By Armando Rodriguez
According to George Will :
Economic hypochondria is also bred by news media that consider the phrase "good news" an oxymoron, even as the U.S. economy, which has performed better than any other major industrial economy since 2001, drives the Dow to record highs.
George F. Will is a 1976 Pulitzer Prize winner whose columns are syndicated
in more than 400 magazines and newspapers worldwide.
Below are some selected emails to a client advocating for a trade strategy that I called "reverse trading". Which consisted, not in trying to beat the trend by selling dollars when the beginning of a drop is detected, but instead, by buying at its lowest relying in that, being this a panic price, it will come back.
Jan 2007
At last we have the amount of valid data necessary to calculate a statistically significant average dolar event.
Here’s the average of the 35 events in which the correlation (CorrP+CorrM) has peaked above 10 since April 2006.
And here’s the negative 35 event with CorrP+CorrM < -10
Observations:
The overshoot effect in positive events had already been observed back in 2006, data was short then (less than 9 events) and a bit unreliable but it was consistent with what has been called Economic Hypochondria (check my email from 10/22/2006). I suggested then that we could profit on this overshoot with “reverse trading”, but at the time the interest was going for the FIX protocol. The speeds of the FIX protocol led us to believe that catching events on the rise was the way to go, but this has proven to be no panacea.
Under the light of these new findings, based on a lot more reliable data, I think we should consider again reverse trading.
Jan 2008
Happy New Year! I have been thinking of all our approaches to the problem (profiting on FOREX) and why each one has failed. You may find interesting to summarize what we learned from this FOREX trading adventure.
Surfing the Market Waves
This was our first approach back in 2005; it was based on the alleged existence of a Market Inertia. We didn’t invent the concept, actually it was a strong trend of thought within the early Analysts and if you Google up “Market Inertia” you may still find people pushing the concept. It took us several months to realize that we were going nowhere with this approach and that it was as much BS as Fibonacci, the neural system and the sort.
Event Catching
It took us long time, to realize that we didn’t have a chance in this either. I spent much of my thinking time adapting to different protocols from new ever faster brokers and providers.
At some point I hit the basic problem that prevented us from profiting with this approach, but even when the truth was found, it still took some time to abandon the approach.
In short; for any threshold that you may set for trading, in order to profit, no matter the closing algorithm, you needed an event peaking over some ever growing price manipulated delta above this threshold value. Peak queries proved over an over, that for every profit peak, there were too many small non profit peaks to make an overall profit. There was no way of knowing in advance if an event was going to be big enough.
Earn Little is More Likely that Losing Big
That’s as true as a landmark, but how much likely? Though I suspected that this sounded like the search of a perpetual motion machine, I developed simulation queries and even a VT (Virtual Trader) to profit on the idea.
By this time we knew that the market behaved Brownian for all practical trade time intervals. This we learned from the frequency spectrum and by the linear behavior of the standard deviation with the square root of time. This meant that the market behaved as a one dimensional random walker with a drift to be noticed only in a yearly scale.
I messed with equations trying to find, theoretically, optimal values for target wins and loss limits. At some point, when I was already on the right track, I found a century old theorem that read.
The probability that the random walk will go up to b steps before going down a steps is a/(a+b)
That was it, there was no a or b that could render an average profit. I told you the bad news immediately, but you refused to listen. You argued that you believed the market was not exactly Brownian and that you this approach had prove profitable for the first time and so on. There was nothing much reason could do against faith, but I don’t expect that your winning streak will hold for long.
Profiting on Media Induced Panic
It has been way more that a year now that I found that in the average, a dollar drop produced an overshoot on the rates, while a raise undershot. Media always forecasting America’s economic disaster may play some mayor roll on this. I have suggested several times that we could profit on this behavior, but neither you nor the investors seemed to give this approach much of a chance, favoring any of the losers above.
I called this reverse trading and even when it was never “officially” accepted, I developed a few queries that showed a profit chance on this idea. The main advantage of this reverse trading over the direct one is that, since trade will happen after the event, then you go for it only when you already know it has been “big enough”.
Of course, there’s this issue with maximum finding that can always be fooled by a temporary back step, but back steps are not that common on the raise of events and not all fooling back steps imply losing. I admit that a proof that you can profit in the average with reverse trading has not been produced, we always have had so many other things with higher priority preventing that, but at least the contrary has not been proved either. There is also the idea of hedging that may work well in this context.
