Not Death-Over Economy but Control Percentage: A Ten-Match Baseline Audit of T20 Bowling

Not Death-Over Economy but Control Percentage: A Ten-Match Baseline Audit of T20 Bowling

Over the past decade in T20 cricket, I keep seeing the same pattern. A death-over spell ends, and people judge from the scoreboard: "Sixteen runs in one over, the bowler cracked today." Or the reverse: "Just twenty-two runs in four overs, outstanding." Both judgments stand on the same trap: they are reading economy rate, not control percentage. Yet the two numbers do not always tell the same story.

When the first T20 international began in Auckland on 17 February 2026, nobody imagined the format would rewrite cricket's analytical language. I was young then, working in radio, writing about cricket. Nearly two decades later, doing Bengali-language commentary at the 2026 ICC T20 World Cup, I understood again how numbers mislead people—and how, once the right baseline is set, those same numbers tell the truth. Around 2026 I used to write weekly data threads on the English Premier League from Rangpur. One thread carried Burnley's PPDA of 12.1 and 38 percent possession. People said Sean Dyche's side played passive football. Sorted out, the numbers showed it was not passivity but calculated efficiency. The Burnley thread looked like noise until I sorted by PPDA. That thread taught me one rule: no verdict before the numbers are sorted. In cricket I use the same rule—I do not open my mouth about a bowler without looking at control percentage, economy rate, dot balls and boundary percentage together.

This piece is a methodological audit. I will not claim who is best or worst. I will show what baselines must be set before a death-over spell can be judged fairly, why one match is never enough, and why economy rate without control percentage tells half the story. This is no hot take; it is a checklist anyone can use.

Context: Baseline First, Verdict Later

My core principle is simple: before judging any performance I set five baselines—format, venue, era, phase, and opposition norm. In T20 this matters even more, because the result changes with every ball. Six runs in one over versus sixteen in another is not only about bowling quality; it is approach, field setting, pitch character and match state combined. So the first question: in which phase, at which venue, in which era was this match played?

Era matters. Early-2010s T20 is not 2020s T20. In the powerplay, 7.5 runs per over was once normal; now many teams push past 9. In the death overs, 9 an over was once good; now 10.5 is ordinary. If you measure today's bowler with an old economy rate without adjusting for era, you are being unfair—not to the bowler, but to time.

Venue is equally vital. Sharjah, Abu Dhabi, Mumbai's Wankhede, Sydney Showground—each has a different scoring pattern. Small boundaries, fast outfields and altitude make the ball travel. The same bowler may average 9.8 in Sharjah and 7.8 on a spin-friendly Chennai surface. Without a venue norm, economy rate is an incomplete calculation.

Phase Baseline: The First Pillar

I read a T20 match in three parts—powerplay (1-6), middle overs (7-15), death overs (16-20). Each phase has its own batting and bowling norms. Below are approximate baselines from my own maintained database for men's international T20 (2026-2026). These are ten-year phase averages, not any single match.

| Phase | Batting SR | Bowler Economy | Control % | Dot Ball % | Boundary % | | Powerplay (1-6) | 125-135 | 7.8-8.5 | 58-62 | 40-44 | 16-18 | | Middle (7-15) | 120-130 | 7.4-8.2 | 60-65 | 34-38 | 12-15 | | Death (16-20) | 150-170 | 9.8-11.2 | 50-55 | 28-32 | 18-22 |

This table is the first pillar. If a bowler concedes 9.5 in the death overs, that is better than the phase baseline. But if someone calls 9.5 "too many runs," he simply does not understand the phase. In the death overs, 9.5 is actually above average, because the phase norm is 9.8-11.2.

There is a subtlety here. The death-over batting strike rate of 150-170 means 150-170 runs per 100 balls—so runs fall naturally in this phase. The bowler's job here is not "not conceding" but "keeping control"—forcing the batter into shots, bringing fielders into play, and bowling more low-risk deliveries. This is where control percentage enters.

Control Percentage: The Hidden Truth

Control percentage is the share of deliveries a bowler lands where he intended, deliveries the batter cannot play freely. In T20, to me, it is the cheapest and most neglected metric. Take an example.

