Latency: A measure of delay.
In a market where speed is of the essence, the spotlight falls on the issue of latency. Latency is the time it takes to get a deal done, cancel an order or know what is happening in the market.
Given its importance in today's trading environment, it will be important to arrive at an industry definition of latency measurement, to ensure that trading venues, systems and infrastructure are all judged by a common standard.
Latency is a statistical function that can be impacted by:
- Market participants' systems and architecture
- The architecture and construction of the venue to which market participants are connecting
- The connection between the two
Market participants are concerned with the end-to-end latency, since latency creates risk, no matter where it is introduced. Although venues often refer to their time to acknowledge the receipt of an order, that statistic on its own does not tell the full story when assessing trading risk caused by latency. Best practice is moving to monitor the following end-to-end statistics with attention to the tails of the latency distribution at the 90%+ percentile as well as the average:
- Time to cancel an order - TTC
- Time to fill on take out when an aggressive order is placed
- Market data distribution speed - MDDS
Latency increases in importance when the time horizon for trading is shorter because its adverse impact on the certainty of filling a trade or price slippage is greater in these strategies. For price takers, any delay will expose them to market risk prior to confirmation the order has been filled. For price makers, any delay can leave their prices in the market at a time when the market has moved and they wish to cancel. Timeliness of market data is also extremely important. If data feeds are latent, the market may have moved, thereby creating an inaccurate view of the current market state which is itself an input into the execution strategy.
As latency may vary with market activity, the most important latency measures are taken at peak times when the systems are experiencing high loads. Loads during these peak times can be much higher than average and it is important for market participants to ensure the performance of the venues they choose is satisfactory during these conditions. Trading systems should be able to proactively monitor latency throughout the trading day to calibrate their models and pot problems before they impact the trading strategy.