From the September 01, 2011 issue of Futures Magazine • Subscribe!

HFT: How fast can you really trade?

The times when the best bid rises above the best ask in a very short interval are few and far between, as Table 1 illustrates. As Table 1 shows profitably “flipping” in and out of a trading position at ultra-high frequencies is much more difficult than it is at lower frequencies on an example of three commonly traded near-term currency futures on a randomly picked trading day, May 25, 2011. Table 1 shows one of the most liquid futures, euro FX, had moved either up or down by more than the spread in only 10.8% of 15-second intervals throughout the day. That is, in nearly 90% of the time from 9 a.m. to 4 p.m. no high-frequency trader facing positive transaction costs had a chance of profiting at 15-second intervals. At higher-frequency intervals, the situation looked even worse: The price of euro FX futures cleared the spread less than 0.1% of all 1-second time slots, with 99.9% of time leaving no chance to any high-frequency trader.

Table 1: For each frequency, the table presents the percentage of time during the 9AM-4PM EST trading session that the best-bid exceeded the best-ask during a period of the specified frequency, recorded on May 25. 2011.

Frequency

Proportion of time intervals where max best bid exceeds min best offer

CAD futures, Sept 2011

JPY futures, June 2011

EURO futures, Sept 2011

10 minutes

100.0%

100.0%

100.0%

5 minutes

80.7%

91.5%

100.0%

45 seconds

13.2%

35.6%

48.6%

15 seconds

0.8%

9.8%

10.8%

1 second

0.0%

0.2%

0.08%

200 milli-seconds

0.0%

0.0%

0.01%

The observed lack of profitable opportunities at narrow time intervals is not an arbitrary happenstance, but rather a function of the speed with which markets move. Table 2 and Table 3 document the average number of times bid and ask quotes change in the given time period. Most/ask quotes of shown currency futures did not change even once during time periods of 1 second and under, rendering profitable non-market-making trading at such frequencies nothing but impossible.

Table 2: Average number of times best-bid of selected currency futures changed during a given time interval from 9 a.m. ET to 4 p.m. ET on May 25. 2011.

Frequency

Average number of times best bid moved within the time interval.

CAD futures, Sept 2011

JPY futures, June 2011

EUR futures, Sept 2011

10 minutes

100.0

25.9

170.5

5 minutes

51.7

34.7

87.7

45 seconds

8.2

5.3

13.7

15 seconds

3.1

1.9

5.0

1 second

0.5

0.2

0.6

200 mille-seconds

0.3

0.1

0.4

Table 3: Average number of times best-ask of selected currency futures changed during a given time interval from 9 a.m. ET to 4 p.m. ET on May 25, 2011.

Frequency

Average number of times best offer moved within the time interval

CAD futures, Sept 2011

JPY futures, June 2011

EUROFX futures, Sept 2011

10 minutes

105.1

70.2

163.4

5 minutes

54.3

36.1

84.2

45 seconds

8.6

5.5

13.1

15 seconds

3.2

2.0

4.7

1 second

0.5

0.2

0.6

200 mille-seconds

0.3

0.1

0.4

The frequency of trading is dependent on the size of the minimum bid-ask spread, as that is often the biggest cost in non-market-making strategies. The maximum frequency of portfolio allocation also is linked tightly with the volatility of the market and the security: Wilder market oscillations increase the number of short-term periods where the best-bid may overcome the best-offer. Days with higher market volatility are naturally more conducive to higher trading frequencies.

Irene Aldridge is a quantitative portfolio manager, investment consultant and managing partner at ABLE Alpha Trading, LTD. She is also the author of “High-Frequency Trading: A Practical Guide to Algorithmic Strategies and Trading Systems.” Irene’s research on properties of high-frequency data applicable to portfolio management of HFT strategies is forthcoming in “Equity Valuation and Portfolio Management” (F. Fabozzi and H. Markowitz, eds.). She also teaches a course in high-frequency trading. To learn more, you can reach her at ialdridge@ablealpha.com.

<< Page 2 of 2
Comments
comments powered by Disqus