I'm hoping someone who has access to super fast hardware can answer this question, because with my technology I can't tell what's happening. I tend to put my exits around a fair value and move a passive order around the price as needed. Occasionally (maybe 15% of the time) my feeds and infrastructure are too slow to cancel and move the order to a better price, so I end up getting filled and taking less out of the trade than I would like. I noticed that this generally happens whenever someone does a massive volume spike, usually through some sweep order. When I look into my low resolution data, I can see that the volume spike happens in a matter of just milliseconds (the best I can do with my setup.) But I see these trades spaced out by milliseconds on the consolidated tape. So the question: At the microsecond time-scale, is it possible for market makers to see the sweep order (tagged as 'F' or ISO) on one exchange and cancel before the liquidity seeking algorithm hits the other exchange? Or are these block trades so close in time that market makers can't avoid them? And are they spaced out by a few milliseconds on the consolidated tape because of reporting delays to the tape, or are they really a few milliseconds apart? I can't tell with my data, because the precision is just too crappy. If I boil down the question even further, I guess I'm trying to figure out if market makers can really avoid IOC sweeps targetting every exchange by observing a trade on another one. I'm interested in the sweep mechanics in particular, and whether I really need to start investing in even better technology than I have as I progress further and further into my trade.
I believe you have the right idea with this question. I think it is safe bet that this stuff shows up micro-seconds apart simply because of reporting to the consolidated tape and that no one could see an ISO that was sent at the same or nearly the same time on one exchange and cancel their orders on the other exchange unless there was something really different about how the ISO sender sent it to one vs the other. It is possible though. Keep in mind also, if you had the raw exchange data you still can't compare times on that scale because you don't know that NASDAQ and ARCA have synchronous clocks. People that dive into this sub-micro second stuff have to keep in mind the source of all the data and who is recording the time.