What Makes a Pass Well Timed?
Quantifying optimal pass timing with tracking data, OBSO, and the PAUSA framework.
The following summary critically reviews the research paper titled “Valuing La Pausa: Quantifying Optimal Pass Timing Beyond Speed” by Minho Lee, Geonhee Jo, Miru Hong, Pascal Bauer, and Sang-Ki Ko. All data, figures, and analysis presented here are drawn from their original work; I do not claim any authorship or ownership of the content. This summary has been written to provide a concise and technically informed synthesis of the paper’s findings, methodologies, and implications, while maintaining fidelity to the authors’ intellectual contributions.
1. Introduction
The paper starts from a simple but important gap in football analytics: models have become strong at evaluating where the ball should go, but they still struggle to evaluate when it should be released. Frameworks such as Expected Threat, VAEP [1], EPV [11], and OBSO [4] help quantify the spatial value of actions or locations, but elite passing is also a timing problem. A pass can be directed to the right area and still arrive too early or too late.
This is where the authors challenge the usual “faster-is-better” logic. Metrics such as Pace of Play [2] reward quick ball circulation and treat time on the ball as something to minimise. That assumption captures part of elite football, where faster leagues often require quicker decisions and execution. But it misses the tactical value of “La Pausa”: the deliberate delay that allows a player to move defenders, open a lane, or wait for a teammate’s run to become valuable.
The paper therefore frames time as a strategic resource. Instead of asking whether a player released the ball quickly, the authors ask whether the player released it at the value-maximising moment. Figure 1 illustrates this idea as a full pipeline: event data and tracking data are synchronized, the potential Off-Ball Scoring Opportunity is calculated across a temporal window, and the pass decision is decomposed into two components, Temporal Judgment and Spatial Selection. This is important because the framework does not simply reward players for waiting. It evaluates whether waiting, or not waiting, improved the value of the eventual pass.
The proposed contribution is named PAUSA, short for Passing Ability Under Spatiotemporal Awareness. The authors present three main methodological advances.
First, they use the ELASTIC framework [3] to align event and tracking data at frame level, identifying exact timestamps for “Pass Receipt” and “Pass Release”.
Second, they use OBSO [4] as a continuous value function and decompose each passing action into a timing component, the “When”, and a target-selection component, the “Where”.
More on OBSO:
Third, they validate the resulting metric using Bundesliga tracking and event data, showing that PAUSA relates more strongly to team strength than traditional speed-based metrics and can separate players who create value through quick circulation from those who create value through strategic delay.
2. Related Work
The related work section positions pass timing as part of a broader sports analytics problem. In elite sport, very small temporal windows can separate successful and unsuccessful actions [5, 6, 7, 8]. Timing has been studied in more discrete settings such as baseball batting and tennis returns [9, 10], but football passing is more complex because the player must coordinate time, target location, teammate movement, opponent pressure, and the changing value of space.
Existing football research has often treated speed as a proxy for temporal quality. Pace of Play [2], for example, assumes that better players minimise time on the ball. The authors argue that this can overlook tactical intelligence, because a delayed pass may be optimal if it manipulates the defence or creates a better receiving option. In this sense, “La Pausa” is not the opposite of efficiency; it is a different form of efficiency.
The paper also builds on spatiotemporal valuation models such as EPV [11] and OBSO [4], which evaluate the value of spatial configurations. However, these approaches usually treat the action timestamp as fixed. They ask what the best option was when the player acted, but not whether the player should have acted earlier or later. PAUSA fills this gap by treating execution timing as a decision variable and evaluating the pass against the best opportunity available within a feasible temporal window.
3. Methods
The methods section formally introduces PAUSA, a metric designed to quantify the quality of a pass decision across both space and time. The authors begin by emphasizing the need for high temporal fidelity. Since the framework depends on comparing the actual pass moment with nearby counterfactual moments, rough event timestamps are not enough. The paper therefore uses ELASTIC [3] to synchronize event data with continuous tracking trajectories and identify precise frame-level timestamps for pass receipt and pass release.
When a player receives the ball, the value of potential pass options evolves over time. At each frame, the model can estimate the best available passing opportunity using OBSO. The player’s actual pass is then compared against the best opportunity available across a feasible temporal window. This allows the authors to separate two different questions.
Did the player choose a good moment to pass?
Given that moment, did the player choose a good target?
