Kicking for Goal or Touch? An Expected Points Framework for Penalty Decisions in Rugby Union
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Statistics. The day in statistics: methods, theory and the inference under them.
Tom: We're looking at the paper: "Kicking for Goal or Touch? An Expected Points Framework for Penalty Decisions in Rugby Union".
Jane: Here's the gist of the paper. Following a penalty in rugby union, teams typically choose between attempting a shot at goal or kicking to touch to pursue a try The authors develop an Expected Points (EP) framework that quantifies the value of each option as a function of both field location and game context This study offers, to our knowledge, the first comprehensive EP-based assessment of penalty strategy in rugby union and outlines extensions to win-probability analysis and richer tracking data Framework Construction The core of the framework involves constructing two context-aware surfaces: (i) the expected points of a possession beginning with a lineout, estimated from phase-level data by field location, and (ii) the expected points of a kick at goal, which integrates a continuous angle–distance model of kick success together with the continuation value of misses These surfaces are then compared to produce decision maps that indicate where kicking for goal or kicking to touch maximizes expected return The decision quantity is defined as the difference in expected points, with positive values favoring a lineout Modeling Lineout Value The Expected Points of a Lineout (EPlineout) is estimated using a multiple linear regression based on possession sequences that begin with a lineout in a given field zone The model is specified as EPlineout = βˆ 0 + βˆ 1 · meter line + βˆ 2 · Card Diff + βˆ 3 · WinPct Diff The negative coefficient on meter line reflects the intuitive decrease in expected next-score value as the lineout moves farther away from the opposition try line Furthermore, a procedure is used to prevent double counting across consecutive possessions with effectively the same state and outcome by randomly retaining a single representative observation from each group of n consecutive lineouts Modeling Kick Outcomes The expected points of a penalty kick at goal are calculated as EPkick(d, θ) = Pmake(d, θ) · 3 + 1 − Pmake(d, θ) · EPmiss(d, θ) The probability of a successful kick is modeled as a smooth function of the kick’s distance (d) and lateral angle (θ), taking the form P(make d, θ) = logit−1 ·β0 + f(d, θ) The continuation value of a miss is approximated by identifying restarts that do not follow a change in score and computing their expected next-score value using the same EP framework as for lineouts Decision Mapping and Analysis The relative value of a lineout versus a kick at goal is determined by ∆EP(x, y; dtouch) = EPlineout(xLO) − EPkick(x, y), where xLO = max
5, x − dtouch: is the resulting lineout x-coordinate Decision maps are obtained by mapping this difference across the field for various assumed meters gained to touch (dtouch) Scenario analyses examine how the indifference frontier shifts with yellow cards and team quality, showing that stronger teams realize larger expected points from lineouts across much of the pitch A case study illustrates that sub-optimal decisions were relatively frequent but typically near the decision boundary, leading to a total regret in expected points that was modest relative to the final score margin Limitations and Future Directions The framework acknowledges several limitations, including combining phase-level data from club competitions with penalty-kicking data from international matches, which introduces potential inconsistencies in player quality Future work could develop fully competition-specific models by collecting phase-level and kicking data from the same league and year Additionally, the analysis restricts team choices to kicking for goal or kicking to touch, suggesting that extending the framework to additional decision branches like scrums would be a natural next step The study also notes that selection bias may influence results, as stronger teams are more likely to generate attacking lineouts in the opposition half Finally, incorporating player- and team-specific models for both goal-kicking accuracy and meters gained to touch could produce personalized decision maps REFERENCES arXiv:2512.00312v2 4 Lock, D.
Tom: and Nettleton, D.
Jane: (2014).
Tom: Using random forests to estimate win probability before each play of an nfl game Romer, D.
Paper summary: Tom: So, looking at the "Kicking for Goal or Touch? An Expected Points Framework for Penalty Decisions in Rugby Union" paper by Watts and Pipping-Gamón, the authors developed this method to quantify the value of a penalty decision based on field location and game context.
Jane: Essentially, they created a system that compares taking a lineout versus kicking for goal or kicking to touch by calculating expected points from both options using phase level data and an angle distance model.
Lu: The framework provides decision maps that show exactly where maximizing the return means choosing one option over the other based on those calculated differences in expected points.
Meng: It’s a unified, data-driven approach for evaluating penalty decisions, but they admit that combining different types of data from club and international matches can introduce inconsistencies related to player quality.
Lalam: The paper suggests future work could include building competition specific models using data from the same league and year, and extending the framework to include other decision branches like scrums.
