This subject comes up fairly often. More than a decade ago forum member Timjet and I ran some real world experiments. My procedure was based on engine prop charts, which track almost perfectly to well established twin engine fuel burn over long runs at constant power. I believe Timjet used fuel flow meters.
Test proccedure. Ocean Alexander 44. Calm day, no wind. current, or waves (unusual day on Lake Michigan). Set GPS speed at 8.2 knots, twin engine. Record engine rpm. Shut down one engine (prop freewheeling). Increase power on remaining engine to achieve same GPS speed of 8.2 knots...steady track. Record rpm and rudder angle. Using the same speed eliminates one variable.
Look up fuel burn for both twin and single engine rpm as recorded (twin engine times two). Compare. I found about 5% savings running on one engine. I don't recall the rudder angle, but the Will Hamm autopilot did fine and appeared to have plenty of excess capability had the ambient conditions been more demanding. The OA has relatively large rudders.
Timjet saw larger fuel savings on his 36 Carver. I do not recall if his test procedures were the same.
For the Ocean Alexander 44 (48' LOA) and a loaded displacement of 30,000 lbs (13.64 tons) the theoretical hull speed is 8.5 knots, with a spec'd speed of 12/16 knots (not sure what the "12" is, single shaft?):
Brake horsepower required to achieve hull speed: 27.3
A total of 500 brake horsepower installed yields 36.7 horsepower/ton...
That's fairly significant... And now my head is starting to hurt and I need lunch...