3.1 Value Function: Off-Ball Scoring Opportunity (OBSO)
The authors use Off-Ball Scoring Opportunity, originally proposed by Spearman [4], as the value function underlying PAUSA. This choice is important because OBSO does not evaluate only the ball carrier. It evaluates how the full spatial configuration of players and ball contributes to scoring potential. In the context of pass timing, that makes OBSO useful because the value of waiting depends on off-ball movement, defensive coverage, and the probability that a team can control the ball in valuable space.
Given a game state D, containing positions and velocities of all players and the ball, OBSO estimates the probability of scoring if the next on-ball event occurs at a target location r. The model combines three components: Scoring S, Control C, and Transition T. The total probability of scoring is written as:
Here, P(S_r∣D) represents the probability of scoring if the attacking team possesses the ball at location r, P(C_r∣D) represents the probability that the attacking team controls the ball at that location, and P(T_r∣D) represents the probability that the next event occurs there. For a specific pass location and time, the value is defined as:
This gives the paper a continuous spatial value surface at each moment in time. The key point is that PAUSA can then track how that surface changes before and after the actual pass release.
3.2 Temporal Judgement (When).
Temporal Judgement is the first major component of PAUSA. It measures whether the player released the ball at the right moment, independent of whether the actual target was the best target. The authors explicitly link this to “La Pausa”, defining the problem as one of identifying whether the chosen timing was theoretically optimal within a feasible window around the action.
Let t_act be the actual time of the pass. The model constructs a temporal search window W, with a backward margin δ_pre and a forward margin δ_post. The player cannot pass before receiving the ball, so the lower bound is constrained by t_receipt:
Temporal Judgement then compares the maximum potential OBSO value at the actual pass time with the maximum potential OBSO value available anywhere inside the temporal window:
This ratio is high when the actual pass was released close to the best possible time. It is lower when the player acted before or after the peak-value moment. Importantly, this component assumes optimal spatial selection at the actual pass time. It asks only whether the timing was good, not whether the player found the correct receiver.
3.3 Spatial Selection (Where)
Spatial Selection addresses the second part of the decision: whether the player selected the best target at the moment they passed. The authors note that a player may time the action well but still miss the highest-value option. Conversely, a player may select the best local option at a poor moment. By separating timing and target choice, PAUSA can identify different forms of decision inefficiency.
Spatial Selection is defined as the ratio between the OBSO value of the actual pass endpoint and the maximum OBSO value available at the actual pass time:
Here, r_act is the coordinate where the completed pass was received. A high value means the player directed the pass close to the best available location at that moment. A low value means that even if the timing was acceptable, the player chose a lower-value target.
3.4 PAUSA (Passing Ability Under Spatiotemporal Awareness)
PAUSA combines the two components into one metric. It is defined as the product of Temporal Judgement and Spatial Selection:
Because the maximum potential value at the actual time appears in both components, the expression simplifies to:
This simplified form has a clear interpretation. PAUSA measures the value of the actual completed pass relative to the best possible pass opportunity available across the feasible timing window. But the authors stress that the decomposition remains the main theoretical strength of the framework. It can distinguish between different mistakes. A player might have found the right moment but failed to select the best target. Another might have selected the best target available at the time but acted too early or too late.
4. Experimental Results
The experimental section validates whether PAUSA captures meaningful football performance rather than merely producing an elegant theoretical score. The authors test the metric at both player and team level, comparing it with established baselines such as Pass Completion Rate, Pace of Play, and Expected Threat.
4.1 Dataset
The dataset consists of 55 German Bundesliga matches from the 2023-2024 season, including the open-source dataset released by Bassek et al. [16]. It combines event data with optical tracking data for all 22 players and the ball at 25Hz. This high sampling frequency is important because the method depends on short timing windows and frame-level changes in spatial value.
The authors apply several filters to make the evaluation methodologically valid. They include only successful passes and crosses, based on SPADL action types, because the actual endpoint must be known and controlled by the attacking team. Set-pieces are excluded because they lack the same dynamic receiver trajectory and make the simulation of the “waiting phase” less meaningful. The analysis is also restricted to cases where the receiver is on the same team as the passer, since assigning realized OBSO value to intercepted passes would be ambiguous in this framework.
4.2 Implementation Details
The temporal search window is asymmetric. The authors set δ_pre=3 seconds and δ_post=1 second. The backward margin is designed to cover most of the holding phase, with the paper noting that it captures over 80% of open-play passes in the dataset. The shorter forward margin reflects uncertainty in simulating player trajectories and pitch-control dynamics too far into the future.
The preprocessing follows ELASTIC [3]. Tracking coordinates are transformed into a standardized 105m×68m pitch with play aligned from left to right. Positional noise is smoothed using a Savitzky-Golay filter before velocity and acceleration are derived. Dead-ball frames are removed so that the analysis focuses only on active gameplay.
