Using Open Interest To Find Bull/Bear Signals

In markets, its the dream of every trader to find a anomaly that can be juiced out day in and day out. Big traders / funds keep trying to find such inefficiencies that they can exploit to their advantage. Those inefficiencies are never heard in the public domain since if everyone knows about it, its sure to be arbitraged away.

One such anomaly is about how using change in price and open interest, you can predict the short term (next bar at the very least) move in the ticker concerned. In fact, Investopedia has a page with the same title as this post explaining the rules of the same.

In fact, I myself have been guilty of believing the same to be true even though my systems do not use either Volume or Open Interest to detect trend change signals. Today, a Tip Seller said the same when queried saying that today’s move was short covering and hence he foresaw continuation of the trend. Being long, I am happy to accept and agree with any logic that matches my position, but I commented that short covering was not a viable trend continuation signal (If you read the Investopedia link, it says and I quote “a declining open interest in a rising market is Bearish”.

The thing I keep learning from Tweeples is that while its okay to attack Journos (questioning their integrity), attack fund managers (ridiculing their picks, returns), don’t you dare attack Tip Sellers. In no time was I questioned about my intentions though I had not even linked / tagged to the original message.

Anyways, being a believer in data than hearsay, I decided to test whether the logic really had predictable value. The table below shows the results from Nifty (since it started).

The results were a surprise even to me since they indicated that regardless of what happened in Price or Open Interest. When Nifty closes +ve, irrespective of change in OI, the next day, you have a 55% probability of markets closing in positive zone. The same is 52% when Nifty closes in Negative Zone.

Given the fact that 52.68% of days are positive, the above stats say that the predictive value of such information is equivalent to what you can get by tossing a coin.  And hey, you don’t need to pay thousands to someone to get that odds.