Hi all,Factor-based EMN stat-arb funds are of mysterious to me.I am trying to guess what they do.How do I do market neutral quant equity based on multi-factor models?Taking FF model as an example, if I regress stock returns on the FF factorsr = beta1 * X1 + beta2*X2 + beta3*X3 + epsilonI then long stocks with positive epsilon and short stocks with negative epsilon.Am I right in understanding the quant funds secret sources?My questions are:1. What are the differences between alpha factors and risk factors? (I saw in risk models it's the same multi-factor model).2. If we regress out all those alpha-factors in the above regression, what actually are we trading based on?3. How do we generally find more alpha-factors and risk factors?Thanks!

you are not regressing anything out.. the equation tells you what the forecast return for a stock - you should be able to construct a portfolio that has no factor exposures and positive alpha. then you leverage that 10 x because the alpha will be too small to make an impression on your investors.

How do you select the factors?You select the factors such that the cumulative factor returns are good, am I right?If the factors are good for returns, why would we want to be neutral to these factors?

QuoteOriginally posted by: mizhaelHow do you select the factors? that's the art of itQuote You select the factors such that the cumulative factor returns are good, am I right? the factors do not affect the expected returns but simply explain them Quote If the factors are good for returns, why would we want to be neutral to these factors?If you want to get exposure to systematic factors then its quite easy to do so. what you are trying to do is to generate alpha.

QuoteOriginally posted by: mizhaelHi all,Factor-based EMN stat-arb funds are of mysterious to me.I am trying to guess what they do.How do I do market neutral quant equity based on multi-factor models?Taking FF model as an example, if I regress stock returns on the FF factorsr = beta1 * X1 + beta2*X2 + beta3*X3 + epsilonI then long stocks with positive epsilon and short stocks with negative epsilon.Am I right in understanding the quant funds secret sources?My questions are:1. What are the differences between alpha factors and risk factors? (I saw in risk models it's the same multi-factor model).2. If we regress out all those alpha-factors in the above regression, what actually are we trading based on?3. How do we generally find more alpha-factors and risk factors?Thanks!I dont follow you 100% but,1. The difference is that if you call something 'alpha' you are implying that your understanding, analysis of and information about that factor is superior to that of the market. i.e. you are implying that you have an 'edge' in that factor. A risk factor is a factor that can cause variation in your returns but where that variation is in line with the broader market. In reality of course, 'alpha' factors also have risk associated with them but the idea is that the profile of your risk in alpha factors is not identical to that of the market.2. You are not 'regressing out', you are building a model of the returns of the security. Eg. in the simple case, CAPM, we simply model the returns of a security as directly dependent on the returns of the market. 3. Risk factors are quite easy to find since they are characteristic of the market. One way is statisitcal analysis of market data (e.g. PCA). Another way is regression of market returns against ecomonic or fundamental factors, eg. interest rates, P/E. Finding alpha factors is of course more difficult! This is the same as finding an 'edge'. In practice it is made more complicated by the ease with which beta can appear to be alpha. i.e. One can create distortions of market beta (through sophisticated strategies and instruments) that can look very much like alpha, until it doesn't and then one realises that it was really just a distortion/misunderstaning of the actual risk.True alpha normally has a very clear and compelling cause. It is of course hard to recognise these causes when others in the market are not.You might want to neutral to certain risk factors because risk factors will only mimic the market. If you want 'absolute return' you have to neutralise risk factors that are not idiosyncratic to your edge (to make the strategy market neutral) and expose your returns only to your edge. Sometimes being neutral these risk factors can lead to an underperformance of the market, but the idea would be that over time the absolute return stategy is superior (due to compounding, control of drawdowns etc...)This ideal is extremely hard, if not impossible, to achieve however (especially in equities).You may have heard to 'smart-beta' (its trendy right now). The idea here is not to provide 'alpha' (as in something that is not already encoded in the broader market) but rather to customise the market risk profile or overweight certain systematic factors that provide excess returns over and above the market-weighted return. The source of these returns are the so called 'anomalies', eg. small cap premium, low volatilty premium etc...