10 Responses

1. buffet

   May 20th, 2008 at 4:50 pm

   1

   I would say, let the facts speak for themselves. Don’t build NN to predict a wishfill thoughts, teach them first to be factual!

2. Tom

   May 20th, 2008 at 9:38 pm

   2

   buffet: I’m not asking you to fight the trend but one day the trend will end and the USD will strengthen. What I’m asking is, what would the data inputs be in your model to alert you that the downward trend is over. Would it be a collapse of Gold? Would it be a 6% yield on the 10 year T-note?

3. dc

   May 21st, 2008 at 8:21 am

   3

   mr. buffet(is that really you?)

   if you are the real buffet could you lone me 50k i’ll pay you back with 20% interest in one year.(ok maybe 3 years)
   Also if you are the real buffet you would probably catch wind from some source conected with lets say, maybe the world bank, that might give you a hint as to which way the wind is going to blow on some of these currencies.(your problem is solved)Us little folks will never know anything before it happens, only as it’s happening.Personally to stay sane i decided to go only technical. i tryed fundimentals/news. i compare it to predicting the weather with a little wishfull thinking/dellusions thrown in.(sort of like a weatherman looking up at the cloads and stepping into some foul smelling,it can get a little unpleasent sometimes) In a NUTshell it’s more frustrating to me then just staying 100% technical.(no punn intended) I THINK IN THE END A PERSON HAS TO STAY WITH WHAT IS PAYING THE BILLS (oh don’t forget about that 50k )

   P.S. on currency futures a person could use buy/short brackets to get in a possition to play the news before it hits the wires ” GOOD TRADING”

4. buffet

   May 21st, 2008 at 11:55 am

   4

   Tom, put in gold price/volume time series and analyse past interaction/correlation with the US$. Was there any in the past? can we learn some lessons to predict the future? If people go out of gold, do they go into dollars, probably not or not anymore? DC, forget your 50k!

5. sc

   May 21st, 2008 at 4:17 pm

   5

   Start with macro factors that make sense. GDP growth, relative interest rates, inflation, shape of the relevant yield curves, CPI, GSCI, trade stats, debt service as a proportion of GDP… the dollar is both and input and an output of these. Thus look at cointegration, covariance and correlation. Build a simple MA model - timing may be that simple in fact. The work on the inputs - do they need to be smoothed? Can you build a simple timing model off the smoothed data? Now go NN… good luck.

6. Tom

   May 23rd, 2008 at 5:33 am

   6

   DC: If buffet is thee Warren Buffet, he’d gotten rich by not giving people $50k!

   buffet: I think in the case of the USD, price and volume is alone and gold is not enough, sc (below your comment) has the right idea IMHO.

   sc: sounds like you’re off to a good start! Thanks for sharing!

7. sherry

   May 23rd, 2008 at 3:21 pm

   7

   Is the high oil prices a function of the declining value of the dollar?

   Can OPEC trade in euros to compensate?

8. SerpentMage

   May 25th, 2008 at 4:01 am

   8

   Hey,

   My answer is you can’t. I am dead serious here. There is no real precedent for this since it is a black swan type event. The only thing you could do is do a series of Monte Carlo simulations that walk the path of the dollar and see where you get.

   And if that fails throw some dice and see where it ends up.

   Here is how I see things playing out from some of my past MonteCarlo simulations. Oil eases back slightly the market is squeezed, but the dollar keeps doing its mambo in a range. OR we have an oil super spike collapse of the equity markets and collapse of the dollar.

   Right now the pyschology is to move into commodities to avoid the dollar drop. I am easing slightly back from my 1.76 prediction, but not much. It really depends on oil.

   As Jeff Mache in Fast Money said, “There is only one thing to invest in, oil. Everything else will get destroyed by it step by step.”

9. Tom

   May 25th, 2008 at 7:37 am

   9

   SM: What kind of distribution did you use for your data to come up with this? A normal dist? You have to remember a BSE is an outlier way off in the “long tail” of a distribution.

   The reason I asked this question in this post is that for all the gloom and doom of the USD, one day it will rebound and rally. The question is, where it would rebound from and what would be the inputs you’d use to give you signal.

   What goes up must come down and what goes down will come up! How’s that for scientific?

10. Christian Gross

    May 25th, 2008 at 11:54 am

    10

    Hey Tom, I am going to post a blog entry on this topic. I think you know my blog URL… If not http://www.devspace.com/~christianhgross