%%{init: {'theme': 'dark'}}%%
xychart-beta
title "Average F-Bombs per Episode by Season"
x-axis [Season 1, Season 2, Season 3]
y-axis "Avg F-Bombs" 0 --> 15
bar [5.6, 8.83, 11.5]
How Often Does Roy Kent Say F*CK?
A Statistical Analysis of Ted Lasso’s Most Expressive Character
Executive Summary
ImportantKey Finding
Roy Kent’s coaching status significantly increases his F-bomb frequency, while his dating status has no measurable effect. This analysis replicates Julia Silge’s findings using bootstrap resampling, permutation tests, and negative binomial regression.
1 Introduction
Roy Kent is a fictional character from the Emmy-winning Apple TV+ series Ted Lasso, portrayed by British actor, writer, and comedian Brett Goldstein—who also serves as a writer and producer on the show. The character is a former Chelsea F.C. midfielder and Champions League winner who joins AFC Richmond in the twilight of his career, eventually becoming team captain, then assistant coach, and ultimately head manager after Ted Lasso’s departure.
Roy Kent is heavily inspired by real-life football legend Roy Keane, the notoriously intense former Manchester United captain known for his fiery temperament and no-nonsense attitude. Like Keane, Kent is a central midfielder who commands respect through sheer force of personality—and an impressive vocabulary of expletives.
This analysis examines how Roy Kent’s use of what appears to be his favorite word depends on his coaching status and/or his dating status.
The dataset was created by Deepsha Menghani for her excellent talk at posit::conf(2023) and is available in the richmondway R package. Our data is stored in SQL Server table DoctorWho.dbo.RoyKent.
1.1 Research Questions
- Total volume: How many F-bombs does Roy Kent drop across the series?
- Trend analysis: Does his swearing increase over seasons?
- Coaching effect: Does becoming a coach change his vocabulary?
- Dating effect: Does dating Keeley affect his swearing?
2 Descriptive Statistics
2.1 Overall Summary
Roy Kent’s F-bomb statistics across all 34 episodes:
| Metric | Value |
|---|---|
| Total F-Bombs | 300 |
| Unique Episodes | 34 |
| Average per Episode | 8.82 |
2.2 Trend by Season
| Season | F-Bombs | Episodes | Avg/Episode | Change from S1 |
|---|---|---|---|---|
| Season 1 | 56 | 10 | 5.6 | — |
| Season 2 | 106 | 12 | 8.83 | +58% |
| Season 3 | 138 | 12 | 11.5 | +105% |
TipTrend Insight
Roy Kent’s F-bomb frequency more than doubled from Season 1 (5.6/episode) to Season 3 (11.5/episode). This coincides with his transition from player to coach.
2.3 Top 5 F-Bomb Episodes
| Rank | Episode | F-Bombs |
|---|---|---|
| 🥇 | S2 E12 (Season Finale) | 23 |
| 🥇 | S2 E5 | 23 |
| 🥉 | S3 E4 | 17 |
| 4 | S3 E2 | 16 |
| 5 | S1 E9 | 14 |
3 Coaching Effect Analysis
3.1 Group Comparison
Observed Difference: +5.16 F-bombs per episode when coaching (+89%)
3.2 Bootstrap Confidence Intervals (BCa Method)
Using 10,000 bootstrap resamples with bias-corrected and accelerated (BCa) method:
| Group | Sample Mean | 95% CI Lower | 95% CI Upper | Bootstrap SE |
|---|---|---|---|---|
| Coaching = Yes | 10.95 | 8.0 | 23.0 | 4.28 |
| Coaching = No | 5.79 | 2.0 | 9.0 | 1.91 |
ImportantStatistical Interpretation
The confidence intervals show minimal overlap — the coaching group’s lower bound (8.0) nearly equals the non-coaching group’s upper bound (9.0). This strongly suggests a real difference in behavior.
3.3 Permutation Test Results
Using 10,000 random permutations to test if the observed difference could occur by chance:
| Metric | Value |
|---|---|
| Observed Difference | +5.16 F-bombs |
| p-value | 0.0001 |
| Significant at α = 0.05? | ✅ YES |
| Permutations | 10,000 |
ImportantVerdict: Coaching Effect is REAL
Out of 10,000 random shuffles, essentially none produced a difference as extreme as what we observed. There’s only a 0.01% probability this difference occurred by chance.
4 Dating Effect Analysis
4.1 Group Comparison
Observed Difference: -0.04 F-bombs per episode (essentially zero!)
