Individuals are able to harness predictions about the likelihood of needing to shift attention to adjust their shift readiness, known as attentional flexibility. However, the nature of these predictions remain poorly understood. In the current study, participants made saccadic eye movements among three rapid serial visual presentation (RSVP) streams of alphanumeric characters in response to embedded visual cues and made button presses in response to targets at the cued location. We manipulated the overall likelihood of receiving a shift cue, known as the list-wide shift probability, across alternating blocks of trials and shifting attention to one of the streams, referred to as the oddball location, was equally likely regardless of the block type. Participants demonstrated smaller target detection shift costs and faster saccade latencies when the list-wide shift probability was high than when the list-wide probability was low and were faster to initiate saccades from the standard location than from the oddball location, reflecting learned modulations in flexibility. Furthermore, in high list-wide shift likelihood blocks, participants were faster to shift attention to the opposing standard target than to the oddball target. However, latencies for shifts to the oddball location in high list-wide shift probability blocks were shorter than those in low list-wide shift probability blocks, demonstrating that attentional flexibility is not yoked to a particular anticipated target location. Our findings provide evidence that moment-by-moment changes in attentional flexibility are not limited to an expectation to shift to a single location, but rather reflect, in part, a location-independent state of control.