Deadline June

I have tried twice before to make it with mechanical systems - once in 1992 and once again in 1998, and now I'm going to give it another go, because the bug never left me.

I've been earning a living as a software developer and I've been playing with Tradestation a lot. My plan is simple: I have £40K risk capital and £15K to live off and the goal is to start trading at the latest at the beginning of June and to be profitable enough to withdraw £1500 per month from then on for subsistence.

Hopefully my career will last a bit longer than June.

I've already done a lot of research and testing - it covers futures and forex – about 25 markets with high liquidity – all on a daily time frame.

I concentrate on break-out algorithms, with medium to high frequency trades – at least 10 per annum per market.

I find set-ups in TradeStation that look good, and then export the trade history into my own risk analysis program to have a look at its behaviour in different markets, with equity curves, correlations, performance stats etc. Most of the set-ups I find are rubbish – but every now and then I find one I think will work.

I've tried several times in the last couple of years to kick off this trading project but until now it has always fallen down because of outside factors in my life taking over. If anyone reading this is at all like me, you'll know that working on your own without outside input leads to going off on tangents. I hope that running a journal here will draw some kind of commentary from other people that will sharpen my focus when things start to go off the rails.

I aim to write a journal entry everyday - so between now and June my entries will be all about the research and preparation I'm doing – but I hope I'll do some exploratory trading within a month or so because I need to test the markets to work out how much slippage I'll be paying, which is probably the most important factor that can make or break the algorithms I come up with.

So here goes. My theory is, I'll get focus and direction from writing here to make public what I do everyday. You want to comment? Be my guest.

For my first 2 days – Feb 1st and 2nd –

I've been composing that little announcement above

I put a project plan together - I have just under 18 weeks to get this show on the road

I've been fighting some crazy problem with my internet connection which makes trade2win.com unreachable about 50% of the time - anyone else ever experienced that? - it's just this website, not any other.

I talked to a hedge fund manager in NY about possible investment with me, and I realise I have to put together some kind of summary of what I plan to do, along with some stats that I've never even considered including beta-adjusted returns, delta-adjusted returns / net exposure, and leverage-adjusted returns. He talked about VaR and I talked about Black Swans – I must find a more convincing way of saying you should give me your money but I'm not going to give you any VaR figures.

but please tell me how you came to the fund manager? for me it is very important, I am looking for an investor, and sell every day from Monday to Thursday. I hope you all get

neronius said:

but please tell me how you came to the fund manager? for me it is very important, I am looking for an investor, and sell every day from Monday to Thursday. I hope you all get

Click to expand...

Just through networking, friend of a friend

No progress yesterday - spent the day trying to sort out my broadband (ironically it's this website that I have problems with) and my printer and a glitch in my risk analysis software.

I can see why it's worth paying someone to sort out this sort of faff for you, so you can get on with the really important stuff.

Anyway, another day gone, 16 weeks and 4 days to go.

i wondered do you mind reveal all your entry and exit technique here?

everyonerich said:

i wondered do you mind reveal all your entry and exit technique here?

Click to expand...

haven't got an entry and exit technique yet, at least not live.

but it's a break-out system. The entry could be based on anything - upper or lower break-out points defined by stuff like a % of the ATR from yesterday's high (for long) or from yesterday's close, or it could be a linear regression of the previous 5 highs. I coded a resistance level algorithm, but the conditions occured too rarely.

The exit strategy at the moment is just to exit on close. I never got further than that because I still don't know what slippage is inherent in the markets I want to trade. But I like exit on close because it means I'm flat overnight and my margins would be day trading margins (I think).

Plus I use filters to stop trading.

Adamus said:

No progress yesterday - spent the day trying to sort out my broadband (ironically it's this website that I have problems with) and my printer and a glitch in my risk analysis software.

I can see why it's worth paying someone to sort out this sort of faff for you, so you can get on with the really important stuff.

Anyway, another day gone, 16 weeks and 4 days to go.

Click to expand...

This site is the one I spend most time on, it is also the most likely to fail to load. Used to have it as my home page in IE8 but had to revert to Google or I would never have got anything done online.

try checking your routers MTU settings, and also the MTU settings in windows. Google around a bit about MTU settings, I cant remember exactly what to do, but this has been the solution to similar sounding problems ive had in the past.

1Pipped said:

try checking your routers MTU settings, and also the MTU settings in windows. Google around a bit about MTU settings, I cant remember exactly what to do, but this has been the solution to similar sounding problems ive had in the past.

Click to expand...

Technically you might be correct but after Googling how to do this as you said I think I'd rather perform surgery on my own leg.

Not the curse of the MTU settings! Arraaaagaghaghghagahahaaaaaaghghghgh .......

Been there, got the burn marks.... I had trouble uploading files because of MTU problems last year, and the hosting provider gave me an MTU setting to sort it out but it turned out to be my broadband modem which had some flakey firewall inside (undocumented of course) which blocked the MTU packets ... or bytes ... or whatever they are.

Bang goes my plan to trade a basket of 25 futures and forex mkts!

Just put together a trading simulation and looked at the margin requirements (IB margins) - x axis is USD - ouch! That might require a bit more capital than wot i've got in my account



Edit: the basket is weighted so that all the markets have roughly the same leverage - hence the first column is 25000.

