Wednesday, 13 February 2013

Trading outposts

Wednesdays are my science and philosophy evenings.  And I'm still on the brilliant thought-provoking Kahneman book.  I want to think a bit about how you might apply some of his identified biases to the realm of trading.

The most important and general, to my mind is the phenomenon of anchoring.  In trading practice, a trader usually finds himself with an implicit or explicit target of profit to reach on a monthly basis.  They also are subject to monthly (and other) stop loss limits which, if breached, trigger a formal review of their book by management and risk teams, with possible further sanctions, including having to sell out of some positions, to stop trading, or in extremis, to leave the company.  There are also more industry-wide targets which they're expected to aim for - like a benchmark index.

All of this anchors and contextualises what they actually do.

How else could it be?  Some traders know their own career average return and draw-down.  But one of the benefits of being a trader is the partial clean slate you get when moving from one employer to another.  A soft reset occurs on their trading record, unless of course the trader in question has a career record he can crow about, in which case he'll go to a special effort to maintain the career numbers.  I would guess many traders have exploited the ability to wash away their previous return datasets.

These are what I'd call paradigmatic anchoring effects.  There are also syntagmatic ones.  Imagine you have a multiple security strategy in play - for example you are long a convertible, short some stock, long CDS protection.  In an important sense this is a singular strategy.  And you would think the trader should look at the strategy p&l atomically too.  But in reality, the trader will be in and out of the individual holdings regularly.  Perhaps these holdings in isolation traders develop anchors,  Anchors at the level of the individual holding can be potentially sub-optimal.  Or to take an even simpler strategy, imagine you're long the stock.  Perhaps your net quantity changes dramatically through the year as you're buying and selling.  But each sell or buy of your stock can have adverse micro-anchors, relating to the p&l of the holding, not the overall strategy.  

Imagine you thought this year being long gold is a good strategy.  Most likely there would be many points during the year where you'd be buying, and then at other times selling gold.  At each sell, there would be a profit or a loss.  If it was a loss, then you might re-anchor it against an overall profitable net performance for the long gold idea.  Conversely, if you're overall wrong on gold, then this might dominate even though your most recent sell was for a profit.  Being net losing on gold, according to prospect theory, would tip you into that quadrant where a person risks more than he would normally.  Perhaps this is the advantage of the monthly target and stop loss - it encourages you to bank your profits metaphorically and to partly wipe the slate clean on a loss, so that you are carrying less anchoring baggage with you going into the next month's trading.

I can see how the 'doubling down' practice is grounded in prospect theory.  A loss causes you to increase risk since you concentrate on the small likelihood of 'winning back' your losing position.  Perhaps you increase your bet size as you do it, a permutation of the classical martingale trading strategy.

In Kahneman's language whilst a rational agent will prefer a broad frame in judging simultaneous decisions, real humans sometimes perform sub-optimally by preferring a narrow frame.  My examples above merely add the syntagmatic / paradigmatic dimension, that is, narrow framing in the context of genuinely parallel collections of decision, and in the context of a close series of related  actions spread in a short period of time.