Should Delisted Stocks Be in Your Backtest? Depends What You Trade.
Dave and I disagree on this one. We're both right. Here's where the line sits.
Bed Bath and Beyond went bankrupt. Then Overstock bought the ticker and became a completely different company trading under the same symbol. If your backtest doesn’t know the difference, your data is lying to you.
Helena asked a simple question on Line Your Own Pockets this week: should you include delisted stocks in your backtesting database? Dave and I came in ready to fight about it. We ended up in a surprisingly clean answer.
If you swing trade, yes. Full stop.
The overnight risk is where it matters. Companies go bankrupt. Stocks get halted and never reopen. Firms get acquired at 200% premiums. If you’re short a stock that gets bought out overnight and that ticker isn’t in your backtest because it was delisted years ago, your backtest never showed you the trade that could have ended your career. That’s not a fringe case. The regional bank crisis of 2023 wiped out names that looked perfectly tradeable the day before.
Norgate Data handles this. They keep every delisted symbol going back to the 1950s, appended with the delisting date. If you’re running S&P 500 constituencies from 20 years ago, Norgate knows what was added and removed. It’s the cleanest solution I’ve found for daily data.
If you day trade, probably don’t worry about it.
Dave’s argument is solid here. When you’re making 30 trades a day and closing flat every night, one edge case where a ticker gets weird is a rounding error inside thousands of trades. His philosophy: build strategies that minimize exposure to these problems instead of trying to solve every corner case. He calls it “a fart in a hurricane.” Direct quote.
The real backstop either way is daily reconciliation. I have an AI process that runs AmiBroker and RealTest against his broker statement every night and writes a report comparing fills. Pass/fail isn’t “did I make money today.” Pass/fail is “did my live trading match the backtest.” That distinction is the whole difference between systematic and discretionary confidence. A discretionary trader measures success by P&L. A systematic trader measures success by how closely reality tracked the model. You can lose money and still pass.
The systems I run at Stats Edge have been reconciled against live fills since day one. When the reconciliation flags something ugly, I publish it in The Drawdown Memo. No hiding.
The free 25-Year Backtest PDF at letters.statsedgetrading.com walks through the data integrity process behind 52,000+ trades, including how I handle survivorship bias. This episode is basically pages 4-5 come to life.
For the real-time alerts built on that clean data, that’s Stats Edge Pro at $149/month with a 30-day money-back guarantee.
— Michael Nauss, CMT, CAIA, CDMS