2 HFTs in a game of Chicken
- Thread starter laserwolf
- Start date
to quote
<i>Panel A of Table 2 is an excerpt from the message file for ticker symbol ADCT
on October 2, 2007 beginning at 09:51:57.849 and ending at 09:53:04.012 (roughly 66
seconds). Over this period, there were 35 submissions (and 35 cancels) of orders to buy
100 shares, and 32 submissions (and 32 cancels) of orders to buy 300 shares. The pricing
of the orders caused the bid quote to rapidly oscillate between $20.04 and $20.05. The
difference in order sizes and the brief intervals between cancelations and submissions
suggest that the traffic is being generated by algorithms that seem to respond to each
other.14</i>
<i>Panel A of Table 2 is an excerpt from the message file for ticker symbol ADCT
on October 2, 2007 beginning at 09:51:57.849 and ending at 09:53:04.012 (roughly 66
seconds). Over this period, there were 35 submissions (and 35 cancels) of orders to buy
100 shares, and 32 submissions (and 32 cancels) of orders to buy 300 shares. The pricing
of the orders caused the bid quote to rapidly oscillate between $20.04 and $20.05. The
difference in order sizes and the brief intervals between cancelations and submissions
suggest that the traffic is being generated by algorithms that seem to respond to each
other.14</i>
1) One reason people have done this at the market is to allow their system to breath.
I know of one Java based system that does this at various times of the day to allow GC to kick in for 1/10 a second. This allows the quote side thread to breath and ignore (although I don't think it is very safe) incoming prices.
2) get a latency read on their line for their pending arbitrage orders - although I would think an off market cancel/replace would be more prudent.
3) quote stuffing - simplying trying to push the other guy's algo engine.
4) trigger - I am not certain on that.
The one thing that makes me concerned with these cancel and replace is one firm and another can easily get together to move a market. Firm A sends a cancel replace in a particular sequence. Firm B sends back an ACK with cancel/replace in another sequence. This would be impossible for SEC/CFTC/FINRA etc to find. Shaiz... not just across firms. If some guy on the sells desk wants to move something and wants to tell his buddy on the prop desk, that would be the perfect way to communicate - without getting caught by gov or internal compliance.
I know of one Java based system that does this at various times of the day to allow GC to kick in for 1/10 a second. This allows the quote side thread to breath and ignore (although I don't think it is very safe) incoming prices.
2) get a latency read on their line for their pending arbitrage orders - although I would think an off market cancel/replace would be more prudent.
3) quote stuffing - simplying trying to push the other guy's algo engine.
4) trigger - I am not certain on that.
The one thing that makes me concerned with these cancel and replace is one firm and another can easily get together to move a market. Firm A sends a cancel replace in a particular sequence. Firm B sends back an ACK with cancel/replace in another sequence. This would be impossible for SEC/CFTC/FINRA etc to find. Shaiz... not just across firms. If some guy on the sells desk wants to move something and wants to tell his buddy on the prop desk, that would be the perfect way to communicate - without getting caught by gov or internal compliance.
if we allow this to continue that's all that will be trading
