I guess my point is that if I'm going to get rolled I'd rather it happen on 10 contracts with inherently less gamma risk rather than 150 OTM contracts just waiting to kill me. ATM premium is already providing more leeway for the additional risk of being wrong.
I'm no option expert, but do have some experience winning and losing money with them. Let's take AAPL JUN 130 straddle vs 115/145 (+- 12%) strangle as an example just keeping vol constant for sake of simplicity. Figures are taken from TOS analyzer and maybe a off by a few decimals (current underlying price: 128.77):
130/130 Straddle, 1 contract:
* 670$ of premium
* Initial margin using reg-t: 2585$
* B/E range 1 week before opex: 123.50/136.50
* B/E range 2 weeks after open: 124.90/135.02
* 5 pts beyond where it was sold at (123.77/133.77), 2 weeks after open: -70$/+62$
* 5 pts beyond where it was sold at (123.77/133.77), 4 weeks after open: +25$/+220$
* 10 pts beyond where it was sold at (118.77/138.77), 2 weeks after open: -481$/-262$
* 10 pts beyond where it was sold at (118.77/138.77), 4 weeks after open: -458$/-210$
* 5 pts past strikes, 4 weeks after open: -108$
* 5 pts past strikes, 4 weeks after open w/ +30% IV: -990$
* 20 pts past strikes, 4 weeks after open: -1335$
* 20 pts past strikes, 4 weeks after open w/ +30% IV: -1756$
* Capital needed to hedge with 100% of the underlying at B/E: ~12,350-13,650$
115/145 Strangle, 15 contracts:
* $690 of premium
* Initial margin using reg-t: 19345$
* B/E range 1 week before opex: 118.83/141.05
* B/E range 2 weeks after open: 123.50/136.74
* 5 pts beyond where it was sold at (123.77/133.77), 2 weeks after open: +46$/+394$
* 5 pts beyond where it was sold at (123.77/133.77), 4 weeks after open: +629$/+680$
* 10 pts beyond where it was sold at (118.77/138.77), 2 weeks after open: -1405$/-495$
* 10 pts beyond where it was sold at (118.77/138.77), 4 weeks after open: -9.82$/+454$
* 5 pts past strikes, 4 weeks after open: -450$
* 5 pts past strikes, 4 weeks after open w/ +30% IV: -8700$
* 20 pts past strikes, 4 weeks after open: -9462$
* 20 pts past strikes, 4 weeks after open w/ +30% IV: -18269$
* Capital needed to hedge with 100% of the underlying at B/E: ~178,245-212,550$
This is of course an over-simplification, but selling OTM and juicing up with quantity to meet that same 1 contract ATM premium seems like using a 1 tick PT with a 10 tick stop-loss in outright trading terms. You better be right 95% of the time (which you of course have a higher chance of being "right" with OTM). You can be right less with ATM but also be compensated more for the increased risk. On top of that you can hedge with the underlying a hell of a lot easier than you can with 15x the OTM contracts. You'll probably also suffer less hedging adjustment costs. In my mind, OTM premium selling relies on probability and statistical chance being the only thing saving you from a very bad loss. Additionally, if you want to hedge it constantly you've got to have much more capital to do it with as well.
Robert Morse's points about selling at the right time (taking advantage of volatility) are of course completely valid. One could also sell ATM legs during high volatility as separate legs during defined S/R areas using the increased window as more breathing room for when to put on/take off hedges (they could also do the same thing with OTM straddles). I think it's a given that hedging with the underlying could/would be used in some form or fashion with any of these strategies.
My main issue with OTM strategies are the horrible risk when wrong. Works until it doesn't, then it really turns out badly. This also seems to be the pragmatic results as seen during black swan events. Things like this come to mind:
http://www.surlytrader.com/gamma-hemorrhage/
and
http://www.surlytrader.com/volatility-selling-strategies/