"Trading Strategies that are Designed, Not Fitted" by Robert Carver from QuantCon NYC 2017

Quantopian
Quantopian
38,281
0
Publicado 2017-12-12
Engineers design stuff. Why do Quants prefer to fit? In this talk, Robert will explain what designing a trading system actually involves, explore why designing might be better than fitting, and introduce some of the tools you could use. He will also take you through the design process for an example trading strategy. Finally, he will discuss how we can have the best of both worlds: strategies that are well designed and also fitted to the data.

Link to slides: www.slideshare.net/secret/1XXl5iDp0MdW2l.

To learn more about Quantopian, visit us at: www.quantopian.com/.

Disclaimer
Quantopian provides this presentation to help people write trading algorithms - it is not intended to provide investment advice.

More specifically, the material is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory or other services by Quantopian.

In addition, the content neither constitutes investment advice nor offers any opinion with respect to the suitability of any security or any specific investment. Quantopian makes no guarantees as to accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.
Todos los comentarios (10)
  • @mimmotronics
    This is an interesting observation for the table at 39:30, when you consider the Nyquist Theorem used in signal processing/reconstruction. The theorem states that in order to reconstruct a given signal, you need to sample it at at least f_max/2 sampling rate, where f_max is the maximum frequency found in the signal. If you extrapolate that thought to trend following, in order to accurately identify a trend of N discrete bars, you need a window_size of at least N/2 bars. In practice, however, it is possible to reconstruct a signal at sampling rates that are below f_max/2, with some error. I presume that's why in your table you still obtain good/decent values at window_sizes that are less than N/2. Also, 47:00 you say there's a 0.05 loss in correlation between forecasts in real data vs. the fake data of 0.90. It may be possible to parameterize the sawtooth waves and additionally randomize that parameter set in order to introduce some variability in the correlation between forecasts. Just a thought...
  • @tedchou12
    I am the first to comment on this wonderful presentation?
  • @alrey72
    We have many instruments (forex - different currency pairs, stocks - different companies, and even a number of commodities), so the variety of real data is there. Also, we have the historical data, we can (for example) use prices of different instrument from 1995 to 2015 to create the model and test it from 2016 up to now (Aug 2022). So why still use fake/random data?
  • @chadyu1551
    they are full of brown sticky stuff XD
  • @itslike123
    Each market is different, you can't design one for all, I tried for 3 years doesn't work or perform poorly. Each market need to have its own, wether you call it overfittin but I'm profitable
  • @mr.kurazsnaj837
    I'm reading your books, listening to podcasts and youtube where you participates. I guess you have someting to say...but I really do not get it...words, words and words...