Math/statistical challange: stock price prediction algorithm

  • Status: Closed
  • Hadiah: $300
  • Penyertaan Diterima: 4
  • Pemenang: kannasanthosh

Ringkasan Peraduan

NOTE: this math/statistical challange requires programming or advanced Excel skills
_______

INTRODUCTION:

As most people know, when a bad news hits a company the stock price of that company falls and the inverse happens if it's a good news.

In the attached two images I show you three recent real-world examples of what happened to the stock price and volumes of NVidia and Amazon when bad news hit them.

As you can see there is a sudden increase of traded volume (bars on the bottom) and the price falls because people sell in panic, both things happening for some tens of minutes, allowing an investor able to act immediately (let's say after just 1-2 minutes of seeing the sudden change in volume and price) to earn money for that fall predicting it, with a financial operation called "short selling" (earn while the price falls), closing the position when the fall is over.
Similar thing can be done in the opposite direction when there's a good news: the investor can immediately buy the stock, selling when the rise finishes, earning money for the rise...
_____

REQUEST:

What I ask you is to find the best algorithm for earning the most money by automatically detecting the best moment to start a position ("short sell" or buy) and the best moment to close it, applied to the attached 12MB big file that contains the real price and volumes of the Amazon stock with 1 minute resolution of last 10 years, for your performance tests.

The algorithm that, applied to this data set, virtually earns the most money, wins my award (and you can use it like me for earning money on the stock market of course;)

A very simple example algorithm could be this:
- start a position immediately after there's a sudden 4x increase of volume together with 0.50% change in price ("short sell" if -0.50% or buy if +0.50%)
- close the position 1 hour later or when the price reverts back of half of the total reached change, the sooner

RULES:
- to calculate the earnings of your algorithm, you must start the calculation with a virtual investment of 10.000$ and subtract a virtual (like real world) commission equal to 0.06% price + 10$ fee every time you virtually start and every time you virtually close a position
- the algorithm can use only the data provided by the 12MB file, that are prices and volumes. Its actions must be only based upon prices and volumes (and time).
- you cannot cheat by overfitting code to specific single patterns in the provided data set. The algorithm must make sense as general rule to predict the future cases, though it can be complex

NOTE:
the headers of the 12MB txt file are these:
date time open(first price) highest lowest close(last price) volume

Kemahiran Disyorkan

Winner
Peraduan Pilihan
Siarkan Peraduan Seperti Ini

Penyertaan teratas dari peraduan ini

Lihat Lebih Banyak Entri

Papan Penjelasan Umum

  • visweswaran1998
    visweswaran1998
    • 6 tahun yang lalu

    Could you explain briefly how the Machine Learning Model failed please, It has predicted the future values of your data set

    • 6 tahun yang lalu
    1. interservices
      Penganjur Peraduan
      • 6 tahun yang lalu

      If I remember well, you took multiple columns as volumes, while there is only only with volume as described in this clarification board.

      • 6 tahun yang lalu
    2. interservices
      Penganjur Peraduan
      • 6 tahun yang lalu

      If I remember well, you took multiple columns as volumes, while there is only one column (the last) with volume, as described in this clarification board.
      Furthermore, please clarity your return over 10.000$ initial investment, considering the specified costs on every buy and sell

      • 6 tahun yang lalu
  • kannasanthosh
    kannasanthosh
    • 6 tahun yang lalu

    Hi mate.. Please make this a sealed contest. I am ready with my entry. Do let me know if I should submit as it is.

    • 6 tahun yang lalu
  • visweswaran1998
    visweswaran1998
    • 6 tahun yang lalu

    See Entry 2 which has 100% of prediction

    • 6 tahun yang lalu
    1. interservices
      Penganjur Peraduan
      • 6 tahun yang lalu

      Hi Visweswaran!
      as wrote yesterday in this public clarification board, the headers of the 12MB file are:
      date time open highest lowest close volume

      where...
      open is the price of the first trade of the minute (at the beginning)
      highest is the highest price traded during the minute
      lowest is the lowest price traded during the minute
      close is the price of the last trade of the minute

      • 6 tahun yang lalu
  • lourdesreynaldo
    lourdesreynaldo
    • 6 tahun yang lalu

    Good evening, could you detail what's in each column? The first is a date, the second is a time, but I do not know if the next four are prices, and the last volume? Thank you.

    • 6 tahun yang lalu
    Lihat 1 mesej lagi Sedang diproses...
    1. interservices
      Penganjur Peraduan
      • 6 tahun yang lalu

      open is the price of the first trade of the minute (at the beginning)
      highest is the highest price traded during the minute
      lowest is the lowest price traded during the minute
      close is the price of the last trade of the minute

      • 6 tahun yang lalu
    2. lourdesreynaldo
      lourdesreynaldo
      • 6 tahun yang lalu

      Thank you.! I start working on the project.

      • 6 tahun yang lalu

Tunjukkan lebih banyak komen

Bagaimana untuk mulakan dengan peraduan

  • Siarkan peraduan anda

    Siarkan Peraduan Anda Cepat dan mudah

  • Dapatkan berjuta penyertaan

    Dapatkan Bertan-tan Penyertaan Dari serata dunia

  • Anugerahkan penyertaan terbaik

    Anugerahkan penyertaan terbaik Muat turun fail - Mudah!

Siarkan Peraduan Sekarang atau Sertai kami Hari Ini!