All tested models lost the vast majority of their alpha when tested out of sample. For example, DeepSeek 3.2's strategy generated +20.7% annualized alpha in-sample, and -1.0% out-of-sample.
Bigger models had bigger drops in out-of-sample performance. More parameters just meant more memorization.
Full writeup: https://t.co/h8rOKjAWnk
A new benchmark tested commercial LLMs' ability to pick stocks in two matched time periods: one inside their training window, one after.
Meaning, your LLM's alpha might be mere memorization.
Serious investors ask every data provider the same question: “Is your data point-in-time?”
When the answer is a credible yes, they lean in, and often pay a large premium.
But why is point-in-time data so important? And why do so many providers who claim to have it, still face deal-killing skepticism and long trials?
In our latest post, we break down why “point-in-time” has become the standard for top-tier data — and why so many still fail this basic test.
https://t.co/ncB9ACRBDs
Excited to share that @validityBase is now integrated with @QuantConnect!
Send QuantConnect signals directly to validityBase and turn them into globally credible live indices with easy-to-share performance dashboards.
A big step toward helping any QuantConnect user showcase the value of the signals, strategies and research they create.
Huge thanks to the QuantConnect team and @jaredbroad for making this possible.
https://t.co/3aeUKmN6b6
Congrats to our intern Quant Research Analyst Mathew Thiel (https://t.co/EKi1kPZ0PL) for winning first place at the 2025 Society of Quantitative Analysts (SQA) Alphathon! 🏆
We’re excited to share that @Forbes featured validityBase in their latest piece on the growing demand for verifiable, point-in-time data and track records.
Proud to see our mission: helping data vendors and investment managers build trust through verifiable data, gaining broader attention.
Link here: https://t.co/GKCZzZJKQM
Next up in our Quant Summer Lunch & Learn series:
Dr. Lisa Borland on modeling extreme market events — from the GFC to COVID — and how quants think about fat tails.
🗓️ June 20, 12pm ET
🔗 Webinar Link: https://t.co/W3YdWq646V
🎙️ Co-hosted with @BlueWaterMacro
All welcome.
Excited to announce our partnership with @HaykGrigor and SharpeMachine, LLC -- an innovative investment manager running systematic strategies with a laser focus on capital preservation!
We’re excited to help bring SharpeMachine’s work to a wider audience through fully transparent, live performance records.
You can now track SharpeMachine's strategies via our live verifiable dashboards, much as you would a stock or an ETF.
https://t.co/9MWgzx70gD -- flagship
https://t.co/Fm9Lf4uSWV
https://t.co/2Uk68YHbQf
https://t.co/z0Y5GG20Zx
https://t.co/dAjQYBYihH
https://t.co/46bgpFsGkA
Read our press release for more details about the collaboration:
https://t.co/d4BdjMe9he
To explore the strategies further, reach out to @HaykGrigor
Thanks @MarcBautis - it was a real pleasure joining your show and sharing how validityBase helps to make investment performance 3rd party verified and globally credible.
How does blockchain ensure data integrity without exposing sensitive information?@DanAverbukh from @validityBase breaks it down in Episode 238 of The Agent of Wealth Podcast. [https://t.co/KJygIgiZG2]
Turn your backtest into a verified live track record, complete with stunning dashboards and insightful analytics. 📈
Showcase your strategies to investors or build internal benchmarks.
Learn more: https://t.co/8awDAbTBBu
Many traders believe that if they can get good results, investors will beat a path to their door.
This can be true but with some caveats. Investors have to trust the results, which can be a major challenge.
Creating a verifiable trading track record is shockingly difficult.
To do it, you have to convince the person on the receiving end that
a) What you're showing is complete and tamperproof. If they suspect you can go back in time to change any element of your record or that you can present an incomplete record (e.g. missing trades, accounts, start times, etc), your track record loses all credibility.
b) You haven't cherry-picked the best of your many outcomes. How do you show that you haven't run other strategies the recipient isn't seeing?
Plus, you have to do this while making the data easy to parse and analyze. Few people will spend hours parsing statements, trade data, etc.
Many aspiring traders and money managers believe they can use their brokerage statements to validate their track record to potential investors.
Unfortunately, this approach lacks credibility and is unpersuasive to most employers and capital allocators.
Why invest blood, sweat, and tears in building a strategy for beating the market only to have no one believe you did it?
validityBase has built a better approach that is lightweight, low-cost and globally verifiable.
https://t.co/AUDd2RlzdG