1 of 18

Sruthi Machina | Kevin Fu | Dhruvi Kothari | Ricardo Jénez

1

“Enable The Individual Investor To Make Better Long-Term Equity Decisions”

2 of 18

2

WHAT ARE WE DOING AT

3 of 18

Problem

3

Problem 1 : Product Issues

Current AI/ Machine Learning efforts geared either towards trading (not long-term investments) or is not built for individual investors

Problem 2: Conceptual Profitability Issues

If there exists a service such as stock-rankings which help individual investors in making long-term decisions, they are not transparent or require a lot of prior market studying.

Current Gap

A product that’s data-driven, automated, and transparent which assists individual investors in making sound and quick long-term investment strategies and gives insight as to what those decisions are based on.

4 of 18

Hedge Funds continue to underperform the SPX by 50%+

4

-WSJ

5 of 18

One year expected stock appreciation and dividends representing total return for the stock as percent relative to invested capital.
These percent returns are then compared to similar annualized returns from a specific index, and then ranked against all other stocks.

5

Solution

Generate a list of five stocks that will outperform the S&P 500 over any one year period that individual investors can use to outperform the market and hedge funds.

6 of 18

Competitive Market Insights

StockPickAI is working in a challenging market area – predicting long-term performance of stocks using fundamental and other market data – that has not been (publicly) addressed by many companies

-Trevor Anderson (Sequoia Capital)

-Joe Mirza, CFA (MIDS ‘22, HawksHead Capital)

6

7 of 18

Inputs

Historical Prices of 7000+ Securities

Financial Statements

(10-K, 10-Q, 8-K)

Outputs

Ranked Stocks

XGBoost

Models

Feature Analysis

Solution: Product Architecture

Fundamental Ratios

7

8 of 18

Our Solution vs. Peers

*HFRX Global Hedge Fund Index, Returns over 20 years: 1998-2018: 4% per yr (vs. ~8% SPX)

https://bookshelf.corestone.ch/wp-content/uploads/2018/11/HF-and-PE-Factsheet.pdf�**https://www.forbes.com/advisor/investing/how-to-invest-in-hedge-funds/, https://smartasset.com/financial-advisor/citadel-advisors-review#:~:text=Citadel%20Advisors%20Minimum%20Account%20Size,minimums%20at%20their%20discretion%2C%20though.

8

	Us (StockPickAI)	Hedge Funds	Quant Funds
Outcome Variable	A Probability	Binary (Buy or Sell)	Binary (Buy or Sell)
Measurement of Success	Outperf SPX every year	Outperf SPX over 10 years	Outperf SPX over 10 years
Time period of prediction	1 year	10 years	Seconds/Minutes/Days
Main Indicators Used	Fundamental	Fundamental	Technical
Returns	≥1.5x SPX returns over 10 yrs	0.5x SPX returns over 10 yrs (HRFX Global HF Index)*	No measurable index equivalent
Transparent	Yes	No	No
Uses Machine Learning	Yes	No	Yes
Min Investment Req	$1200	$100,000 - $2M**	$1-10M (Citadel)**

9 of 18

9

10 of 18

User Guide / Example for Illustration

10

3 Easy Steps:

StockPickAI picks 5 stocks to buy every month (60 stocks per year). Investors buy those 5 stocks every month
Investors sell each “class” of 5 stocks one year after initial purchase.
Investors reinvest proceeds from the previous 5 stocks into the 5 new recommendations.

Example for Illustration:

Kevin

Thanks Dhruvi! So just to illustrate how all this is going to work, we have a simple example:

Lets imagine today is 1/1/2020. Our model picks 5 stocks: AAPL GOOGL etc. Since most of us are graduating here, let’s call those “The Class of Jan 2020”. An Investor will invest $100 in each of those 5 stocks.

Next month, its 2/1/2020. Our model again picks 5 stocks: AMZN SQ, etc, and that will be “The Class of Feb 2020”. Again the Investor invests $100 each of them as well.

Repeat this same process every month until Dec 2020.

It is now 1/1/2021. The Investor will now sell “The Class of Jan 2020”, which is the AAPL, GOOGL, etc. This time, the Investor will reinvest those proceeds into our model’s 5 new picks (WMT, Target).

Repeat this for Feb 2021 - Sell the 5 we picked on Feb 2020 (AMZN SQ), and reinvest into the new picks.

Repeat this process unemotionally for 20 years and you should see something like….

—---------------------------------------------------------------------------------------------

If a potential investor has $1200 to invest. Divided by 12 months, you have $100 total per month. Divided by our top 5 picks, its $20 to each stock.