Suppose one bowler delivers four overs in the death and concedes 42—economy 10.5. Another concedes 34—economy 8.5. The scoreboard says the second did better. But what if the first's control percentage is 62 and the second's is 48? Then the story flips. The first kept the batter under pressure on 62 percent of balls, yet luck or fielding gaps leaked more runs. The second released pressure on 48 percent of balls, but mistimed shots did not fall to fielders, or catches were dropped.

Here is my core argument: control percentage shows whether a bowler is doing the process right; economy rate shows what the outcome became. Process persists, outcome fluctuates. Over ten matches, a bowler with control above 60 may show a bad economy in one or two games, but in the long run it corrects. The reverse—low control, good economy—is usually temporary.

So I look at control percentage, dot-ball percentage and boundary percentage together. This triangle draws the true picture of a death-over spell.

The Ten-Match Threshold: Why One Match Is Never Enough

When I wrote the Burnley thread in 2026, I imposed a rule on myself: no tactical claim without at least ten matches of data. In cricket I follow it more strictly, because T20 is a small-sample game. A bowler gets roughly two overs per match in the death. Ten matches means twenty overs, that is 120 balls. Even 120 balls is not a large sample.

But over ten matches, some things begin to stabilise. Control percentage, dot-ball rate, and strike-rate-versus-expected-runs become relatively reliable across ten matches. In one match a bowler may deliver six dot balls, but that is an event, not a pattern. If the average of six dots holds across ten matches, that is a pattern.

A confession here. I do not apply the ten-match threshold blindly. In some cases it shifts. For condition-specific bowling, such as bowling only in the powerplay, chances are fewer, so I push the threshold to twenty innings. For a death specialist, ten matches suffice because the role is fixed. The threshold is a methodological caution for me, not scripture.

One thing I state clearly. Across ten matches I see some patterns, but they are not proof—they are signals. A signal must be tested across different opponents, conditions and match states. Without these three layers of verification, I do not write the word "trend."

The Death-Over Myth

Now to the death overs. The biggest myth here is that "the bowler who concedes least is the best death bowler." Data says otherwise.

In the death, a bowler must do two things at once—stop runs and take wickets. These two often conflict. The bowler who concedes less with a mix of yorkers, slower balls and bouncers often takes fewer wickets, because batters take singles without risk. The bowler who attacks for wickets may concede more.

So I measure a death bowler on a composite index—runs per over, control percentage, and "over-specialist wickets" (wickets that change the game's tempo). No verdict without combining the three.

An illustration. The yorker is now T20's most valuable delivery because it kills the batter's swing time. But keeping the yorker has a cost—slightly full and it becomes a six. So a yorker-reliant bowler may have a higher boundary percentage while his control percentage is also high. Economy rate alone can never show this combination.

In my maintained data, death bowlers with control above 55 have seen their economy drop by about 0.6-0.9 runs over the next ten matches—cricket's version of regression to the mean. Those below 50 control rarely sustain a good economy. This is where control percentage's predictive power lies hidden.

Spin and the Middle Overs: A Different Truth

The death overs belong to pace. But spin in the middle overs has a different baseline, and this is where I see the most errors.

In the middle, a spinner's job is to choke runs. Here control percentage touches 65—because a batter taking a big shot carries more risk, so he takes singles. If a spinner concedes 6.5 in this phase, that is better than the middle baseline (7.4-8.2); if he controls 68 percent, the number is exceptional.

The problem is that people do not credit this work. A spinner conceding 26 in four overs and taking two wickets is called a hero; conceding 24 with no wicket is called "helpless." Yet in the data the second often controls more and helps the team more—he keeps the batter under pressure, creating wicket chances at the other end.

On Bangladesh: on our spin-friendly wickets, middle-over spin control works. But on foreign pitches where the ball turns less, a spinner's control drops to 55. Then economy suddenly looks bad. The bowler did not suddenly become worse—the conditions changed. That is the lesson of a baseline audit.

The Batter's Side: The Strike-Rate Trap

The same method applies to batters. In the death overs, a strike rate of 170 is outstanding, 150 is normal. But this number alone says little. The question is at what risk he did it—how many boundaries, how many dots, how many field-saving runs?

