What If Option Closing Prices Were Trustworthy? A Machine Learning Approach

Hi! Here's Iván with this week's exciting newsletter, brimming with insights and discoveries on building robust investment strategies and risk models using Machine Learning.

In this edition, I am presenting the following sections:

• 🕹️ AI-Finance Insights: I summarize two must-read academic papers that mix cutting-edge ML/DL with Asset Pricing & Quant Finance:

  • What If Option Closing Prices Were Trustworthy? A Machine Learning Approach

  • Forecasting Stock Crash Risk with Machine Learning

• 💊 AI Essentials: The section on top AI & Quant Finance learning resources: Today, I'm introducing one of the best free courses on Deep Learning (by Yann LeCun & Alfredo Canziani).

• 🥐 Asset Pricing Insights: In this edition, I share a novel method that enhances factor investing by applying machine learning prior to the sorting process.

Today’s Sponsor

These daily stock trade alerts shouldn’t be free!

The stock market can be a rewarding opportunity to grow your wealth, but who has the time??

Full time jobs, kids, other commitments…with a packed schedule, nearly 150,000 people turn to Bullseye Trades to get free trade alerts sent directly to their phone.

World renowned trader, Jeff Bishop, dials in on his top trades, detailing his thoughts and game plan.

Instantly sent directly to your phone and email. Your access is just a click away!

Subscribe for Free

“What If Option Closing Prices Were Trustworthy? A Machine Learning Approach”

The paper in a nutshell here... 🔻 

➡ The paper proposes a machine learning model to improve the accuracy of equity option closing prices, addressing issues like stale information and manipulation.

➡The model uses underlying stock prices from closing auctions to create a more reliable benchmark for option prices.

➡Significant deviations were found between traditional methods and the machine learning benchmark, averaging 35% to 47%.

➡The study suggests that adopting closing auctions for options, similar to equity markets, could enhance price efficiency and reliability.

“Forecasting Stock Crash Risk with Machine Learning”

The working paper in a nutshell here... 🔻

➡ The study explores the use of machine learning models to forecast stock crash risk, aiming to improve the identification of firms likely to face financial distress in both developed and emerging markets.

➡ The study leverages traditional financial indicators, like distance-to-default, combined with more granular financial data, to capture complex, nonlinear relationships that traditional models might miss.

➡ The results show that portfolios based on ML-predicted distress signals significantly underperform the market, with developed markets returning just 2.3% and emerging markets incurring a 1.5% loss compared to average market returns of 10.0% and 11.6%, respectively.

➡ The study suggests that incorporating ML techniques into stock selection processes could help investors enhance returns by effectively avoiding high-risk stocks, as demonstrated by improved portfolio performance when excluding stocks with the highest distress probabilities.

Sponsor 👇

Get value stock insights free.

PayPal, Disney, and Nike recently dropped 50-80%.

• Are they undervalued?

• Can they recover?

• Read Value Investor Daily to find out.

We read hundreds of value stock ideas daily and send you the best.

Subscribe with one click.

AI-Essentials

Arguably the best free course on Deep Learning, based on YouTube videos along with code examples. It caters to everyone from beginners to advanced students... and most importantly, it's designed and taught by Yann LeCun & Alfredo Canziani.

Full course here: https://atcold.github.io/NYU-DLSP21/?utm_source=theaifinancefrontier.beehiiv.com&utm_medium=referral&utm_campaign=identifying-market-regimes-with-machine-learning

Asset Pricing Insights

“Uncertainty-Aware Lookahead Factor Models for Quantitative Investing”

Instead of sorting by factors, let's forecast future factors using deep learning and sort based on those forecasts.

Here's a quick sum up of the paper in under 2 minutes: 👇

➡ This research breaks new ground by showing that if we could pick stocks based on future financial data (think earnings, debt, etc.), those portfolios would massively outshine those selected using today's standard models.

➡ By training advanced deep learning networks, the team successfully predicts these future fundamentals based on the past 5 years of data, creating what's termed as "lookahead factor models."

➡ These models not only forecast stock performance but also incorporate risk aversion through innovative uncertainty estimates, significantly boosting portfolio returns and risk-adjusted performance (Sharpe ratio).

➡ With an industry-grade simulator, the team proves that these enhanced models can deliver an impressive annualized return of 17.7% and a Sharpe ratio of 0.84, outperforming traditional models by a significant margin.

➡ The insights challenge the conventional wisdom in quantitative finance, pointing towards a new era where predictive analytics and risk management converge to optimize stock portfolio performance.

If you're enjoying our newsletter and want to support us, please recommend it to anyone you know who's interested in AI and Finance. Your referrals are the biggest compliment and help us grow! 🌟🤖💼