I wrote a bot and gave it 25k to trade

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

  1. _terminus_

    _terminus_

    Agreed. However, market conditions do change over time so you can be confident only up to a certain point that your system will keep working, even if it tests ok over a very long period.
     
    #211     Oct 8, 2020
  2. d08

    d08

    Depends on the timeframe. Data from 2008-9 is still relevant to today, there have been some regulatory changes but it was the same electronic market as before. Now if you're going back to 1987 and looking at the execution issues back then, those might not be repeatable - market was structurally different.

    But if I tested a long strategy from 2010 to 2019, I could easily find strategies that produced large profits with low effort. Of course they would have all collapsed in 2008-9 and now 2020.

    The more I trade, the more I focus on losses and loss-making periods rather than profits.
     
    #212     Oct 8, 2020
    They and _terminus_ like this.
  3. d08

    d08

    I assign data a weight, accepting lower performance the further into the past I go. Yet I still consider catastrophic losses even 20 years ago significant. For example I believe currently we're in 1999 heading into 2000. The same proliferation of dumb retail money and everything tech getting bought up.
     
    #213     Oct 8, 2020
    ValeryN and _terminus_ like this.
  4. I was thinking about that this morning. For example, long strategies post-GFC are easy to find. I've found those same strategies don't work as well in the bull markets leading up to the GFC.

    Seems like the markets were choppier back then. Has anyone noticed something similar?

    To your other point, the long strategies go from good profitability post-GFC to breakeven or slightly positive pre-GFC. Not like they turn into complete losing strategies.
     
    #214     Oct 8, 2020
    d08 likes this.
  5. _terminus_

    _terminus_

    Generally speaking, the older we get, the more we value safety.

     
    #215     Oct 8, 2020
    d08 likes this.
  6. d08

    d08

    I've witnessed the opposite. I was trading an event based strat up to 2008, before that it produced impeccable results. It stopped performing in 2008 and ever since has been sideways, it's long-only.
     
    #216     Oct 8, 2020
  7. henry76

    henry76

    Yeh I'm certainly not criticizing Terminus , his attitude to producing honest results is better than most , probably better than mine as I know of this tendency in myself. Though rather like d08 implies , the long only part of Terminous' strategy is one thing that bothers me a little as regards reliability.
     
    #217     Oct 8, 2020
    _terminus_ and d08 like this.
  8. _terminus_

    _terminus_

    Thank you. That's why I'm still working in developing a viable short strategy. But:

    1) I'm not going to use anything that I'm not confident with.
    2) As long as my long only strategy is able to cut off and go flat when the price is going down, I'm ok.

     
    #218     Oct 8, 2020
  9. henry76

    henry76

    Fair enough , but if you could develop even a slightly less reliable strategy , but which could trade both long and short , well there's is a good chance you could use it in other markets and timeframes ( thereby increasing available trades , thus more frequent trading and for similar returns lower amounts of capitol and hence risk.
     
    #219     Oct 8, 2020
  10. henry76

    henry76

    Just wanted to reply to this , I think if your using daily data then yes , maybe true , but using tons of intraday , say 3 minute bars etc , there's so much data especially across markets that if you use enough of it , and it works, you have something much more universal and I suspect less dependent on the "era " of the data.
     
    #220     Oct 8, 2020