Thanks for the advice Doobs. The real heart of my question is what are some modeling approaches people actually use to form these opinions? I guess I started the thread because I am trying to narrow the modeling universe somewhat -- I feel like I have too many choices based on all the reading I've done. In this thread and others, I see people talking about simulation approaches, maybe some regression based forecasting, and using structural inconsistencies in the market to look for opportunities (e.g., regulatory arbitrage, overpriced tail risk, etc.). I am somewhat surprised, but it seems like the conclusion of this thread is that basically no one who actually trades their own account for profit uses much in the way of econometric time series methods (not counting hedge funds, MM's, prop desks and the like). I guess that's all for the academics!
Time series methods are occasionally useful as initial inputs into your decision process. Specifically, you can generate signals using time series methods that could tell you whether something is historically rich or cheap. However, that can only serve as the starting point for further in-depth analysis.
Gotcha. And then the simulations begin in earnest! Martin, another question for you about sims. I am pretty familiar with running simulations in general, but what about the specifics here? I assume you'd want to run some scenarios over the range of (per time period) historical vols, and probably include some outlier stuff also. What about news/events/jumps? Is there a good way to include those in the simulation? My guess would be you could sample from past ones and fit some sort of (possibly time/vol varying) Poisson process to it...
Yeah, you want to perform "stress" simulations, as well as go through the more probable scenarios. As to news/events, the only way I know how to incorporate those is by looking at the past reactions to such events. Once you have that, you really don't need more simulations. It's just offers you a measure of stress, which then allows you to tailor your size/leverage.
Cool. Thanks. To extrapolate this a little bit, it seems important to recognize when going to the trouble of formal modeling actually generates any insight. In the case of news, it modeling effort might just be wasted.
What are your thoughts on doing hedged calendar trades on cointegrated ag futures, thru FOPs? Would that negate the gamma risk? Or does it become a wash? I am hoping you can have volatility and mean reversion work for you.
I am not sure 100% follow the trade you are proposing, so I am not sure I get what you are trying to accomplish. Could you give a detailed example? I like the idea of trading vol against terminal distribution using mean reverting underlyings... but I think mean reversion may explain why markets are thin in ag options. There's also the federal crop insurance program removing any demand of farmers for ag puts. As a result any edge you have would have to be large enough to overcome substantial transaction costs. For me, personally, there is also the more practical issue that I am not working with size where I could trade futures and fopts comfortably, even without the transaction costs. I am guessing this is why you brought up spreads (lower margin requirement), but like I said above, I am not sure I understand the angle on the trade you proposed.