Dow 96 point entry

[jpwone]

Hijack away

Forex data (amongst others) available for free here.

http://www.dukascopy.com/english/rdata/longterm/

I too would love to see a fuller explanation of JonnyT's Forex method if he does not mind sharing a bit more.

 

[JonnyT]

Hi jp,

My Forex method is staying confidential but I have given you the money management bit.

Have a look at the times of day when trends tend to start (There are two) and work around those.

75,

The daily cash is a figment of the spreadbet companies imagination. You cannot trade it in real life. You really do have to test systems using futures data as the SBs quotes are based on the underlying futures market.

JonnyT

 

[alp1us]

good tool

thanks man

 

[mmillar]

There are two ways to skin a cat...

I know a number of you are still working on this, so I thought I would try a different approach to see what happens.

As well as using long term backtesting to try and find a formula that will give you the best trades going forward you can also backtest on very short term data and then over-optimize the results to tell you what to do next (the idea is that what happened several years ago isn't relevant but what happened last week/month is). So, for instance, you optimize on January's data to find the best entry price in January, then you use that entry price for February. Then at the end of February you optimize on February and use that entry price for March etc etc. Effectively, you are trying to see if the best entry price for one month has any relationship with the best entry price for the following month.

I've attached a spreadsheet that shows the results of optimizing for each month. The most important worksheet is the Summary sheet. I have split the results into Longs and Shorts. The stop is 50 points from entry. In Column A is the Entry Level (so 70 means 70 points above yesterday's Close). You can see that, for example, the best Long Entry Level for December 2002 was 70 (giving you $107 profit), but if you had used 70 as your Entry Level in January 2003 you would only have made $381, whereas the best entry level in January was at 10 (giving $619 profit). If you then used 10 as your Entry Level for February you would have lost $356. I hope this makes sense. :|

Unfortunately, I couldn't find a relationship between one month and the next. You can see that months like February and August on the long side produce losses whatever Entry Level you have. The best results come from using 60 on the Long side and -60 on the Short side right throughout the year. But you have no guarantee that these levels will work going forward.

 

[rglenn]

Thanks for that Mmillar

That has been extremely helpful for further studies.

The Dow has definitely been cracked but we need to fine tune things.

I will post my full system when I have refined it enough.

All the best

 

[bansir]

crazy thought

Just been reading this thread, a lot of the old problems have surfaced regarding the backtesting.

1. The Dow cash data is only daily OHLC so some triggers are brushed under the carpet.

2. The SB prices track the futures (more or less) rather than the index, this means the SB bid/offer will typically be a distorted, or amplified version of the data being used for the back test.

It was the second point which led me to an interesting question.

Assuming that historical intraday cash data is available courtesy of GS ( or was it BB??), could an amplification factor+offset be applied to it so that we are in effect modeling the SB price?

Thus the actual input data used in the SS OHLC coloumns would be P*X+Y where P is the intraday price data and X and Y are scaling and offsets which could be experimented with to see how tolerant the system is to increased bias+volatility.
Perhaps some techniques could then be devised to increase the
tolerance of the system.

With X set to 1 and Y set to 0 then the backtest results would be unchanged, but what if X were increased to say, 1.1 ?

Just a thought, probably been done before.

Cheers,
Neil

 

[bansir]

re-crazy thought

I've just been playing around with the fixed +/- 75 version of the backtest.
I inserted extra coloumns for the OHLC data and applied a various multiplying factors varying from .8 to 1.5(!).

I can't make the system actually lose money at all.
I thought I may be able to find levels of volatility where the stops keep getting triggerd more thus taking out some of the gains.

In fact the higher the multiplying factor, the more net gain at the end of the period. More triggers being hit, also the larger swing on the close price bumps up the end of day reward.

Reducing the factor gradually reduces the net gain, but you would expect this because the triggers start to reduce.

I think this does suggest that the real problem lies with the intraday swings.
I'll try and do the same with intraday data-I only have the cash prices but I'm curiouse; can I make this thing lose somehow

See Yah,
Neil

 

[darrren]

Hey Bansir

Keep up the good work.

'I can't make the system lose money at all' - music to my ears!

Darren

 

[rglenn]

Bansir

I think you will find that there are many different ways to use the data on this thread to create a very viable system.

There are lots of free data and information on this site that people can use to create a profitable system- GEMS in fact.

But so many people walk on by without realising the benefits.

I have adjusted and re-adjusted my system using some of the data on this thread and came up with a even more profitable system.

I have no need to sell it nor will I be putting the details on this BB as I know there are some people just scouring the boards for details of systems to sell on etc.

Anyone that can't put a system together and would like more information of my system - please feel free to PM me at anytime.

Looking forward to your further research.

All the best

 

[bansir]

Egg on face time for me I; if my re-think is correct then
for a system using one buy stop threshold (Yest. close+75) and one sell stop threshold (Yest. close-75) to enter the market, the order of the intraday highs and lows is irrelevent.

I was mislead by some work I had carried out with a bracket trading system in which the order of the peaks and dips does matter as OCO orders were used.

