Asian T20 Baseline Audit: Powerplay, Middle-Over Spin and the Real Death-Overs Arithmetic

Sharjah, last January. A side chased 197 with 11 balls to spare in the ILT20. The scoreboard called it dominance. My delivery log disagreed. After the 15th over their boundary rate was 14.2 percent against a Sharjah baseline of 22.8. The win came from six full tosses and three half-volleys in the opposition's death overs; 34 runs came off those nine balls alone.

Three days later, same venue, the reverse picture. Chasing 188, the side was above baseline through 15 overs with seven wickets in hand, then scored 32 in the last five and lost. The scoreboard tells two different stories about these two matches. My notebook says the process sat in the same place both times: death-over ball selection and the middle-over spin matchup.

I sat down to run this audit because an uncomfortable pattern is accumulating in Asian T20. Teams buy big powerplay hitters, the market prices those hitters, and yet matches are decided between the seventh and fifteenth overs, where finger spinners bowl and lower-priced batters make the calls.

Context: how the baseline was built

In 2026 I built the K League xG baseline at Footballist because the goals were lying. Jeonbuk scored 2.11 goals per game against 1.84 xG, and the market kept overpricing them away from home. The lesson was singular: scoreboard and process are different objects, and separating them happens at the desk, not on the replay screen.

I carried that discipline into cricket. My current baseline rests on 14,200 deliveries: Asia Cup 2026 in the UAE (19 matches), ILT20 2026 (34 matches), and 11 selected domestic and bilateral T20 matches from India and Sri Lanka. For every ball I log five things: phase, bowling type (right-arm pace, left-arm pace, finger spin, wrist spin), line-and-length zone, batter's hand, and venue cluster.

| Phase | UAE | India | Sri Lanka | |---|---|---|---| | Powerplay (1-6) | 8.1 runs/over | 8.9 | 7.8 | | Middle (7-15) | 7.4 | 8.2 | 6.9 | | Death (16-20) | 10.3 | 10.8 | 9.6 | | Spin share (7-15) | 54% | 41% | 58% |

One caveat before reading the table: no cell has fewer than 1,100 deliveries, and the standard error on run rate is plus or minus 0.4. You can speak about 7.4 against 7.8; you cannot speak about 7.4 against 7.6. I trust a number only after I can reproduce it on a quiet Tuesday.

The model is plain. I have no ball-tracking data, so I make no claim about the xR of a specific delivery. What I do is fit a logistic model on line-length zone, phase, bowling type and batter's hand, outputting boundary probability and dot probability. Run value falls out of those two. It is not xG; it is xG's poor relative. It is enough to detect whether a baseline has broken.

Core: the powerplay is a false signal

Across 64 matches, one thing is clear. The side that wins the powerplay run-rate differential wins 58 percent of matches. The side that wins the middle-over differential wins 71 percent. A 13-point gap is not coincidence.

The reason is arithmetic. The powerplay is six overs, 36 balls. The middle phase is nine overs, 54 balls. Fielding restrictions inflate powerplay scoring artificially, so the differential it creates reflects field settings rather than batting or bowling skill. On 36 balls, variance is high and signal is low.

My second observation is more awkward. Of the sides that lead by 20 runs or more at the end of the powerplay, 39 percent go on to lose. Powerplay-heavy sides typically spend two or three wickets buying those 20 runs, then walk into the spin web in the seventh over.

This is the real geography of Asian T20. In England or Australia the spin share in the middle overs sits near 34 percent; in the UAE it is 54, in Sri Lanka 58. The phase with the most deliveries is handed to a bowling type most teams budget as a fourth or fifth option.

There is variation within spin, but less than the noise suggests. In UAE middle overs my log gives wrist spin an economy of 6.6, finger spin 7.2, pace 8.4. That is a 1.8-run gap between spin and pace, and only 0.6 between wrist and finger spin, which in some cells still sits inside my noise band. Anyone building a big narrative around wrist spinners is pricing bowling quality, not bowling type. Rashid Khan, Wanindu Hasaranga, Varun Chakravarthy, Noor Ahmad, Ravi Bishnoi: their success is not the success of wrist spin. Their success is that they are the best bowlers in that phase.

