Guide / Market data
Adjusted close vs. raw prices in a backtest
Choose prices that match your backtest: a worked stock-split example, dividend adjustments, and checks that catch double counting.
Explanations & project notes
Blog
The ideas behind the research, explained in everyday language. Commentary revisits what we learned; dated updates record the work as it stood.
Follow new research and guides with RSS
Guide / Portfolio accounting
Ex-dividend date vs. payment date in a backtest
A dividend claim is not spendable cash. Follow a worked share-and-cash ledger from the ex-dividend date through payment and reinvestment.
Guide / Research methods
Look-ahead bias in backtesting: check when data was known
Find future information in a backtest with a worked release-time example, revision checks, and clear boundaries for training and test data.
Commentary / Market data
The dividend records we rejected and why
A lesson from a historical dividend audit: check that an event belongs to the security before using it to calculate returns.
Lab update / Model development
Vesper-400M: Implementation Status Before Training
The transformer is implemented, but corpus training has not started. Software-check results, the five-file dataset pilot, and the remaining training prerequisites.
Lab update
Building a Public Research Record
Our first two technical notes, the engineering questions behind them, and how we are documenting the work at Vesper Applied.