Linear Regression Forecast

[cd173]

Hello

Does anyone use Linear Regression with price action to catch longer term price swings.

I got the idea from www.linnsoft.com/charts. I've been looking at a revised system on the DAX 10min which has a 6 point stop and 5 point trail stop with entry on a price break. Glancing over the last few months it appears to work pretty well, catching big price moves. Not sure in the long term though as it would need further testing!

Anyone using something similar?

C

 

[Pippppin]

Dynamic Linear Regression

Hello

I have spent over eight years developing a trading system applicable to all markets which uses Linear Regression. It is also applicable to all time frames.

Because the system measures changes in the regression characteristics from one bar to the next I use the term Dynamic Linear Regression.

My goal is to make a significant contribution to financial analysis and system trading development. Time will tell..........................

 

[cd173]

Hi Pippppin

From, what I understand LRF doesn't have the 'lag' that a ma crossover does?
If you wouldn't mind, I'd be interested in hearing about your methods, but understand if you want to keep them under wraps!

Cheers
C

 

[Pippppin]

MA versus DLR

Hello C

It is good to know that other people recognise LR as a potentially profitable trading tool.

Funny you should mention the "lag" of MA. I began my research in 1996, and I spent a couple of years doing some pretty intensive research re: moving averages. My eventual conclusion was that you are always "behind" the market ie reacting to what it has done, rather than what it is going to do. Which is not to say, of course, that it cannot be used as a successful trading tool.

It was March 1999 before I had an insight as to how LR could be used to quantify a specific recurring situation in market behaviour. Nothing since then has led me to deviate.

I am a little "coy" about revealing my findings, at least at the moment. But that may change, and if it does I wil be happy to give you some more information.

Meanwhile, good luck with your trading and system development

 

[Bigbusiness]

I have used MA envelopes in the past and I still use them on daily charts to try and find good times to trade when the indicies are oversold or overbought. My charts are in Excel and it took me a long time to work out how to make them. I agree that the lag is a problem and I want to use Linear Regression. My brain takes a long time to process these mathematical equations and work out how to plug them in to excel but I will give it a go.

Here are a couple of links that I found interesting:-

http://tinyurl.com/59t7x

There is a free 350 page pdf book here. It is about biological data but I presume the methods can be applied to stock market data:-

http://www.curvefit.com/

 

[cd173]

Bigbusiness

Thanks for the links. Do you intend to use LRF? If you follow the link it gives a detailed breakdown for the ESmini with backtest results. From what I gather, LRF works more effectively over longer time frames.
Suppose it's down to personal preference!
I'll probably do some testing over the weekend and post the results

Cheers

 

[theknifemac]

Any reason for picking linear rather than non linear regression apart from the added complexity ?

Cheers

Stew

 

[Bigbusiness]

Don't know if I will use LFR. I just want to make some charts in excel and play around with them to see if there is anything I can use. I am too busy to do much at the moment, so it will have to wait.

 

[cd173]

Been playing around with LRF/price action and have found that a setting of SMA1 with regression period of 26 can produce some pretty reliable signals. Only tested it the last few months, but the results look promising. On choppy days it's awful, but on a 1min chart with a bit of range, such as the DAX, seems ok.
Needs more testing me thinks.

Any thoughts anyone?

C

 

[ronfalcone]

LRC - the best kept secret

I've been researching and presently using linear regression channels with some success.

A common mistake traders make, IMO, is to confuse linear regression channels (LRC) with fixed zones of support and resistance. They're not! They are simply dynamic, moving lines which create a 3 dimensional view of price deviation. Really not different, in concept, from Bollinger Bands, only that LRC's are straight lines and remain straight, creating a more cogent view of price movement.

I'm combining both LRC's, technical analysis and fundamental analysis to help me accurately plot correct deviations and expected probabilites of price movement back to the mean. I use Stock Fetcher to find stocks trading at channel lows or highs. So far, results have been impressive but I'm not finding a lot of good candidates (don't know if that's due to current market conditions, or is normal); nor am I making large profits (though my account at this time is less than 5k). We'll see.

By from the US of A.

Ron

 

[Pippppin]

it will be a beautiful irony...............markets, often viewed as random events, contain strong linear characteristics, and DLR could yet prove to be the best method for predicting price movements

 

[blackcab]

Pippp - as Mac the Knife asked above, have you done much work on non-linear regression, or any non-linear methods come to that? If not, is there a particular reason?

 

[Pippppin]

my research project has lasted for about 6 and a half years, and has concentrated exclusively on applying linear regression, rather than non-linear regression. Given the strong linear characteristics of recurring chart patterns, I have not felt any need to apply non-linear regression.

 

[blackcab]

Thanks. I hope it's all going well, and maybe see the results on here one day.

 

[Pippppin]

thanks blackcab.................i will keep you posted

 

[Grey1]

Regression line of either type can help trader to have a more mechanical approach to trading provided the following questions are objectively answered

1) At what point a change in the slope of RL is considered as change of trend
2) What time frames should be examined in identifying a change in LR slop ?

These two questions are kinda simple and crucial in trend spotting and those who have done enough research on them can honestly say they have an edge over other traders..

grey1

 

[Pippppin]

i am pleased to see other T2W members either using, or considering applying, linear regression to their trading analysis. The potential is highly significant

 

[Grey1]

i am pleased to see other T2W members either using, or considering applying, linear regression to their trading analysis. The potential is highly significant

Pippini ,,
As you are well aware RL is another useless indicator if the above two criteria's I mentioned above are not objectively defined.. Please correct me if your research has discovered some thing new ..

Grey1 does see any edge obtain solely by using LR to have a future price projection ..

Grey1

 

[Pippppin]

Grey1

I am aware that most existing trading software which offers a linear regression tool is extremely limited. My research takes the application of linear regression to financial markets much further, and applies to all timeframes and all markets....