initiate a trade. That's waaaaaayyyyyyyy off though. I have absolutely no problem building the required positions on a vanilla retail platform and the big retail guys are getting large enough now to really provide good liquidity so I may never really have to take that jump... We'll see.

Firs things first; I've got to save my lunch money and open a small micro account and just get back to trading and getting a feel for the application and process again. I've trade rust I need to brush off with a flap wheel!








bizWiz
Sep 26 2009 at 13:56
397 posts
looks like you have a clear plan how to move on.

i agree that metatrader 4 isn't powerful enough for advanced computing algorithms, but i think the new metatrader 5 which should be released soon is going to change that (hope so anyway).

so you're saying that something that was developed several years ago is still working in today's market?

Sleep is for the weak.
Please login to comment .