Quote from makloda:
Premium prices skyrocket on big upward price swings. You may keep your bet size small to minimize potential losses, but in turn you limit your potential profits should your hopes for a 10 sigma event eventually come to fruition.
I'd love to see a theoretical backtest for your "risk 3% premium annually on OTM back month put premium" using a vanilla Black-Scholes approach in Excel on the Nasdaq 100 index during the period of 1995-2003. Let's assume it was painfully obvious tech stocks were a bubble in the summer of 1995. How would this approach have fared until 2003, well after the bubble burst?
Then, one could see how the "enter the put buying strategy only on breach of 200 SMA" would change results. I would assume the risk reward characteristics change massively to the positive side, since fat tail events are historically statistically more likely in down- rather than in uptrending markets.
Assuming the above two tests play out as I expect, this brings up the question: Does "knowing" something is a bubble help you (IMO, it is rather irrelevant)? Or is it enough to disregard if something is a bubble or not and one simply keeps speculating in the direction of the prevailing long-term trend?
Usually in stock indices, premium drops on rallies. In 2007 implied vol was at about 10-15 near the highs. For 2000 it was higher but not abnormally. I haven't looked at it for thing like Asia in 1995-97 or Japan 1988-1990, it would be worth investigating.
Regarding the MA filter, I haven't done a comparison but I would expect returns to downside momentum strategies like this to do much better than normal once bubble valuations/sentiment appear. Even if a MA approach 'works' normally, it should do better in bubble situations. Again, definitely worth a more comprehensive test.
I wouldn't say the NASDAQ in 1995 was a bubble, and the s&p definitely wasn't. I could see 1997 maybe, and 98-2000 definitely. A bubble doesn't mean 50 pe when growth is 50% per annum, or the S&P at 18 times forward pe estimates, that's just somewhat pricey. Bubble means values that can't remotely be justified even on optimistic forecasts.
Another filter I use is negative newsflow which results in bad price reaction, in the leading bubble sectors e.g major earnings misses, or downturns in leading indicators. It's not foolproof and is somewhat subjective but raises the odds IMO. In crashes I've seen, it always took sustained negative newsflow to really hammer prices down. In 2000 you had things like the MSFT antitrust trial loss, lockups expiring, fed hiking rates, and the big telcos in Europe making ludicrous 3g pice bids, with very negative market reaction. In 1995-99 by comparison, TMT company newsflow was good even during the 1998 mini-crisis. More recently, housing-related stocks started having bad news in late 06 and early 07, this spread to financials by late 07.
I am pretty sure that combining bubble values/sentiment, momentum indicators, and news flow can reduce the risk of being too early. Apply limited risk with either stops or puts and it should be promising. It's hard to backtest because some elements are discretionary, and the ones based purely on technicals aren't as good risk/reward IMO. But even something as simple as doubling size on downside momentum methods could potentially boost profits noticeably. Maybe it's not possible to build a quant system out of that, but it definitely is possible using discretionary judgement.
Another possibility is wait for significant extension above long term MAs befor buying puts. And potentially booking some profits on volatility/fear spikes.