Missing stocks that should be targeted

• Coarse-universe filter never checks if DollarVolume is positive or if the symbol is actually trading in the current backtest window; fundamental records with stale data (e.g., VENA) can slip through, so non-existent or delisted symbols appear.
• Every symbol passing the coarse filter is blindly added to self.closing_prices, even if no valid recent close exists for the current backtest date; this stores phantom or historical prices, enabling gap calculations on invalid tickers.
• self.closing_prices retains these stale or never-traded symbols and their last price indefinitely (only cleaned when dropped from the universe), so a non-trading ticker can satisfy gap logic many days after its last activity.
• SetupTargets samples ActiveSecurities only at 09:31. If a legitimate symbol (like KIDZ or FRGT) has not yet produced valid data (HasData is False), it is skipped and never reconsidered that day, causing genuine IPOs/late openers to be omitted.
• The algorithm never schedules a second pass, so any late-trading or IPO stocks can never make it into the trade candidate list after 09:31, while phantoms persist in memory.
• The line accessing first_minute_high = history.loc[symbol]['high'][0] assumes a single-level index, which often fails on QuantConnect’s multi-index DataFrame (may quietly skip valid triggers due to error/NaN).
• The for-loop building entry_triggers is mis-indented and only runs if final_targets exists, but the block is not actually part of the conditional; as written, the trigger creation logic may never execute, so no legitimate or phantom symbols make it to trading, which can explain missing expected tickers and erroneously tracking/storing phantoms.

Sources: Legacy Fundamental Universes - QuantConnect.com, US Equities - QuantConnect.com, Fundamental Universes - QuantConnect.com, All Shortable Symbols Coarse Selection Regression Algorithm, Coarse Fine Universe Selection Benchmark