From Hand Counts to Model Estimates: How Cricket's Pressure Map Was Built

Sitting in the press box at the Khulna stadium in 2026, I was stunned by a scoreline. Abahani Limited Dhaka versus Sheikh Russel Krira Chakra ended 1-1, yet the model I had built showed 2.7 xG for Abahani and 0.8 xG for Sheikh Russel. The scoreboard told one story, the numbers another. That day I decided to read every cricket match as a dataset rather than a story. Standing in August 2026, with preparations for a new Bangladesh Premier League season underway, the conversation around cricket data has grown louder. The problem is that much of that conversation is about heatmaps, wagon wheels, and graphics. The metrics that now float on screens were built long ago from hand-counted work. Before the model had a name, I counted chances by hand. The first thing to understand is that pressing logic from football cannot be copied directly into cricket. In football, PPDA tells you how many passes an opponent completed before each defensive action. But in cricket, pressure is not a continuous state; it is a series of discrete events. In 2026, I applied PPDA to Germany's 0-2 loss to South Korea, because football already had a language for measuring pressure. Cricket has to build that language differently. In my dossier I break pressure into four countable events: the density of dot balls in the powerplay, the run-rate drop in the five overs after a new batter arrives, the ratio of wicket-taking balls, and the ability to suppress boundaries in the last ten overs. Combined, these give a Pressure Index that sits two different venues' defensive pressure within the same frame, after adjusting for weather and pitch. The second most common mistake is failing to strip out cross-venue comparisons. Dew at the Sher-e-Bangla National Cricket Stadium in Dhaka behaves differently, and Chattogram's pitch behaves differently again. In 2026, when sport stopped, I analyzed 83 matches in empty stadiums. Home win rates fell from 43 percent to 33 percent, and goals per game dropped from 3.2 to 3.0. From that data I built an empty-stadium adjustment coefficient, adding 0.15 xG to away teams. In cricket I apply the same principle with separate pitch coefficients for Dhaka, Chattogram, and Sylhet, because even within one venue a day match and a night match behave differently. Here is the real conflict. Many cricket data analysts open with a heatmap, then a boundary run map, then tell a story. But a heatmap misrepresents a player's role. A finisher's job is short, high-damage innings, which a heatmap shows only as small density. An anchor's job is a long stay at the crease, which can disappear in a heatmap without a single boundary highlight. This is why, in my view, two numbers must precede any visualization: runs per 90 balls, and the drop in strike rate when under pressure. These two numbers are far more reliable than a play-off performance narrative. The full framework rests on three layers. The first is manual count: hand-counted shots, hand-counted dot balls, hand-counted pressure overs. The second is verification against tracking data, with any divergence recorded. The third is a standardized dossier built from that verified baseline, applicable across batters, bowlers, teams, and venues. These three layers can answer the same question differently, and that is normal. The model and the eye witness are both needed, but when the witness becomes the judge, the data stops working. For Bangladesh specifically, three factors must sit in every analysis: capital constraints, the absence of performance-data infrastructure, and squad-pool size relative to opponents. A low-scoring match at Sylhet cannot be compared to a high-scoring match at Dhaka. Yet many discussions skip that correction, and then build big claims on small samples. One good spell from a bowler does not tell the story of a ten-year career, and one match of high xG is not a season's picture. A caveat is needed on manual counts. If hand counts are treated as morally superior to machine data, the verification function of the model is lost. My own rule is to publish both numbers side by side: hand-counted and tracking-derived. Where they diverge, that is curiosity, not suspicion. This lets readers see where the model and reality part ways. Another issue is forcing PPDA knowledge into cricket. Football pressing is a continuous state; cricket pressure is discrete. So the phrase 'pressing root' needs a cricket-specific definition before it can be used. Placed inside three defined events (dot-ball clusters, wicket-taking balls, boundary suppression) the word pressing regains meaning; without that, it is just a word. The signal to track most closely in the coming season is the condition-corrected defensive rate. Across the tournament, comparing each team's middle-over dot-ball density with the opponent's run-rate difference reveals which teams are creating real pressure and which are simply scoring slowly. That difference is the real message. A match scoreline gives readers excitement, but a process scoreline gives a signal for what the next season may hold. Cricket data becomes real power when it makes defensive pressure countable even when it is outside the eye. And that language is still being built, not only on screens, but standing in the middle of the field.