Why discretionary Trading?

A young reader sent me an email and in one section asked me why I switched from mechanical trading to discretionary trading and I thought it was a very interesting question, especially considering the recent popularity of “algo” driven strategies. For one I think there are many misconceptions regarding black box trading and the biggest one is that in some way your work is done once you discover the ‘right’ algo. From that point you simply switch on the money maker, sit on a beach, and drink mojitos; RenTech has over 95 PhD’s full time on staff, I worked with another individual full-time, 10 hours+, 5 days (sometimes 7 on my spare time) at a hedge fund, researching….researching….researching. The quest is never ending, and in the end your left searching for a grail that does not exist.

Perhaps you are ‘enlightened’ and realize that truly successful algo trading is based on portfolio construction, risk management, and position sizing. You’ve cut back on work considerably but you will eventually and inevitably come to a point where your skill in deciphering a market inflection point will be put to the test.

I think this paragraph from “Against the Gods” summarizes why much better then I could:

“The past seldom obliges by revealing to us when wildness will break out in the future. Wars, depressions, stock-market booms and crashes, and ethnic massacres come and go, but they always seem to arrive as surprises. After the fact, however, when we study the history of what happened, the source of the wildness appears to be so obvious to us that we have a hard time understanding how people on the scene were oblivious to what lay in wait for them. Surprise is endemic above all in the world of finance….

If these events were unpredictable, how can we expect the elaborate quantitative devices of risk management to predict them? How can we program into the computer concepts that we cannot program into ourselves, that are even beyond our imagination?

We cannot enter data about the future into the computer because such data is inaccessible to us. So we pour in data from the past to fuel the decision-making mechanisms created by our models, be they linear or nonlinear. But therein lies the logicians trap: past data from real life constitute a sequence of events rather then a set of independent observations, which is what the law of probability demands…..Once again, resemblance to truth is not the same as truth. It is in those outliers and imperfections that the wildness lurks.”

Your model works until it doesn’t. How do you know it’s not working? Well you can wait till you have enough trades and analyze the distribution versus the past distribution or against a back test but by then your broke. No, at some point one has to make a discretionary decision, has the market changed??

So instead of chasing my tail, I’ve decided to focus on learning and analyzing the intangibles in the market. How to read the tape, when is it right to have a bullish bias, bearish bias, no bias. Re-read “Reminiscence of a Stock Operator” and apply that versus say Advance Statistical Analysis. I just need to be good enough to be consistently profitable. I have learned the skills to test my ideas before implementation and an understanding of behavioral finance so I have a chance to inhibit biases while trading or formulate a method (thru stop losses, position sizing, etc) to work around them. Only time will tell I guess, if it doesn’t work out I guess I can sell the old systems online for $$$ 🙂


~ by largecaptrader on March 24, 2010.

7 Responses to “Why discretionary Trading?”

  1. I dislike your discouragement of that “young reader”. I too have found myself at a place where system trading was a depressing money losing enterprise. Pouring tons of energy into a losing battle is disheartening. I found my way out. I wrote in Matlab a walk forward algo to test my ideas completely out of sample. This has completely changed my ability to understand and quantify the risk I am assuming with trading. Lots of good came from the exercise which at a minimum was learning Matlab. I am an mechanical engineer by training and found Matlab powerful yet easy to learn. Matlab also opened up my trading to utilizing basic indicators which are fed into a fuzzylogic engine with outputs back to TradeStation/MultiCharts. Perhaps, the “young reader” already has access to Matlab (academic copy) and could have worked collaboratively with him. I actually began farming out my Matlab programming in chunks to students in foreign countries to make better use of my time. All of which created lots of positives for all.
    I cannot speak to discretionary trading as I am an engineer and I at core just a numbers guy. However, the fact that “RenTech has over 95 PhD’s full time on staff” surely indicates something lucrative coming out of them. Or is your point that it is not possible to compete with 95 PhD’s? I heartily disagree with that point as I have access to 100’s of PhD’s ready/willing and able to code up anything I want at a ridiculously low cost.

    Best of luck to you and your discretion.

    • Tom? Jeff?

      Thanks for your comment. The point is not that money can’t be made from systematic trading; like value investing, merger arbitrage, volatility arbitrage, etc systematic trading is just another discipline. And like those other strategies it takes persistence and dedication to become profitable. Walk forward, fuzzy logic, etc are tools to decipher new information presented by the market, analogous to what an experienced discretionary trader does with their brain.

