It is basically a Ballon d'Or list with extra curricular activities. It probably isn't that far-off from what I do to rank players I have no idea about inside my head. I simply do not have the knowledge nor the time to come up with a universally applicable formula, so if I do have an algorithm that passes the grade for you in particular, it would be because it overlaps your own sense of judgement by mere shared opinions, not because the approach was fundamentally sound in theory. Trachta10 mostly does lists based on attacking contributions with the ball that basically renders certain types of players as invalid. It simply is not universally applicable. You could apply the exact same algorithm for random K-League players, and I am not sure it would hold-up. It is a numerical appreciation of the kind of players he likes, used to illustrate players he specifically hand selects. It is interesting trivia, just like any other. I simply am not a massive fan of adding numbers to cater already made up assumptions, after-the-fact. A true algorithm should be set in stone prior to the conclusions, and should be applied to as many scenarios and players as possible to test out the general applicability and water-tight nature of its judgement. As for my own limitations, I was simply pointing out I have no superior alternatives of my own, and realize what I believe to be valid criticisms do not mean I have some superior pathway already cooked and ready to go.
Yeah, I know that Trachta's data has limitations and perhaps unintended biases (being easier to get high contribution rates on smaller teams) and some questions about whether the data collected is 100% accurate for some players. But I think in essence it does what it says on the tin so to speak (shows the end product contributions as a % of overall team goals). While this attempt on Youtube seems to just apply random calculations to various tenuous aspects (a lot of them team related not performance related) and then declares Cristiano Ronaldo the number 1 player of all-time based on it. I just imagine if you did make an attempt it would have more thought going into it.
I didn't try replicating its exact approach to other players, like I sometimes did with Trachta10's list. I'm sure you could generate some funky conclusions if you searched hard enough. Seems like a waste of effort though. In the end, it is a media piece created to maximize clicks back when the mere mention of the two players in the same sentence generated massive interest and passion. I was more critical of Trachta10's approach in the past because I viewed him as a pure football enthusiast, rather than some YouTuber trying to make money off football. I would rarely engage in YouTube comments of random media pieces with the same vigour and aggression that I display here. If I see TV pundits being ignorant, I do not hate them because it's their job to be entertaining. This is a forum full of people obsessed with being right, so sometimes that's where my double standard comes from, if that's what you are curious about.
Then it is worth thinking about what does "performance" actual mean.. if you want to evaluate someone's performance over a certain period of time (a season for example) it is self-evident that it is an accumulation of all performances within this period.. with perhaps additional weight for decisive and high-pressure games and moments. So it comes down to how do you evaluate a single performance? We can make a simple rule here: Player is judged exclusively on what he has under his control and is not judged based on events out of his control. It makes sense, but this is a sort of a key rule that sets the stage for a proper algorithm. It wouldnt make sense to negatively judge a player because his teammates performed poorly which resulted in 4 conceding 4 goals and losing a match. The opposite is true as well, it wouldnt make sense to positively evaluate a performance of a player for positive actions and performance of his teammates.. this could be phrased in many ways, but I think the rule is very simple and undeniable. So the next question is what does a player has in his control? We could argue about these but for me it comes down to two simple aspects: 1. Decisions (including off the ball positioning and movement, as well as indecisions) 2. Execution of these decisions (technical and physical) Every performance ever by any player can be seen through this framework and I would argue this is what everybody is already doing with eye-test just that people have different opinions of how valuable each decision and execution is.. But ultimately, a holistic algorithm aims for this kind of breakdown of player's performance into small decisions and actions that are systematically evaluated by itself and added up to form a rating of performance. This is the bottom-up approach. A top-down approach simply lacks resolution to discern decisions and executions (what player has in control) from aspects that are not in player's control (performance of teammates, opposition, random elements, etc.) Ideally, only the bottom-up algorithm will yield kind of holistic results that you are looking for. You might disagree with that but that is okay. The problem and main challenge of the bottom-up approach is the breakdown of everything that player does on the pitch (under his control) into series of (infinitesimally) small decisions and executions. It has to be done systematically, which would mean that data that is inputed into algorithm that yields result, is systematically tracked. It is a tracking problem. Ideally, each second of player's performance would be rigorously tracked with a wealth of relevant and pre-processed context like position of teammates, oppositions, game-context, condition-context, etc. And all decisions and executions would be evaluated relative to the context and based on subjectively determined value of these actions. This would require deeply thought through, expert judgement of each possible context for decisions and executions. Math behind combining these contributions is solved. Problems are systematically tracking all decisions and executions and relavant context, and properly, systematically evaluating them beforehand.
