Deseasonal Forward Curves

Discussion in 'Commodity Futures' started by Speculate, Aug 7, 2015.

  1. Speculate

    Speculate

    Hello,

    I mostly trade futures spreads. Im always trying new methods, models to add to my existing ones.

    So I've been doing some research around forward curves.

    Although there seems some merit in my tests in spreading with the forward curve, the same can be said for trading with a deseasonal curve. The problem with that is that they are often in conflict: the forward curve (black) may be in carry, while the deseasonal curve (red) is in backwardation. Example below is Corn from 2011 fyi.

    Has anyone done any research in forward curves and deseasonal curves and have any useful comment? Perhaps the steepness of the curve is of more significance, or SD between months...???

    [​IMG]

    Here's a paper (not mine) on constructing a deseasonal curve should you be interested:

    http://www.kcl.ac.uk/nms/depts/mathematics/research/finmath/seminars/geman2008.pdf
     
  2. Vyki

    Vyki

  3. Speculate

    Speculate

    Thanks for that! I'll take a read later - looks interesting.
     
  4. Trader13

    Trader13

    The general case for time series (price series) decomposition includes components like trend, seasonality, and a random component. This is useful for historical analysis of price when you're trying to understand price movement in hindsight. But I don't see any practical value of this technique for predicting future price movements, which is what trading is all about. To make something useful from this, you would have to develop a prediction model for each component (trend, seasonality, random, etc) and sum them to get your notional price forecast. That doesn't sound easy.

    I'm interested to hear opposing viewpoints. Interesting topic.
     
  5. Speculate

    Speculate

    I see your point T13.

    Currently, my 'testing' of the deseasonal curve has involved getting a load of past spreads - both the winners and the losers that were generated from a variety of my methods; mostly seasonal, but not all. These go back several years and Im not done yet.

    Im then simply looking at both the forward curve and the deseasonal curve at the time/day of entry and seeing if there is anything significant there in terms of either the forward curve or deseasonal curve, or both being in or out of alignment with the trade result.

    I'm excluding results where the spreads consisted of adjacent months on the curve such as NG X15-V15

    Ive enough outcomes to be be statistically significant, however, as you can probably tell, my stats method is pretty basic - Im much more of an 'art' than 'science' when it comes to speculation!

    So far, there seems to be a slight bias to the deseasonal curve as suggestive of whether the spread will widen or tighten.