      So the two situations are actually quite similar, lose time and money learning systematic trading. OR lose time and money discretionary trading. The choice is up to you. My preference for discretionary trading is anchored to the idea that our brains can still assimilate information much more efficiently and effectively then the machine (at least at this point and with my current knowledge).

      Shops like Rentech, Getco, etc are market-makers and I would venture they spend a majority of their time trying to reduce latency or reverse engineer execution algorithms. There’s only so many ways to sell strength and buy weakness. But in any case the point is they have 95 PhD’s full time on staff because it IS A FULL TIME JOB. Thus my point also is to dispel an all too common misconception that systematic trading is a ‘holy grail’ or solution to the markets. It isn’t.

      Also keep in mind that I still believe the tools, matlab, tradestation, learning to analyze and understand statistics are extremely important and useful skills. It’s just that I would rather be the one that ultimately pushes the buttons as opposed to the machines.

    • Tom/Jeff

      Im the young reader in the article. Ive been planning to build and backtest a stat arb strategy and feel that matlab is the way to go (which i have access to), however i have difficulty knowing where to start as there are many things i need to learn to do i.e aligning data when dates are missing from yahoo, testing/ranking many instruments for cointegration, optimisation etc.
      If you have experience performing this sort of thing in matlab and wouldnt mind pointing me in the right direction it would be greatly appreciated.

      my email is lewisbowker@gmail.com


      • Ernie Chan wrote a good book on Quantitative Trading systems called (funny enough) “Quantitative Trading”. It includes full MatLab code and examples how to align data and a full stat arb strategy.

        P.S. Aligning data is a pain the ass sometimes…

  2. Gentleman,
    Thanks for the reply.
    First, I agree with most of what “LargeCap” detailed. I guess how comfortable are you with no ability to backtest/walk-forward test your brain? I forgot what I had for dinner two days ago. I again am an engineer by training and so I am wired perhaps differently. It is like looking at a design/blueprint of a bridge. The discretionary bridge builder has seen many good bridges and bases the new bridge on ALL of the long standing bridges he has seen before. Needless to say I would structurally analyze the bridge because that is what I am most comfortable doing. Note the strong looking bridge below.

    I too agree with the sentiment that it takes a lot of effort and is a full time job (at first). Once you have a system that works. I always crack up too when watching the ads for ETrade or Schwab with their basic assumption that you can trade stocks between 9’s at the country club. And the heat mapping thing is just crazy funny. The “action” might be there, but can you make any money with that knowledge alone? I think not.

    Next, A strategy I planned to use to trade @nq works much better on Russell2K and the MidCap. So the knowledge/hard work is leveraged.
    Anyway, I do not mean to come off as “Mr. having all the answers”. That surely is not the case. Lost plenty of money and was humbled by it. I just kept plugging away. Nothing more nothing less. The big appeal to me is that I can be on a boat in Miami or a condo in Vail making money. Nothing else comes close to the freedom trading allows you in that regard.

    There is a great webinar through Matlab that will not only point you in the right direction, but juice you up for what is possible in that environment.

    The code is also available for the above at the Matlab site. Do a search and if you cannot find it I can email it to you. Matlab is awesome (fast/easy/scalable). One can find programmers to do some of the coding too if your time is better spent elsewhere (boating/skiing/etc.).

    Quick fyi on my setup. I gather historical data/indicators etc. in TradeStation and do the analysis in Matlab. I have a little dll that moves data back and forth from Matlab to Tradestation. If you want to be frugal about it, you can get Tradestation for a month and get all the data you need (store as csv or similar) and then suspend the account while your cpu cranks on Matlab. The key for me was an automated walk forward routine that I use 10 years of data. The best surprise- is no surprise eh’?

    Good luck to you both.


    Just to be clear

  3. Jeff,

    I’m recently trying to learn how to develop algorithm using TS. From above, I was wondering how you are able to extract the data from TS? I wrote codes on how to extract it from Yahoo. Any direction on where to start? I appreciate your input. Thanks.

    • Hi Jeff,

      Not sure what you mean by extracting the data, for use in Excel? I THINK there might be a TS to Excel API on the message boards somewhere but otherwise you can use a print command in the indicator/strategy to generate a csv file. Create a new indicator with something like: print(file(“C:\spy.txt”), eldatetostring(date), ” “, time, ” “, open, ” “, high, ” “, low, ” “, close);

      That will output a txt file you can open in excel, use txt to columns and you should be good!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: