Tradevo Data

Survivorship bias and lookahead bias get lumped together constantly. They're not the same failure mode. One deletes losers from your history. The other lets your model know things before they happened. Different bugs, different fixes. tradevodata.com/blog/survivors…

Point-in-time means: what did the market actually know, and when. Not the restated 10-K/A filed years later. On our 40-company sample, the average gap between fiscal period end and first filing was ~43 days. That gap is where lookahead bias hides. tradevodata.com/blog/point-in-…

Where lookahead bias actually hides: not in your model. In your merge. tradevodata.com/blog/lookahead…

Built this by pulling 10-Ks straight from EDGAR's XBRL API, not caching a fundamentals vendor's numbers. Wrote up the whole pipeline, warts included. tradevodata.com/blog/building-…

Survivorship bias and lookahead bias get lumped together constantly. They're not the same failure mode, and fixing one doesn't fix the other. Broke down the difference. tradevodata.com/blog/survivors…

Wrote up how lookahead bias hides in plain sight in fundamental backtests. It's not exotic. It's a join on the wrong date column. tradevodata.com/blog/lookahead…

40 companies, 7 concepts, first-filed value + restated flag, pulled straight from SEC EDGAR. Free, no login, no api key. Go find the lookahead bias in your current data provider yourself. github.com/christianpichi…

Point-in-time means: what did the market actually know, and when. Not the restated 10-K/A filed years later. On our 40-company sample, the average gap between fiscal period end and first filing was ~43 days. That gap is where lookahead bias hides. tradevodata.com/blog/point-in-…

Most fundamental backtests join price data to the fundamentals available today, not the fundamentals available on that date. That's lookahead bias. Part of your Sharpe ratio is fiction and you don't know how much. tradevodata.com/blog/lookahead…

40 companies, 7 concepts, first-filed value + restated flag, pulled straight from SEC EDGAR. Free, no login, no api key. Go find the lookahead bias in your current data provider yourself. github.com/christianpichi…

Your backtest 'knew' Q3 revenue the day the quarter ended. It didn't. Most fundamentals vendors timestamp by fiscal period, not by when the 10-K actually hit EDGAR. That's not data. That's a leak. tradevodata.com/blog/lookahead…

Your backtest doesn't know what today's date is. Join fundamentals by fiscal period-end instead of filing date, and you're feeding the model numbers it couldn't have seen yet. Not a bug. The whole backtest lying to you politely. tradevodata.com/blog/lookahead…