While doing some analytics last week I realized that there is a pure (not statistical) arbitrage situation in VXX options. By the time I noticed it was too small of an edge to try to trade manually so I did not do anything, and since this is something that is very unlikely to happen again I'll disclose all the details that came up in my analysis. All data obtained from Nanex(nXCore) - a very high quality datafeed. This feed has all messages from exchanges, although to create the charts below I took 1 second snapshots.
When VXX split in Nov 2010 OCC decided to leave the "old" unadjusted strikes, so one can trade both after-split VXX and before-split VXX which has quarter of its value. On August 30th some options were trading out of line with the others, I guess because some market-maker did not properly adjust volatility skews that day. Because options are so far OTM early exercise is not a consideration, and the mispricing is a pure arbitrage.
For example (pre-split) VXX 21 Jan 2012 Call should trade at 1/4 value of VXX 84 Jan 2012 Call, and they do most of the time. The first chart plots bid of VXX 21 Jan 2012 Call times 4, versus ask of VXX 84 Jan 2012 Call. Around lunchtime bid exceeded ask, and traders had an opportunity to capture as much as 12 cents by selling four 21-strike contacts and buying one 84-strike contract. Using conservative estimates of costs for institutional trader this would give 0.12-5*0.01=0.07 profit on 5*0.25 haircut (margin), or 5.6%. Given that at that time there was 143 trading days until expiration, this translates into 5.6 %* (143/365) = 14.3% annualized of riskless profit. The volume on the quotes was sufficient run the spread at least 25 times (0.07 * 100 * 25 = $175) by the most conservative estimate.
Similar opportunity presented itself between (pre-split) VXX 22 Jan 2012 Call and VXX 87 Jan 2012 Call, which is the highest strike available for regular contracts. Because 22*4 = 88 > 87, 22-strike contract bid multiplied by four should be smaller than ask of 87 strike contract. Here the spread also widened as much as 12 cents. These are of course not common opportunities, but provide an interesting lesson. There were obviously no low-latency arbitrageurs that brought the spread in-line as mispricing persisted for many minutes.
Thanks to Jason S. Williams of Front Run the Delta for his invaluable feedback on this post.