Year-to-date, before the decimation that the Knight x1000 Algo wreaked upon the market, volumes had trended lower YoY but had not cratered. As the charts below suggest - in somewhat stunning technicolor - that since the Knight-algo was put-to-death, NYSE volumes have coincidentally plunged by 40%! Today's run-rate with an hour to go was the lowest of the year. For those that hang on the consideration that this is due to high-priced stocks and USD-volumes are stable - err, wrong answer - futures volumes cracked in half also (and that is a stable USD volume); The summer doldrums explains it - err, wrong answer - we are 20 percentage points below a normal summer-drop-off. The simple fact of the matter is, with retail suddenly the smart-money and exiting stage left (unable to trade this ridiculous market), it seems that losing one market-maker algo has almost halved trading volumes; what happens if GETCO ever goes down?

NYSE Volume is down 40% since the Knight Algo Slaying relative to the year-to-date prior...

[5]

and S&P 500 e-mini futures volume (lower pane) has cliff-dived (as has average range per day - middle pane) coincidentally as Knight's Algo died...

[6]

We guess the proof will come soon as the next few weeks of events are sure to encourage more than just the algo-bots to battle it out as manager's 'years' are made or lost on a red-flashing-Bloomberg headline...

Uncoinvinced that the market is entirely run by algos - the following stunning drill-down by Nanex [7] should dispell any mystique...

Below One Second

The SEC has stated on more than one occasion that it isn't worth the effort to look at data in resolutions under one second, because it's just noise. Follow us "down the rabbit hole" as we drill down from 1 second to 1/1000th of a second and see what we find. Note: more examples on this page [8].

The charts below drill down from 1 second intervals to 1 millisecond intervals in the stock symbol FXS on the morning of August 30, 2012. Each chart shows the National Best Bid/Offer spread (NBBO) as gray shading. The top of the shading is the best ask, and the bottom is the best bid. Changes to the NBBO are shown as triangles color coded by exchange. Triangles pointing up represent best bids, triangles pointing down represent best offers.

1. FXS - 1 second interval, showing about 17 minutes of data.
Others might think the column of color in the middle is just noise, or a glitch, or fat-finger. Let's look closer and see.

2. FXS - 25 millisecond interval, showing about 26 seconds of data.
As we drill down, what looked like single quotes (triangles) above, now appears complex and involves multiple exchanges (multiple colors).

3. FXS - 5 millisecond interval, showing about 5 seconds of data.
At this resolution, note the white spaces between the gray lines. This indicates that the NBBO isn't stable, it actually flutters quite rapidly at many different prices. We couldn't see this fluttering on a resolution as fine as 1/40th of a second (25 millisecond). Note how many different exchanges are involved . This might be the work of one HFT shop plugged into at least 4 different exchanges.

4. FXS - 2 millisecond interval, showing about 2 seconds of data.
A pattern starts to emerge. This is anything but noise. 

5. FXS - 1 millisecond interval, showing about 1 second of data.
The pattern looks very familiar. This appears to be an algo testing out its ability to rapidly back away from the market from multiple exchanges when a large, informed order appears. Based on this chart, it looks like HFT is able to withdraw from multiple markets in a few milliseconds.

Charts: Nanex and Bloomberg