Algo gone wrong – managing systems

Viewing 2 posts - 1 through 2 (of 2 total)
  • Author
    Posts
  • #87236 quote
    jebus89
    Participant
    Master

    Walking into the extreme volatility we have been seeing im wondering if there are many here that have had systems break on you?

    So far it seems like its only 1 so far that has gone crazy. Luckily I managed to stop it when i saw the 2 big drops happen in equity curve , which of course happened just days after activating it..

    The algo tries to enter when market is about to break up towards a new high, but if market does a big move up but instantly drops this is gonna cause fake signals and the algo will buy every single time. With low volatility the algo looked good, but with the crazy volatility i saw the algo do 4-5 losing trades in 1 day hitting stop loss every time, causing the “huge” drops in equity curve, i realized that no matter how the algo would perform in the future, this is not the right type of algo for me. It caused slight panic both times seing it go into 4-5 trades very quick one after one and hit stop every time. It managed february well, but the volatility and choppy stuff after the drop killed this one.

    I feel very lucky that i quit this when i did and i missed out on the crazy drawdown. I have a fixed amount of € for each algo and when the limit is done, the algo is done as well. This was about 50% towards that limit.

    I also want to point out a very important point: Another obvious point to why this algo stopped in my opinion is curvefitting. This was created before i developed my process for designing new algos in the hopes of minimizing the risk of overfitting strategies, which i have described in another post. This is potentially overfit and the huge recent drop would certainly speak towards that. Who knows how this equity curve looks in 10 years time lol. Might do fantastic as soon as things settle down, might continue down to 0. Im not gonna stick around with real money to see hehe.

    System is for US tech 100 (nasdaq futures) 1€ CFD contracts.

    nqdisaster.png nqdisaster.png
    #87249 quote
    Vonasi
    Moderator
    Master

    Yes this is one of the down sides of only being able to back test on data back to August 2010 when using tick by tick. Most Indices have been in raging bull markets since then and so it is very easy to curve fit an idea to a line that goes straight up with no major deviations. We have entered a different world now. The weekly close has been below the 52 week low for a couple of weeks now and historically this has normally happens before major recessions or at least before major sideways market wobbles. Any algo curve fitted to the tick by tick data will likely fail now. It is always best to back test and include a major recession and the odd market wobble in the test which is why I rarely use tick by tick data. Having said that future markets will always surprise us eventually by doing something that we have never seen before!

Viewing 2 posts - 1 through 2 (of 2 total)
  • You must be logged in to reply to this topic.

Algo gone wrong – managing systems


General Trading: Market Analysis & Manual Trading

New Reply
Author
author-avatar
jebus89 @jebus89 Participant
Summary

This topic contains 1 reply,
has 2 voices, and was last updated by Vonasi
7 years, 3 months ago.

Topic Details
Forum: General Trading: Market Analysis & Manual Trading
Language: English
Started: 12/17/2018
Status: Active
Attachments: 1 files
Logo Logo
Loading...