Note: A version of this article appeared on the Trading Markets site 10/9/06.
In a recent post, I showed how the 11% of largest trades in the S&P emini futures controlled essentially all of the market's price action. While large trades of 100 contracts or more make up only about 3% of all trades executed in that market, they account for well over half of all volume. Conversely, about half of all executed trades are one- and two-lots, but they only account for several percent of the market's total volume. When you recognize that volume and volatility are highly correlated, it isn't difficult to figure out who controls the marketplace.
Below is an unusual Market Delta chart in that it tracks the number of trades that are occurring at each 15-minute bar, and it is only counting those trades that are 50 contracts in size and larger. When the bar is green, we know that more large trades are occurring at the market's offer price. When the bar is red, those trades are predominantly transacted at the bid. By limiting what we look at to trades of 50 contracts and higher, we gain a level of transparency regarding the sentiment of large traders. When they're bullish, they will be willing to take the market's offering price; when they're bearish, they will bail out at the bid.
Tracking large trades over time also allows us to see if the participants who move the marketplace are entering or leaving the market. Notice, for example that the number of large trades expanded as we moved higher during Friday's trade at the 10:00 AM bar. Much of this expanded participation was at the market bid as well as the offer: sellers as well as buyers were attracted to the rise. Over the next three bars, the participation of large traders declined and we could not make new price lows. This invited selling and increased participation of large traders to the downside (at the market bid).
Many technical indicators treat each time unit and price change in the market as equivalent. But not all trades and traders impact the market equally. Perhaps reducing the data we look at could help us filter signal from noise.
Brett N. Steenbarger, Ph.D.