Introducing 

Prezi AI.

Your new presentation assistant.

Refine, enhance, and tailor your content, source relevant images, and edit visuals quicker than ever before.

Loading…
Transcript

A

Equities Research

C

3

Questions to Answer

-Expected return on a stock given revenue?

-Risk Metrics i.e. VAR and investment thresholds

Uncertainty in EPS, guidance, Revenue|estimate

-What stocks are suseptible to an earnings surprise?

-How could we tell if the edge is dissapearing?

Experimental Setup

Setup

Create a Subtopic

Customize the Cover

3

Insert Your Content

Questions to Answer

Modeling Approaches

ML Approach

-Gather features related to a stock (size, price history, fundamentals history, sector, etc.) and use these to predict price given revenue

-Cross validate using groups of stocks and groups of time periods as holdout to ensure proper generalization

-Predict each quarter for a range of possible revenue values

Time Series Approach

-Tends to be more theory driven, or particular treatment of a given stock

-Time series for each individual equity

-Testing Specific hypotheses for a given

Recommender Approach

-Loosely based on baseball's PECOTA

-Determine similarity metrics based on

History of revenue/price

Fundamental qualities of the company

-Pick KNN to the target equity

-Average reaction for target stock based on historical reactions of similar stocks

Important Factors

Past surprises

Immediate Reaction vs Drift

Other important factors

Lit Review

Past

Studies

Data Sources

  • Stock Prices/ Trade Volumes
  • Yahoo Finance
  • Fundamentals
  • Edgar/Finstr in R
  • Estimize reported actuals
  • Matched to 'ground truth' for 10Qs
  • Expectations
  • EPS- Zacks, Estimize
  • Revenue- Estimize
  • Outlook
  • Earning Reports and Transcripts from Seeking Alpha

Data Sources

Stocks Included

Every Equity Currently Listed on AMEX, NASDAQ, or NYSE

concerned about survival bias

Only included firms in 3 sectors & 20 Industries

Market Capitalization >=$100M

Data Preparation

MySql DB to integtrate sources

X% had info from All Sources

X% Missing Estimize (fixable with expecation model)

X% Missing prices (dropped)

X% Missing SEC Filing

X% Missing Transcripts

X% had truncated time period

Data Integration

Revenue and EPS

Expectations

Estimize where available

Time Series Model of Revenue

Change in Stock Price

Actuals

Agreement between sources

Accuracy of Expectations

Surprise Index

Prediction Targets

Look at multiple Periods

-2 days to +1 day

-2 days to +7 days

Movement in Stock Price

Raw Return

Adjusted for Volitility

Simple -1,0,+1 for Positive, Negative, Neutral return

Total Volume of Trades

Connect

Results

with Clients and Partners

Learn more about creating dynamic, engaging presentations with Prezi