When Data Goes Silent: Lessons from a Failed Analysis Pipeline

At the end of a match, I usually rewatch the footage and cross-check it against the stats sheet. But this time, what I received was not numbers that spoke, but a blank space. No title, no source, not a single data point. Every model I built collapsed from the first line. The question was not which team won, but: what happens when our information system fails before the match even begins? In volleyball, we often talk about gaps on the court. The gap between two outside hitters when the opponent plays a high ball. The gap behind the block when the setter delivers a quick set. But there is another gap that few notice: the gap within the analysis process itself. A good analyst does not just read the match; they must also read the reliability of the data they have. When the input data is empty, every conclusion built afterward is built on sand. I once witnessed a coaching staff meeting where the analysis team presented a 40-page report, but when the head coach asked about the team's perfect-pass rate in the third set of the previous match, no one could answer. It turned out the stats software had a failed update, and all the defensive play data was taken from an unofficial friendly match. The model was still beautiful, the charts still smooth, but the foundation was hollow. That is a type of systemic failure: not a player hitting out, not a coach choosing the wrong lineup, but the information supply chain having broken at its root. In professional volleyball, every tactical decision depends on a continuous data chain. The ace-to-error ratio indicates serving pressure. Effective blocks per set reflect the block's ability to read attack direction. Perfect-pass rate is a direct indicator of first-touch system quality. When one of these links is missing, the entire tactical picture becomes distorted. But worse than missing data is not knowing we are missing it, and still proceeding to make judgments. The coaching staff I once worked with had an unwritten rule: never make a judgment about a player without watching at least three matches under different match-up conditions. This rule may sound extreme, but it stems from a simple reality: volleyball is a sport of linked action chains. A successful attack can originate from a perfect pass, an accurate set, and a smart run. If we only look at the final spike, we only see the tip of the iceberg. When the analysis process fails, the consequences do not stop at one wrong report. They spread into how the coach reads the match. A coach who believes their team has a high block rate will play more aggressively on the wings, leaving the middle open. A coach who believes their team has a high serving ace rate will take more risks on decisive serves. These beliefs, if based on wrong data, lead to wrong decisions. And in volleyball, one wrong decision at set point can decide the entire match. There is a paradox in how we treat systemic errors. When a player hits out, we can see it immediately, analyze it, and fix it. When a data process fails, that failure is often invisible. It does not appear on the scoreboard. It is not mentioned in the press conference. It only surfaces when the results on the court do not match what we predicted. And by then, it is usually too late. In the 2026 season, when the pandemic postponed matches, I had time to rebuild the expected goals model for the team I was working with. The data came from 45 matches of the previous season. But upon review, I discovered a problem: 8 of those 45 matches were recorded manually, with errors that could reach 15% in position-related metrics. If I had not caught this, the entire model would have been biased toward overrating the team's defensive capability. That was when I learned an important lesson: verify first, conclude later. No data is immune to error. No model is perfect. The only thing we can control is the verification process. Every number needs to be cross-checked against at least two sources. Every conclusion needs to be validated against footage. And every report needs a section dedicated to its own assumptions and limitations. A loss is more like a puzzle than a verdict. But to solve that puzzle, we need enough pieces. When the input data is empty, the puzzle is no longer about tactics. It becomes a puzzle about our own process. And sometimes, the most honest answer is to admit we do not have enough information to answer. In volleyball, people often talk about decisive plays. But there is another type of decision, less often mentioned: the decision not to conclude when there is not enough data. That is a difficult decision, because it goes against the pressure to have an answer immediately. But it is also the right decision. When the whole world believes in the champion, I only look at the link that is cracking. But to see that link, I must first be sure I am looking at the right machine.