Eleven Runs in the Final Over and the Arithmetic of the Baseline: A Method for Auditing a Cricket Match

Eleven runs were needed off the final over. A set batter at the crease, five wickets in hand. The word "pressure" appeared three times in the commentary box. I wasn't watching the scoreboard; I was watching a table — over-by-over expected runs, expected wickets, and a fielding-pressure index. First ball to long-on, two runs. Second ball, a cutter, bowled. The commentary said momentum had shifted. My table said that most of the match's deviation had been booked in those two deliveries — and that the deviation had actually been building for the previous four overs.

I have seen this scene many times. From the press box at the Sher-e-Bangla to the media tower at Old Trafford, the place changes but the question stays the same: which over, which matchup, which fielding residual decided the result? I have spent several seasons working across Bangladesh and UK cricket feeds, and I keep returning to the same place — the baseline.

Context: what a baseline actually is

A baseline is the expectation you can compute before the match begins. How many runs an opener should be expected to make in a T20 powerplay, what a bowler's economy should be before and after a wicket falls, what a finisher's strike rate should be at the death — these are baselines. They are not opinions; they are numbers standing on a sample.

When I first built an xG model for football in 2026, as a statistics student in Manchester, on 380 Premier League matches, I learned that the gap between expectation and outcome is the real story. Manchester City's 18-game winning run produced 56 goals from 44.3 xG — an overperformance of +11.7. I later pulled that model across to cricket, changing only the units: goals to runs, shots to balls, assists to matchups.

The first xG model I built did not predict football; it predicted my patience. Building a table is easy; trusting it is hard.

I break baseline construction in cricket into three steps. First, split the innings into phases — powerplay, middle, death. Second, compute expected runs and expected wickets in each phase. Third, measure the deviation from the actual outcome and show whether it belongs to batting, bowling, or fielding.

Expected wickets are harder than expected runs, because the value of a wicket depends on the phase. A wicket in the powerplay and a wicket at the death are not the same. So I weight every wicket by its phase context — overs left, wickets in hand, the bowler's line. Without that weighting, a figure is just a number, not a story.

The eye test is a witness; the data is the cross-examination. What commentary calls a great catch, the fielding residual calls two runs saved on a probable catch-drop edge. Both are true, but one can be measured and the other cannot.

The data chain: every feed is a block

When I place Bangladesh and UK feeds side by side, one thing becomes clear — data is not a neutral truth; it is a chain. Every delivery is a block, every label a verification. Who logged it, when they logged it, where the ball of a given over went — if these details are missing from the feed, the whole analysis is weak.

The Sher-e-Bangla feed and the Old Trafford feed are not the same. One tags ball-by-ball quickly; the other carries more spell-level data. One records fielder positions; the other sometimes only holds the final score. This is where my real work sits — identifying missing data and refusing to guess where none exists.

I publish my code and raw data, because a model that cannot be verified is not a model — it is a claim.

Core analysis: the Shakib puzzle of the 2026 World Cup

At the 2026 World Cup, Shakib Al Hasan scored 606 runs, the highest by a Bangladeshi in the tournament. Commentary explained it with "form." In the table I saw something else. His strike rate sat close to the tournament's best batters, but the internal shape of his innings was different — he scored in the middle overs against spin, and that matched Bangladesh's batting baseline.

That match matters. A batter's runs are not only his own; how well they fit the team's phase structure tells you whether the innings is sustainable.

Same logic for Mustafizur Rahman's cutter. Commentary says "unfamiliar, therefore effective." The data says his death-over economy was then better than the league average, but his reliance on a wide line means mis-execution risk. When a two-over spell becomes a four-match series of suddenly good figures, I question the baseline itself — because a cutter's effectiveness runs in inverse relation to the opposition's batting experience.

I do not chase narratives; I build a table and wait for them to arrive.

To catch deviation from the baseline I use four indices — expected runs, expected wickets, fielding pressure (the cricket version of PPDA), and matchup residual. The last is the most neglected. How far an off-spinner's economy against a left-hander sits from the league average — that gap is the matchup residual.

