CME’s Globex oil-trading platform failed due to technical issues, and trading was halted shortly after 2:00pm yesterday. The CME was mum about the incident, and would not comment regarding the glitch or the shutdown.
However, Nanex’s Eric Hunsader quickly got to the root of the issue:
On February 13, 2012, starting 13:59:57, quotes for crude oil began queuing. At 14:00:35, all of the queued quotes were sent at once. Again at 14:01:08 the same 38 second block of quotes sent earlier was sent again — old timestamps and all plus a few new quotes. Again at 14:01:18, all quotes since 13:59:57 were sent again. This repeated 12 times.
From a programmer’s perspective, it looks like a system problem caused a blast of quotes that corrupted a memory queue causing the software to believe the queue was full all the time. The same block of quotes just kept getting transferred by the system. Then I think someone pulled the plug.
An HFT algo blasted quotes into the CME, and then repeated the same blast 12 times. Boom. CME down. But in case you thought that was the end of the story, Erik would like to direct you to the resulting action in the USO ETF, which at the same time displayed a massive spike in quotes and trades, as HFT’s piled into USO to manage risk.
This was a quiet Monday in general on Wall Street, and we are thankful for that. You see, this CME tech glitch, and shut down, and the transfer of “risk” over to an ETF/ETN- well it reminds us of May 6th, 2010, where something went awry in the futures, and spilled over to ETF’s and then individual stocks. We are thankful indeed that it was a low-volume slow day, because if this happened during a Greek riot or worse, we would fear another major price crash based not on market fundamentals, but on market microstructure.
Automated quote-generating HFT has an element of systemic risk that cannot be ignored. Glitches will always happen, and that must be an assumption by regulators and exchange operators. The question is how much do we really want to rely on “liquidity” and pricing from market participants who can only provide it if they have uninterrupted and immediate data? Too frequently, relying on these participants is like Charlie Brown relying on Lucy not yanking the football away at the last minute.