4.3 Results
The results are organised around two validation questions.
How does PAUSA relate to existing player-level metrics?
Does team-level PAUSA correlate with match performance?
Figure 2 provides the key correlation analysis. On the left, the paper compares player-level PAUSA components with Pass Completion Rate, PoP, and xT for players with at least 150 completed passes. On the right, it compares team-aggregated metrics with Points Per Game. The figure is central to the empirical argument because it shows that PAUSA is not just another expression of speed, completion, or xT. It captures a different dimension of passing quality.
4.3.1. Player-Level Evaluation: Correlation with Baseline Metrics
At player level, Pass Completion Rate has a positive correlation with Temporal Judgment, r=0.307. The authors interpret this as partly positional. Centre backs often receive the ball in lower-pressure build-up contexts, which gives them time to execute stable passes at suitable moments. However, because these actions usually target safer and lower-value areas, they tend to score lower on Spatial Selection and therefore do not necessarily produce high PAUSA values.
Pace of Play shows a weak positive correlation with Temporal Judgment, r=0.200, but a negative correlation with PAUSA, r=−0.286. This is one of the paper’s most important findings. It suggests that having more time on the ball can help players find a better moment, but delay without spatial gain becomes costly. The paper is not arguing that slow play is always good. Instead, it argues that the value of time depends on what the delay creates.
Expected Threat has a moderate positive correlation with Spatial Selection, r=0.404, but a negative correlation with Temporal Judgment, r=−0.484. This makes intuitive sense. xT is primarily spatial, so it rewards passes into valuable zones, but it does not assess whether the player acted at the best possible moment. The authors also find a strong negative correlation between Spatial Selection and Temporal Judgment, r=−0.499. This reflects a genuine football trade-off. High-value spaces are often more pressured and harder to access at the perfect time, while low-pressure situations allow better timing but usually offer less spatial reward. PAUSA is designed to identify the rare players who can overcome that trade-off.
4.3.2. Team-Level Evaluation: Correlation with Match Outcomes
At team level, the authors compare aggregated PAUSA with Points Per Game calculated only from the matches included in the dataset. This avoids comparing a partial tracking sample with full-season league performance.
PAUSA shows the strongest positive correlation with Points Per Game, r=0.733, outperforming Completion Rate, r=0.469. Pace of Play, by contrast, has almost no relationship with Points Per Game, r=−0.033. This result supports the paper’s central hypothesis: the speed of circulation alone is not what matters. A team benefits from slowing down only when that delay reflects tactically valid regulation. Similarly, a team benefits from playing quickly only when speed captures the correct window before it closes. PAUSA is therefore presented as a better indicator of collective attacking quality because it evaluates timing relative to opportunity rather than treating tempo itself as the target.
5. Application
After validating the metric, the paper turns to interpretation. The authors use PAUSA not only as a performance score, but as a way to profile player styles. By decomposing the metric into Spatial Selection and Temporal Judgment, they show how different players create value through different forms of spatiotemporal awareness.
5.1 Elite Player Profiling and Team Performance Analysis
Figure 3 plots players and teams according to Spatial Selection on the x-axis and Temporal Judgement on the y-axis, with PAUSA represented by marker size. The most interesting region is the upper-right quadrant, where players combine high spatial value with strong timing. The paper identifies Exequiel Palacios and Granit Xhaka as examples of effective playmakers in this area. These players are described as bridging stability and creativity, combining “Strategic Vision” with the “La Pausa” required to release the ball at the optimal moment.
The team-level plot gives this interpretation broader tactical meaning. Bayer 04 Leverkusen cluster strongly in the high-spatial, high-temporal region, which the authors connect to their historic undefeated Bundesliga campaign. The paper interprets this as evidence that Leverkusen’s dominance was not simply about circulating the ball quickly, but about repeatedly identifying and exploiting optimal spatiotemporal windows.
The same figure also reveals positional structure. Centre backs tend to cluster in the upper-left quadrant. They often have high temporal freedom because they operate under lower pressure, but they usually select safer, lower-value targets. Midfielders and forwards, by contrast, operate in denser and more pressured zones, which can push them into lower-temporal regions. The most valuable players are those who can still produce high spatial value without losing timing quality.
5.2 Temporal Judgement Style
Figure 4 extends the analysis by examining timing bias. The left plot compares PAUSA with Time Difference, defined as the deviation between the actual pass time and the calculated optimal timing. Players closer to zero with high PAUSA are generally closer to the model’s ideal timing. But the distribution also separates different elite styles.