That's less scary. Basket contains 17 futures and forex mkts.

If I got my hands on an extra 50K I could probably trade this - assuming it didn't go straight into drawdown

That's 2500 days and only about 200 of them require more margin that I'd have (with 100K in my account).

Exactly what I would do on a day when all my orders got hit, I have no idea.

One option would be to just cancel the remaining unfilled orders once the account is maxed out.

I could also exit positions that didn't look promising or were in drawdown. Not exactly system trading though

16 weeks to go exactly - and here's the current situation:

Decided that getting other people's money to trade with can't be a deciding factor, so I'm going to concentrate on a trading approach that doesn't require so much capital.

That means I'm ditching LTTF with a big basket of markets in favour of Pardo style backtesting and optimization on 5 year / 6 month walkforward windows.

Actually I originally planned to use Pardo style back-testing / walk-forwarding but it requires so much more work to do the research and record the results and all that, I thought LTTF with a big basket of currencies would be a much quicker solution. I also had more capital then

So I'll look at the German Bund to start with and to get a good idea of the effectiveness of the technique at least in simulation, I'll start with the following:


Back-test window - - - - - - - - Walk-forward
2002-07-01 - 2007-06-31 - - - 2007-07-01 - 2007-12-31
2003-01-01 - 2007-12-31 - - - 2008-01-01 - 2008-06-31
2003-07-01 - 2008-06-31 - - - 2008-07-01 - 2008-12-31
2004-01-01 - 2008-12-31 - - - 2009-01-01 - 2009-06-31
2004-07-01 - 2009-06-31 - - - 2009-07-01 - 2009-12-31


If I can get to the point where I choose an optimised system in back-testing that remains profitable in the 6 month walk-forward period, I'll be happy enough to start trading it with optimisations from 2005-01-01 to 2009-12-31

This though is not just an exercise in TradeStation system optimization - I have to identify the boundary in the complexity of my algorithms where future-proofing fails and curve-fitting kicks in - and then make sure I keep the right side of it.

Here's what I came up with for my optimisation window:



That was as smooth as I could get it. At that point I already knew this method was doomed since it was flat for 2 years and I wanted something to give me profits in the walk-forward. Fat chance! Spent the night dreaming of jagged mountain ranges.

But I was shocked at the walk-forward result when I looked - the equity curve for the whole data I've got is rubbish:

So after that disgraceful display of curve-fitting, I'm going to try it again. This time however I need to use a different window. I want something smaller, perhaps 2 years or 18 months, so that a successful walk-forward carries on the same sort of returns that I get in the optimisation window.

5 years was too long for the kind of algorithms I'm using. I only got one decent winning streak and that only last for 25% of the optimisation window. I don't know whether it can work with the algorithms I'm using, but
cutting it down to 2 years might be worth a try.

So I'll also cut down the curve fitting with all the different switches and filters I did yesterday and stick to 2 filters and 2 range variables.

Degrees of freedom mean nothing to me - I thought they were markings on the dial outside Ronnie Biggs' prison cell.

Code:

Optimize                             Walk-forward
Start               End               Start                End
2005-07-01      2007-06-30    2007-07-01      2007-12-31
2006-01-01      2007-12-31    2008-01-01      2008-06-30
2006-07-01      2008-06-30    2008-07-01      2008-12-31
2007-01-01      2008-12-31    2009-01-01      2009-06-30
2007-07-01      2009-06-30    2009-07-01      2009-12-31

Computer wouldn't turn on this morning - it was a short day for me anyway due to other commitments but I lost the whole day essentially working out that it was the power supply, buying a new one and installing it.

But before the computer died, the back-testing I was doing looked fairly interesting.

Simple stuff gives decent returns - haven't forward tested it yet, but it consists of a break-out system with a trailing stop, so I can optimise the size of the band being broken out of, a filter, and the position of the trailing stop.

Hopefully I'll be able to get some more work done tomorrow, unless another something else in my pc dies.

15 weeks and 2 days left.

equity curves at last

three days down the pan struggling with java and tradestation problems but finally managed to do an optimisation run, get something looking half-useful, and the walk-forward worked out too.

Optimisation window 1



Walk-forward window 1



Ignoring all the wasted time faffing with java and tradestation, that only took about 2 hours. So extrapolating that and assuming I get a bit faster at it, I think I could do a 5 window series of optimisations and walk-forwards in a day and get something tradable in the end.

Except the doubt creeps in when I look at that optimisation equity curve - it only goes up for half the time.

not what I was looking for....

Optimisation window:



and the walk-forward:



I sensed that something was wrong with this one because (1) varying the stop caused big jumps in profits and (2) 90% of the profit was from the short side. Should have listened to that little voice and looked a bit further in the optimisation, but I was getting impatient.

Next optimisation window should be better if I stick to the straight and narrow hopefully.

I am not getting far with the optimisation in the next window. I want an optimisation where the biggest drawdown was max 3 or 4 months.

Considering that I'm going to walkforward for only 6 months, what are the chances of the walkforward being profitable when I have a 6 month drawdown in my optimisation?

Even if the optimisation is super-profitable, if it contains a 6 month drawdown, a repeat of such a drawdown in the walk-forward will just kill it. No point really bothering.

I guess I need some new algorithms, because I can't get anything more out of the current lot without curve fitting it.