Nooby McNoob becomes a quant

Discussion in 'Journals' started by nooby_mcnoob, Mar 24, 2017.

  1. Last 10 days profit*:

    Trade closed 370
    Trade closed 2030
    Trade closed -300
    Trade closed 840
    Trade closed 470
    Trade closed 1450
    Trade closed 800
    Trade closed -700
    Trade closed 600
    Trade closed 1180
    Trade closed 570
    Trade closed 690
    Trade closed 240
    Trade closed 2200
    Trade closed -40
    Trade closed -240
    Trade closed 120

    * I use 10 lightly, just have a report that does last 10 days. May not actually have been trading for 10 days.
    #141     Feb 8, 2018
  2. Just sharing thoughts... Not really an update.

    The algo has been running for a while now, mostly successfully. Every week so far has been profitable, except when things crash due to whatever. A few days ago, I had a multiple of $1000 loss because the algorithm crashed and did not put in a stop in time. When I say crashed, I mean the broker disconnected.

    That problem is fixed, most problems are fixed, so my aim is to have a smooth 1 month of profits with no unexpected behaviour before I make my next move.

    I've been reading various threads here and I wonder if I'm the only one pissed off when they read this: "The market is going up/down because algos are driving prices higher/lower to take out the weak stops and then it will crash/moon."

    How the fucking hell do you know that? Are you front running the entire market? No, you're just making shit up to explain why you're wrong. Even a broken clock is right twice a day, so eventually you'll be right, but is that really the kind of life you want to live?

    God fucking damnit.
    Last edited: Feb 23, 2018
    #142     Feb 23, 2018
  3. I want to apologize for the language in my last journal. It takes all kinds and I was a little hyper.

    Progress on the paper trading is good. Building in multiple tests to validate backtest, paper trade vs live. I "tweaked" my system to work on a different timeframe without backtesting (haha) and it went insane so I used my "stop trading" button. Lost about $300 there

    The one thing I noticed this week was that (paper) trades were being entered into live that would not have been entered into in the backtest. This is worrying me. The stops are handling it, the profits are still good, but I need to figure this out. I can't recall whether this was during the time where my algo went insane though.

    The progress on my trading platform is feeling very productive. I've got the algo logging to a cloud logging monitor on which I set alerts that sends me notifications to my private Slack instance. This is what led me to use my "stop trading" button above when I changed the algo to run on a different timeframe without backtesting! I was on the phone chatting and started getting alert after alert. "Hang on" *Push stop all trading button* "OK what were you saying". It was cool.

    I've also got each datapoint being used in my algo automatically being logged to a database so I can run things like:

    $ sqlite3 -batch -cmd '.headers on' logs/dbs/backtest.sqlite "select sum(pnl) from trades where datetime >= date('now','start of month');"
    $ 4120.00000000012

    All in all, I'm feeling very good about things. Things I have still to do before I can switch off paper trading:

    1. Detect when the algo is malfunctioning due to technical reason or otherwise and automatically restart
    2. Have a full trading week without touching the algo at all, for maintenance, tweaks or otherwise
    3. Compare the full trading week in the following manner:
    1. Backtest
    2. Paper trade
    3. Actually executed trades in IB

    I think I can get (1) done this weekend and so have (2) for the upcoming week and then perhaps (3) for the next journal entry?

    After that... Let it paper trade for a month completely untouched. That will be the real test.
    #143     Mar 3, 2018
  4. Looks likes you’re on top of it. The list of failsafes gets longer and longer as you trial and error your way through it.

    Have you considered any off the shelf solutions? Might help fix many of your issues, and allow you to focus on the stats and competitive advantage.
    #144     Mar 3, 2018
    nooby_mcnoob likes this.
  5. Thanks! A workflow for managing these failsafes that isn't haphazard is important. So I'm trying to keep it simple.

    I have, and I think we discussed this in my journal earlier or another thread here. I can't find it but the gist of it is that as a techie, I am more productive building my own platform over the long term. It's not like I'm doing it from scratch though, I'm using a variety of tools written by other people to do so. I'm just putting it together in a way that works for me.
    #145     Mar 3, 2018
  6. It looks like you have achieved quite some progress. Nice to see.
    #146     Mar 3, 2018
    nooby_mcnoob likes this.
  7. #147     Mar 4, 2018
  8. @sle gave you some good advice in that thread, especially in regards to establishing an intimate connection to your front end. Trading off that intimacy for efficiency gains, just isn’t worth the headaches in most cases.
    #148     Mar 4, 2018
  9. Since that thread, I have indeed developed an intimate connection to the front end so it was good advice indeed.
    #149     Mar 4, 2018
  10. I have figured out the issue with my algorithm entering trades live that it didn't in the backtest. It was indeed at the time where I changed my parameters without backtesting. As a result, I also found a bug that exists at all timeframes, but only manifests at a specific one.

    This is getting to be a ton of fun. I better make money, lol.
    #150     Mar 5, 2018