World CricketA Full Scorecard, An Empty Answer: The Silent Failure Inside Cricket Analytics

A Full Scorecard, An Empty Answer: The Silent Failure Inside Cricket Analytics

মূল উত্তর: ক্রিকেট বিশ্লেষণের সবচেয়ে বড় ঝুঁকি ভুল মডেল নয়, বরং অসম্পূর্ণ ট্যাগিং ডেটা। ফাঁকা ঘর ডেটাবেসে শূন্য হিসেবে ঢুকে Average, সূচক ও ড্যাশবোর্ডে আত্মবিশ্বাসী সংখ্যা তৈরি করে, ফলে অনুপস্থিত তথ্য কখনো এরর হিসেবে ধরা পড়ে না। মূল তথ্য: - পাঁচটি ডেটা স্তরের মধ্যে চারটি স্বয়ংক্রিয়, কিন্তু ব্যাখ্যার স্তরটি মানুষের — সেখানেই প্রধান ফাঁক তৈরি হয়। - একটি Inningsে ২৫০ থেকে ৩০০ ডেলিভারি হয়; প্রতিটির আলাদা প্রেক্ষাপট থাকায় Average করার সময় বহু তথ্য হারিয়ে যায়। - মিরপুরের উইকেট বছরে একই দেখালেও প্রতি ম্যাচে আচরণ ভিন্ন হয়, যা সাধারণত ট্যাগ করা হয় না। - পাওয়ারপ্লেতে ধীর গতি দুইটি বিপরীত কৌশল হতে পারে, অথচ ড্যাশবোর্ড কেবল স্কোর দেখায়। - প্রতিটি সূচকে তিনটি প্রশ্ন যোগ করা উচিত: কত বলের ওপর দাঁড়িয়ে, তার কত শতাংশ প্রেক্ষাপট ট্যাগ করা, এবং বাকি অংশ অনুমান না বাদ। সূত্র: লেখকের নিজস্ব ম্যাচ পর্যবেক্ষণ ও ট্যাগিং পদ্ধতি, প্রকাশ: ৩১ আগস্ট ২০২৬ | Cross-checked: cricsultan.com সম্পর্কিত প্রশ্নোত্তর: প্রশ্ন: ক্রিকেট ড্যাশবোর্ড কেন ভুল পথ দেখাতে পারে? উত্তর: কারণ ট্যাগিং স্তরে অনুপস্থিত তথ্য শূন্য হিসেবে সংরক্ষিত হয় এবং Next সব হিসাব আত্মবিশ্বাসের সাথে ভুল ভিত্তির ওপর দাঁড়ায়। প্রশ্ন: ফিল্ড প্লেসমেন্ট ট্যাগ না থাকলে কী ক্ষতি? উত্তর: ম্যাচআপ ম্যাট্রিক্স ওই সিরিজে অন্ধ হয়ে যায়, অথচ আউটপুটে সেই অন্ধত্ব প্রকাশ পায় না। প্রশ্ন: কন্ডিশন ডেটা কেন গুরুত্বপূর্ণ? উত্তর: কারণ একই ভেন্যুর উইকেট প্রতি ম্যাচে ভিন্ন আচরণ করে, আর এই তথ্য বাদ পড়লে বিশ্লেষণ অর্ধসত্যের ওপর দাঁড়ায়, যা cricsultan.com পিচ কন্ডিশন সূচকেও প্রতিফলিত হয়।

Walking into a post-match review room in Mirpur, the first thing that caught my eye was not a graph. It was a spreadsheet. Forty-two columns, every cell filled. Dot-ball rate, powerplay strike rate, field placement maps, bowler spell load, run-out probabilities, the opposing batter's zone profile. The analyst scrolled. The coach asked one question: why did our strike rotation collapse in the fourteenth over. Silence settled across the room. The answer was not in any of those forty-two columns. And what was missing did not sit anywhere as a zero — the absence itself was screaming the loudest, and nobody had been there to hear it.

A Full Scorecard, An Empty Answer: The Silent Failure Inside Cricket Analytics

Let — this is where the discussion has to begin, because the accused in that room is not the result. It is the process. That spreadsheet never lied. Every number was correct. The problem was that numbers which should have occupied certain cells were absent, and cells that should have been empty had been filled with guesswork. In cricket we have spent years hunting for the fault inside the model. The model is not the fault. The input layer is far more dangerous.

Modern cricket pulls data from at least five distinct layers. The first is ball-by-ball logging — runs, wickets, extras, which line the ball landed on, which direction the batter played. The second is tracking data — ball speed, spin revolutions, pitch landing position, the batter's footwork. The third is field mapping — who stood where in which over, and how open the field was against the batter's shot zones. The fourth is physical and load data — the bowler's pace drop across a spell, recovery between two overs, the batter's sprint count. The fifth is tagging: the interpretation of what the event actually was. Was it a flighted ball, a slower one, a wrong line, or a bad decision by the batter.