Every match audit carries four questions. First, how large is the deviation, and does it sit inside normal sampling noise? Second, in which phase did it accumulate? Third, is it repeatable or one-off? Fourth, was the baseline itself appropriate for this match? If those four do not align, I do not write a conclusion.

Cross-sport translation: football to cricket

I bring football's expected-value logic into cricket because the structures differ but the question is the same. In football: what is the goal probability from this shot? In cricket: what is the run or wicket probability from this delivery? Both answers come from a baseline, and both decisions come from deviation.

I measured what happens when the crowd is zero, in 2026. Across the first five rounds after the Bundesliga returned, the home win rate fell from 43.2% to 21.1%, and home goals per game from 1.65 to 1.08. That was my Empty Stadium Index.

Every empty stadium was a controlled experiment we never asked for. That experiment showed that a large part of what we call home advantage is crowd noise, pressure on officials' decisions, and player habit — not just pitch or weather.

That index cannot be translated directly into cricket, but the logic travels. In international cricket, part of home advantage comes from pitch familiarity and travel fatigue, and part from the crowd and umpiring decisions. Fail to separate the two and the analysis goes the wrong way.

At the 2026 World Cup in Adelaide, Bangladesh beat England by 15 runs to reach the knockout stage. The story of that match is usually written as a "historic win." In my table it shows something different — in the middle overs Bangladesh's spinners cut deep into England's middle-order expected runs, and England's death-over baseline broke under that pressure. The result arrives first; the cause arrives later.

Contrarian: the baseline is also on trial

This is where I stand against myself. Baseline auditing is attractive and dangerous in equal measure. If the baseline is wrong, the deviation story is wrong too. A baseline can fail for four reasons — era shift, competition standard, pitch character, and data provenance.

Suppose I compare Bangladesh's first Test win in 2026 with a 2026 match. In January 2026, Bangladesh won their first Test against Zimbabwe in Chittagong. The scoring baseline of that match is not comparable to today's — different pitch, different ball, different fielding setup.

So is measuring deviation pointless? No. But every deviation must carry a condition — in which era, in which competition, on which pitch.

There is another danger: dismissing narrative entirely. I do not treat narrative as an enemy; I treat it as a hypothesis to operationalize and test. Before rejecting the word "momentum," I ask — what is its operational definition? Over how many overs do we measure it? Under what conditions would it be falsified?

That balance matters in South Asian cricket culture. Emotion and storytelling are vast here, and data culture is still under construction. My work sits in that gap — not breaking someone's story, but telling it in a way that can be verified again.

Data provenance: Bangladesh versus the UK

I work across feeds from two countries, and each time I learn the same lesson — however good the model, the pipeline is only as honest as its source. Some information exists in Bangladesh's domestic cricket that the UK feed lacks, and the reverse is also true. Nobody says which field came from where.

I follow one rule — before writing any number, I must know where it came from, who labelled it, and how large a sample it stands on. Without that chain, the analysis is a guess, and passing a guess off as truth is the greatest sin.

That is why I keep a methodology box in every match analysis, stating sample size, limits, and confidence level. Readers may skip it, but its presence matters. Because where limits are admitted, analysis survives.

In my university years I published the Germany–South Korea autopsy within 12 hours, and it taught me one thing — show the gap between outcome and expectation correctly and the story writes itself. Germany lost with 74% possession and 2.7 xG; South Korea scored twice from 0.9 xG. In cricket, possession is not run-rate but ball count, and the value of that ball count is priced by wicket risk.

What it comes down to

Back to that final over. Eleven runs, five wickets, a set batter. Commentary had one story — failure under pressure. The data had another — the previous four overs had delivered 22 runs fewer than expected, and that happened because of a spin matchup, not a final-over mistake. The last-ball cutter was the result, not the cause.

In the coming tournament cycle I will watch three signals — how far a death bowler's economy sits above or below his own baseline, how differently a middle-order batter strikes against spin and pace, and how many rest days the international calendar allows for travel. If a team looks good on those three numbers, then whatever the commentary says, I will put a green tick beside its name.

And a question for you — next time someone says momentum shifted, will you look for the number, or will you believe the story?