The "Spot Play" Rabbit Hole 20 years ago or so I was doing spot plays successfully. It takes a lot of research and some thinking outside the box. The payout is the primary consideration. ALWAYS track the payout as the prime consideration. You must find long-term profit.
I was working the tedious way. Take a basic premise. Does it have any hope at all? If so, try tightening it up until it was showing a profit. Then, does it continue to show a profit or does it regress to the mean? Most will regress. Divide into "sessions"... months or whatever. Does the spotplay show a profit in several consecutive sessions? Stay on top of it all the time.
I usually had about 20 spotplays under development. I'd use maybe 5 or 6.
I would stay away from maidens at first.
Some possible fertile ground you might try:
1) form cycles
2) pace extremes
3) claims... if you TRULY understand claiming
4) improvement
5) incorporate class
Don't hesitate to incorporate computations that are available "out there", such as Scott's PCR.