It's a tradeoff, more variance vs lower transaction costs. Agreed, 3- 4 is extreme, but transaction costs depend on each participant's infrastructure, so who knows?
Sure but how to execute that as a home gamer? Lets say I'm trading an ETF like EWP. Two stocks make up 40 percent of its holdings, both liquid. The top 10 make up 75 percent of holdings. So if I create a basket out of those 10, then odds are the etf and basket are cointegrated. Beggars can't be choosers and all that...
Well, index arbs for example don't hedge with the entire S&P500. But I have never heard of anyone doing less than 50. In the example he gave which doesn't sound like index arb, that theory says 25 symbols are required to capture all the statistics. Sure less is better if the risk tolerance is kept well within bounds. But practice almost always rules that out.
Yep. The way to do it is to keep your risk very small (trade fractional shares), work at a job, and continuously add to your account size by putting money into it. Forcing a trade to pay bills is a sure way to the poor house.
What if you could figure out a way to leg into a position keeping risk tolerable, at different times adding/removing the legs? Lead/Lag allows that. Trading ETF components against the ETF is a tough game, imo. I tried something that I thought would be a sure money maker, the ETF options against the ETF components options. Couldn't make it work, at least not to the risk tolerance allowed. I also tried a model of S&P components against the ES, a sort of very poor mans slow index arb. Couldn't make it work (although, I am excited to bring that model back into a far more sophisticated version). Trading is as much strategy as tactics. All retail traders are tactical.
Wow...I never thought of that. Goes back to what you were saying initially about the trading firm. I agree about trading ETF components. If I were to do it, I'd try and do it in Europe or Asia, like with the spanish ETF.
I don't think a modern-day trader should be allowed anywhere near a buy/sell button without at least knowing one high level programming language, and at least a year of Calculus and a year of Prob/Stats and a course in Psychology. In other words, an undergraduate degree.
So, i am looking at a lot of lead/lag correlation data. Etfs vs. Etfs and spreads of Etfs vs. spreads of ETfs. Trying out cross correlation data and standard correlation data. Basically i am looking for significant (6 of 10?) where for example , a risk spread will follow another risk spread, lag 1 to several days. Probably a good thing to know. It will be educational if nothing else. I am challenged as well regarding the math so.....