Correlation Analysis

This is mostly just "new-boot-goofing" and not anything serious. It's a lot of films but for anything interesting to be said the films would need to be selected based on a defensible criteria. As with all the other sections of this site this site is not here to answer questions, but for me (or you) to browse and notice patterns that merit further investigation (once I can afford more runs!).

Response Length vs Errors

Spearman's ρ 0
n = p = 1

Insufficient data

Film Age vs Errors

Spearman's ρ -0.0112
n = 2490 p = 0.5749

Film age has little impact on AI accuracy

Average Error Score by Decade

1920s
4.0
1930s
7.7
1940s
9.8
1950s
6.7
1960s
7.7
1970s
6.4
1980s
6.7
1990s
7.6
2000s
6.8
2010s
6.4
2020s
7.7

Search Grounding Impact

With Search 4.04 1711 reports
vs
Without Search 13.59 816 reports
↓ 70.3% Reduction in errors
p = 0.0 Significant

Search grounding reduces errors by 70.3% (statistically significant)

← Back to Statistics Overview