>>11088794If the effect of the action is seen later, say a time equivalent to ten or so latencies, then the intended timing of the action could be estimated from subsequent actions and the model could be corrected to reflect a different timing than recorded from the user, but it would still be closer to what the user meant. This would only correct the inaccuracy of the users reflexes, nothing really would happen before the user initiated the action, but it would seem to the user, who gets information about the consequences of their actions significantly later, that the timing had been more on point.
