May

18

What are the most basic market states traders might need to model?

1. Going up
2. Going down
3. Reversing

Ranges, trends are subsets of the 3.

Next step is modeling what simple mechanism causes the 3.

Hernan Avella writes:

There are no “simple mechanisms”. But I would start with the microscopic dynamics of “the turn”. Yesterday [2 May] was a good day to study.

Big Al offers:

An interesting thought experiment is to imagine that you have a chart of a random walk but you still have to trade it. Money management, trade sizing, stops, limits - could you still trade it?

Zubin Al Genubi responds:

Random walk with drift would be the default basic state (S) with random factor u say with sd2. What simple rules might model market activity. Like ants and bees following simple rules but building coordinated complex structures. Adam Smith first mentioned emergence in his invisible hand.

Hernan Avella responds:

Isn't this the basis for most uniform trading that occurs?. While the other big chunk of participants "think" they have a model, "think" they have patterns, but are essentially doing a version of the same?

This reminds me of the infamous Kirilenko paper:

We examine the profitability of a specific class of intermediaries, high frequency
traders (HFTs). Using transaction level data with user identifications, we find that high frequency trading (HFT) is highly profitable: 31 HFTs earn over $29 million in trading profits in one E-mini S&P 500 futures contract during one month. The profits of HFTs are mainly derived from Opportunistic traders, but also from Fundamental (institutional) traders, Small (retail) traders and Non-HFT Market Makers. While HFTs bear some risk, they generate an unusually high average Sharpe ratio of 9.2. These results provide insight into the efficiency of markets at high-frequency time scales and raise the question of why we don’t see more competition among HFTs.

Zubin Al Genubi adds:

Yes HFT guys probably have done it at market maker level. Chair says yes you can trade random walk with drift with buy and hold due to drift. MM and HFT may also have order flow info they buy which may or may not be a different process.

Adam Grimes writes:

Absolutely and of course… that's why the hurdle rate for any test has to be the baseline (unconditional) drift in the sample.

[Re the "thought experiment"] Unless I'm missing something, not profitably (over a large sample size). All these other things are important, but they, at best, keep you at breakeven in a RW environment (i.e., no "signal" or "edge" possible). In real life, a comparable approach keeps you paying the vig with consistency. As for the thought experiment, correct?

Big Al responds:

For me, the thought experiment doesn't have a correct answer but forces me to think more rigorously about issues such as money management, trade sizing, stops, limits.

Andrew Moe writes:

Chair often advised that rather than considering just up/down or above/below a given threshold, one might look at "up big"/"up small"/"down small"/"down big" as classifiers. This is particularly salient in information theoretic calcs (ie, entropy) but interestingly moving to deciles offers little or no improvement.

Zubin Al Genubi adds:

I've been interest in agent based modeling of complex systems using simple rules. A new wrinkle would be adding a random factor following power law distribution in tails which stock data displays.

Jeff Watson responds:

That sounds like a perfect task for ChatGPT.

Gary Phillips adds:

Absent from the previous post on modeling was any mention of time frame. There is greater model risk the shorter the time frame you’re trading in, because price action is more random. Realized volatility, liquidity, gamma, and 0DTE options can and will, shape the trading environment. And, has been demonstrably evident the past 10 weeks, each day has its own ecosystem and market structure. This makes modeling in a short time frame a fool’s errand, and its participants useful liquidity providers.

As one moves to a higher time frame, positioning, money flows and sentiment become most important. Fund flows and positioning, along with cross asset flows, target dated funds, corporate buybacks, seasonal factors, and factor flows take on more meaning.Yet, even if one could ascertain the above factors with certainty, he wouldn’t know if the data was priced into the market or not.

And finally, while there may be a lag or even a disconnect on a long term time frame, macro economic factors, geo-political factors and CB policy, will inevitably exert its influence on the market. But, we don't know if we have experienced the event(s), nor know how traders will react to the event(s), that will finally move the market out of its current trading range.

A pragmatist's model then, is to know the market one’s trading, and to have a well defined process. Then one can make (bias free) observations and accurate, probability-based assumptions.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search