Quantitative Arbitrage in its most literal application always struck me as pairs trading devoid of common sense. I always visualized the Quant handing the trader a report which proclaims, for conversation's sake, a very high correlation between General Mills and Intel. And the trader says thanks. And he continues to make jokes about the Quant and he continues to pair General Mills with ADM, and Intel with QQQ.
In equities alone there are 7000x7000 possible pairs, roughly. Thats a large number, 49 million, and especially with the short term risk monitoring (trading as I like to call it) available to all of us, can you truly say that its dead? Maybe you haven't check for a pulse in the right places.
pairs trading works in markets that aren't going anywhere. in momentum driven markets, the return is diminished because you get killed on one side of the pair, make a little on the other, but you have twice the commission expense. why not just scalp on the RIGHT side instead. when the market is going up, I'd rather just scan and buy strong stocks at opportune times than being a smartass, pairtrading and shorting the strong, buying the weak and getting my ass handed to me. It just doesn't make the best sense anymore. Last year, yeah, it worked, this year, I think scalping is a more viable strategy. also, every tom, dick and harry knows how to pair trade........
This is a very interesting discussion here about how difficult the math for pairs trading. And what is better â correlation or cointegration? But what are we looking for here, anyway? Neither â correlation nor cointegration could give you entry point or exit point. As a result from applying these functions to the stocks you just get a list of suspected pairs of stocks. And then you apply some other functions or indicators to that list to determine if one stock is oversold compared to the other one. Like MACD or ADX or RSI or Stochastics or whatever else you like and pleased. So, what is this suspected list of pairs we got? Isnât it just a list of pairs that during some period of time behave similar under some (different) market conditions. Why are we looking for such pairs that most of the time behave similar? Just to be market neutral or am I missing something? Just to be protected against market moves. So, to me, whatever function is easier will do. Unfortunately some simple scans (like stocks in the same industry or sector) do not work, because you can see a lot of companies in the same industry with negative correlation.
I am looking into MSFE programs, but I'm wondering if this whole concept of Fin Engineering is a bunch of hocus pocus, as used in propietary trading. It is so heavily invested in by the best and the brightest in finance in the world (namely the Wall St big firms) - that their guys are going to invent ways to "beat the markets", using complicated mathematics and statistics, etc.. Until you can quantify human behavior, this doesn't make alot of sense. We still have no damn clue how the brain works, same with decision making and that IS the financial markets. I sense what is going on is the scientific types have convinced the business types that this is necessary, like the tech people in the 90s convinced business leaders they needed an Intranet portal or they'd be out of business. Nonsense!