A 4-for-1 split turns a $500 stock into a $125 stock overnight. Without adjustment your backtest sees a 75% crash and either triggers a stop or generates an enormous fake mean-reversion signal. Dividends are smaller but relentless: an unadjusted index series understates long-run equity returns by a couple of percent a year.
The standard approach is a back-adjusted adjusted-close series for return calculations, kept alongside raw prices for anything that depends on the actual traded level, such as whether a stock was above $5 or under a per-share commission threshold.
Spin-offs are where most data breaks. The parent's price drops by the value of the distributed entity, and vendors vary in how well they capture it. Check any large single-day drop in your universe against a corporate action calendar before treating it as a signal.
Related: adjusted-close, point-in-time-data, data-quality-check, outlier