I wrote a bot and gave it 25k to trade

Discussion in 'Automated Trading' started by _terminus_, Feb 23, 2020.

  1. _terminus_

    _terminus_

    This particular algo just long buys/sells the MNQ future.
     
    #161     Jul 14, 2020
  2. Trader200K

    Trader200K

    Exact same thoughts/questions here. Terminus' bot impressively nailed the blow off top yesterday. Hands down. But that was also quite the 400 point short opportunity a couple hours later. (not being critical there because I didn't get any of it).

    I have to admit it became really obvious to me in the last week of Feb that I had a total bias to the long side from all the years riding the market up. Everything I was doing shut off like a light switch. I put on some discretionary 3X shorts for a little upside, but it was a wake up call to see 95% of several hundred watchlist instruments turn into hard decliners for weeks. That has me on the path of rethinking my processes to get more nimble/symmetrical given the state of the world.
     
    #162     Jul 14, 2020
    _terminus_ likes this.
  3. _terminus_

    _terminus_

    Yes, the bot did a good job in catching yesterday's top.

    And it did an even better job by getting out of the losing trade early.

    As I previously wrote, all my backtests show that the short side is more difficult to trade than the long side. I must be missing some info, because many systematical and discretionary traders successfully play both sides. I have yet to get there.

    So I try to leave greed aside and settle for only half the game. For now...

    For example, the trades below are from the short-only algo that I disabled in the live bot, still active on the IB paper trade account on a separate computer:

    short.png

    Started quite well yesterday, but it's currently losing more than it gained.
     
    Last edited: Jul 14, 2020
    #163     Jul 14, 2020
  4. ValeryN

    ValeryN

    Here are few ideas -
    If you are looking into shorting stocks then try to change your short universe. To start with don't short under 10$ for example. See if index membership helps to smoothen your equity. Try to group trades into bins by industry groups, average volume, price range to name a few. See if there is any of those bins tend to have higher concentration of loosers or biggest loosers. No matter what your backtest says never put more than 10% of capital on a single short bet.

    PS. I short a lot, ~500 trades this YTD, most of PL came from them too.

    Val
     
    #164     Jul 15, 2020
    _terminus_ likes this.
  5. 931

    931

    Perhaps its related to input data being biased upward most of the time.
    Do you use separate or single entry threshold parameters for long and short?
     
    Last edited: Jul 15, 2020
    #165     Jul 15, 2020
  6. ValeryN

    ValeryN

    To followup on my previous reply - just finished reading your journal and realized you are doing a single instrument. So not very relevant for your current strategies.

    Good to see a systematic retail trader succeeding with a highly liquid futures. This is a difficult market.
    Great progress! Keep it up.

    Val
     
    #166     Jul 15, 2020
    _terminus_ likes this.
  7. _terminus_

    _terminus_

    In fact, any stock index is mostly an average of liquid and valuable stocks. To me, stock indexes are easier to trade than single stocks.

    Congratulations if your P&L mostly comes from shorting stocks.
     
    #167     Jul 16, 2020
  8. _terminus_

    _terminus_

    Well, stock indexes are upward biased. Any sound trading strategy, I suppose, must have different parameter sets for long and short, if that's what you mean.
     
    Last edited: Jul 16, 2020
    #168     Jul 16, 2020
  9. _terminus_

    _terminus_

    After some modifications I reactivated the previous shorter-term algo. Now it can open up to 2 positions simultaneously. So the system is running again off two distinct, long-only algos.

    Since reactivation the modified algo did 14 trades in a couple days, netting +266$ (0.6%). I'm not sure how it could be classified. Probably not scalping nor day trading, since it can take positions overnight.

    Four positions currently open for the trend following algo, up +498$ as of now. These four where gaining more than 2600$ yesterday night. Message from my son: "Dad, is there any problem/bug with the bot? Yesterday for a few minutes I saw my share rise above 10200$". I patiently explained that I gave myself the rule of never override the computer. Me inside: wanted to kill someone. He: "Ok, let's hope for the best". I worry more about his money than he does.
     
    Last edited: Jul 22, 2020
    #169     Jul 22, 2020
    931 and ValeryN like this.
  10. 931

    931

    Why not code drawdown kill switch at level where you are comfortable.

    Then computer does the worrying for you.

    If switch triggers then could send notification to phone and start to evaluate situation.
     
    Last edited: Jul 22, 2020
    #170     Jul 22, 2020
    _terminus_ likes this.