A batter makes 40 off 20, strike rate 200. Sounds brilliant. But if 30 of those 40 come from three sixes in two overs and the other 18 balls yield 10 runs, he took risk for the team but gave no consistency. Conversely, someone makes 34 off 20, strike rate 170, but with five boundaries and two dots—he gave more at lower risk.

So I measure a batter's "death-phase impact" in three numbers—strike rate, boundary percentage, and "ball-sink" (how many balls he wasted). Without this triangle I do not write a verdict on a death finisher.

Venue, Home-Away and Empty Stadiums

One baseline I always add—home versus away. T20 has a home advantage, but it is not always equal. After the 2026 Covid hiatus I logged empty-stadium matches separately. Empty stadiums changed home advantage. Data did too.

The reason is simple. A large part of home advantage comes from crowd pressure—on umpiring decisions, on batter mentality, on bowler adrenaline. Without a crowd, that pressure drops. In my data, during empty-stadium periods home teams won less often than their historical average, and away teams played more freely.

In T20 this effect is phase-specific. In the death overs, where pressure peaks, the crowd's presence or absence makes a big difference. So when judging a death-over spell I check—was there a crowd, was it home, how important was the match. Comparing economy rates without this context is comparing apples to oranges.

Opposition Quality Adjustment

The biggest trap is opposition quality. If a bowler concedes 7 against a weak side and 10.5 against a strong one, where is his "true" economy? Answer: between the two, but weighted by opposition quality.

I divide opponents into three tiers—elite batting line-ups, middle, and weak. Then I look at the bowler's numbers separately by tier. If a bowler concedes 9.8 against elite line-ups and 6.5 against weak ones, you need both to understand his real value. An average alone loses half the story.

Here the ten-match threshold and opposition tier work together. I only conclude if at least three matches came against elite line-ups. Otherwise I say: "sample insufficient, verdict withheld." That is not weakness; that is honesty.

Contrarian Angle: Correlation Is Not Causation

Now I write a caution against my own method, because without it the piece is incomplete. There is a relationship between control percentage and good economy—but a relationship is not a cause.

Consider a bowler with high control and low economy. This may have several causes. One, he really is bowling well. Two, he is bowling in a phase where batters are defensive. Three, a partner at the other end is taking wickets, so batters avoid risk against him. Four, fielders are taking brilliant catches, so wickets fall even with lower control. Without distinguishing these four, you cannot know a bowler's true contribution.

So when I say "control percentage is predictive," I understand it is a correlation, not proof of cause. To find cause you must look at match state, field setting and partnerships together. A number opens a door, but you need a map to see what lies behind it. Modric ran twelve kilometres, but the map showed where the game turned—not the total, but the phase and position told the real story. Cricket is the same: a spell's story lives over-by-over, not only in the total.

Another trap—survivorship bias. We look at successful death bowlers' control percentages and think high control is the cause of success. But bowlers who failed with low control and were dropped—their data we do not see. So a baseline must be drawn from the whole population, not only the stars who survived.

Method Note (Appendix)

Here I open my method so anyone can verify it. First, data sources: ball-by-ball data from the ICC and media, plus my own logged control tagging. Second, cleaning: rain-affected matches, Duckworth-Lewis-influenced innings and injury replacements were separated out, because they distort baselines. Third, phase split: overs 1-6, 7-15, 16-20; super overs separate. Fourth, weighting: opposition tier, venue type and home-away.

I excluded single-match control outliers, as with single-match xG, because an extreme number in one match is usually noise, not signal. Likewise, I keep spells of under 12 balls per innings separate—they inform but do not join a trend.

This appendix is no formality. It is my core contract as a writer: I show the path to what I claim, so you can recalculate yourself. Without a reproducible method, data analysis is just arranged opinion.

Takeaway: What to Watch Next Series

Next series, when you see a death-over spell, ask three questions before glancing at the scoreboard. First, what is this bowler's control percentage? Second, what is the baseline for this phase, this venue, against this opposition? Third, does this spell match the ten-match pattern, or is it one event?

If you have answers to these three, a bad night's spell cannot fool you—you will know it is noise, not signal. And if you do not have them, at least wait ten matches. Patience is this game's cheapest, most neglected metric.