Gold, from Victor Niederhoffer
Many bearish things about gold lately. That it doesn't go up with no inflation, that we're in recession. That the dollar is going up. That there is great overhand of stocks. I am reminded of a question that I always ask when we hear rumblings that we are going into recession and someone suggests that it is bearish for stocks. I always ask, "what does that have to do with the likely outcome of the stock market? Will the drift be lower or higher?" Oh, I haven't tested that is the unspoken answer. Same for gold. I have not been averse to considering speculative buying of it on all the dips and one is not averse to upholding the spirit of Gavekal idea that it is good to consider things of that nature when caught in Africa by natives, or in large deposits by flexions. One notes a 20 day minimum and is not averse to considering expectations thereafter even before Dr. Zussman runs it on small tab.
Kim Zussman writes:
Using ETF "GLD" daily closes (12/04-present), new instances of 20 day lows were defined as the first 20 day minimum in 20 days. For these new 20 day lows, the return for the next 5 day interval was positive but N.S.:
One-Sample T: next 5D
Test of mu = 0 vs not = 0
Variable N Mean StDev SE Mean 95% CI T P
next 5D 32 0.0012 0.0317 0.0056 (-0.0102, 0.0126) 0.22 0.828
However 7 of the last 10 instances of new 20D lows have been followed by 5 day periods which were down:
Date next 5D
02/11/13 -0.027
12/04/12 0.007
10/15/12 -0.005
06/28/12 0.018
05/08/12 -0.040
02/29/12 -0.004
11/21/11 0.021
09/22/11 -0.067
06/24/11 -0.009
01/07/11 -0.007
07/01/10 0.011
03/24/10 0.025
01/27/10 0.020
12/11/09 -0.003
06/22/09 0.017
03/10/09 0.022
01/12/09 0.047
10/16/08 -0.109
07/30/08 -0.032
03/20/08 0.022
08/16/07 0.010
05/10/07 -0.014
03/02/07 0.008
12/15/06 0.011
08/17/06 0.012
06/01/06 -0.026
02/13/06 0.026
12/20/05 0.050
10/20/05 0.026
08/30/05 0.031
07/06/05 0.002
03/22/05 -0.001
Anatoly Veltman writes:
Fantastic work, as always. Now, I will ask a few skeptical questions:
-
So you test a historical period which saw the price move from $400 to $1600. Wouldn't you expect bullish historical results of a purchase made just about any random day?
-
So we're having a market in 2013, bouncing around on any piece of planted news from Cyprus, from EU, from Putin, from Japan, from Fed, from WH, from investment banks, from fund characters (the ilk of the upside-down), etc. How will one adjust one's timing of statistically catching the falling knife - given that the timing of such leaks (releases) has significantly changed from the test years?
-
Also, the market mechanism has changed in those 8 years, on two fronts:
-the increased weight of ETFs vs. bullion/futures
-the increased prolifiration of HFT exploratory orders
My gist: it's good to have a study, but there are plenty of caveats that call for increased amount of discretion.
In fact, here is my idea: I've observed this to work at an increasing rate since the transfer of investment capital from public into the coffers of the banks and funds has been initiated by the Central authority.
So Gold drops too quickly from $1600 to $1563, which rightfully piqued the Chair's interest in the wee hours. So this is what investment banks, playing with unending public capital, do (for a 24-hour play): they buy momentary cheap Gold and sell Oil against it (got to get the quantity mix right). Oil could not be considered cheap following last week's straight rise. Works plenty of times. And when it doesn't (really, once in a blue moon), a short term spread position becomes a longer term hedge, then the books may get cooked, then a rogue trader is disclosed, etc. who knows…But a good statistical trade to be sure. I like it.
Jason Ruspini adds:
If it seems like HFT is degrading certain strategies over time, there might be testable differences between different futures exchanges that support different order types. For example CME supports stop-limits without any additional software, but Eurex and TSE do not. ICE natively supports ice-berging, most don't. HKFE and SFE only support limit orders natively. Does the performance of benchmark momentum or reversion systems on equity contracts differ between these exchanges (without applying slippage assumptions)? They aren't apples-to-apples of course but if HFT has polluted the microstructure for certain strategies, it seems like something should show-up here, even if many participants have ways to create the other order types.
CQG Order Types Supported by Exchange
Ralph Vince writes:
Interesting points Jason. Timely too, I believe.
When market meltdowns occur, the technologie du jour is the scapegoat. In 1929, it was margin accounts. In 1987, program trading. Tomorrow, HFT.
Not that HFT caused the meltdown, but the fact that they stepped aside and enormous air pockets formed in the faveolate theatre of perceived liquidity.