All HF relay systems were horizontally polarized.....

I wonder about the time of day and freq? I'd almost bet many were in
the daytime,
and using fairly high frequencies as far as HF. IE: 31,25,19 m, etc...
Seems the choice was as much a receiving/noise consideration rather
than absolute
signal strength.
I think the choice is much more complex than any theoretical gains seen
in modeling,
ect. In the daytime, I don't think it really matters much. So in that
case, it
would probably make sense to use horizontal to reduce local noise
pickup. That
would improve the receive s/n. As far as transmit strength, probably
not a whole
lot of difference either way.
But at night, it seems to be a different ballgame. I think the
differences in
propagation skew things towards the vertical on the low bands at night.

The farther the path, the better the advantage.
It could be stated that most horizontal wire antennas are lower to the
ground in
terms of wavelength on those bands. This is true. But you still have
cases where
people have tried the high antennas on the low bands, and still see the
verticals
usually win on long paths.
I've never tried it, but any interested could model my 36 ft high
dipole, and then
model my 10 ft center loaded mobile whip, on a ford truck.
I'd almost bet the dipole creams the mobile antenna in the model at low
angles
as far as the gain numbers shown.
But I know in the real world, that mobile beats the 36 ft high dipole
from Houston
to Jacksonville Fla at 2 AM. Yes, even I was surprised the first time I
saw it.
But I tried it over, and over again, and it was not a fluke of nature.
If you could have two 160/80/40 m antennas at 1 wave up, both with the
same exact
gain, IE: one a 1/2 wave vertical with any radials needed to equal the
ground loss
of a horizontal dipole, I'd bet money the vertical would win on long
paths 95% of
the time. It's not just a pure "gain" thing.... I think even verticals
with less gain
will win over the dipoles once the path becomes long enough. Note my
mobile...
I know for a fact from real life, if you are going to run a dipole, and
expect to
equal my 36 ft high ground plane, you better plant that puppy *WAY*
high, or you won't
have a chance. I'm talking over a 1/2 wave up. More like a full wave,
and even then
you might lose, once the path gets to about 4k or so...
BTW, these days in Houston, local noise has just as good a chance being
horizontal
as vertical...Most is powerline noise...So with my vertical, I never
really noticed
any extra noise. The s/n ratio was always better on the vertical, for
long haul.
IE: if the noise comes up 1 s unit, but the desired signal 2 s units,
the noise
is a non factor...Many times I saw no extra noise on the vertical.
MK