Disclaimer at both ends, OMG: the best money I've made has been trading this sort of thing, but I DON'T trade it at the moment, and won't until I can get stats I'm confident of from proper back testing that'll give me long term outcomes. Seat of the pants stuff like this is how people blow themselves up. This is NOT a system, I just wanted to show you a possible way gap-filling could be exploited if properly developed.
---
Sorry big post, becuase I've been thinking about this stuff a bit lately.
The way I've been tinkering with this gap trading on a day-trader basis (open to +1 hour or so) - feel free to smack me down, people who know more.
In my experience there are stocks that fill gaps more than others, and in this area I think it makes sense - if enough of the regulars "know" that the stock fills gaps, they'll be reluctant to play against it until it's filled.
So you can search for a stock that gaps-and-fills, get a percentage of times that it has filled a gap of X amount (some stocks rarely fill big gaps, some usually do) - on up days and down days - and get a number for the amount it moved with the gap after the open on those winning days before it turned and filled the gap.
Find a stock that fills gaps at a high percentage, preferably in a short time, and that tends to a low "with the gap" movement before filling, and then wait for it to gap enough to be profitable.
Then you can get an entry (either on the open or even just with the pre-open indicative) from the size of gap that you've calculated likely to be filled, and you've got your stop from the usual maximum movement from the open. Quite testable and sizable. Adjust %winners vs win/loss more towards your comfort that full optimisation (over optimisation being, of course, Certain Death, especially with such a lot of variables).
Further refine your exit by testing the time of day that a contrary move will tend to peak - it's been surprisingly consistent for the few stocks I've got working. Can also further refine entry / exit by waiting for a particular pattern or confirmation movement (eg enter after a simple move of x from the open in the direction of a gap-fill; exit with a trailing stop of y after the fill; etc). Anyway, you get the point.
I think the important part is that every stock is different, and some lend themselves to this much more than others. Crucially, I've found some stocks go into this mode of behaviour after they've had a bit of a panic - so a big fall or big jump. There are regulars (I've heard AAPL is mentioned for this stuff, though I've not looked at it - not in a position to day trade the US open), but a lot of the better targets seem to be "excited" stocks. They're just a lot more risky, because they'll probably stop soon.
So get those figures above on several timeframes. Shorter and more recent is more relevant, but much less reliable. Something that gives good results in all time frames are great, but there are also interesting opportunities on smaller timeframes if you want to chance them. That's where (even my rudimentary skills in) reading the DOM helps.
This is where I'm trying to get better at stuff like Amibroker. Gap-filling is an area where I think an individual stock's character is the important factor, not how patterns apply universally to all stocks.
---
PS: the best money I've made has been trading this sort of thing, but I DON'T trade it at the moment, and won't until I can get stats I'm confident of from proper back testing that'll give me long term outcomes. Seat of the pants stuff like this is how people blow themselves up. This is NOT a system, I just wanted to show you a possible way gap-filling could be exploited if properly developed.