Vietnamese Esports and the Data Problem: When Analysts Must Learn to Say 'Not Enough Information'

I remember that evening. After the grand final of a domestic esports tournament had ended, I didn't turn off my machine. I reopened the recording and started the stopwatch. Not to count kills — the official stat sheet had already done that for me. I counted something else: how many times a team repeated the exact same engagement pattern, the silence between two attack rhythms, and how long the opposing lineup took to react. Four hours later, I had a dense handwritten data table. But when I put the pen down, I noticed something uncomfortable: most of what I needed to analyze did not exist in any public data source. If I had not counted it myself, I would have had nothing to say beyond empty commentary. That was when I understood the problem did not lie in the match. It lay in the analytical frame itself. CONTEXT Vietnamese esports has moved past the stage where anyone could comment on feeling alone. Ten years ago, a match analysis could consist of a few lines about the most spectacular play. Today, audiences are used to stat tables: damage per minute, kill participation, gold difference at minute ten. These tables have value. The problem is they answer only half the question. The other half — the harder half — lies in what is not recorded. Why did the same team, in the third game, switch its attack direction to the opposite flank? Who called the signal? How long did the opposing lineup take to notice the change? These details usually exist in no official API. They only surface when someone sits down and counts. I came to esports from a different foundation. In 2026, I sat in the stands at My Dinh, running a stopwatch for the 4x400m relay at a national youth athletics meet. Hanoi finished second, 0.8 seconds behind the champions, and the cause was not speed — it was a botched baton exchange on the third leg, where the incoming runner started 2.1 metres earlier than the standard. That gap slowed the trajectory, and 0.8 seconds was the consequence. I logged the exchange rhythm step by step, built a handwritten data table, and wrote a long analysis. An editor shared it. That was the first time I saw raw data I had collected myself generate a real argument. In esports, I carried the same habit: splitting a play into segments, counting repetitions, and comparing. But I quickly hit a wall that athletics never had. In athletics, every run leaves an absolute number: time. In esports, data depends on who is measuring, how they measure, and — most importantly — whether they record the measurement process at all. That dependency creates a paradox. Esports has more data than athletics in quantity, but less data in reliability. A 400m lane has only one way to measure time. A teamfight in a game can be measured in dozens of different ways, and each way yields a different number. ANALYSIS When I build an analytical frame for an esports match, I split it into layers. The first layer is public data: patch information, schedule, match results, starting lineups. The second layer is semi-public data: stat tables from tracking sites, pick-and-ban rates, average match duration. The third layer is self-collected data: what I time myself, count myself, and tabulate myself. The problem appears when the first layer is empty. I have run into this more than once. After a tournament, I sat down to analyze a specific match, opened the archive, and found that the minimum baseline did not exist: no note of which game version was used, no detailed schedule, not even a fully published official roster. Not because the tournament was small. Because the record-keeping stage was skipped. At that point I had two options. Option one: write. Use memory, use feeling, use whatever I vaguely remembered, and fill the gaps with plausible-sounding speculation. Option two: stop, and say plainly that there is not enough information to conclude. Most people choose the first. It is easy, it is fast, and it produces something readable. But it violates the first principle of any decent analysis: every conclusion must trace back to a specific source. I start with a self-counted data table, because memory does not know how to yield to error. That sentence sounds like a slogan, but it is a technical rule. Human memory tends to smooth over anomalous details — we remember the best play, and forget the ten boring repetitions before it. A self-counted table has no such tendency. It records the boring ones too, and it is precisely in those boring ones that tactics get etched in. When a team repeats the same pattern seven times, they are not hoping for luck, they are engraving tactics into muscle. I have verified this many times. In one match I watched, a team kept funnelling play to one flank in its engagements. The official stat sheet recorded only 'right-side attacks: 12'. But when I counted them in waves myself, I found that 7 of them happened immediately after a major objective control phase — meaning they were not random, they were a rehearsed script. That 7 is a signal, not a coincidence. This is why I never begin with aggregated stat tables. Aggregate tables answer 'what'; I need 'how' and 'how many times'. HOW TO BUILD A SELF-COUNTED TABLE My process has four fixed steps. Step one: choose a single question. Not 'how does this team play', but 'how does this team open engagements in the first 15 minutes'. A data table only has value when it answers a narrow question. Step two: define the event before watching. If I am counting 'left-flank attacks', I must write out the definition of 'left-flank attack' before hitting play. Otherwise I will quietly adjust the definition mid-way to match what I am seeing — a form of bias that is very hard to detect. Step three: watch twice. The first pass for overall observation, the second for timing and note-taking. Often the second pass produces a result opposite to the first impression. Step four: cross-check against an independent source if one exists. If none exists, I note clearly that this is self-collected data, not cross-verified. These four steps sound simple, but they turn a match-watching evening into a disciplined data-collection session. THE RISK OF BAD DATA There is a type of error more dangerous than missing data: bad data presented as real data. I have seen stat tables circulating on social media with numbers nobody verified. One article claimed a player had an '85% teamfight win rate', but did not say how many matches the sample came from, over what period, or by what definition of 'teamfight win'. The 85% sounds convincing. It may also be meaningless. My principle is clear: if I cannot cite a source, and if