I agree with Steve that flash crash may not be the right phrase to use. That was a market-wide glitch. That said, it is legitimate to consider the many occassions when individual stocks do take big 20%-30% or so instantaneous whacks.
That’s an important point and in fact many might take it further ans suggest it’s a buying opportunity, for a quick and temporary bounce if nothing else. Realistically, though, this is one area of the market where conventional wisdom can be most dangerous. For one thing, this sort of advice is all over the place. “Ride your winners and sell your losers” and “Buy when others are fearful and sell when others are greedy” are both often repeated and are obviously contradictory. So the bottom line is that you have to address this as just one element of a coherent (and on p123, hopefully reasonably tested) strategy.
Diversification is an obvious measure (more than five holdings).
Another important consideration is the kinds of stocks your models are designed to attract. Some companies are more solid than others. Quick and steep price hits can strike in an impossible or almost-impossible to predict manner. More predictable are the underlying fundamental characteristics of the companies. So you can design your model such as to increase the probability that the stocks have superior bounce-back potential. Assuming I don’t trade margin (an accurate assumption – I don’t), I’m much more comfortable living for a while with a stock that suffered a big paper loss if I’m fine with the company’s essential characteristics than I would be if they were in the portfolio, for example, for no reason other than that size, liquidity and technical characteristics are similar to those that popped every now and then during my simulation test period.
Another very important but seldom discussed (on the forums) consideration is the rebalancing interval you choose. The coming together of Hollywood (which, for Wall Street type films, usually can’t generate exciting visuals through gun-fights or car chases so does this instead by having crowds of traders screaming, running around and gesturing wildly as their terminals flash things at warp speed), modern technology (that lets the average person do things faster than ever before), and garden-variety human impatience has combined to create a sense that faster rebalancing is better, almost more prudent, than longer rebalancing periods. Actually, though, don’t be so sure. Just because we now have the technology to transmit information in nanoseconds doesn’t mean we have the capacity to digest and properly assess it that quickly. The, too, some events simply are longer lasting. Hence many things take longer than one or five days to get reflected in stock prices. For me, for example, I’ve tended to focus on the four-week period. I do have an R2G I actually trade real money with (cherrypicking the blue chips) that’s done well with a five-day rebalancing. But that’s a rarity. Pretty much everything I’ve done has tested, and performed better, with four-week rebalancings (and I’ve taken and recovered from more big hits to individual positions than I can count). I can’t remember how my cherrypicking model worked with four weeks models and given my tendencies, am fully prepared to accept that its decent performance with a five-day rebalancing may be analogous to the chimp that accidentally types great poetry if you leave it in front of a keyboard long enough. Recent work I’ve done is actually suggesting I may want to lean the other way; for example change the analyst 4-week estimate revision formulas/factors I’ve been using into 13-week revisions. Bottom line: quick rebalalancings feel macho and used to be great at funding the college educations of the offspring of the account execs at the old-time full-service brokers, but for us, it really doesn’t give ideas much of a chance to play out nor does it give oddball market events much opportunity to start moving back toward normalcy.
As to stop-loss, which is based on one of those who-knows-if-it’s-really-right adages, I’ve never had any success using it to improve performance. We added it to p123 because we know there is demand for it. If you can make it work for you, fine. But don’t feel you need to use it just because it’s there.