OK, you asked for it!
There are four things at work here:
-
Pulling two power tubes does reduce power output because there are fewer “pistons” pushing the load.
-
This also creates an impedance mismatch. For example, four EL34s typically operate at 1750 ohms plate to plate. Two EL34s operate at roughly 3500 ohms plate to plate. That mismatch affects how much power the tubes deliver to the load and how perceived volume has changed. So, this is the first reason to reject the “3dB” reduction argument. That doesn’t happen.
-
A 100W output transformer, besides being wound to match 4 power tubes, also has sufficient mass to support a low frequency response at rated power and therefore will saturate less easily at full power under 50% of maximum load. In short, the perceived volume will be louder than that produced by a 50W OT with a 3500 ohm primary.
-
Because of the roughly 50% reduction of current draw on the power supply, the plate voltage on the remaining pair will be higher. The result will be a greater headroom and output power potential than an actual 50W power amp could produce.
Before someone jumps in and says “But you can correct the mismatch by changing the impedance!”, keep in mind that this only changes one of the four conditions described above and is not a perfect correction because that depends on the character of the actual load which varies with frequency.
In other words, you can calculate for any static condition you like, but you can’t ignore the reality of dynamic behavior - meaning the real world playing experience. Math is great for designing a circuit. It sucks when trying to account for what happens on stage when other loud musicians, cute girls and stiff drinks are introduced into the equation.