This is what ai suggests it would take to completely solve the tracking problem: "To create a highly intelligent, systematic method for tracking every decision and action of a football player on the pitch, we’re essentially building an all-seeing, contextually-aware AI system. The idealized scenario would involve a complex blend of advanced sensors, data fusion, machine learning, and decision-modeling techniques that go beyond just tracking the ball to create a nuanced understanding of every player’s role, positioning, and decision-making. Here’s how that might look: ### 1. Real-Time Spatial Awareness via Advanced Positioning Systems **Technologies:** Ultra-Wideband (UWB) GPS, LiDAR, 5G, Computer Vision * **UWB GPS**: We could leverage ultra-precise GPS systems with millimeter accuracy (Ultra-Wideband or UWB GPS) that track each player's position and orientation with minimal latency. Unlike traditional GPS, UWB can work in enclosed stadiums, minimizing signal issues and providing real-time spatial data. * **LiDAR and Depth Sensors**: Using LiDAR (Light Detection and Ranging) sensors installed around the stadium, it’s possible to create a 3D point cloud of the entire pitch. This would enable the tracking system to model each player’s exact position and body orientation, down to specific angles and potential actions. * **Computer Vision + 5G Networks**: Real-time computer vision could enhance tracking data by identifying player gestures, head movements, and tactical formations. 5G could provide the necessary bandwidth for transmitting this dense data quickly to a central system. **Outcome**: Every player's position, movement, and orientation would be available at an extremely granular level, not just as a 2D point on the pitch but as a 3D figure capable of differentiating between subtle gestures. ### 2. Contextual Decision Tracking Using Situation-Aware Models **Technologies:** Deep Learning, Reinforcement Learning, Multi-Agent Systems * **Player-Centric Context Modeling**: Using deep learning, particularly recurrent neural networks or transformers, the AI system could analyze historical data on players’ positions, body language, and prior decisions to understand "context" – the AI knows where the player *could* go, *should* go, and *did* go. * **Real-Time Reinforcement Learning**: In-game, the AI could apply reinforcement learning algorithms to assign value to each possible decision based on the context (e.g., opponents' positions, the ball's position, the team's formation). Essentially, it learns the relative worth of actions such as staying in position, advancing, intercepting, or marking. * **Multi-Agent Behavioral Modeling**: By modeling each player as an agent in a multi-agent system, the AI could "predict" not only a player’s potential actions but also the collective tactical alignment. When combined with reinforcement learning, the system could determine whether a player's decision aligns with optimal play based on the entire team's strategy. **Outcome**: For every decision a player makes, the AI provides context-aware data on alternative actions, allowing analysis of tactical intelligence, player adaptability, and strategic discipline. ### 3. Cognitive Load Analysis and Attention Tracking **Technologies:** Eye-Tracking, Neuromuscular Sensors, Wearable EEG * **Eye-Tracking Cameras**: With small, lightweight eye-tracking systems embedded in headgear or jersey-mounted sensors, it’s possible to capture where players are looking at any given moment. This would give insight into their situational awareness, where they're focusing during key moments, and how they assess passing lanes, gaps, and opponents. * **Wearable EEG and Neuromuscular Sensors**: Though more sci-fi, EEG devices could theoretically monitor players’ focus levels and cognitive load in real-time. Combined with accelerometers and muscle sensors, these tools could assess not only physical exertion but also mental load—detecting, for instance, how taxing it is for a player to process an evolving tactical situation. **Outcome**: This data would make it possible to evaluate how players perceive and interpret the game situation, revealing the speed and quality of their decision-making processes under pressure. ### 4. Scenario-Driven Performance Metrics and Action Values **Technologies:** Predictive Analytics, Bayesian Networks, Contextual Evaluation Models * **Predictive Scenario