Aug

22

One would imagine the Sunday open to close in Israel might be predictive of the open in the US on Sunday night, and possibly the open to close of us on Sunday. By Israel open on Sunday, the US has already passed Friday close. And Israel would be catching. Of course the US Open is not a predictive thing since it can't be acted upon, but a descriptive one. The whole subject of the influence of Indian, European, Asian, and mideast markets on the US is an interesting one and calls for much counting, correlation, and finesse.

Anatoly Veltman writes:

I'd be the first one to stress the equities "rolling wave" over the timezones, as well as inter-market influences (as in currency-gold-stocks-bonds-oil, etc). Being said, there are two clear new ingredients that make historical statistics less than meaningful: central meddling and modern algos.

1. What can possibly be the use of percentile correlations and sequences observed over any historical duration, if current market interventions and near-global ZIRP are unprecedented.

2. Modern algos thrive on constant change/adjustments. To paraphrase Jim Simons: what feeds "our" fascination is that our former immersion into discoveries (within pure science) would eventually yield an ever-lasting law or theorem — while (market) discoveries we achieve today will only live a blip of time, and so you have to journey on (almost daily) to your next discovery and implementation.

So in consideration of the above major influences, my current MO would be: do not rely on hard stats. Do rely on your instincts, understanding of the new world financial order and good occasional privileged information — and trade discretionary.

Chris Cooper adds: 

I can accept Anatoly's "two clear new ingredients" but reach different conclusions. My conclusions are:

1) Trade at a higher frequency so that you can get enough recent stats to be meaningful.

2) Trade fully automated, not discretionary, so that you don't fool yourself about your alpha. Also, it's the only sensible way to trade at a higher frequency.

"Relying on your instincts and understanding the new world financial order" are important only at the meta-level.

Paolo Pezzutti adds: 

I think:

1. Cycles are ever changing. Today it is because of ZIRP, tomorrow it will be because of new rules or products coming on that influence market structure. I don't know if cycles will be shorter or longer. You trade them until they work. Counting still works.

2. Frequency depends very much on commissions. Some regularities at shorter time frames cannot be traded if your commissions are too high. Frequency depends also on technology you have available. Also, one should trade a frequency where you have less competition.

3. New cycles means new patterns to come up and old patterns to die. Keeping track of ongoing patterns is important and also establishing criteria to determine a pattern has stopped working. Early discovery of new patterns is vital for your performance. But how much data and evidence do you need to validate a new pattern? More importantly on the tech side is how you implement the search of new patterns. A continuously running search can scan the data according to certain criteria and propose pattern to be evaluated further.

4. Trading should be fully automated to trade higher frequencies, more markets simultaneously, and decrease stress.

Newton Linchen writes: 

Dear Paolo,

You said: "Keeping track of ongoing patterns is important and also establishing criteria to determine a pattern has stopped working."

I once asked this question (how to measure the "death" of a trading strategy) to the List, and the answers were disappointingly vague. ("They work until they don't anymore", and such kind of answers).

To my knowledge, this is a vital question.

Recently, I backtested a strategy a colleague was trading, to discover that in the last 6 years you would lose your entire wealth trading it. But he kept trading it, due to an anchoring with an event when "it worked", plus a kind of empirical testing of only few months.

This means he was caught by the siren song of a series of "lucky strikes" within a larger distribution of years of losses.

This behavioral concept ("anchoring") is quite interesting, and we smile at the poor guy who don't count.

But what concerns me is that we can behave the same way, (although counting), when we face a regime shift (ever-changing cycles) and keep trading the defunct strategy… Until when?

Perhaps a rough answer would be to establish a drawdown metric related to the maximum historical drawdown? (i.e., we trade it until a drawdown x% larger than the greatest historical, and then quit?)

Or maybe the reason to trade a strategy must be quantitative whether the reason not to trade it anymore should be qualitative? (i.e., acknowledgement of the regime shift…)

A final thought would be a strategy based on market microstructure — in the way it is present in ALL regimes.

Any thoughts?

Newton.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search