Compare & Understand Your Backtests

Welcome to Weekly Roundup #5 at ForecastOS! We're glad to have you here, following our journey :)

This newsletter is split into three sections: 1) thoughts / musings relevant to ForecastOS, 2) what we worked on last week, and 3) what's coming next.

1. Thoughts / Musings

Compare & Understand Your Backtests - Our New Screen

Performance often differs across backtests.

It can be challenging to understand why, as differences are introduced by a myriad of sources, including:

• Your production feature set
• Model hyperparameters / tuning
• Portfolio engineering / optimization methods, costs, and constraints
• Your investment universe
• Data revisions
• Etc.

With the above in mind, backtest performance differences are easy to understand with the right tools! Our new comparison tool allows anyone to easily understand and analyze dataset, feature, forecast, and portfolio engineering differences across backtests:

Email us to set up a 30-minute demo; we'll show you how easy it is to compare and understand your backtests!

2. Work: What Got Done

ForecastOS is a new startup. As such, we try to spend 100% of our time either building or selling. We log that effort here.

This past week, as it pertains to selling, we:

• Continued the outreach process for our unified AI management (AIM) platform; if you know anyone we should talk to, please let us know! We have capacity to onboard 1 additional client in Q4 2023

This past week, as it pertains to building, we:

• Added side-by-side backtest analytics functionality, as highlighted above

Other notable activities from this past week:

• Recovered from Salmonella food poisoning; cook your chicken y'all!
• Prepared for our first engineer - he starts today! More to come on Tom soon :)

3. Work: What's Coming Next

It's important to keep velocity high. We keep ourselves accountable by sharing what we hope to finish over the next week.

This week we will:

• Continue outreach for our unified AI management (AIM) platform
• Build, and prepare to deploy, our open-source (PyPI) package and associated API for consuming features and datasets through ForecastOS in less than 10 lines of code
• Start planning our first (financial) forecasting competition

Until next week!