If you ever find another investor, please talk him into this approach, which is the only one that could possibly render a profit.
During this entire FOREX quest, I have been trying to find some cause-effect phenomenon that will allow systematic profiting, but unfortunately price manipulation, spread and fees compensated any edge we found. Now, what I called “panic effect” is real, happens most of the time, yet when it doesn’t, you may lose all you’ve won. Still, hedging and some algorithm improvement may find a way to a profit, but this involves long and hard work that no one seems to be willing to pay for and I can’t afford to do on my own.
If you manage to find support for the above mentioned development effort, I will gladly join the team, but I refuse to keep working on the “SystemR” even if offered payment to do so.
Sincere regards,
February 2008
What I have learned in all this FOREX adventure can be summarized as:
· There’s no such thing as market inertia, no point in surfing the market
· There’s no profit in randomness, so much for long term trading
· Prices are manipulated during scheduled events, no chance of profiting in direct event trades
This leaves us only with reverse event trading. The American media, which is nearly the world media, profits on alarming news about everything, bad news seem to sell better than good ones and this produces the effect observed time and again on this overshooting of the quotes every time the dollar drops.
There’s a definite potential here, as a matter of fact, the bug that predicted those huge winnings, was not entirely wrong, it just happen to trade at the peak and not at the time the simplistic algorithm rendered. The peak is not attainable, but I’m sure we can get near by sophisticating it. My approach will be to use the concept of digital filtering on the FOREX signal, not just a EMA, with is a first order IIR (Infinite Impulse Response) filter, but a more sophisticated one that can separate noise from signal, pretty much the way it is done in electronic circuits using Chebychev or Butterworth higher order filters.
Today I settled for a 4th order Bessel filter with a cut period of 33 samples. Since this description tells you near to nothing, here is the response of that filter to an example event with a panic effect:
All filter types, the Chebyshev, the Butterworth and Bessel, isolate pretty well the panic effect from the noise, but they all introduce delays. The advantage of the Bessel filter is that is the one with smaller delay.
Even with this delay, the Reverse VT (Virtual Trader) would have made a juicy profit with this one, but reducing the delay is of utmost importance. The smallest delay here is of 91 samples.
The filters above are belong to the IIR (Infinite Impulse Response) category, these show the sharpest frequency cuts with the lowest orders, this is for using recurrence. From CPU usage point of view the IIR’s are the most efficient. There’s another category of digital filters, called the FIR (Finite Impulse Response). These filters use no recurrence, the delay that these filters introduce in samples, equals the order of the filter, meaning that for, say order 30, the delay would be just 30. With the same order, the FIR’s cut is less sharp that the IIR, but since we have no CPU shortage, maybe a 30th order can do the trick with less delay… that would be really exiting.
Tomorrow, I will try with a 30th order inverse cosine FIR.
May 2008
As I have told you this reverse trade thing was not going to be easy, if it were, lots of people would have profited and screwed the business long ago. I know it’s there, but it is not easy to pull that moment when the doomers come out of their panic out of the data stream.
First I thought that I could find a max just by filtering the data from one currency, then I tried all the currencies, that came near, now I’m going to include the quote density into the equation. Quote densities (quotes per minute) are event driven. I there were no events and quotes were perfectly random, the number of times you would find a particular minute count should follow the Poison distribution. Guess what... it doesn’t (see the chart below). This means that quote densities are not random, during quiet times they probably follow that distribution, but then, during events, densities can acquire values unthinkable in a purely random process. This means, that monitoring the density, together with the movement correlation and the currency behavior, could give us a clue to when the quotes are stopping their surge and starting recovery.
Armando
MH:
Just for the record (because I can see there’s no way I can make myself understood):
This said, OK, I will modify your SP (SQL Server Stored Procedure). I say yours, because don’t remember ever writing an SP under that name. I will work on request 1, not in request 2, since I don’t understand the concept of temperature oscillation. Please, describe what you mean with the term “oscillation”, for instance, when the market cools fast, you regard this as market instability? When you say “small period of time” is that in the seconds range? Or maybe minute range.
My client was just "too slow for the West" and never got to make a profit with Reverse Trading. By the time it was implemented, the panic effect overshoot, that showed strong during 2006, had attenuated to unprofitable levels by 2008. It is not that the media had changed its ways or that people don't panic any more, it is that the market learns, nothing stays profitable for a long time, as soon as operators start profiting on some effect, the rest follows smothing it out until it disappears.