If anyone wants to put me right feel free.
If I am right, we don't need to worry about intraday data; daily ohlc will do
Looking at the action on the dow today the system was triggerd at the 10460 spike low (D4F) but on SC data the system remained untriggerd, the low being 10466.
I think I will re-format the spreadsheet to just strech the minute bar ranges without altering the O and C values, see if I can cause
some adverse trades. On paper I lost -58pts today due to that 10460 spike down.

bye4now
neil

 

[mmillar]

Bansir,

For an intra-day system you need intra-day data - OHLC data makes the system look really good but ignores all the intra-day stops and losses. e.g. (assume you triggers are at +/- 75, stop is at 50) If the market started at 10000, then down to 9025, then upto 10000, then closed at 9000. On paper it looks like you made 25 points, when in reality you would have been stopped at -50 points.

When you are trying to model the SB numbers, why not just use the futures data? It's much more accurate then trying to second guess the SBs.

re your post above about your system being triggered at 10460 when the market only got to 10466. If I understand correctly shouldn't you base your triggers on the Index and then just take the SB price? That way you know you are getting triggered correctly and just have to assume a certain slippage.

'I can't make the system lose money at all' - sorry to be a damp squib but that's not the issue. It's how it made that money. If a system takes 10 trades, makes $1000 on the first one and then loses $50 each on the other 9, you still have a system that 'makes $550!'. But in reality it's a terrible system that is untradeable. You have to look deeper into the numbers to work out what is going on.

Cheers

 

[Dow Dog]

Go long on the Dow when it is up 96 points and short when it is down 96 points in any day. Stop if it reverses back through the previous day close and always close at the end of the day (no trades carry over and there are plenty of days with no trades entries). If you back test this you will find it is very profitable. I have not forward tested it. It might not work in the real world but at least it wont break you finding out.

Does the Dow 96 point system actually work for anybody ?

The current ATR is less than 96 points and even if it weren't, if the DOW went up by 96 points and you then went long, isn't it more likely to go back down - having exhausted its run - rather than keep on going up ?

How many points per trade is this system looking for ?

Maybe I am missing something here ?

 

[JonnyT]

Agreed, too many novices here think they are on to a good thing without proper testing.

You absolutely have to test these systems with intraday futures data.

For example I have backtested a system using daily cash intraday data that were so good it was unreal. The same system using the futures data was not tradeable.

YOU CANNOT TRADE THE DAILY CASH

People here seem to think they can, sorry to disappoint.

JonnyT

 

[jpwone]

Hi Dow Dog and JT

the entry level has been reduced to +/- 75 points to take account of the recent reduced ATR.

When I was looking at the data I had exactly the view that you have. If it has gone up by x points then surely it would run out of steam and you should trade in the opposite direction. Backtesting showed that this lost money consistently. As an experiment I opened the trade in the market direction. This was surprisingly succesful but was based on EOD data.

The work done subsequently by the many contributors on the thread has revolved around testing using intraday data and modifying the entry point to better account for the more recent ATR.

It is still very much a work in progress but I now have some good data and need to find the time to properly test this. The initial indications are that it does actually work but (and its a big but) I feel that I need to do a lot more work before I will forward/live test with a good degree of confidence.

The bottom line for myself is that one way or another I have a high degree of confidence that the basis for tradable system is here. It needs some work and it needs some testing but I have not changed my view that there is a profitable system here.

I am unconcerned about the debate with regard to cash or futures data. I will continue testing with the cash and if I find I am wrong I will change. Until that time I am sticking with the cash. It provides a good data set for the purpose of testing and if I can then transfer the system to futures data and improve it by doing so then I will. The basic arguement in favour of the futures prices is that the SB company price for cash is based on the futures price and it is difficult to get the entry price using cash. In the few experiments I have done I have always got the entry price give or take a few points. The significant advantage of the cash is you are sure to get the exit price at the end of day with no slippage and no spread. So, there are arguements in favour of both cash and futures. If one is more profitable than the other with all factors considered then I will change but picking and choosing amongst the relevant factors should not be a basis for change at this time in my opinion.

 

[JonnyT]

YOU CANNOT TRADE THE CASH INDEX

 

[bonsai]

jonnyT
perhaps you are taking this too literally ?

JP has always implied that this system would be traded using sb's.

(I think that is right)

 

[jpwone]

LOL JT

I'll agree to diagree and will happily accept I could be wrong

 

[Mr Chill]

Hi

I've been following these trades using fins only at £1 pp to see how things go ,so far i've had two triggers 29/12 =+42 ,
5/1 =+54 ,early days i know but interesting all the same , both using daily cash .Good work jpwone .

 

[ChartMan]

JT, when you say you cannot trade the cash index, can you be more specific ?. Clearly you can with SB account, clearly you can't on direct access.

 

[jpwone]

thanks Mr Chill

I picked up those triggers as well. One of my next tasks is to look properly at the win loss ratio and see if i can identify any consistent factors for the losing days which will reduce these false triggers.