Aug

15

A VeltmanToday, I want to open the following idea for discussion: I noticed that Bond O.I. has diverged bearishly in recent days. (Definition is simple: O.I. is declining, while Price is moving into new highs.). I eyeballed a daily chart covering about 40 trading days, since June Bond left the board June 19th . It appears that on 75% of those 40 trading days the Price and the O.I. went either both up or both down! To flip this observation: on only 25% of the days the Price and O.I. diverged! What surprised me: Price since that date is 3%+ higher, while O.I. since then is 3%+ lower! (O.I. has now fallen to its lowest in almost a year!). Does that mean that one should fade current rally, once one gets Sell signal from one's other indicators? Note: I'm also intrigued by the fact of continuing bullish pattern of "O.I. down on down-days, up on up-days." How does one reconcile the two?

Christopher Tucker asks:

Shouldn't one fade any rally when one gets a sell signal from one's "other" indicators?

Nigel Davies extends:

GM NigelWe've this kind of issue with cycles. The parameters are intuitively obvious to the human mind but the very devil to explain in a way that a computer can understand.

There is the same problem on the chessboard, for example in understanding positional elements such as pawn structure. Humans are able to divine what is important in a position whilst the computer will assign the weights it was programmed to do even when these things are unimportant. The problem is that it cannot take a holistic view, it can only work on already disected parts.

Besides the Senator's book I think it's worth reading Dee Belveal on this. But once again it's not going to be something that lends itself readily to 'testing' via quant methods because the parameters are very difficult to define. I suggest instead that one adopts the approach chess masters use, and that is to play through all the games by hand in order to acquire a 'feel'.

BTW, I'm indebted to Anatoly for posting in the way that a games player can finally understand.

Manuel Bravochico adds:

I just got back from my monthly luncheon with my friend. He used to manage a restaurant before trading. About all he knows how to do on a computer is flip through charts looking for momentum. He amassed a small fortune and has compounded at the highest rate of return — although with some 50% drawdowns — that I’ve seen, north of 60% since 1999. Verified.

I keep asking him how he does it. He always gives me the same answer, “the chart just looks good.” He has a “general” set of rules that I have never been able to program over the years. At the elite level — Rentech excluded — most trading I think cannot be programmed, i.e. a holistic process unable in this age to be programmed.

Shmuel Layla writes:

The way I deal with false divergence signals that start occurring in the course of a Trend is not by looking at other confirming indicators, but rather by looking for the occurrence of a divergence on a higher time frame corresponding to a different set of peaks and valleys that has yet to resolve itself. I then find that the “local” divergence has more reliability. This is fractal trading for the mathematically challenged such as I. This works on volume charts of the ES contract. Maybe other issues with stronger trending properties require a more sophisticated solution. For the time being there is sufficient liquidity for me in the ES.

Rocky Humbert muses:

Anatoly’s post is indeed thought-provoking. It hits on many different issues. One approach might be to adjust position size to reflect confidence: i.e. you don’t have to go “all in” on every hand that you play. Certainly, this is how some gamblers might handle the problem.

Taking his question literally, however, I continue to argue that increasing open-interest reflects a coincident indicator of trend persistence, rather than an indicator of absolute direction. For an illustration of this phenomenon, look at the bond contract from March 2006 to May 2006 when the contract fell from 112 to 106 and the aggregate open interest increased from about 0.6m to 0.9m.

As the bond contract has been range-bound for the past year, this would be consistent with a declining open-interest. Of course there are other fundamental deleveraging explanations for declining open-interest too.

Anatoly Veltman responds:

Rocky, you picked a remarkable 2006 example! It really helps to understand one of your points: that the reason for currently stagnating O.I. is that bonds have been mired in a boring range (as opposed to the clear chart breakdown in spring of 2006). However, I will always take issue with outright scepticism about using O.I. to derive signals, at times. If not used properly, as in my own hand-picked current bond example, it sure may be of little use. But I know, and have utilized throughout my career, so many fantastic opportunities where I was able to make use of O.I. for prediction/confirmation.

Archives

Resources & Links

Search