A positive Time Difference means the player tends to pass later than the calculated optimal moment. High PAUSA in this region reflects effective “La Pausa”, where the delay is not wasteful but productive. The authors mention Exequiel Palacios and Alejandro Grimaldo as players who can deliberately wait to stabilise play or manipulate the defensive structure.
A negative Time Difference means the player tends to release the ball earlier than the baseline. High PAUSA in this region reflects rapid anticipation rather than delay. Florian Wirtz and Atakan Karazor are used as examples of players who generate value by exploiting fleeting windows before they close. This is a useful nuance in the paper: PAUSA does not glorify slowness. It values the correct temporal adjustment, whether that means waiting or accelerating.
5.3 Spatial Selection Style
The Spatial Selection analysis focuses on Pitch Control Difference, defined as the difference between the control probability at the actual pass endpoint and the optimal location. Figure 4’s right plot compares this spatial-control difference with PAUSA.
High-performing players tend to be closer to zero on this axis and higher in PAUSA. This indicates that their actual passes are directed to locations where their team can realistically control the ball, rather than simply to theoretically valuable but inaccessible areas. The authors describe this as “Spatial Stability”. Palacios, Grimaldo, and Wirtz are highlighted as players who generate value by finding open spaces that preserve possession and attacking potential.
Larger negative values indicate passes to areas with lower control probability than the optimal option. The paper mentions Waldemar Anton in this region, suggesting a different spatial profile where the selected target may be less secure relative to the best available controlled space. The broader point is that PAUSA can help distinguish whether a player’s value comes from timing, target selection, or the combination of both.
6. Case Study
The case study section gives the metric tactical interpretability. Rather than only presenting aggregate correlations, the authors show two specific situations where the optimal decision requires opposite temporal behaviours. One scenario rewards waiting. The other rewards playing earlier.
6.1 Slower vs Faster
Figure 5 presents a spatiotemporal comparison between actual execution and optimal timing. In the first scenario, a red team counter-attack begins after a recovery in the middle third. Red 6 plays immediately to Red 27 at t_act, producing a PAUSA value of 0.523. At that moment, the OBSO heatmap shows value mainly in wide areas. But the counterfactual at t_act+1 shows that waiting one second would have allowed Red 2 to exploit emerging central space. The delayed scenario raises PAUSA to 0.690. This is the paper’s cleanest example of “La Pausa”: a short delay would have aligned the release with the peak of opportunity.
The second scenario shows the opposite. Blue 10 holds possession while Blue 5 runs into open space. The actual pass at t_act produces a very low PAUSA value of 0.008 because the optimal window had already passed and defenders had consolidated. The counterfactual at t_act−2.12 shows that an earlier pass would have reached Blue 5 in a more advantageous situation before the space closed, increasing PAUSA to 0.011. Although both values are low in absolute terms, the example demonstrates the same conceptual point: optimal timing is context-dependent. Sometimes the best decision is to wait. Sometimes the best decision is to release before the opportunity disappears.
7. Conclusion
The conclusion presents PAUSA as a shift away from evaluating passing through speed alone. The authors argue that the relevant question is not “how fast” a player acts, but “how well-timed” the action is. By combining tracking data, event synchronization, and OBSO, the framework decomposes passing into Temporal Judgment and Spatial Selection, allowing analysts to evaluate whether players choose the right moment and the right target.
Empirically, PAUSA correlates more strongly with team performance than Pace of Play and other baseline metrics. It also identifies distinct player profiles, separating those who create value through rapid anticipation from those who use strategic delay to manipulate defensive structures. The paper’s main contribution is therefore not simply a new passing score, but a broader analytical framing: time on the ball should not be treated as inherently good or bad. Its value depends on whether the player uses it to improve the opportunity.
Presented at the MIT Sloan Sports Analytics Conference
This research was also presented at the MIT SSAC 2026, where the author broke down the models, visuals, and implications in more depth.
If you want to see the framework explained directly by the person who built it, the full talk is worth your time:
Learn More
Here you have a direct link to get the free xG Mini-Guide, suited for those who want to fully understand the model in 5 minutes.
References
Lee, M., Jo, G., Hong, M., Bauer, P., & Ko, S.-K. (2026). Valuing La Pausa: Quantifying optimal pass timing beyond speed. Submitted to the 2026 MIT Sloan Sports Analytics Conference Research Papers Competition.
https://www.sloansportsconference.com/research-papers/valuing-la-pausa-quantifying-optimal-pass-timing-beyond-speed