4.2 Permutation Test Results
| Metric | Value |
|---|---|
| Observed Difference | -0.04 F-bombs |
| p-value | 1.00 |
| Significant at α = 0.05? | ❌ NO |
| Permutations | 10,000 |
WarningVerdict: Dating Effect is NOT Real
The p-value of 1.0 means the difference is pure noise. Whether Roy is in a relationship or single, his F-bomb output stays constant.
5 Negative Binomial GLM Regression
5.1 Why Negative Binomial?
F-bomb counts are:
- Count data (non-negative integers)
- Overdispersed (variance > mean)
- Best modeled with Negative Binomial rather than Poisson regression
5.2 Model Specification
\[\log(\mu_i) = \beta_0 + \beta_1 \times \text{Dating}_i + \beta_2 \times \text{Coaching}_i\]
Where \(\mu_i\) is the expected F-bomb count for episode \(i\).
5.3 Model Results
| Predictor | Coefficient (β) | Std. Error | z-statistic | p-value |
|---|---|---|---|---|
| (Intercept) | 1.604 | 0.195 | 8.24 | < 0.001 |
| Dating | 0.236 | 0.196 | 1.20 | 0.230 |
| Coaching | 0.721 | 0.205 | 3.51 | 0.0004 |
Model Fit Statistics:
- AIC: 204.96
- Theta (θ): 4.97
- Observations: 34
- Converged: ✅ Yes
5.4 Exponentiated Coefficients (Rate Ratios)
| Predictor | exp(β) | Interpretation |
|---|---|---|
| Intercept | 4.97 | Baseline: ~5 F-bombs when not dating & not coaching |
| Dating | 1.27 | 27% more F-bombs when dating (NOT significant) |
| Coaching | 2.06 | 106% more F-bombs when coaching (SIGNIFICANT) |
NoteModel Equation
\[E[\text{F-bombs}] = 4.97 \times 1.27^{\text{Dating}} \times 2.06^{\text{Coaching}}\]
6 Comparison with Julia Silge’s Analysis
This analysis replicates Julia Silge’s blog post from September 2023.
6.1 Results Comparison
| Aspect | Julia Silge (Poisson + Bootstrap) | Our Analysis (NB + Permutation) | Agreement |
|---|---|---|---|
| Coaching Effect | CI: [0.24, 1.04], excludes 0 | p = 0.0004, significant | ✅ |
| Dating Effect | CI: [-0.35, 0.45], includes 0 | p = 0.230, not significant | ✅ |
| Conclusion | Coaching ↑, Dating → no effect | Coaching ↑, Dating → no effect | ✅ |
TipReproducibility Confirmed
Same data, same conclusion, different tools. That’s reproducible science!
7 Final Summary
7.1 Complete Results Table
| Factor | Effect Size | Permutation p-value | GLM p-value | Significant? |
|---|---|---|---|---|
| Coaching | +5.16 F-bombs (+89%) / 2.06x rate | 0.0001 | 0.0004 | ✅ YES |
| Dating | -0.04 F-bombs (0%) / 1.27x rate | 1.00 | 0.230 | ❌ NO |
7.2 The Story Arc
%%{init: {'theme': 'dark'}}%%
timeline
title Roy Kent's Linguistic Journey
Season 1 : Player, Pre-Keeley
: 5.6 F-bombs/episode
Season 2 : Player→Coach, Dating Keeley
: 8.83 F-bombs/episode
Season 3 : Coach, Post-Keeley
: 11.5 F-bombs/episode
ImportantThe Bottom Line
Professional stress transforms Roy Kent’s vocabulary; romance does not.
The transition from injured footballer to assistant coach is the pivotal moment that unleashes Roy Kent’s full expletive potential. Managing Jamie Tartt, watching matches from the sidelines, and shouldering coaching responsibilities doubles his F-bomb output.
Meanwhile, Keeley Jones—despite being a calming, positive influence in his life—has zero measurable impact on his linguistic choices. Roy Kent is consistently, authentically, and statistically Roy Kent, regardless of his relationship status.
He’s here. He’s there. He’s every-f***ing-where. And coaching makes it worse. 🏈💢
8 Technical Appendix
8.1 Data Source
- Database: SQL Server (ZEN)
- Table:
DoctorWho.dbo.RoyKent - Original Source:
richmondwayR package by Deepsha Menghani
8.2 Statistical Methods
- Bootstrap Confidence Intervals: BCa method with 10,000 resamples
- Permutation Tests: Two-sample mean difference with 10,000 permutations
- Regression: Negative Binomial GLM with log link
8.3 Tools Used
- SQL Server for data storage and querying
- MCP Statistics tools for statistical analysis
- Quarto for document generation
Analysis inspired by Julia Silge’s blog post on the same dataset.