We had a breakthrough with the Alex Scott report. It was the first I ever fully trusted the process to produce end to end. I edit each report by hand before publication, but this one did not need rewriting. Every post is scored and checked more than once.

For readers, this means sharper questions, stronger evidence, fewer repeated ideas, and conclusions that are more useful in real football work.

How the research process works

We recently overhauled how Tactics Journal Research makes reports. We have revised the research process dozens of times, learning from what worked, what did not, and which questions had to stop before drafting.

1. Gather the source material

The process begins with sources, not a draft. It gathers public web and RSS articles, interviews, and video transcripts, including material published in different languages. It keeps the author, publication, date, link, and relevant passage with every observation. Those details stay attached throughout the work, so we can trace a later claim back to the material behind it.

2. Find report questions

The process groups repeated observations before looking for report questions. Several articles repeating one interview count as one line of support, not several independent accounts. It first looks for grounded ideas emerging in specialist football sources before they reach broad coverage, then searches more widely. Each promising idea becomes a question to test, rather than a conclusion we have already accepted.

3. Score and filter each question

Every question is compared with ideas under consideration and work already published. The process scores novelty, specificity, relevance to a real football decision, source quality and diversity, and lessons from earlier editorial decisions. Stale, duplicate, weak, or irrelevant questions stop here. They remain recorded as stopped work, rather than being reshaped and recycled as apparently new ideas.

4. Define what would count as enough evidence

For each surviving question, the process defines one narrow claim. It records what would support or challenge that claim, whether the sources are genuinely independent, and the minimum evidence needed before drafting. “Worth investigating” means a question is clear and relevant enough to research. “Ready to publish” means the research has met the evidence standard for making a careful conclusion. One does not imply the other.

5. Research the question from several angles

The process creates a research plan and examines several angles in parallel: direct support, counterevidence, other possible explanations, important limits, and practical relevance to football work. If a specific gap can be resolved, it can run one bounded extra round of source gathering or research. If that round still leaves the central claim unsupported, the work stops instead of moving into a draft.

6. Set original-research rules before viewing results

When public sources cannot answer a question, we set original-research rules before viewing any results. The process records the required data, exclusions, comparison groups, minimum usable sample, and stopping rule. It first checks that enough suitable data exists. If the minimum is missed, the work stops. We never lower a threshold after seeing results simply to produce a finding.

7. Draft, verify, and revise

Drafting begins only after the evidence gate is passed. Each source must support the exact sentence beside it. The process checks source accuracy, tactical logic, counterevidence, competing explanations, and claims that one thing caused another. A revision can make wording clearer or narrow a conclusion. It cannot invent missing evidence. If a required check fails and cannot be resolved, the report does not proceed to publication.

8. Human review and learning

An editor reads and edits every completed report, assesses the football interpretation, and decides whether it should be published. We remain accountable for that decision. Edits, rejections, and publication outcomes return as lessons for later question scoring and review. That feedback explains why we have revised the process dozens of times: each outcome helps tighten what future work admits, checks, or stops.

Steps 1 through 7 run automatically. At that scale, the process can read, compare, and check more material than one person could continuously track, including material in different languages. This makes wider research and consistent checks practical.

Step 8 is human. Automation provides scale, but it does not make the final interpretation or publication decision. An editor does both and remains responsible.

Why two planned reports stopped

Two recent studies began as report ideas. Each asked a clear football question and was meant to become a Tactics Journal Research report if the evidence was strong enough.

Neither reached drafting. The corner report lacked enough reliably timed sequences. The finishing report lacked enough eligible players across adjacent seasons. Minimum-evidence checks set before analysis stopped both reports. The available evidence could not support either proposed report, but it did not prove either football claim false.

The corner report asked whether defenders follow particular attackers or hold spaces before a corner. It used the public IDSSE dataset. Tracking data record the positions of players and the ball repeatedly over time.

A corner was usable only when those positions could be reliably matched with its delivery. Otherwise, movement before the kick could be confused with a reaction afterward.

The study required usable matches for 80% of corner sequences. Only 30 of 57 were usable, giving 52.63% coverage against the fixed 80% minimum. The minimum was not lowered after that result. The study stopped before classifying movements or calculating results.

The data therefore cannot support the planned comparison as designed. Even 80% coverage would only allow classification to begin. It would not make every classification correct.

The finishing report asked whether finishing above or below expectation persists into the adjacent season. It used StatsBomb Open Data and expected goals, or xG. This model estimates a shot’s chance of becoming a goal using details such as location, angle, body part, and assist type.

The comparison matched each eligible player-season with the same player’s next eligible season. This kept the question focused on adjacent seasons rather than unrelated samples. It required at least 30 pairs, but the available data produced 6. The study stopped before calculating an outcome.

Those open competitions and seasons cannot answer whether above-expected finishing persists. Even 30 pairs would remain a limited sample requiring further checks.

Passing a minimum only starts the next check. It does not prove a finding or make later measurements certain.

Why we waited to charge

This announcement remains free so readers can understand what changed. We waited to charge for Research until the work could reliably distinguish publishable conclusions from duplicate questions, unsupported claims, and insufficient evidence.

The trusted Alex Scott report and the recorded stops showed those standards working in different circumstances. Research reports will move to Tactics Journal Pro. Opinion and the Weekly Recap will remain free.

Subscriptions will fund deeper research that analysts, coaches, scouts, sports scientists, recruitment staff, and other football professionals can use in their work.