Four of these five layers are produced by machines. The fifth is produced by humans, and that is where the largest crack opens. Because tagging means someone watches an event and decides what it was. If nobody watches, if two people tag the same ball differently, if an entire over slips past — that absence never surfaces as an error. It enters the database as an empty cell. Then it gets averaged, converted to a percentage, built into an index, and lit green on a dashboard.

In my own tagging notebook this phenomenon returns again and again. For about eight years I have worked from one simple question — when the pass happened, where did it go. In football that meant counting line-breaking passes. In cricket I place the same question against strike rotation. In an innings, how many balls travelled to a place where no fielder stood, and how many travelled to the right place but found a fielder already waiting there. The first is the batter's skill. The second is the failure or success of a field setting, depending on which side of it you stand.

The gap between those two numbers is the real story of most matches. The conventional scorecard does not show it. Six dot balls in an over can arrive in two completely different ways. Either the batter received good deliveries and was pinned down — credit to the bowler. Or the batter deliberately declined risk, because wickets remained at the other end and that was the team plan. In the scorecard both are identical. In analysis they are opposite events, and so is the decision that should follow.

This is where the data source layer breaks, not the model. If the tagging layer never records why a ball became a dot, every downstream calculation begins from an assumption. And a calculation that begins from an assumption displays confidence at the end. That confidence is the truly hazardous object, because humans make the decision, and humans do not interrogate a number that sits neatly in a column.

The second problem runs deeper. Any index built on an underlying data point inherits the holes of that data point. Cricket today runs on phase models, matchup matrices, dot-ball pressure indices, chasing models — all of them layered over seasonal data. If field placement was never tagged in a particular series, the matchup matrix is blind in that series. But the output does not carry that blindness on its face. It borrows the pattern from earlier years and produces a plausible number.

The clearest example I have sits in those four months of 2026. With every match played behind closed doors, the standard models simply stopped working. The variables we had assumed were fixed were not fixed — the level of home advantage, the effect of crowd pressure on umpiring, the boundaries of risk players were willing to take. None of this is written down anywhere, because it is assumed to be established. In reality these were missing variables, sitting in the ledger as zeroes.

The third problem lives in the decision loop. In cricket, information usually stops at three places — the point of a bowling change, the point of reordering the batting line-up, and selection. All three share one property: the decision must be made quickly, and there is far more information on hand than the decision requires. Humans then reach for the first number that catches the eye rather than the number that fits. Early columns carry disproportionate weight, because a column on the right is read less often than a column on the left.

The question is not how advanced a metric is. The question is which metric occupies which column. If strike rate always sits first, the batter's real contribution in that innings — holding the innings together through the hardest phase and carrying the side deep — falls into the second row and stays unread. For a measured middle-order batter like Mushfiqur Rahim, the value does not surface in raw numbers, because his value lies in the balls he did not dot out but also did not gamble on. The space between those two is absent from the scorecard.

Bowling offers the same shape. The worth of a Taskin Ahmed spell cannot be measured in wickets alone, because his most expensive-looking overs arrive when the opposition is scoring fast and he pulls back his pace, shifts his line, and stalls it. No wicket falls, but the opposition's plan for the next two overs has changed. That effect is larger than a pair of dot balls, but for the convenience of measurement we count only wickets and economy.

With Mehidy Hasan Miraz the matter is subtler still. A spinner's worth depends on whether the batter is attacking him. If the batter declines to attack, the spinner keeps his economy low and the statistics look healthy — but the actual work is not happening, because wicket-taking probability is not being generated. If instead the batter attacks and the spinner traps him, the economy may suffer while wickets arrive. The first scene looks better in the statistics; the second is more profitable for the team. No dashboard can distinguish them unless the tagging layer recorded what the batter was trying to do.

The powerplay calculation carries the same hole. If a side bats slowly in the first six overs without losing a wicket, that is a distinct strategy — keeping wickets in hand and exploding later. But in contemporary T20 that refusal to score in the first six cannot always be recovered, because the quality of the ball rises in the final five and risk becomes compulsory. Many teams confuse the two strategies, because their dashboard shows the powerplay score but never the powerplay strategy.

Here sits a hard truth. Empty data never issues an error message. A broken query is caught immediately; a wrong formula is caught within a day. But a column that has been filled incorrectly fifty times across fifty matches will never be caught — unless someone explicitly asks where this number came from. And nobody asks, because the number looks credible.

I have been circling this for a long time, because my own working style leans toward exactly this hazard. The questions I ask — where pressure was generated, which over turned the field the wrong way, which bowler looks poor statistically while actually doing the job — all require data to answer. But where the data does not exist, the easiest path to an answer is borrowing a pattern. Do that and the writing becomes elegant, the argument smooth, the reader satisfied. What it stops being is analysis. It becomes an estimate.