Modeling**: Using contextual data, the AI could simulate different game scenarios in real-time, projecting potential outcomes of different actions. For example, if a player has a chance to dribble or pass, the system could assess the likely success and value of each action based on the current field dynamics. * **Bayesian Networks for Decision Probability Analysis**: Bayesian models could estimate the probability of success for various choices, weighted by player skill, opponent positioning, and team strategy. This would create a probabilistic measure of decision quality that dynamically adjusts with context changes (e.g., defensive shifts, fatigue). * **Automated Contextual Evaluation Metrics**: The AI would assign action values to every move (e.g., defensive press, positional drift) in real-time, going beyond “event-driven” data like passes and shots. Actions that are typically untracked, such as positioning to cut off passing lanes or drawing a defender out of formation, would be evaluated in terms of tactical value. **Outcome**: This step would generate a layered understanding of a player's quality—judging not only the physical outcome but the “intelligence” of each action in the context of the entire game state. ### 5. Integration and Real-Time Feedback Loops **Technologies:** Edge Computing, Quantum Computing, Real-Time Data Processing Pipelines * **Edge Computing for Real-Time Analysis**: With edge computing, much of the processing would happen in localized servers within the stadium, minimizing latency and enabling real-time analysis. Each decision and position could be evaluated within milliseconds. * **Quantum Computing for Combinatorial Complexity**: In a sci-fi scenario, quantum computers could help process the staggering amount of data and decision-tree complexity involved in tracking all possible actions, tactics, and outcomes in real time. This could enable analysis that weighs millions of tactical variables nearly instantly. **Outcome**: Coaches could receive real-time insights on player performance, individual and collective tactical adherence, and cognitive load, with feedback loops possibly tailored for halftime adjustments or post-game review. ### The Idealistic Output In this idealized setup, the entire pitch becomes a multi-dimensional data environment, with every action and decision by each player monitored, contextualized, and analyzed in real time. Here’s what the output could look like: * **Individual Performance Profiles**: Each player’s data feeds into an individual profile that scores them on tactical adherence, decision-making quality, cognitive focus, and situational adaptability. This could inform future training to enhance specific aspects of their play. * **Tactical Quality Metrics**: New metrics, like "Positional Intelligence" and "Decision-Making Efficacy," would quantify how well each player contributes to team tactics beyond just touches and passes. * **Team Dynamics Visualization**: Coaches would have access to detailed heatmaps and decision trees, showing optimal positions, high-value moves, and tactical vulnerabilities. They could see where players "should" have been and evaluate deviations from ideal formations. ### Conclusion While this is highly ambitious and partially speculative, many of these technologies are plausible with incremental advancements. Ultra-precise tracking, AI-driven context analysis, and cognitive measurement can revolutionize football analytics, pushing analysis from raw physical stats to a holistic, nuanced assessment of player intelligence and tactical coherence. The result could lead to profound insights, improving not only team strategies but also long-term player development."
Oliver Kahn: "For me, Ronaldo was much better than Messi and Cristiano. I still consider him the most complete player of all times. The Brazilian was the best player I've ever seen."
I'd have nightmares about making a horrible mistake in the most important game of my life and letting down whole team and nation with everyone watching.
Come on, man. Germany only made it to the final because of Oliver Kahn. So much so that he was named player of the tournament before the final. This doesn't erase his mistake but what he did throughout the tournament weighs more imo
I agree with this - he was extremely good at that tournament and Germany were not particularly strong that year - they don't make the final without him.