We buit a program for use by our clients that looked for the bid/ask spread to widen to at least 3 ticks (ask- bid = 3 ticks). Upon discovery of such a spread it would then submit 2 orders, one on the bid+1 tick and one on the ask-1 tick becoming the new inside market. Feedback has been: Trades are often too fast see. Overwhelmingly most trades are fast and profitable. When only one side is taken losses can be large. Managing such a trade is like riding a bucking bronco, but if you know how to do it it can be quite rewading.
This sounds pretty wreckless to me, depending on the market. Do you have any backtested performance figures on it?