Skip to content

Implemented a LightLGM model 💫 capable of predicting the closing price movements for hundreds of NASDAQ listed stocks 🌱 using data from the order book & the closing auction for Optiver's Trading Competition 🚀

License

Notifications You must be signed in to change notification settings

beingamanforever/Optiver-Trading-at-the-close

Repository files navigation

Aman Behera --- Submission for VLG-Open-Project-2023/24 ---

Optiver-Trading-at-the-close

I implemented a simple LightLGM model capable of predicting the closing price movements for hundreds of NASDAQ listed stocks using data from the order book and the closing auction of the stock.for Optiver's Trading at the close competition.

image

In financial markets, an auction is a mechanism for determining the price of a particular asset by allowing multiple buyers and sellers to interact directly in a controlled, regulated environment. There are many different kinds of auction, but for this competition we are particularly interested in the concept of a closing auction.

In a closing auction, orders are collected over a pre-determined timeframe and then matched at a single price determined by the buy & sell demand expressed by auction participants. For Nasdaq Closing auctions, the exchange begins accepting orders at the start of the trading day and begins publishing the state of the auction book at 3:50pm ET for 10 minutes before the market closes at 4pm ET, at which point the orders are matched instantly at a single price.

image

The closing price is determined as: The price at which the maximum number of shares can be matched. In the event the auction would have equal number of matched lots at different levels, Nasdaq uses a proprietary algorithm which takes into account the last traded price, the price-time priority of orders, and the available liquidity at different price levels. It is quite rare that the price at which the maximum shares can be matched is not unique.

Dataset description

Understanding the terminologies mentioned in the dataset's description became quite essential for attempting the problem statement. This dataset contains historic data for the daily ten minute closing auction on the NASDAQ stock exchange. The challenge is to predict the future price movements of stocks relative to the price future price movement of a synthetic index composed of NASDAQ-listed stocks.

dataset_description

11

12

Special thanks to this discussion for providing a very simple explanation for the terminologies sued in contest.

Explanation of some essential terminologies

Many different terminologies like order book, auction book, market maker (like optiver, jane street), wap required explanations. Although theoritical explanations can be accessed from google or any other source, the following images serve to give a working insights into how trading at close happens. Lastly, since I had used quartiles for outlier detection an image explaning the same has also been added.

1

2

3

4

5

6

7

8

9

Evaluation metrics

15

Essential Resources and References used

Here are some of the resources I personaly used to get myself familiar with the terminologies and to understand the workings of trading at the close.

  1. Market Making
  2. Order types and order books
  3. Auction price decision framework
  4. Trading at close terminologies

Refer to the official documentation of Nasdaq to understand the terminologies associated with close

  • Like what is close?
  • Why doesn't trading happen 24x7 ?
  • Why market makers bring liquidity to the table?
  • What is buy sell disparity?
  • And other questionable terminologies

For understanding how stock market works do refer the following resources:

Papers for Reference

About

Implemented a LightLGM model 💫 capable of predicting the closing price movements for hundreds of NASDAQ listed stocks 🌱 using data from the order book & the closing auction for Optiver's Trading Competition 🚀

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published