The Morning of the Qualifier: When 47 Deliveries Said More Than the Scorecard

The scorecard is a compression artifact, not a fact.

On the morning of the tournament's knockout qualifier I was not looking at the scorecard — I was reopening my ledger of the 47 deliveries from overs 7 to 14. The scorecard will tell you a team made 168 in 20 overs and won by 12 runs. My ledger says this match was actually decided at 11.3, when a batsman pushed his front foot out for a slog sweep and was given LBW — and at that moment the DRS graphic's 'impact in line' was more than three degrees outside.

I replayed that ball from three camera angles. The very first frame of the ball-tracking was wrong — the pitch-mapping had shifted the stump line by half a metre. This is the thing I have been fighting for 23 years: the broadcast graphic is an estimate, and presenting an estimate as evidence is professional negligence.

Context: Tournament Pressure and the Limits of Data

A knockout qualifier in Asian cricket carries a different kind of pressure. In the group stage you escape through net run rate arithmetic; in a qualifier, one over's error ends a season. The data available at this stage of this tournament arrives in three layers — and all three have different reliability.

The first layer: the official scorecard. It is built from the on-field umpire's signals, and it does not contain everything that actually happened. The second layer: the ball-by-ball feed. It is fast, but it carries no field placement, no batsman's posture, no bowler's line-and-length context. The third layer: the broadcast tracking system — DRS, wagon wheel, pitch map. It is the most seductive and the least audited.

I am logging every qualifier of this tournament ball by ball, because earlier this season I saw three occasions where the tracking overlay and the actual ball position diverged. Three occasions in more than 400 deliveries — not a large number, but in a knockout, one ball is enough.

Core Analysis: Rebuilding 47 Deliveries

I re-logged every ball from overs 7 to 14 — bowler, line, length, field change, batsman's shot, and outcome. 47 deliveries, 8 overs, one team scored 41 runs and lost two wickets. The scorecard summary will say: 'slow batting in the middle overs.' My ledger says something else.

Of those 47 deliveries, 28 were dot balls, and 19 of them were balls the batsman did not leave — he blocked them. That is a different thing entirely. A left ball is the bowler's win; a blocked ball is the batsman's hesitation. The scorecard converts both into dots.

I sorted those 19 blocked balls by bowler: 11 off the spinner, 8 off the medium-pacer. The spinner gave 18 runs in 4 overs in that phase, but his economy read 4.5 — which looks good. In reality, 9 of those 18 runs came in the one over where the batsman did not block him even once, but stepped out three times and once nearly got stumped.

The real measure of a spinner's success is not wickets, it is 'the direction of the batsman's feet.' I calculated the percentage of times the batsman left the crease against spin in that match: 41 percent. That is 13 points above the season average of 28. Those 13 points explain why runs did not come between overs 7 and 14 — because the batsman was 'reacting' to spin rather than 'playing' it.

Now the target calculation. According to the scorecard, the chasing team needed 58 off the last 5 overs — 'difficult but possible' by record. I re-derived it with a wickets-in-hand model: the fifth wicket had fallen, the sixth batsman was on 14 off 22, and the number eight had faced just 19 balls on the net in three days. My model says the win probability was 23 percent; the broadcast showed 47 percent.

That gap is not a cricket gap, it is a modelling gap. The broadcast model decays batting average as wickets fall, but it does not see which batsman is left. Since 2026 I have given 'quality of remaining batsmen' a separate weight in every model I build, and in this match that weight was decisive.

The Other Side: Where Technique and Numbers Diverge

The most interesting moment for me was in the second innings. The fourth ball of the 14th over — the bowler sent down a slow bouncer, the batsman pulled, caught inside the boundary. In the scorecard this is a 'wicket.' In my ledger it is 'wrong ball, right shot, wrong outcome.'

Five years ago I built a model that measured 'how right the ball was for the batsman.' I ran it on three matches and the result came out opposite to my expectation — the numbers called a shot 'bad' that the captain called 'match-winning.' It took me a week to understand that the model had no batsman's 'posture,' only the ball's line and length.

I abandoned that model, because a model that claims to measure a batsman's intent is not a model, it is a bias. Many decisions in cricket are taken not from numbers but from a player's eyes — and that is not bad. What is bad is that we use numbers to declare that decision 'right.'

A Counter-Thought: Was This Match Won by Data?

Here I disagree with my own thesis. The analysis above stands on 47 deliveries — that is one sample, and applying it to the whole match or the whole tournament would be a serious error. There is another reason for slow middle-over batting that my ledger does not capture: dew. On that ground, the ball starts to wet after the 14th over in the evening — I have seen this in 11 matches over the last two seasons. If the dew had begun from the 13th over, the batsman's blocking was strategy, not hesitation.

My ledger does not measure dew. That is the limit of my model, and I will not hide it. My win probability for this match was 23 percent, but I will not call it 'precise' — I will say it is an estimate built on a specific pattern of 47 deliveries, confidence level moderate. If dew data were added, that number could change.

What to Watch Next

In the next qualifier I will record three things: first, the direction of the batsman's feet on every ball against spin; second, the exact over of dew time, which this time I will ask the ground staff rather than read off the scoreboard; third, the stump-line calibration of the first DRS tracking frame, which was wrong in this match and which I have recorded above.

This is not my claim that numbers win matches. It is my claim that the number being shown to you was built by someone — and it is your right to ask that builder what he left out.

Before the result there is a ball, and someone has to count it.

---

Correction log (public): In the first version of this article I wrote the win probability as '21 percent.' On re-checking the remaining-batsmen weighting in my ledger, it came to 23 percent. The original figure of 21 percent remains visible; the correction is recorded with a date — 23 percent is the updated value.