It kind of reminds me of people that think that we should use the exercise bikes in gyms to power the lighting; yes, brilliant, but actually run the numbers and you realize that humans output very very little energy even at peak exercise. The amount of waste on generators would be phenomenal.
There are some cases where pumped hydro makes sense, but there's limited scalability.
For example, 1000 kilograms of water (1 cubic meter) at the top of a 100 meter tower has a potential energy of about 0.272 kW·h (capable of raising the temperature of the same amount of water by only 0.23 Celsius = 0.42 Fahrenheit).
https://en.wikipedia.org/wiki/Pumped-storage_hydroelectricit...
Given the much smaller height of a rooftop tank, every 1000 kg of water pumped would support a number of minutes of normal household usage (a decent refrigerator would need several thousand kg per hour).
I'm sure compressed air in other forms would be a good store. You could also raise and lower weights into a mine, it doesn't matter so long as you can convert between electric energy and kinetic potential of some kind.
1000L tank on a roof which is ~4m high.
E = m * g * h = 1000 * 9.8 * 4 = ~40,000J
Which is about 12Wh not taking into account any losses.
You could maybe improve this by having the water fall into a well but still nowhere near what a battery can provide for a fraction of the cost/complexity.
A standard 100Ah battery will provide 1200Wh.