Hello--I'm just barging into this old thread I found by accident. I'm a Ford flathead V8 person, and have followed this trail on those engines... As noted above, any visual method is only accurate within about three degrees. I believe the water method is going to be essentially just another visual path...you still have to deal with that 3 degrees with no motion.
I was flabbergasted when I discovered this for myself many years ago, when I bought a dial indicator and discovered the broad results. I really did not believe my own eyes until I found a Smoley Yunick explanation decades later.
Here is a workable interference method for a Ford with plugs over the valves and no direct access: Take a fairly big tie-wrap, the sort of plastic belt thing used to bundle wiring. You want one that is maybe a foot long and has a hefty buckle, close to a quarter inch thick. I find mine at the base of a telephone pole near my house...lineman seems to drop more than he uses.
Slide the buckle in through the plug hole and get it up against the far cylinder wall...wiggle it a bit to try to find that farthest part of wall. Have a sober assistant hold it in place and rotate the engine til it stops against the buckle in both directions, marking each point on the pulley and taking the center point as TDC. Repeat the test several times to control against false results from accidentally moving the stop. Within the following limits, you have now located TDC and can mark any required degrees from that point.
The limits have to do with the offset crank...travel in each direction is slightly different in distance versus degrees, and so there is a slight error findable by math that is beyond me. The difference is trivial, I believe. It can be reduced by moving the stop point farther down the bore, but of course you cannot with head on. I believe this method as I have posted to be a satisfactory approach to accuracy.