Ernie Chan claims that statistical arbitrage is better than momentum like trend

An illustrative example just for reference is SWN vs CNX (U.S. utility stocks) during 2009.

 

but in 2008 drop CNX dropped much more, then they never get close to each other and keep that distance.

 

Yes, indeed. Cointegrated pairs often experience a price "jump" which is sometimes referred to as a regime change or decoupling. The portion of the price series with well behaved cointegration usually exists only for a finite period of time where you make your money. Then it breaks down and you take losses until you realize the price relationship has changed and you make adjustments or move on to something else.

I think your comment explains why you are not finding any pairs worthy of trading. You are looking for a strongly cointegrated pair with extended persistence over time. In other words, you are looking for a beautiful setup. If they existed, everyone would be trading them and probably arb out any edge very quickly. Stat arb is not a perfect trading method with perfect setups. It's messy. That's what the "stat" means ... there is a statistical probability distribution you are trading.

 

I see, then a short term trades with cost lose approach may be profitable in long term.

 

Long term profitability is more likely with a portfolio of diversified trades.

 

I just google trend vs arbitrage and this post is number 2 rrsult ha ha

 

Negative skew.

 

Chan lives in fantasyland. He can come up with endless strategies that simulate well but get killed in reality.

Stat arb works but Chan is far from knowing how to monetize it. Sitting on an incomplete or complete basket vs an ETF for weeks is going to be bad for most everyone. Holding XOM vs CVX and the like for weeks waiting for your convergence is going to lead to a blowup. Doesn't matter what your backtest shows in this case, you can never trust the performance because you'll need to size up to generate meaningful returns and then you'll hit a swan and it's over. Intraday opportunities are hypercompetitive among HFTs. It's harder than it's ever been and that's the way it should be.

The retail trader should forget about stat arb unless he is content to trade illiquid ETFs and make a couple hundred bucks a month.

 

Would you say that trading spreads, based on either stat arb (convergence) or relative strength (divergence), is generally not a sustainable strategy for retail traders?

 

I wouldn't recommend it. I think the retail trader is best off trading names "in play." Large volatility, lack of correlation, and ample liquidity (for retail purposes).