11 of 18

Results: ~20 Year Performance (2004 - 2022)

11

StockPickAI

Avg Return:

+19.9% per year

S&P 500 Avg Return:

+8.9% per year

Assumes no compounding.

Results also published on website

StockPickAI: +19.9%

S&P 500: +8.9%

StockPickAI:

+$4227

S&P 500:

+$1891

Kevin

-Headline = StockpickAI generates roughly Double the S&Ps returns over 20 years. 20% annual returns (on avg) vs. 9% for the S&P.

-This Number is higher if you include compounding, but we excluded it to just make the math easier.

-Our Goal: Drastically Outperform the S&P during Up periods, Match/Tie S&P during Down periods.

-We want to emphasize that in order for this to work, we need to do this consistently over the long term. Eg. If you bought in 05 and sold in 08, your results are not so good, but if you did this over 20 years, your results are likely very good.

—------------------------------------------------------

1990-2002 train (rolling 10 yrs), 2002-2003 validation (1 yr), 2004- test set (1 month predicting on next year).

Rolling 10 Yr Train Period: 1/1/2010 - 1/1/2020, 1 Yr Validation Period: 1/1/2020 - 1/1/2021, 1 Month Test Period: 1/1/2021 - 2/1/2021, Predict top stocks for 3/1/2022

Model Avg Test Acc: 63.18% (where the model actually outperforms and beats the SPY)

12 of 18

Results - Recession: 2008 Performance of Top 5 Picks

12

13 of 18

Results - Recovery: 2009 Performance of Top 5 Picks

13

14 of 18

Results - Top 10 Features per SHapley Additive exPlanations*

14

~50 out of 130 features were significant

<1% difference in accuracy
3x higher run time

Top 10 Features are:

Historical Price (Momentum)
Price/Operating Earnings (pre-tax operating earnings)
R&D as a % of sales
GICS Sub-Industry
GICS Group
Free Cash Flow/Operating Cash Flow
Volatility of Stock

* SHAP -https://towardsdatascience.com/shap-explained-the-way-i-wish-someone-explained-it-to-me-ab81cc69ef30

15 of 18

Results - Detailed

Data

Daily stock prices, monthly fundamental ratios, quarterly & annual financial data from 1990.

5400+ stocks with varied lifetime on US stock exchanges (market cap greater than $300MM)
~19.5M rows (daily dataset) -> ~700K rows (monthly dataset) -> ~225K rows remain for training over 10 years, ~2K test/evaluation size for each month period.

Model

XGBoost Classifier model using RandomSearchCV to identify top model with Shap Analysis

Identifies most important contributors to model performance over a 10 year period
Top features change over every 10 year period (business cycle is ~7-10 years)
Rolling 10 Yr Train Period, 1 Yr Validation Period, 1 Month Test Period to predict top stocks for next year

Predict top 5 stocks that are expected to outperform the S&P 500 over the next year from current date

Based on probability derived from model

15

16 of 18

Feedback

“Knowing the critical factors for each stock’s performance is essential for analysts and investors…StockPick’s model with explainability for each stock pick would be insightful for the experienced and novice stock picker.”

-Trevor Anderson (Sequoia Capital)

“Where StockPickAI differentiates itself is the ability to explain the choices offered to investors. With a proven track record of performance with their models, a solution like theirs would be a valuable asset for retail, institutional, or fund investors.”

-Joe Mirza, CFA (MIDS ‘22, HawksHead Capital)

16

17 of 18

Conclusion

Product-market fit (ready for market)? √

Something people want to use? √
Productively addresses a specific need? √

Transparent and explainable? √
Net positive towards helping investors (ESG)? √

Equal access for all √
No redlining √

Clear pathway to profitability? √

17

Kevin

-Thanks Ricardo! So in conclusion, we think have something that has solid product-market fit, something people want to use, and addresses a specific need.

Stockpickai clearly addresses the investors’ overall Desire to outperform the market.

-In terms of ESG, we also have something that is Transparent and Explainable so that investors arent just throwing money into a black box.

Our model clearly explains what variables are causing which stocks to outperform.

In addition, our product is truly for all – unlike hedge funds, no gates, no minimum account size, no redlining - Everyone actually gets an Equal opportunity to participate.

-Thank you all for your time and attention today – we hope you all invest with us! (go to nxt slide)

—------------------------------------------------------------

Next Steps:

-Sharpe ratio

-Exact numerical figures that correspond to what each feature’s change means for actual financial statement movement

-Vertical Expansion Opportunity: Fundamentals/Financial Statement Prediction

18 of 18

Thank You!

stockpickai.com

18

Sruthi Machina

Associate Data Scientist

Ricardo Jenez

Startup Advisor

Dhruvi Kothari

Data Scientist

Kevin Fu

Senior Associate