Portfolio model using history data

[liziyuann]

Hi all,
I am using WorldQuant Websim (websim.worldquant.com)to simulate portfolio using apha consisting of history data, for example, daily open, daily close, high, low, and etc. There are various functions. However, being a student majoring in Math, I lack the financial knowledge. Do you have any advise on where to find good models? Or would you mind introducing some to me?

Best regards.

 

[Alexandre Chretien]

I would start with Markowitz. Google it.

 

[liziyuann]

I would start with Markowitz. Google it.

Thank you.

 

[Ian Kaplan]

For my Masters project I looked at alpha predictors, using historical data. You can find this work here:

http://www.bearcave.com/finance/thesis_project/index.html

 

[Ken Abbott]

I posted a very simple QP model here a few years ago. I'll try to find it.

 

[liziyuann]

For my Masters project I looked at alpha predictors, using historical data. You can find this work here:

http://www.bearcave.com/finance/thesis_project/index.html

Really nice of you to let me know this. I would tell you if I get any wonderful ideas from your study. Thanks a million.

 

[liziyuann]

I posted a very simple QP model here a few years ago. I'll try to find it.

I would like to study your model if it can be found out. Thank you anyway

 

[diegosanaz]

@Ian Kaplan, I just gave a quick browse to your thesis. It reminded me of a paper I read on JoF. In that paper it shows that the predictability of a variable depends a lot on the underlying model. For instance, if a constant mean constant volatility model is assumed, is more likely that the variable will have much less predictability value as if you were to assume a stochastic volatility model.

The name of the paper is Sequential Learning, Predictability, and Optimal Portfolio Returns by Michael Johannes, Arthur Korteweg and Nicholas Polson.

I hope you find this useful in the case you seek to use this in a professional setting

 

[themarketprofessor]

Hi all,
I am using WorldQuant Websim (websim.worldquant.com)to simulate portfolio using apha consisting of history data, for example, daily open, daily close, high, low, and etc. There are various functions. However, being a student majoring in Math, I lack the financial knowledge. Do you have any advise on where to find good models? Or would you mind introducing some to me?

Best regards.

Have a look at articles related to optimal growth portfolio or the like.

 

[Ian Kaplan]

When you purchase an asset you do it in the belief that it will/may appreciate. This means that you are making some estimate of future return (even if the estimate is as simple as "I think that it will 'go up'"). Portfolio optimization takes your estimates for future return and the historical volatility (risk) and estimates the optimal portfolio in terms of estimated return and historical risk.

The hard part is estimating the future return. What ever method you choose, run back tests to see what the correlation is between your estimated future return and the actual return. As it turns out, mean return is an OK predictor, but it is frequently negatively correlated with future return (pick a bunch of stocks and try out the correlation).

 

[Vinayak Maheswaran]

I would like to point out a good resource to first gather this history data (i.e. tick data). It is called QuantGo and it is available on QuantGo.com. The data is cheaper than some of the resources available online. I would go right now and download the data from the site. It is available on the cloud.

 

[Vinayak Maheswaran]

When you purchase an asset you do it in the belief that it will/may appreciate. This means that you are making some estimate of future return (even if the estimate is as simple as "I think that it will 'go up'"). Portfolio optimization takes your estimates for future return and the historical volatility (risk) and estimates the optimal portfolio in terms of estimated return and historical risk.

The hard part is estimating the future return. What ever method you choose, run back tests to see what the correlation is between your estimated future return and the actual return. As it turns out, mean return is an OK predictor, but it is frequently negatively correlated with future return (pick a bunch of stocks and try out the correlation).

Sorry if this comes across as rude. But most of what you have said is freely available online. I would prefer if we could talk about 'new things' on forums like this. This is the reason that I come to QuantNet.