One thing keeps returning in my notebook: the dot-ball rate of an off-spinner against a left-hander runs above his rate against a right-hander. At Asia Cup 2026 I watched from the stands as captains reshuffled the field before and after a left-hander walked in, adding a slip, removing cover. Captains know this matchup. The market prices openers on powerplay strike rate instead.

Core: overs 13 to 16, the squeeze window

The most compressed four overs in my data are 13 through 16. Run rate there is 7.1, and 0.41 wickets fall per over. Sides that keep the scoreboard moving through this window reach the death overs with seven or eight wickets in hand.

Sides that stall here face simple arithmetic: score below a 7 run rate between overs 13 and 16 and you must chase at 13.5 an over in the last four, which in Asian conditions is a tail event, not a baseline.

Spinners deliver 68 percent of balls in this window. Captains hold their best spinner back for it, and batters often play with a touch of caution because the death overs loom. The squeeze produced by that mutual caution is where the match is actually priced.

In the death overs I count two separate things: yorker attempts, and yorkers that land where intended. Only the second correlates with death economy; the first does not. Over the last two seasons I have watched plenty of bowlers attempt six or seven yorkers an over and land the fourth ball on a short length. The market still pays for the reputation of the attempt.

Dew is its own chapter. The UAE death baseline is 10.3, but in matches with heavy evening dew, bowling economy in the second innings rises to 11.4. Same bowler, same length, different ball after it leaves the hand. Grips change, yorkers become over-pitched. That is why toss value is higher in Asia than elsewhere, and if it is not inside your model you will hunt edge in the wrong place.

Core: the closing line and my model

My relationship with the market is simple. The closing line is the market. The opening line is an opinion; the closing line is money.

Last season my log holds nine matches where a side led by more than 20 in the powerplay and still lost, and in the closing line those sides were favourites. One specific case: a side closed at -160, an implied probability of 61.5 percent. My model, on middle-over differential and spin matchup, gave that side 48 percent. A 13.5-point gap does not appear every match, but it returns through a season for the same reason: the market prices powerplay run rate and does not price middle-over spin control.

Kazan in 2026 sharpened that lesson. Germany was 78 percent implied at -1.5; my model saw 7.8 passes per defensive action and only 0.11 xG per possession. Kazan reminded me that a model can be right and still lose, and that is why I now pre-register outcome ranges instead of writing explanations afterwards.

Contrarian: correlation is not causation

Now I argue against my own conclusion, because otherwise this piece is incomplete.

Spin economy is low in Asian middle overs. True. But attributing all of it to spinner skill is wrong, because three confounders sit inside the number.

The first is selection effect. Captains bowl spin in the middle overs because those spinners are their best bowlers, saving pace for the death. What I may be measuring is "best bowlers are cheap", not "spin is cheap".

The second is batter intent. A side that loses two or three wickets in the powerplay plays defensively in the middle by necessity. That 7.1 run rate is circumstance, not credit.

The third is environment. Dew, night matches, ground dimensions. When the K League stadiums emptied in 2026, home advantage stopped hiding behind the crowd: home win rate fell from 46 percent to 31, home xG dropped 0.28. The same caution applies to Asian spin data. Are spinners cheap, or are scheduling and dew making them cheap? Those must be separated.

My own biggest trap is over-adjustment. In a fatigue-adjusted model I keep adding travel, back-to-backs, day-night splits, until the model is so controlled that no effect size survives. So the rule stays hard: at least 20 matches per new control, and report effect sizes, not just significance.

Takeaway: what to watch

The T20 World Cup 2026 is in India and Sri Lanka. Across the first twenty matches I will log three things: middle-over run-rate differential, scoring rate between overs 13 and 16, and death economy in dew-affected second innings. I will not move the baseline before twenty matches.

The question is yours now. If you bet on the beauty of the six-over powerplay, and the market prices exactly that, then who is paying the bill for the 54 quiet balls between the seventh over and the fifteenth?