Tag Archives: algorithms

liquidity and stock price changes

January 8, 2019 By in algorithm, algorithmic trading, liquidity, Uncategorized Tags: , , , 1 Comment

daily liquidity and price movements

Liquidity has a lot of different meanings.  Right now, though, I just want to write about what I think is making stocks yo-yo to and fro on any given day.

 

The default response by market makers–human or machine–to a large wave of selling of the kind algorithms seem to trigger is to move the market down as fast as trading regulations allow.  This serves a number of purposes:  it minimizes the unexpected inventory a market maker is forced to take on at a given price; it allows the market maker to gauge the urgency of the seller; the decline itself eventually discourages sellers with any price sensitivity, so the selling dries up; and it reduces the price the market maker pays for the inventory he accumulates.

A large wave of buying works in the opposite direction, but with the same general result: market makers sell less, but at higher prices and end up with less net short exposure.

 

From my present seat high in the bleachers, it seems to me the overall stock market game–to make more/lose less than the other guy–hasn’t changed.  But we’ve gone from the old, human-driven strategy of slow anticipation of likely news not yet released to violently fast computer reaction to news as it’s announced.

Today’s game isn’t simply algorithmic noise, though.  Apple (AAPL), for example, pretty steadily lost relative performance for weeks in November, after it announced it would no longer disclose unit sales of its products.  Two points:  the market had no problem in immediately understanding that this was a bad thing (implying humans were likely involved)   …and the negative price reaction continued for the better part of a month (suggesting that something/someone constrained the race to the bottom).  As it turns out, decision #1 was good and decision #2 was bad.  Presumably short-term traders will make adjustments.

my take

On the premise that dramatic daily shifts in the prices of individual stocks will continue for a while:

–if investors care about the high level of daily volatility, its persistence should imply an eventual contraction in the market PE multiple.  Ten years of rising market probably implies that this won’t happen overnight, if it occurs at all.

–individual investors like you and me may have more time to research new companies and establish positions, if the importance of discounting diminishes

–professional analysts may only retain their relevance if they actively publicize their conclusions, trying to trigger algorithmic action, rather than keeping them closely held and waiting for the rest of the world to eventually figure things out

–the old (and typically unsuccessfully executed) British strategy of maintaining core positions while dedicating, say, 20% of the portfolio to trading around them, may come back into vogue.  Even long-term investors may want to establish buy/sell targets for their holdings and become more trading-oriented as well

–algorithms will presumably begin to react to the heightened level of daily volatility they are creating.  Whether volatility increases or declines as a result isn’t clear

 

 

 

 

machines vs. humans

January 7, 2019 By in algorithm, algorithmic trading, Uncategorized Tags: , , 3 Comments

…a financial Industrial Revolution?

I remember reading, years and years ago, an analysis of changes in the nature of work that happened during the Industrial Revolution.  The general idea is that, say, candlesticks had been made as one-of-a-kind items, out of precious materials and ornate decoration, worked for months by an artisan who had spent years learning how to do this.  Yes, the end product was useful, but it was also very expensive, meant for a niche audience, and acted as a sign of the owners’ superior wealth, taste and privilege.  In contrast, the “new” candlestick was made, fast and cheap, out of ordinary stuff, by a guy who knew how to operate a machine.

Today we find it hard to imagine the possible appeal of most pre-IR objects.  Yet they were once the norm.

 

The macro/microeconomic research-based stock market investment reports of the kind I used to create were made by people, like me, who served long apprenticeships under masters of the craft.  The work tended to only start to approach minimum standards after the author had, say, five years of practical experience in an investment management firm.  Buy-side portfolio managers like me also used the voluminous output of internal or brokerage house analysts who spent their careers studying a specific industry group.

By 2019, most of the experienced buy- and sell-siders have either retired or been laid off,  and have been replaced in many cases either by computer-controlled index-tracking products or by algorithms.  The main forces in today’s daily stock market trading have become machines, some programmed to carry out the wacky theories of the academic world, others to react to signals from the patterns of trading itself (i.e., technical analysis) or to news stories (typically written by reporters trained mostly as writers) or to extrapolate from the patterns of past business cycles.

progress or free-riding?

Are the research reports of a decade or two ago analogous to the candlesticks of the Pre-IR era?  Are algorithms like early industrial machines?  Are they a better and cheaper, although different, way of dealing with financial markets than having a very expensive group of human craftsmen?  Does this mean those who decry algorithms are simply Upper East Side-dwelling Luddites?

I don’t know about “simply.”  My feeling is that algorithms are here to stay.  And my experience as an investor is that it’s very dangerous to think that just because you don’t like or understand something that it serves no purpose.

Still, my suspicion is that as it stands now, there’s a healthy dose of free-riding to algorithmic trading.  In other words,  it looks to me as if some algorithms rely on reading the signals of human professional investors as they move in and out of stocks in response to their research findings.  As those humans are displaced by machines, however, those signals will disappear–implying algorithms will have to evolve if their raw material is to be something other than random noise.