Tom: It’s a solid foundation for understanding penalty strategy, especially since they used phase level data which is a unique way to look at this kind of sequence of events.
Jane: What this means for us is that we can use these kinds of models to evaluate complex on-field decisions in sports where the outcomes are not just simple scores but a series of connected events.
Conclusion: Tom: So, we're wrapping up our look at this paper on penalty decisions in rugby union, "Kicking for Goal or Touch? An Expected Points Framework for Penalty Decisions in Rugby Union."
Jane: It really boils down to taking a lineout versus kicking for goal or kicking to touch and seeing which option gives you the best expected points based on where you are on the field.
Tom: Right, so they built this whole system of expected points, comparing a lineout's value against a kick at goal's value as a function of distance and angle.
Lu: The core idea is creating two surfaces, one for the lineout expected points based on where you are in the field zone, and another for the kick at goal that smooth function of distance and angle, plus accounting for what happens after a miss.
Meng: From an engineering standpoint, it's interesting how they handle those inputs—they use phase level data to estimate lineout value and integrate a continuous angle distance model for kicks. But I wonder about the quality of that data input.
Lalam: The study shows how the expected points difference between a lineout and a kick changes across the field, creating these decision maps that tell you where you should aim your penalty choice.
Jane: And they found that stronger teams seem to realize bigger expected points from lineouts across more of the pitch, which is something I hadn't fully thought about before.
Tom: The conclusion is that this framework gives a clear way to map out which choice maximizes your expected return in those high-pressure moments.
Lalam: They also pointed out some limitations, like mixing data from club competitions and international matches, which could introduce inconsistencies regarding player quality.
Meng: Exactly, if you're trying to deploy this for real-time decision making, you need cleaner data sets that don't mix different levels of play or team standards.
Jane: So while the framework is a solid way to look at these decisions, the next step seems to be building models that are specific to one competition and use player-specific data instead of just team averages.
Tom: We'll take that into our next segment when we discuss how we could potentially extend this framework beyond just penalties, maybe looking at other high-stakes choices in a game.
Wharton School, University of Pennsylvania
stat.AP
Submitted: 2025-11-29
Updated: 2026-11-14
Comments: 20 pages; 9 figures, 8 tables. Submitted to the Journal of Quantitative Analysis in Sports (JQAS). Code & replication package: https://github.com/WhartonSABI/rugby-ep (data from a public source; mirrored in the repo with attribution). Preprint licensed CC BY 4.0
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 75/100
The gist: Following a penalty in rugby union, teams typically choose between attempting a shot at goal or kicking to touch to pursue a try The authors develop an Expected Points (EP) framework that quantifies…
Key concepts
- Expected Points (EP)
- A mathematical tool used to quantify the value of different rugby decisions. It estimates the total future points a team can expect from a given situation, considering factors like field location and game context, helping teams choose between lineouts or kicks.
- EPlineout
- The expected points gained from starting a possession with a lineout. This is modeled using regression based on where the lineout occurs on the field and other game variables like card differences, reflecting how valuable different starting positions are.
- EPkick
- The expected points from attempting a penalty kick at goal. This calculation combines the probability of making the kick (based on distance and angle) with the value of missing it, using a continuous model for kick success and continuation value.
- Decision Map
- A visual representation showing the difference in expected points between two options—kicking for goal versus kicking to touch—across the entire rugby field. Positive values indicate that kicking for goal is favored over kicking to touch at that specific location.
