Friday, December 23, 2016

Regulating the bots

Automated Trader on bot regulation:

Supplemental Proposal to Regulation AT: What Traders Need to Know

The CFTC's original proposed rule making called for pre-trade risk control and other requirements, such as order cancellation systems. 
These would have had to be implemented at three levels - the DCM, the FCM and the trading firm. After comments that such a system would be too burdensome and cost-prohibitive, the CFTC is now proposing a two-tiered structure. 
The first level of risk controls would take place at either the FCM or the trading firm, while the second level would occur at the DCM as a backstop to further reduce the possibility of trading disruptions. Whether the first level of risk control is implemented at the FCM or the trading firm depends on whether the order originates with AT Persons or the FCM. 

And so on.  We can do much of this with python import libraries to the pit, but once the regulators have code in the system, verification and bit error control go to hell.


Here is what the regulators need to know.  The pit boss gets cycles and compresses bets making the bit error mostly bound.  Traders pay for cycles on the graph and when the bot that jams will spend itself to death.  Finally, the regulators can do themselves a big favor by eliminating unpriced end point risk, complete the sand box.

The regulatory approach taken will mostly fail when it is in line.  Open architecture, bet compressing pits ordered by inverse probability (longer search path for rare bid/ask). Then focus on the centering variable, risk by red/green; with end point closed that is computable.

No comments: