Cavalry and chariots against infantry

[Domen]

Also, Ammianus, Vegetius, and Orosius all emphasize Gothic archers or missile-throwers.

Archer or missile throwers are good to "soften" enemy formation before close combat, but not deadly enough for armored troops and troops with large shields (such as Roman heavy infantry) to decide the outcome of battle. Also check the info about the battle of Milvian Bridge I posted above.

Macedonian infantry at Gaugamela had much less protection against arrows and other missile projectiles than Roman infantry at Adrianople, while Persians were relying on archers and skirmishers (both infantry and cavalry) to much greater extent than barbarians at Adrianople. Yet Macedonian casualties in the battle of Gaugamela were negligible - which testifies to low actual lethality of missiles.

Also, Ammianus has the battle begin with a Roman cavalry charge.

But later Roman cavalry escaped from the battlefield (and most of it survived the battle).

And when describing combats which took place long after the desertion of Roman cavalry from the battlefield, when Roman infantry was fighting alone, after describing dead infantry, he once again mentions "mounds of fallen horses" which "filled the plains with corpses" - so in my opinion this refers primarly to Gothic horses, much more of which died in this phase of the battle, than of Roman horses.

He seems to have the cavalry flee early in the battle, and again late in it.

There is nothing inconsistent in this. Some of escaping Roman cavalry could rally and return to the battlefield - and then flee once again. Ammianus wrote that escaping Romans "often rallied".

It's worth noting that both Roman cavalry commanders live and both Roman infantry commanders die, as does the emperor. It's likely that the cavalry got away and the infantry were trapped. If the Romans simply fled the field, then the Gothic cavalry could do the pursuing. If, as most interpretations assume the Roman infantry find themselves surrounded, then the Gothic infantry can also cut down the Roman infantry. It's like Cannae or Zama rather than Jena-Auerstadt.

But Ammianus also describes Gothic pursuit of escaping Roman infantry.

Apparently not all of Roman infantry was trapped in such a way that it couldn't escape.

It's worth noting that both Roman cavalry commanders live and both Roman infantry commanders die

Which confirms that Roman cavalry suffered relatively low casualties. And thus the "mounds of fallen horses" which "filled the plains with corpses" had to be Gothic - at least many of them.

And I don't think that skirmishing and keeping at bay can produce "mounds of fallen horses".

Nor does charging enemy cavalry which quickly escaped without much resistance produce them.

So Gothic cavalry had to charge and smash into enemy infantry, suffering losses in horses.