I cannot state the method, that number does not exist in my analysis. This is why I log every reference and every calculation whenever I make a prediction. In 2026, when the pandemic halted most events, I did not sit idle. I built a database of 40 Vietnamese track-and-field athletes, tracking injury recovery times and competition frequency. With help from a sports-medicine postgraduate on physiology, I built an index called 'record-reproducibility capacity'. In early 2026, that index predicted athlete Nguyen Thi Oanh would break the national 3000m steeplechase record. It happened, with a time of 10:05.23. The point I want to stress is not that the prediction was right. The point is that I recorded every source and the calculation method, so that if I were wrong, others could point out where. That transparency is what separates a model from a rumour. MINIMUM INFORMATION PAYLOAD Back to esports. The industry has an advantage athletics lacks: nearly every match is recorded. But it has a parallel disadvantage: data is fragmented across too many sources, and no source is accountable for completeness. A decent esports analyst must work like an investigator. They must first define the 'minimum information payload' — the smallest set of facts required for a conclusion to be valid. For patch analysis, that payload is: game title, version number or update date, the specific changed element, and at least one independent data source such as official patch notes or a win-rate delta. Missing any of these, the conclusion should not be drawn. For tournament analysis, the payload is: tournament name, organiser, format, series length, participating region and dates. For roster analysis, the payload is at least one named individual plus the nature of the event: transfer, renewal, retirement, injury, or coaching change. This sounds obvious. But I have seen many analyses written missing nearly all of it. They open with a general claim, then fill the rest with phrases like 'from my observation' and no data table attached. The minimum information payload has a function more important than quality control: it separates 'no risk' from 'no information'. The two states look identical on paper but mean entirely different things. If an article mentions no problem about a team, that may mean the team has no problem, or it may mean the journalist did not look. An information gap is never a confirmation. PIPELINE INTEGRITY One concept I borrow from software engineering: pipeline integrity. If one stage in an analytical chain fails, every downstream result loses value, however perfect the remaining stages are. In esports analysis, the stage most likely to fail is the first: baseline data collection. When this stage is empty, every later layer — however elegantly written — is speculation dressed in technical language. This is why I spend more time on preparation than on writing. A solid baseline table matters more than a good opening line. And when the baseline is absent, the right move is to stop and say so, rather than continue with whatever is at hand. 0.8 seconds is never just 0.8 seconds; it is where the trajectory breaks. In esports, the equivalent interval may be 0.3 seconds of reaction delay in a teamfight, or 0.5 seconds a team hesitates before committing to a tower dive. These intervals are not in the stat sheet, but they decide match outcomes more than damage numbers do. To count them, I must build the system before I start watching. I split the match into fixed time windows, log events per window, then build a comparison table across games. This is exactly how I split segments in a 400m race: every 50m is a segment, each has its own time, and the segment that loses the most time is rarely the sprint finish. THE VIETNAMESE ESPORTS PARTICULARITY There is a specific difficulty Vietnamese esports analysts face. Most deep tracking sites operate in English, and they cover major international tournaments well. For domestic events, data usually stops at the basics: match results, rosters, and occasionally a few damage figures. To get more, the analyst must collect it themselves. This creates a gap between two groups of writers. The first group writes about international teams, where data is abundant and sources can be cited clearly. The second writes about domestic teams, where data is thin and requires more collection effort yet comes with fewer tool supports. Players who have competed internationally, such as SofM or Levi, have fuller data, but that is the exception, not the rule. That gap is not merely technical. It shapes how audiences understand the level of Vietnamese esports. Without detailed data, people fall back on feeling, and feeling is dominated by spectacular moments rather than repeating patterns. I once spent weeks reconstructing a season from recordings alone, because no stat table aggregated enough. That work gave me something ready-made tables do not: an understanding of rhythm. When you log every play by hand, you begin to see the patterns an aggregate table would flatten out. CONTRARIAN ANGLE There is a widespread belief in sports analysis generally, and esports specifically: silence is failure. If you do not make a prediction, you are not an expert. If you say 'not enough information', the audience will turn away. I think the opposite. Every match is a wager that can be counted. You just have to be willing to observe. But 'countable' does not mean 'count at any cost'. There are matches where the baseline data is so thin that every conclusion becomes speculation. In those cases, the most professional action is to say plainly: I do not have enough data to conclude. Deliberate silence is not weakness. It is a quality signal. An analyst willing to say 'no' to unfounded conclusions is protecting the credibility of every other conclusion they offer. The problem is that sports media culture does not reward caution. It rewards decisiveness. And decisiveness without baseline data very easily becomes propaganda. I do not demand that audiences immediately trust that caution. But I do demand that analysts be transparent about their own limits. When I make a prediction, I always attach an uncertainty range. Not to blur the conclusion, but to make clear which facts it depends on, and that if those facts change, the conclusion changes with them. OPEN CONCLUSION Vietnamese esports will not advance through more prize money or more views alone. It will advance when more people are willing to say 'not enough data' instead of saying anything. Good data does not make analysis dry — it makes it trustworthy. And in a market where noise outnumbers signal, whoever can separate the two will always be the one listened to the longest.