Overs 7 to 15: Where Asian T20 Cricket Is Actually Lost

In the 13th over of the Dubai final last September, a spinner came back on and the noise in the stands dropped a notch. My notebook was working the other way. The previous five overs had produced 28 runs and two wickets. The journalist in the next seat used the word “pressure” three times. My spreadsheet did not record pressure. It recorded a crack in the baseline, and after the tournament I priced that crack at 11.6 runs.

The highlight reels arrived from overs 16 to 20. The bill was written between overs 7 and 15. Of the 11 completed matches in that Asia Cup, nine were won by the side that took more wickets in the middle phase. You can call that coincidence. Coincidence can still be priced, and once priced, it can be re-run.

Method: what I measured, what I dropped

I pulled ball-by-ball data for 96 T20Is played in the UAE, Bangladesh, Sri Lanka and Pakistan between January 2026 and October 2026 from a licensed feed. After removing rain-shortened innings, DLS-altered targets and venues with fewer than eight matches, 81 innings pairs remained. Every figure carries a 95% confidence interval. I publish the raw files and the code each time, because a number you cannot verify yourself is not analysis; it is a recommendation. The logic I use for xG in football is the logic I use here for expected runs and expected wickets. The first xG model I built did not predict football; it predicted my patience. And Germany did not lose to South Korea; they lost to 28 shots and no goals. Since that autopsy, my editorial rule has been fixed: expected score, ball map and pressure index first, narrative second.

Asia Cup 2026 was a 13-match slate, six teams, three venues. In recent editions the draw has placed India and Pakistan in the same group, so the fixture is guaranteed at least once—that is calendar arithmetic, not cricket politics. Three venues, evening dew, and pitches that slow under floodlights form the conditions my baseline is built on.

T20 has no equivalent of possession, so the eye test parks itself in two places: sixes and dot balls. Both are outputs. To find the input I chose a 45-ball window, overs 7 to 15.

The baseline: 158 first-innings runs, and the price of a wicket

In my compiled set, the first-innings baseline at these venues is 158 (confidence interval 154–162). Phase run rates: powerplay 8.2, overs 7–15 at 7.1, overs 16–20 at 9.8.

I define wicket cost as the difference in expected final total between a state with one wicket in hand and a state with one fewer, holding over, score and the set batter’s balls faced constant. The output: a powerplay wicket costs 9.4 runs, a middle-overs wicket costs 11.6 runs, a death-overs wicket costs 6.2 runs.

On a slow Asian surface, a wicket in overs 7 to 15 is worth 2.2 runs more than a powerplay wicket and 5.4 runs more than a wicket in the last five overs.

The mechanism splits into two channels. First, the buy-in: on a gripping pitch, a new batter’s first ten balls in this set go at a strike rate of 105, against 138 for a set batter in the same window. Second, the death ceiling: losing the set batter pulls the expected rate in overs 16–20 down by roughly 1.3 runs per over, because the job of holding the rate passes down the order. That gives 11.6 = 4.9 (buy-in and rotation) + 6.7 (death ceiling). Both are team-controlled inputs. A six is not under your control. Bringing a spinner on in the 12th over is.

Applied: how the final was settled in near silence

The side batting first in the Dubai final finished below the 158 baseline and still did not lose. Their spinners took three wickets inside overs 7 to 15, which forced every incoming batter into that window. The opposition’s death-overs strike rate finished above baseline—the part television showed—but the capital that funds a death-overs surge had already been spent.

I was in Dubai for a fortnight. From the press box it was clear the spinners were releasing wide of slip and the ball was not turning so much as stopping. On that surface the batter’s problem is not spin; it is pace. With the ring two yards closer than usual, the arithmetic of the single changes too.

One detail stayed with me. A team kept its frontline fast bowler on a “week-to-week” listing throughout. His share of overs bowled between 7 and 15 came in nine percentage points below baseline. On paper he was fit; in the numbers he was protected.

This index is not one team’s story. India hold the most Asia Cup titles, Sri Lanka six, and in this region trophies are built from bowling structure rather than pace hype. Afghanistan reached their first ICC Men’s T20 World Cup semi-final in June 2026, and they got there by squeezing spin through the middle overs, not by out-hitting anyone. Bangladesh show the reverse picture: below-baseline expected scoring in the same window, with wicket loss above it.

The data gap: who signs the credit line

There is an uncomfortable part I could not skip. Not every venue in this set carries the same grade of tracking data. The Dubai and Abu Dhabi feeds include ball speed, spin rate and tracking labels; several Sharjah matches do not, or only partially. The effect: in the Dubai subset the middle-overs premium is 11.6 runs with a confidence interval of 8.9–14.3, and at low-sample venues the same figure spreads from 4.1 to 17.2.

Data whose provenance you cannot inspect can explain someone’s luck, but it cannot prove anything.

So every index I publish carries a trace: which innings, which over, which feed, who labelled it. A ledger you can walk backwards through, one number at a time, is the minimum condition for ball-by-ball work. In 2026 I counted the silence and found it had a home advantage; the same discipline applies to a data pipeline. The laziest part of this beat is the DRS wait. I stopwatched 11 third-umpire reviews during the tournament; median dead time was 94 seconds. Ninety seconds is enough to dismantle the moment people paid to attend.

Contrarian: momentum autocorrelation, and auditing my own baseline

On television the story of this tournament was death-overs momentum and team temperament. I tried to operationalise both. Momentum equals the residual after subtracting the phase baseline from the 12-ball rolling run rate. Then I checked how much of next over’s run rate that residual explains—11%.

Momentum, as defined here, carries almost no information from one over to the next (r = 0.11).

The second question is more uncomfortable and points at me. Is the 11.6-run premium a law, or an artifact of the period? I re-ran it on the 2026–2026 subset and the premium fell to 9.8 runs, with the confidence interval crossing zero at some venues. The story may belong to the pitch, the era, or the ball-change regulation. Run this index without venue and pitch reports and it will fail at the next tournament.

Takeaway: what I will watch next cycle

My column for the next tournament cycle will carry a new figure: the middle-overs wicket-cost index. Who is building a spin pairing for that window, who is shielding a set batter to protect the death ceiling—answer those and the dot balls and sixes fall into their proper place. One question stays open. Where the fixture is guaranteed in advance, is the decision to bat or bowl made on pitch evidence, or is the slot itself priced by the broadcast market?