Terminology
Summary
Following a penalty in rugby union, teams typically choose between attempting a shot at goal or kicking to touch to pursue a try The authors develop an Expected Points (EP) framework that quantifies the value of each option as a function of both field location and game context This study offers, to our knowledge, the first comprehensive EP-based assessment of penalty strategy in rugby union and outlines extensions to win-probability analysis and richer tracking data
Framework Construction
The core of the framework involves constructing two context-aware surfaces: (i) the expected points of a possession beginning with a lineout, estimated from phase-level data by field location, and (ii) the expected points of a kick at goal, which integrates a continuous angle–distance model of kick success together with the continuation value of misses These surfaces are then compared to produce decision maps that indicate where kicking for goal or kicking to touch maximizes expected return The decision quantity is defined as the difference in expected points, with positive values favoring a lineout
Modeling Lineout Value
The Expected Points of a Lineout (EPlineout) is estimated using a multiple linear regression based on possession sequences that begin with a lineout in a given field zone The model is specified as EPlineout = βˆ 0 + βˆ 1 · meter line + βˆ 2 · Card Diff + βˆ 3 · WinPct Diff The negative coefficient on meter line reflects the intuitive decrease in expected next-score value as the lineout moves farther away from the opposition try line Furthermore, a procedure is used to prevent double counting across consecutive possessions with effectively the same state and outcome by randomly retaining a single representative observation from each group of n consecutive lineouts
Modeling Kick Outcomes
The expected points of a penalty kick at goal are calculated as EPkick(d, θ) = Pmake(d, θ) · 3 + 1 − Pmake(d, θ) · EPmiss(d, θ) The probability of a successful kick is modeled as a smooth function of the kick’s distance (d) and lateral angle (θ), taking the form P(make d, θ) = logit−1
·β0 + f(d, θ) The continuation value of a miss is approximated by identifying restarts that do not follow a change in score and computing their expected next-score value using the same EP framework as for lineouts
Decision Mapping and Analysis
The relative value of a lineout versus a kick at goal is determined by ∆EP(x, y; dtouch) = EPlineout(xLO) − EPkick(x, y), where xLO = max[5, x − dtouch] is the resulting lineout x-coordinate Decision maps are obtained by mapping this difference across the field for various assumed meters gained to touch (dtouch) Scenario analyses examine how the indifference frontier shifts with yellow cards and team quality, showing that stronger teams realize larger expected points from lineouts across much of the pitch A case study illustrates that sub-optimal decisions were relatively frequent but typically near the decision boundary, leading to a total regret in expected points that was modest relative to the final score margin
Limitations and Future Directions
The framework acknowledges several limitations, including combining phase-level data from club competitions with penalty-kicking data from international matches, which introduces potential inconsistencies in player quality Future work could develop fully competition-specific models by collecting phase-level and kicking data from the same league and year Additionally, the analysis restricts team choices to kicking for goal or kicking to touch, suggesting that extending the framework to additional decision branches like scrums would be a natural next step The study also notes that selection bias may influence results, as stronger teams are more likely to generate attacking lineouts in the opposition half Finally, incorporating player- and team-specific models for both goal-kicking accuracy and meters gained to touch could produce personalized decision maps
REFERENCES
arXiv:2512.00312v2 4
Lock, D. and Nettleton, D. p. 1 In that setting, WP and EP estimators are typically built using
statistical machine learning models (e. , Random Forests and gradient boosting), as is done
in Lock and Nettleton (2014) and Baldwin (2021).Where the paper says it
Romer, D. (2006). Do firms maximize? evidence from professional football
Burke, B. (2009). The 4th down study – part 3. p. 0 Our results provide
a unified, data-driven method for evaluating penalty decisions and can be tailored to
team-specific kickers and lineout units. This study offers, to our knowledge, the first
comprehensive EP-based assessment of penalty strategy in rugby union and outlines
extensions to win-probability analysis and richer tracking data.Where the paper says it
Baldwin, B. (2021). Nfl win probability from scratch using xgboost in r. p. 0 Our results provide
a unified, data-driven method for evaluating penalty decisions and can be tailored to
team-specific kickers and lineout units. This study offers, to our knowledge, the first
comprehensive EP-based assessment of penalty strategy in rugby union and outlines
extensions to win-probability analysis and richer tracking data.Where the paper says it
Martinez-Arastey, G., Datson, N., Smith, N., and Robins, M. (2025). Foundations of expected points in rugby union: A methodological approach. p. 1 , Random Forests and gradient boosting), as is done
in Lock and Nettleton (2014) and Baldwin (2021). In rugby union, public research is only beginning to close this methodological gap, which
can be largely attributed to a lack of publicly-available data.Where the paper says it
Monpezat, B. (2023). Predicting penalties and conversions success: Rugby world cup 2023 kicks probability insight. p. 1 , penalty kicks) and providing a reproducible benchmark rather than a deployable decision tool. Separately, public-facing analyses model goal-kicking success as a function of
distance and lateral angle, offering practically-useful (albeit context-free) estimates (Monpezat, 2023).Where the paper says it
