Try
percentage = ((float) value / ((float) gAVG * 2.0))*100.0;
This will take more time and code space for the floating point emulation, but it'll work.
You could also try
percentage = (value * 100) / (gAVG * 2);
which is mathematically equivalent to your first expression but reorders the operations in such a way to avoid the problem you see. It will still have integer rounding issues, but it avoids the computational penalty for floating point calculations.
-j