General Trading Chat

[timepass]

Ok, plain simple old maths. Let's see how the new approach helps us in trading.

Yes, correct it is just the simple old maths...statistics has been around for a long time now and many of the formulas we use these days are based on equations defined long back. Only recently this field is becoming popular with fancy terms like Data science, Data Scientists, Machine Learning, Artificial Intelligence etc...

The reason for popularity now is due to the huge amount of data we generate these days and the computer processing power that is becoming cheaper day by day. As more and more people are working on it we see new approaches and technologies being developed based on it such as deep learning, AI applications and many research happening in different domains like medical, finance, hospitality etc

These days you may be hearing data is money and AI is the future, that is because with so much big data we can derive data driven prediction formulas. i.e we have the advantages of make the law of large numbers/data samples to build our models with higher accuracy which was not possible earlier.

 

[DanPSup]

Ok, plain simple old maths. Let's see how the new approach helps us in trading.

What about: There is no new approach, as the machines already do the most fastes work when it comes to bid and ask differences in any market. But this will not have to much affect for real strategy traders.

 

[Arindam Bhattacharjee]

What about: There is no new approach, as the machines already do the most fastes work when it comes to bid and ask differences in any market. But this will not have to much affect for real strategy traders.

Guys

Well what was explained about regression was right and this is what I am doing....
First for NIFTY - I look at the following predictor's trend of last 45 days.
OIL Prices, DAX, Dollar Value, NASDAQ, S&P, UKX, Euro, Pound, Yen...

Then I make a Co-Variance Matrix out of the 45 X 9 set of data I have from 45 days and 9 sets of predictors but subtracted by their mahalonobis distance of the latest index.
And then I build an inverse co-variance matrix out of the results which I will multiply later with the Matrix made up of Mahalanobis distances which is another 45 X 9 matrix

Then I multiply each row of the matrix with the inverse covariance matrix I had built and this gives me another Matrix. In this Matrix I multipy again each column of the matrix with the transpose of matrix made of Mahalanobis distances.

After this - the value I get is the values of X1, X2 ,X3 required to build the regression equation.
Then I sort it as rank in ascending order and then assign weights to each parameter...
Then multiplying by weights give me y = w1X1 + w2X2 + w3X3 + w4X4 + w5X5 equation

So aggressive is when I consider top 5 most weighted results - it usually shows the trend where the market might go...so this is where we get y = w1X1 + w2X2 + w3X3 + w4X4 + w5X5

Conservative is when I consider 10 days...
and Stable is when I sum it over 45 days....

Maths itself is tricky, doing it in real time is crazy programming

 

[ncube]

Ok, plain simple old maths. Let's see how the new approach helps us in trading.

Yes, Even I am curious to see the performance as this topic interests me and I work on these technologies. This is a vast domain and it is always good to learn new approaches/advances and gain perspectives from experts in this domain.

 

[timepass]

What about: There is no new approach, as the machines already do the most fastes work when it comes to bid and ask differences in any market. But this will not have to much affect for real strategy traders.

New to me

 

[DanPSup]

Guys

Well what was explained about regression was right and this is what I am doing....
First for NIFTY - I look at the following predictor's trend of last 45 days.
OIL Prices, DAX, Dollar Value, NASDAQ, S&P, UKX, Euro, Pound, Yen...

Then I make a Co-Variance Matrix out of the 45 X 9 set of data I have from 45 days and 9 sets of predictors but subtracted by their mahalonobis distance of the latest index.
And then I build an inverse co-variance matrix out of the results which I will multiply later with the Matrix made up of Mahalanobis distances which is another 45 X 9 matrix

Then I multiply each row of the matrix with the inverse covariance matrix I had built and this gives me another Matrix. In this Matrix I multipy again each column of the matrix with the transpose of matrix made of Mahalanobis distances.

After this - the value I get is the values of X1, X2 ,X3 required to build the regression equation.
Then I sort it by descending order as rank in ascending order and then assign weights to each parameter...
Then multiplying by weights give me y = w1X1 + w2X2 + w3X3 + w4X4 + w5X5 equation

So aggressive is when I consider top 5 most weighted results - it usually shows the trend where the market might go...so this is where we get y = w1X1 + w2X2 + w3X3 + w4X4 + w5X5

Conservative is when I consider 10 days...
and Stable is when I sum it over 45 days....