Quarrie, K. L. and Hopkins, W. p. 3 5 5 1 -7
67 1 Harlequins Sale 1 Away -7 926 0 0 31. 3 Kicking Data
Our kicking data come from the analysis in Quarrie and Hopkins (2015), which studied 582
international rugby matches played between 2002 and 2011. Using phase-level data from the 2018/19 Premiership Rugby season
(35,199 phases across 132 matches) and an angle–distance model of penalty kick success
estimated from international records, we construct two surfaces: (i) the expected points
of a possession beginning with a lineout, and (ii) the expected points of a kick at goal,
taking into account the in-game consequences of made and missed kicks. We then
compare these surfaces to produce decision maps that indicate where kicking for goal
or kicking to touch maximizes expected return, and we analyze how the boundary
shifts with game context and the expected meters gained to touch.Where the paper says it
Where the paper says it
Yurko, R., Ventura, S., and Horowitz, M. (2018). nflwar: A reproducible method for offensive player evaluation in football. Journal of Quantitative Analysis in Sports. p. 0 Our results provide
a unified, data-driven method for evaluating penalty decisions and can be tailored to
team-specific kickers and lineout units. This study offers, to our knowledge, the first
comprehensive EP-based assessment of penalty strategy in rugby union and outlines
extensions to win-probability analysis and richer tracking data. This
means that, once within distance to make a penalty kick ∼ 60m from the opponent try line,
the decision following a penalty is routinely between taking a lineout or kicking at goal. While the choice is often straightforward near the opposition’s corner, given a certain point
difference, or late in a match, many game states are ambiguous and governed by convention,
1
arXiv:2512.Where the paper says it
Where the paper says it
Brill, R. S., Yurko, R. p. 1 perceived kicker ability, and the expected quality of a team’s lineout. In contrast, American
football has long relied on Expected Points (EP), Expected Points Added (EPA), and winprobability (WP) frameworks to guide fourth-down decision-making, translating field position
and context into expected returns and actionable recommendations (Romer, 2006; Burke,
2009; Yurko et al. In contrast, American
football has long relied on Expected Points (EP), Expected Points Added (EPA), and winprobability (WP) frameworks to guide fourth-down decision-making, translating field position
and context into expected returns and actionable recommendations (Romer, 2006; Burke,
2009; Yurko et al. In that setting, WP and EP estimators are typically built using
statistical machine learning models (e.Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Develop a decision-ready Expected Points framework for penalty decisions by comparing two context aware EP surfaces: (i) expected points of a possession beginning with a lineout, and (ii) expected points of a kick at goal, to produce decision maps that indicate where kicking for goal or kicking to touch maximizes expected return.
-
Implement an EP model for the lineout that uses multiple linear regression incorporating field location, Card Diff, and WinPct Diff, while mitigating bias from repeated states by retaining a single representative observation from groups of consecutive lineouts sharing the same run id.
-
Construct an EP surface for a penalty kick at goal by combining the probability of success estimated via a generalized additive model with the continuation value of a miss, where EPkick(d, θ) is defined as Pmake(d, θ) times 3 plus 1 minus Pmake(d, θ) times EPmiss(d, θ).
-
Create decision maps that show how location and assumed meters gained shape the optimal choice by calculating the difference in expected points: DeltaEP (x, y; dtouch) = EPlineout(xLO) - EPkick(x, y), for various assumed translation distances dtouch.
-
Implement scenario analysis capabilities to construct hypothetical decision surfaces under different game contexts by varying covariates such as manpower advantages and team strength differentials, demonstrating how the indifference frontier shifts with these factors.
Abstract
Following a penalty in rugby union, teams typically choose between attempting a shot at goal or kicking to touch to pursue a try. We develop an Expected Points (EP) framework that quantifies the value of each option as a function of both field location and game context. Using phase-level data from the 2018/19 Premiership Rugby season (35,199 phases across 132 matches) and an angle-distance model of penalty kick success estimated from international records, we construct two surfaces: (i) the expected points of a possession beginning with a lineout, and (ii) the expected points of a kick at goal, taking into account the in-game consequences of made and missed kicks. We then compare these surfaces to produce decision maps that indicate where kicking for goal or kicking to touch maximizes expected return, and we analyze how the boundary shifts with game context and the expected meters gained to touch. Our results provide a unified, data-driven method for evaluating penalty decisions and can be tailored to team-specific kickers and lineout units. This study offers, to our knowledge, the first comprehensive EP-based assessment of penalty strategy in rugby union and outlines extensions to win-probability analysis and richer tracking data.
Sources
Related papers
- Unraveling time-varying causal effects of multiple exposures: integrating Functional Data Analysis with Multivariable Mendelian Randomization
- Strategic Causal Policy Learning with Manipulable Eligibility
- BayesJudge: Uncertainty-Aware Bayesian Meta-Evaluation of Human and LLM Judgments
- Automating structural reliability analysis with a multi-agent large language model framework
- Improving Minority Population Sampling with BISG Probabilities: Evidence from a Survey of Jewish Americans
- When Do Sports Leagues Stop Being Competitive? Bayesian Detection of Competitive-Balance Eras from Match-Level Data