I've run the numbers and under the following assumptions, Horsemen and Panzers have the same expected damage output (assume for simplicity that the horseman's team is mono-element).
1) H-team has 6.5K total attack (a bit generous?).
2) The probability of rolling the correct element in a cell is independent of other cells, with probability 1/6 that a given cell will contain the correct element (this is almost certainly not how DE is coded, but gives an expected value consistent with testing).
Under these assumptions, the probability of success on, say, line 1 is roughly 7.3% when horseman is activated. If 21 lines are rolled over two turns, this gives an expected value of roughly 35,200 damage.
Some comments:
1) The standard deviation on this estimate is roughly 11K if one assumes that the lines are independent. Since they are in fact highly dependent, the true standard deviation is much higher.
2) If a team has multiple horsemen, this value does not increase, since it is calculated within the window of activation.
3) This equality of expectation only holds if both attack the same number of targets. Since a huge portion of the horseman's value comes from very high numbers, this is a realistic assumption.
4) The only circumstance under which H-team has a clear edge (in expectation) is in the presence of, e.g., Lilith. Whether this is a clear advantage in reality is highly debatable, as the horseman's attack is over two turns and is quite unreliable.
In conclusion, one must say that panzers are indeed far stronger than horsemen, and the only circumstance I can envisage in which H-team would have a commanding superiority would be in combination with Lilith against a raid boss with 5 million HP or so (not guild war). Claims that they are equal in the case of, say, guild wars is completely unsubstantiated, given both the generous assumptions needed to have equal expectation, and the diminishing returns of the GW-curve.
That said, those who claim that they have the same expected damage output are certainly not wrong, though perhaps that assertion should be qualified.