My understanding is that many algorithms float "test trades" to probe the other market participants (human and machine). It would be very costly for the algori

My understanding is that many algorithms float "test trades" to probe the other market participants (human and machine). It would be very costly for the algorithms to risk too much money on probing at an inaccurate price, so chances are the "hot potato" was not very deep at all. Nonetheless, your point stands that it could be an artifact of the traders expecting human-like decisions to underly the phenomena they saw. (it does seem implausible to me, but oh well)