Distance Latency
Speed of Light and the Problem of Distance Latency
Markets have become more sophisticated and so have trading strategies. Models no longer receive market data from ECN's located in a single trading center. In order to be competitive, price discovery is being conducted on as many venues as possible. But if an ECN is separated from a customer's algorithm by 40 to 100 milliseconds of transfer time, the market data sent by the ECN could be severely outdated by the time it arrives.A typical scenario from the foreign exchange markets looks like this:
Figure 1: A localized trading strategy
After starting with local exchanges, customers begin to connect to market data feeds in other trading centers.
Figure 2: A local trading strategy receiving pricing data and attempting to deal on distant exchanges
It soon becomes apparent that trading on remote venues could put the trader at a tremendous competitive disadvantage. The trader in New York is seeing prices that are 40ms seconds out-of-date. Chances are the prices have already been acted upon.
A common solution is to replicate the trading strategy and run it from both locations, especially if New York and London trading hours overlap as they do in currency trading.
Figure 3: Same strategy, multiple locations
While both strategies now enjoy 1 to 2 millisecond latency to their local exchanges, they are uncoordinated and can easily end up conflicting with one another. In this case they are not only paying the spread but also forgoing the opportunity to make a pip.
Figure 4: Models do not share information, conflict with one another
"You can't change the laws of physics. Everyone in the market is subject to the speed of light... "
Figure 5: Comparative latencies visualized
Distance latency has become the dominant percentage of total round-trip time, with inter-regional transfers consuming 56% to 96% of total transaction time.
With this in mind, MarketFactory created Whisperer to help traders overcome these issues.