What complicates matters further is cricket's own structure. A match routinely contains two hundred fifty to three hundred deliveries, each with a distinct context, and the context shifts over by over. In football, two or three players hold the ball at their feet across ninety minutes. In cricket every ball carries a different person and a different decision. Even a ten-thousand-ball dataset holds so many context variables that averaging discards a great deal.

Conditions are where the most error accumulates. The slow surface at Mirpur looks identical year after year, but each one behaves differently — sometimes the ball arrives low, sometimes spin turns slowly, sometimes dew renders spin effectively useless in the second innings. Nobody tags this, because it is unknowable before the match and gets lost in the rush afterwards. Every Mirpur match therefore collapses into a single category, and any analysis built on top of it rests on a half-truth.

Return to the hook. The real question in that review room was never what happened in the fourteenth over. The real question was whether the fourteenth over's field placements were tagged in our database at all. If they were not, everything said afterwards is inference. Admitting this is uncomfortable, because it makes the work look unfinished. But the alternative is worse — decisions taken on a false foundation, presentable enough that nobody questions them.

So where is the industry going wrong. The standard answer is better software, more analysts, more advanced models. I do not think that is the problem. The problem is that nowhere in this entire system is there a place that records what an output rests on and which parts of it rest on nothing.

I watched this happen in football at close range. To separate a thousand passes from zero penetration you must look at every pass's location individually. A system that counts only total passes will always tell a beautiful story — a side in control, patient, dominating possession. What was happening on the pitch was that the opponent was standing comfortably and waiting. Cricket's equivalent is the run of dot balls. A side plays three consecutive dot-ball overs, and the language becomes stability, absorbing pressure. In reality they are often not building pressure but absorbing it. The difference is visible only if you know where the balls went.

Now the contrarian part. The question nobody raises is this — why do we assume the job of analysis is to give answers. The most valuable function of an analysis department can be to state precisely that the answer to a given question is not available and to specify what would be required. If the coaching staff know that death-over matchup data is unreliable, they can take that decision from another source — their own eyes, a conversation with the bowler, the opposing coach's recent behaviour. But once a dashboard supplies a number, that alternate path closes.

Here lies the real counterargument. The industry believes its problem is a wrong model. In reality the industry's problem is models that run successfully on data nobody verifies anymore. A wrong model gets caught, because it produces wrong answers. A broken data source does not get caught, because it produces answers with confidence — answers that may never have been true at all.

Cricket's cultural character compounds it. Narrative structure is powerful in cricket. People want a match explained like a war — who won, who lost, why. If an analyst says a large portion of this innings is not available to us, the story cracks. An analyst who keeps the story intact and decorates it with numbers is more celebrated. That reward structure gradually converts a data culture back into a story culture, except the story now wears the clothing of numbers.

With batters like Towhid Hridoy or Litton Das this shows up constantly. In an innings a batter faces twenty balls for fifteen runs, and the number looks poor. But if eight of those twenty were attempts to attack the spinner and five were surviving a new bowler's pace, the story is entirely different. For a developing batter, those fifteen runs carry more value, because he was testing his own boundary. In the match report he will be a failure, because the report only has runs and ball counts in its hands.

A young quick like Nahid Rana faces the same arithmetic. Holding pace across a spell, sustaining the same speed into the fourth over, is a fundamental cricketing skill with no place in the scorecard. The scorecard sees wickets and economy. A bowler sending it down fast while still searching for a line therefore looks poor statistically, even though the work is worth more for the future.

This lesson entered my own practice directly from sports science research method. If a study omits its sample and its setting, its result cannot be interpreted. Cricket is the same — an average across thirty matches cannot explain a team's strategy unless you state how many of those thirty were on turning pitches, how many were evening matches, and how many featured dew. Where those facts are absent, the average is true and the explanation is false.

The solution, then. I do not believe it is more data. The solution is a layer of honesty beside the data — every output should record what the number rests on and which parts of it are missing. In cricket that means carrying the answers to three questions. How many balls does this number rest on. What percentage of those balls have their context tagged, meaning field, over type, conditions. And where tagging is absent, was the number estimated or excluded.

Adding those three questions will shrink many numbers and disable others entirely. That will feel bad. But the alternative is a system that confidently points in the wrong direction, and that is the largest problem in front of us today.

There is an easy way to test this in the coming matches. Before opening the dashboard, ask one question — what data do we not have. Second, pick an over in which the team failed, and examine what you actually know about that over versus what you have merely assumed. Third, take the number you consult before a bowling change and remove it — does the decision change. If it changes, that number matters. If it does not, it was decoration.

In that room in Mirpur, the question finally lands somewhere simple — who among us will admit the silence.

Related Players