A VWAP for All Seasons

As I tally up our Asia Pacific clients’ algo performance, I’ve realized that last year’s market volatilities likely injected more noise into these results, making like-for-like comparisons difficult. This probably holds true for buy-side firms assessing their brokers as well. As I sifted through the rows of data, one pattern that jumped out was that performances against the trusty old VWAP benchmark exhibited much less variation than performances against other “point-in-time” benchmarks such as the arrival price. Since trading desks, even with the help of algo wheels, may have trouble sending similar orders to different brokers’ algos, the outcome is that the varying performances may be due as much to the nature of stocks and markets as to the algos’ own logic. Giving consideration to once-abandoned VWAP performance, at least for the past year, as part of the evaluation process may help the normalization between different sets of brokers’ data.

Over the past 12 months, our APAC algos displayed vastly different ranges of performances against the various benchmarks. Against the arrival price benchmark, the range of performances for the orders was multiples of spreads even when the average slippage itself was just about 1 spread. This contrasts with Interval VWAP performances, which saw the lowest amount of variation (just around 1 spread) across all strategies as the benchmark restricts measurements to just within the time an order is live. The PWP benchmarks had variations somewhere between the extremes. As expected, performances varied to a larger degree than they did in 2H 2017, suggesting the difference is likely caused by market volatilities in 2018 (and less so from algo logic!).

Using interval VWAP as one of the benchmarks to evaluate an algo suite’s performance makes sense logically too; if an algo is capable of keeping in line with market prices, the trader has mitigated part of the risk. I know it might seem an additional burden for buy-side desks to spring a new benchmark on their brokers, but the once-in-a-decade market last year does warrant desperate measures!

Tony Cheung, Head of Quantitative Analytics, APAC

Source: Liquidnet internal data of APAC algo orders for FY 2018.

Ellen Gordon