Some stats... about to run live ?

Discussion in 'Automated Trading' started by savagemp5, Jan 28, 2012.

  1. Hi,

    My previous posts and threads are on qns regards to Microsoft Excel performance and Automated trading. I'm (I hope:)) about to run live on my IB TWS from paper trade to Live.

    Say I'll start with 30,000k. That gives me 100,000k intra day SMA margin, enough to trade my strategies.

    - US Stocks during US daytime with top few most Volatile / highest volume equities.
    - US futures. Index & commodities (Asia to US)
    - Global Index. Minis of Hang Seng, N225m

    Results with Law of large Numbers:
    - 100-200 trades a day
    - 35%-45%
    - Avg Win Amt. per trade => Avg Lose Amt. per trade (which means somethings I may lose due to the win/lose %)
    - Largest win trade > largest lose trade
    - Sharpe Ratio -0.3-0.8

    Hoping to reach:
    - 10,000 - 20,000 trades as statistical track records (For applying a job at HF or Props hehe)
    - win rate >40%
    - Avg Win Amt. per trade > Avg Lose Amt. per trade
    - Sharpe Ratio >3 ? (according to Quant Job postings)

    Somehow the Sharpe Ratio's Risk factor cannot be applied here, since my risks includes anything from 3 minutes to 20 minutes, pretty much riskless if you'll ask me.

    Any other optimization can I do ? I'm extending to more Stocks, perhaps maxing out the IB TWS 50 quotes, and collecting stats before deciding how to continue.

    Any suggestions ? Bet most successful people have found the holy grail. I wish I could trade mobile, like visiting Europe and US, bringing along my Credit card / Laptop and travelling for my global operations.

    btw im frm Singapore :)
  2. jcl


    Hmm, why do you think the Sharpe ratio does not apply to your strategy? It's normally a pretty good performance indicator. On the other hand, 200 trades a day and still 45% profit sound quite good and should give a far better Sharpe ratio. How have you tested it? Excel is good for quick developing strategies, but useless for testing.
  3. well.. this 45% was very very hard work of sleepless nights, and it was only 25% back then. So much algorithms have been thought thru and put in place just to push this number higher, ie. taking lesser trades or trading only during specific part of the time for example.

    I'm not sure, in Sharpe Ratio, I just use the (Avg PnL of 100 trades) / STdev( PnL of 100 trades) The numbers are wierd even when I have none losing trades the Sharpe ratio doesn't give me 3. hmm...

    Excel -> twsdde.xls -> IB tws. Using forward testings with live data only (heck i can't be bothered with back-testing).

    My automated dream.
  4. jcl


    Normally the Sharpe ratio is taken not per trade, but per bar, and is annualized. The formula is

    sqrt(bars per year) * Mean(bar results) / StdDev(bar results)

    The ratio of Avg and StdDev is dependent on the number of samples, so for 200 trades you would get a different result than for 100 trades. That's why the Sharpe ratio is normally annualized.
  5. If that is in fact your sharpe ratio, your leverage will be very, very limited (assuming you don't want to blow up). Personally I would also strongly question any system or discretionary trades with that low of a rating, no matter how good your other evidence is.

    A 0.3 rating really is terrible quite frankly, that's like saying you're going to take on Stanford GSB having completed a tough 1.7 undergraduate GPA.
  6. heh maybe im using the wrong details. In a typical aualized sharpe, you take the 12 x month of performance in %, for the ROE.

    Instead here I have a string of small wins and loses per trade, not anualized, and not in % on my ROE. In fact I dont even watn to have a ROE, its not calculated as per AUM.

    If u have a string of results from $-10 to say $50, and with the stats I mention, pump in Avg and STDEV, the AVG will always be lower than the STDEV, hence the low sharpe.

    Even my PDF curve looks so-so.

  7. 200 trades a day?
    How will you ever find the time and energy to analyze them on a daily basis?
  8. heh. it's 200, not 20 million trades a day.

    Why not ?

    Run a program on my logs to determine avg bars per buy/sell, etc and formulate stats ? chart out graphs ?

    Automate every human actions... Not possible ?

    I have 51 hours, from US EST 5pm to GMT +8 Asia time 8 AM before nikkei and WTI starts trading, to do my analysis.

    What do you do during weekends ? Play and enjoy ?
  9. the1


    200 trades per day is A LOT of trades. At one point I was doing 40-50 per day by taking just every trigger the market offered me. Nowadays I choose wisely on my triggers. I wait for the best the market has to offer and take roughly 5-10 trades per day. In my case, less is definitely more. Your trade frequency is probably what's driving your Sharpe down but if you're profitable I wouldn't worry too much about the Sharpe unless one of your goals is to manage money. If so, then the Sharpe is important.

  10. its 200 trades a day over 20+ products, or around 5-10 trades per day per product.

    If I put it across to 50 uncorrelated products then the numbers increase.

    Yes I figured out Sharpe Ratio won't be my target anymore for high frequency. In fact my Sharpe increases when I took away those few high profiting trades (trades that were 2 Stdev out of the Norm)
    #10     Jan 29, 2012