KKR's 9.1 Powerplay Rate Was an Illusion: The Story of Runs Built on Edges

Kolkata Knight Riders lost to Mumbai Indians in the 2026 IPL playoff, and the report that formed on the desk afterwards was the easiest one to write: a 9.1 run rate in the powerplay, then a collapse through the middle overs. The number was correct. The story was not.

I was building a ball-by-ball pressure index at a Kolkata digital sports startup at the time — the 2026 version that later became the foundation of my expected-runs model. I spent three nights tagging every powerplay ball: bat middle, edge, mistime, and fielder position. What came out was the most uncomfortable result of that season — 41 percent of KKR's powerplay fours and sixes came off the edge or a mistimed shot, and their phase-adjusted powerplay rate was not 9.1 but roughly 7.3. Their strongest-looking phase was their most fragile.

This piece is the changelog of that model, not its output. Because a number you cannot reproduce is just an opinion wearing a decimal point.

Context: Why the powerplay number lies

The T20 powerplay is four overs long — six in 2026 — with only two fielders outside. Every run scored in this phase is not worth the same. A four through a clean middle is a decision won against the bowler; a four through the edge is just an event that cannot be repeated on the next ball. The commentary box calls both a 'boundary', and that is where the confusion begins.

The traditional scorecard's powerplay run rate is a simple division: runs ÷ overs. It does not record where the ball came from, what shot the batter intended, or what the fielder's position was. My index's first version was making the same mistake — I added up boundaries one by one without asking what kind of boundary it was. The first draft was calling KKR the most aggressive powerplay team of that season. In the second draft I weighted edges at 0.5, and the ranking changed. In the third I added mistimes at a coefficient of 0.7, and only then did the model tell me the truth.

A confession is needed here: every metric is a translation, and every translation loses a dialect. The xG borrowed from football — weighting shot quality by position and angle — has to change its accent when brought into cricket, because cricket carries wicket risk alongside every ball, not shot quality alone. Losing a wicket and hitting a four in the powerplay are two sides of the same coin, and any model that leaves wickets-in-hand out of the equation is weighing baseball, not cricket.

Core analysis: Sunil Narine's cameos and the hidden cost

The biggest story of KKR's powerplay that season was Sunil Narine — his quick cameos as an opener changed the first over of the match. On television it was thrill. In my phase-adjusted data it was telling a different story: Narine's raw strike rate as an opener was showing roughly 20 percent higher than his phase-adjusted strike rate, because a large share of his runs flew off edges over slip and third man.

I need to be clear here that I am not calling Narine a bad batter. I am saying his 'impact' was as unstable as raw strike rate does not show. An opener whose 40 percent of runs come off edges can be out for zero with the same intent the next match — and that was the early signal of the middle-overs collapse. Against Mumbai in the playoff it happened: in the 14 overs after the powerplay, KKR's phase-adjusted run rate fell below 7, while the raw scorecard calls it 'middle-overs sluggishness' and moves on.

The second part of the model came in useful here — a pressure index that combined required run rate, wickets-in-hand and fielding intensity on every ball. I found that KKR's real crisis was not in the powerplay but between the sixth and tenth overs, where their rotation against spin slowed — and the raw numbers hid this gap behind the 9.1 powerplay figure. Pressure is not a feeling; it is a ratio you can measure ball by ball, provided you decide the context you are measuring in.

Contrarian angle: When the model was right and still wrong

Here is the part I have never written before. My index had correctly flagged KKR's middle-overs stall before the playoff. But it did not say why that stall would come. That inference I made later, and not during commentary, but after the match watching clips.

Two things stayed outside my model. First, the fielding set-up: Mumbai pulled the boundary rider in during the middle overs, forcing KKR's batters to attempt the big shot, and fielder position was a static variable in my index, not dynamic. Second, the wicket type: the Eden Gardens surface in that playoff was slow for spin, and my model was weighting all pitches by the same average. The model was right about the average and wrong about the moment.

This is where I have to be most careful with my own method. Once you have built a metric, it is easy to hold the illusion that what is not in the data was not on the field. The regional commentary box had mentioned the pitch's slowness before that playoff — I dismissed it as 'anecdote' because I had a three-decimal index. The eye test has to be treated as a data source, not as a rival. My index that day was accurate to three decimals and just as wrong.

Takeaway: What to watch next time

My suggestion for the coming season is simple but uncomfortable: never read the powerplay run rate alone — read it alongside the percentage of edges, the ratio of mistimes and wickets-in-hand. A team building a run rate of 9, if more than 40 percent of its output is coming off the edge, is not 9 but 7 — and will find itself at 5 the next match.

My index is now on its third version, and I have kept its changelog public. Do not trust a metric until it has embarrassed you once.