HFT's Flash Crash part 2

I have to admit, I know people who work on black box systems and the main concern always seems to be latency. They were always upgrading and adding features to get the fastest executions possible.

I also have to admit that I don't know how long it takes or how many pings are needed to figure out if a an iceberg (a large order that you are only seeing the tip of through 200-300 share transactions) is likely. It may be 3C is picking up on distribution before the stock catches the attention of the HFTs, I just don't know, what I do know is with this exceptionally fast version of 3c, literally order to order, it seems to be identifying distribution before the HFTs cause a flash crash.

Again, AAPL and LQD on 1 minute time frames.

 I wouldn't think that the action in AAPL above qualifies as a flash crash, but it certainly is a pretty decent drop pretty quickly. In both cases, 3C FAST VERSION picked up distribution, both into the gaps higher (see orange arrows)

Above LQD is showing about a day of distribution before the next morning's flash crash. The continued negative stance of 3C, suggests to me (at this very early stage) that there's more downside coming on LQD.

Unfortunately I have only about 2.5 days of data on a 1 second time frame and PGN's flash crash was longer ago, so here's a 10 min chart, again showing a relative negative divergence before the occurrence.

I don't think we can identify flash crash stocks, but once we find the stock, we may be able to identify the course of action that comes afterward to develop an edge in this phenomena of HFT induced flash crashes. So anytime you hear of one, notify me immediately and we'll see how it behaves afterwards on a 1 min chart and see if there's a pattern that develops. To me, uncovering icebergs alone tells us something worthwhile, such as someone has a position in AAPL they want to exit. We may not be able to catch the flash, but we may be able to catch the resulting crash.