Thursday, September 06, 2007

NASDAQ order flow

This is part 4 of the series on order flow begun on Sept. 1st and covers data available on the Nasdaq. The NAZ is an 100% electronic marketplace, which gives it a decidedly different character than the NYSE. The NAZ is driven by bid/offer spreads provided by market makers, which the NAZ does track and which numbered about 892 on a daily basis for August. But the reality is that with the anonymous electronic market, anyone can be a market maker, just submit a bid/offer.
While the NAZ does not segregate retail and program trading, it does identify "block trades", which it defines as trades over 10,000 shares. August averaged 8,783 daily block trades, which accounted for only 0.15% of all trades but 15% of total volume. Being home to most of the high beta stocks, the NAZ is the favored venue for day traders, so maybe the piddling NYSE retail numbers just mask the real action on the NAZ. But wait a second, looking back to our NYSE data we see an August daily average of 5.8M trades as compared to the NAZ's 6.0M trades. We also see NYSE average daily volume at 1.86B and Naz at 1.88B. Given the magnitude of the numbers in question, the correlation between them is rather surprising.
Discounting the block trades, we see that the remaining 99.85% of the trades account for 85% of the volume and yield an average of 265 shares/trade, a number that is midway between the NYSE retail and program trading trade share average. These numbers strongly suggest that program trading and algorithmic trickle filters are accounting for the bulk of trading volume but the nature of the electronic marketplace creates a veil that is difficult to penetrate.
As was the case with yesterday's study of WM on the NYSE, as I look at EBAY on the NAZ this morning I note 3.9M shares traded during the first 90 minutes and as I watch the L2 time and sales go streaming buy, the vast majority of the trades are 100 shares, a few 200s, a few 400s and an occasional 1000 share order, or about 14,700 orders = 163 trades/minute. The WM study produced 12,500 orders and 140 trades/minute and this further supports the contention that orders are being algorithmically partitioned and queued.
How awareness of these trickle filters can benefit the average trader will be the subject of upcoming research. My next series of studies will focus on NYSE volume tells and will begin in October.
Anyone wishing to further investigate NYSE or Nasdaq data can refer to the links on the right side of this blog under "Data and Charting links". Then click either "NYSE Data" or "Nasdaq internals". NYSE gives you 14 days of info free, Nasdaq gives you 5. I tried to collect AMEX data also since this is the domain of the large volume sector SPDRs like XLE, XLF, XLB, XLU, etc, but AMEX charges $30/day for a peek at their numbers so I will have to defer.
As I mentioned in Part 1, I'll be glad to forward my study as a consolidated Excel spreadsheet to anyone who wants to massage it further.

No comments: