Important Disclaimer

This tool presents statistical analysis of NTSB accident data for experimental amateur-built aircraft. The data is provided for informational and educational purposes only.

This data should not be the sole basis for any aircraft purchase, build, or flight decision. Aviation safety depends on many factors not captured here, including pilot training, recency, weather decision-making, maintenance practices, and individual judgment.

Key limitations of this data:

  • Accident rates are fleet-level statistics — your personal risk depends heavily on how you fly, not just what you fly.
  • Small sample sizes for some aircraft types make their statistics unreliable. A single accident can change a type's fatal percentage significantly.
  • No per-type flight hour data exists for homebuilt aircraft, so true exposure-adjusted rates cannot be calculated.
  • Accident narratives were individually read and classified using an LLM, then fully audited with 614 corrections applied. This is not an official NTSB classification.

The analysis uses an initiator-based approach — identifying the first event in the accident chain rather than the NTSB's probable cause. This is an analytical framework, not an official NTSB classification.

If you are considering building or purchasing an experimental aircraft, consult experienced builders, flight instructors with type-specific experience, and your local EAA chapter. No website can substitute for hands-on guidance.

We believe safety data should be free and accessible to everyone in the experimental aircraft community. This tool is free to use and always will be. If you find it valuable, consider supporting the project.

Experimental Aircraft Safety

About & Methodology

How this data is collected, classified, and what it means

Data Sources

All accident data comes from the NTSB Aviation Accident Database (bulk download at data.ntsb.gov). Fleet registration data comes from the FAA Aircraft Registry, with per-type fleet sizes derived from Ron Wanttaja's annual registry analysis (2000-2025). The dataset covers 1982-2026 and includes only aircraft flagged as Experimental/Amateur-Built (E-AB) operated as personal aircraft.

Initiator-Based Classification

The NTSB probable cause often emphasizes pilot actions downstream in the accident chain. For example, when an engine fails and the pilot stalls during the forced landing, the probable cause may focus on "failure to maintain airspeed" rather than the engine failure that created the emergency.

This tool applies the initiator-based classification methodology developed by Ron Wanttaja in his "Safety Is No Accident" series for Kitplanes Magazine. His decades of work analyzing homebuilt accident data established the framework this tool automates. The approach identifies the first major event in the causal chain. In the example above, the initiator is the engine failure, not the subsequent stall. This reframing typically increases the proportion of accidents attributed to engine and mechanical factors.

Each of the 8,817 accident narratives was individually read and classified by an LLM (Claude), not by regex or pattern matching. The full dataset was then audited — every classification checked against the original narrative — resulting in 614 corrections. Fifteen cases that remained ambiguous after the audit were resolved through manual human review. Many additional classification judgment calls were made during the audit process. The final classifications were cross-checked against published research and NTSB benchmarks for consistency.

Key Metric: Fatal Accident Percentage

The primary comparison metric is fatal accident percentage— when an accident occurs in this aircraft type, how often is it fatal? This avoids the flight-hours data gap (nobody tracks how many hours each homebuilt type flies per year). A type with a high fatal percentage means its accidents are more likely to kill, regardless of how frequently they occur.

Limitations & Caveats

  • No per-type flight hours. We cannot calculate true accident rates per 100,000 flight hours. Fatal percentage is a proxy.
  • FAA registry may undercount homebuilts. Some homebuilts have blank airworthiness codes and may not be counted in the experimental category, meaning actual fleet sizes could be larger than reported.
  • Make/model normalization is imperfect. ~61% of accidents matched to canonical types. The rest are one-off designs.
  • Small sample sizes for rare types. Types with fewer than 20 accidents should be interpreted cautiously.
  • NTSB narratives vary in detail. Preliminary reports may lack probable cause text.

References

  • Ron Wanttaja, "Safety Is No Accident" series, Kitplanes Magazine
  • Ron Wanttaja, annual FAA Registry fleet size analysis (2000-2025), personal correspondence
  • NTSB Safety Study: "The Safety of Experimental Amateur-Built Aircraft" (NTSB/SS-12/01, 2012 — covering 1984-2009 accidents)
  • NTSB Bulk Data: data.ntsb.gov/avdata
  • FAA Aircraft Registry: faa.gov/licenses_certificates/aircraft_certification/aircraft_registry

Support This Project

We believe safety data should be free and accessible to everyone in the homebuilt aircraft community. This tool will always be free to use.

Every accident narrative in this dataset was individually read and classified — not by pattern matching, but by careful analysis of what actually happened. If this tool helped you make a more informed decision about building or buying an experimental aircraft, consider supporting the project.

Support This Project

100% of donations go toward data updates, hosting, and keeping the tool free.