Maths itself is tricky, doing it in real time is crazy programming

Did you ever compare your results with existing old proved math models like Stdv, Pivot Points and others?

 

[Arindam Bhattacharjee]

Did you ever compare your results with existing old proved math models like Stdv, Pivot Points and others?

See I am not a trader and just built this model as a hobby.
My real work is in Weather within Climate Predictions where I use KNN and NHMM techniques extensively. So I have done high resolution time series forecasts for monsoon at district scale and total rainfall forecasts. To understand this - we need to understand the physical equations that govern the atmospheric dynamics...

So I tried to implement this with stocks - and found that it has been predicting well...But never had the money to invest - so never looked into other options...Open for suggestions

 

[DanPSup]

See I am not a trader and just built this model as a hobby.
My real work is in Weather within Climate Predictions where I use KNN and NHMM techniques extensively. So I have done high resolution time series forecasts for monsoon at district scale and total rainfall forecasts. To understand this - we need to understand the physical equations that govern the atmospheric dynamics...

So I tried to implement this with stocks - and found that it has been predicting well...But never had the money to invest - so never looked into other options...Open for suggestions

Ok. I understand so far. But prediction weather is also a game of probabilitys and specialy you which comes from those field will be aware about this as all weather predictions are made on such old data bases which are on probabiliy and not on "It will be like this for 100%". Any way: Loved to talk to you and I wish you all the best. Kindly move on with your ideas to expand traders knowledge in any way. My respect you have at any time.

 

[Arindam Bhattacharjee]

Also doing tercile is difficult - because this is not probability forecasts....So let me do an exercise with you to understand if this makes sense....I will give you five stocks which are throwing up now as top options for tomorrow - which one would you choose based on prediction...


Stock Rank,Sharpe,LastVal,Close,Aggressive Forecast, Conservative Foeracst,Stable Forecast, P/E ,Volatility, Short if Market Up, Short if Market Down
1 Sadbhav Engineering 3.10, 0.03, 186, 178 , 219, 215, 232, 150.95, 2.72, 232, 215
2 Aditya Birla Fashion 2.69, 1.61, 203, 202, 209, 204, 211, 48.93, 2.22, 211, 204
3 Praj Industries 2.69, 1.38, 129, 126, 132, 130, 136, 34.62, 3.39, 136, 130
4 Bata India 2.50, 2.13, 1360, 1357, 1405, 1362, 1406 , , 1.80 , 1405, 1362
5 BEL 2.47, 0.38, 103.3, 103, 110, 106.5 , 110, 13.34, 2.79, 110, 106.5

Now these stocks will behave differently tomorrow - but are best stocks now from 160 odd stocks I run....
So usually the stocks makes a run from Down to Up or Up to Down
Sadbhav has gone from 178 to 186 today - but it can't reach that point - and many times traders will bring it down purposely.
So you use Graph Theories when any of the state is reached - and can maximize profits between these states...

Now the question is - which stock will you choose if I give you this data on run time...
Remember you are also doing day trading - and you have to manage these five different behaviors of stocks....How will you maximize your profit as a trader...

 

[Arindam Bhattacharjee]

Ok. I understand so far. But prediction weather is also a game of probabilitys and specialy you which comes from those field will be aware about this as all weather predictions are made on such old data bases which are on probabiliy and not on "It will be like this for 100%". Any way: Loved to talk to you and I wish you all the best. Kindly move on with your ideas to expand traders knowledge in any way. My respect you have at any time.

Thats not true - predictions are made on cutting edge datasets....and probabistic forecasts are done mostly in datasets which show Gausian figure - then you can make tercile predictions....
Stock behavior is Gausian on a long run - but it is completely biased in short terms....So you can't have tercile probablistic forecasts - when you are talking of different states. The occurance of these states is not 33% each....They are dependent on Nifty's behavior...