Pretty accurate summary of the limitations of blindly applying Sharpe ratios. The Sortino ratio should have taken care of the issue you mentioned of counting upside volatility as bad. Additionally, you can certainly throw outliers out like that when calculating your ratios.
Yeah, I know the margin calculations are somewhat quirky with IBKR, but that sounded odd to me too. FWIW I couldn't replicate the suggestion with success in my account.
Yes -- I parsed his comment similarly to you, basically a diagonal spread, where he had the fortune of getting in for a credit and also getting out for a credit, due to the spread flipping in between him entering and exiting. That's great if you can get it to play out that way, but not typical IMHO, though of course at least a possibility with a diagonal spread. -- Starting out with a diagonal does give you, er, more options .. .. for management than putting on just a straight vertical spread, at the cost of collecting less premium initially. It's always a tradeoff of risk vs reward: take on less risk and/or gain more flexibility up front, then you will be paid/collect a smaller reward. Take on more risk up front and with less flexibility, then all else equal you should expect to collect a larger reward/premium. It's your job to manage that process, and ensure you always collect enough to make the risk worth your while, and ensure you have the flexibility you desire.
Whats interesting is "everyone" is buying calander type spreads at super jacked vols,assuming/hoping they can keep on rolling the shorts at crazy high vol... Never thought I would be buying calanders at with the long wing trading over 250 vol
Well, all of these metrics are subject to statistical significance. In fact, the best way to think about long-term performance is by multiplying both by sqrt(time) - that gives you rough t-stat of the strategy and that's your "staying power". For example, if you were running a 0.55 Sharpe for 1 year, that's a T-stat of 0.55 which is equivalent to random noise, but if you were running it for 100 years that's t-stat of 5.5 which is "very significant".
A high Sharpe ratio over a short period may not be as reliable as a modest Sharpe over a long period. Thanks for pointing this out. I was looking for a way to give time some weight. Looks like the t-stat sqrt(time) is a nice candidate.