A system with good potential, how would you improve it?

rockford

Newbie
Messages
3
Likes
0
Hi,

Although I have been a forum member and occasional visitor since 2003, this is my first post, so I realise that you will almost certainly view my comments with a lot of scepticism. I'm here to ask for your advice, not to pretend that I'm a trading expert (which I'm not).

I have uncovered a system for day trading the Dow Jones Index, which I have recently started trading for real. So far it's very early days, but it seems to be going to plan so far.

It's a very raw system, and from what I've read on here, most traders would look at it, shoot it down with flames, then run a mile. However, to me the figures are just too good to ignore, so I am running with it.

If you have any tips on how I could make it more tradeable, I will be grateful to hear them.

Without revealing everything, here are the basics of the system...

I came up with the system after looking at a caption of about a months historical data. It looked promising so I ran a test which showed that it appears to have worked for many years. It has nothing to do with charts, or any other complicated system. I just noticed that the Dow did something regularly, which is very obvious. It’s not linked to any part of the day.

Using a simple moving average formula I can come up with a with a buy signal, as soon as I know the opening price. The signal is normally triggered 4 days out of 5 on average. I use a different moving average formula to produce a sell signal, which also gets about 4 trades every 5 days on average.

Each signal is traded independently of each other. I place buy orders on the Wall Street Daily Cash, and the sell orders on the Wall Street Daily Future.

Some days there are no bets, some days both orders are triggered, and others just a buy or a sell signal.

With just the raw system 'as is' there are no stops. I place a buy order, if it gets triggered that's it. It's left to run until the end of the day. This fact alone will make most people reading this dismiss what I have to say. However, as I've said, I'm here to ask for advice, not tell you how wonderful my system is.

Doing exactly what I said above, I download the historical daily data from Yahoo, and ran my system through a spreadsheet test.

I've been using Finspreads, because of their narrow 4 point spread on the Dow. For every trade that was triggered, I subtracted 2 points to account for the spread. Here are the profits the spreadsheet returned at a £1 a point.

------------------------------------------------
When I first posted this, I cut the last part of the post off by mistake, after pasting it in from Word.
------------------------------------------------

The bit I originally forgot to add...

2005 4491.89
2004 7287.91
2003 5643.02
2002 8697.15
2001 11635.64
2000 20858.05
1999 18883.59
1998 11946.76
1997 6742.31
1996 5687.80
1995 4990.30
1994 3301.12
1993 3740.79
1992 2739.86
1991 1976.99
1990 1653.89

If I fiddle around with my moving average, and a fee other things, the above results can be optimized and increased by about another 40%.

Without any stops in place, this is obviously hard to trade with a large bank. The way I've been approachng it is to split my bank into more points. So for example, instead of say having a £5000 bank, and betting £10 a point, I would have a 5000 point bank and bet £1 a point.

The biggest losing run since 1990, was about 950 points over a ten day period. In the last 4 years, the worst run has been about 450 points. Both runs recovered quickly.

I have tested putting in a stop, and set it so that I would lose double what I could potentially win on any given day. This reduced the above figures to about 40% of what they are now, but with the added comfort of knowing what to expect.

What would you do with a system like this, to maybe get the most out of it, and make it more tradeable?

Thank you in advance.
 
Last edited:
rockford said:
...... I download the historical daily data from Yahoo, and ran my system through a spreadsheet test.

I don't trust yahoo data. You seem to be using their EOD data? If so, check it against data from here:-

www.eoddata.com
 
I appreciate your reply. I didn't realise that Yahoo's figures were not trusted. I have done some random checks against other sources for the end of day, and Yahoo's data appeared to be OK. It wasn't an extensive check though.

I have just registered with eoddata.com, and will see how things look with their data.

Thanks.
 
rockford said:
Hi,
What would you do with a system like this, to maybe get the most out of it, and make it more tradeable?
Thank you in advance.

Hi Rockford,
Without knowing your system I cannot tell you how to tune it.

However, you could impose some limits if you have thoroughly tested it.
For example, if it is 80% accurate on 4 trades per week, then you might say
that if it dips below this result for 2 consec weeks you stop trading until it comes back up again.

Goodluck
 
would it be a better option to use outright options to predifine your risk against some of the drawdowns ? you need to look into it, lowest option cost per % point, not sure though just an idea,you'd need the advice of someone who could tell you the best option to get based on time, volatility, out of the money etc.
 
I just downloaded the historical data for 2005 from eoddata.com. If they are as reputable as you imply BigBusiness, I would like to say a big thank you to you.

Eoddats's figures do differ quite considerably from Yahoo's. Their data shot a huge whole in my system. I must have been lucky with my real time trading so far.

With a few tweeks it does show a profit, but much smaller.

I've got more analysis to do for sure.

Thanks, you've been a great help, and saved me a lot of time and money.
 
rockford, If the system places trades at the openning price it may be an issue as the cash openning may be different from the futures. Good luck, Tuffty.
 
As Tuffty has said, when you trade your values will be linked to the futures prices, not the index itself. You should test your data using futures data- it may or may not blow your system.

Try using your parameters on the futures data itself. Also try using the index data to provide the triggers, and the futures data for the buy and sell values. You may get better / worse results again.

Also, as every trade is opened and closed, you need to apply the full spread to each deal - ie 4 pts.

If you want EOD futures data for this (until Dec2004), PM me and I'll send it on.

Hope this helps,
UTB
 
Are you saying you have only " tested " this system by backtesting ?

If so , sorry to rain on your parade , but that ain't the real thing and in trading 1 and only 1 thing counts - real profits in real markets with real risk control over a real long time .
 
Top