Predictions: how the desk frames a match
Our predictions desk publishes scenario reads, not guaranteed outcomes. This page explains the language we use, the conditions we attach to each scenario, and how we update the call after the toss — and how we audit our calls after the result.
What a prediction is on this desk
On Fantasy Cricket Live, a prediction is a probability-weighted scenario read with explicit conditions. We do not name a winner; we name a most-likely outcome under stated assumptions, and we name the alternative scenarios that would change the call.
Each prediction is timestamped and archived verbatim. When the fixture ends, we publish a post-match review that cites the original wording and names whether the scenario played out. That archive is the desk's accountability surface — readers can audit any call we have ever made.
Scenario framing
A typical prediction page lists three scenarios: a base case (the most likely outcome if pre-match assumptions hold), an upside case (the outcome if our specific conditions reads are correct), and a downside case (the outcome if our conditions reads are wrong). Each scenario has an associated captain-case and a risk label.
Why we frame in scenarios
Single-outcome predictions are a confidence trick. No probabilistic model produces one number that survives contact with a live match. Scenario framing makes the uncertainty visible to the reader, which is the only way to use the prediction responsibly.
Risk labels we use
Each prediction carries one of three risk labels. Low risk means the call rests on multiple converging signals (form, role, conditions, opposition matchup). Medium risk means one signal is contested or the sample size is small. High risk means the call rests on a single contested signal — usually a debutant, a returning player, or a venue where the conditions read is genuinely unclear.
| Risk | Confidence basis | Sample floor | Update cadence |
|---|---|---|---|
| Low | 3+ converging signals | Last 6+ matches | Reviewed only after toss |
| Medium | 2 signals, 1 contested | Last 3-5 matches | Reviewed after toss + confirmed XI |
| High | 1 contested signal | Last 1-2 matches | Reviewed continuously until toss |
How we update after the toss
Once the toss lands, the prediction page is updated in two stages. The first update reflects the playing XI and the bat/bowl decision. The second update reflects conditions evidence (whether the pitch is playing as the desk read, whether dew is forming). If the second-stage read materially changes the call, the original scenario block remains visible and a revised block is appended.
The prediction archive
Every prediction we publish is archived at the same URL with the original timestamp and wording preserved. The archive is searchable by team, by venue, and by risk label. The post-match review link is appended to the archive entry when the fixture ends, so a reader can audit the chain from pre-match to result in one place.
How to use the archive
The archive is not a leaderboard — small samples mislead. Use it to audit our pattern of calls, not to compute a hit rate over a handful of matches.
Reading a prediction card
A prediction card contains four elements: the base-case scenario (one paragraph), the upside and downside cases (one paragraph each), the captain-case table with risk labels, and a list of three or four signals that would force a re-read. The card is short by design — the goal is to give the reader a defensible working model, not a wall of caveats.
What we don't predict
We do not predict player injuries, selection surprises, or last-minute withdrawals. We do not predict toss outcomes (we report them). We do not predict individual player scores. We do not run 'best XI' picks without an attached conditions rationale. The boundaries keep the desk's predictions auditable — every call can be traced back to a stated set of signals.
When we revise after the toss
The desk revises a prediction after the toss in two stages. The first revision reflects the playing XI and the bat-or-bowl decision. The second revision reflects the conditions evidence — whether the pitch is playing as the desk read, whether dew is forming, whether the square boundaries are behaving as expected. Each revision is appended to the original prediction page, not substituted in. The original wording remains visible so a reader can audit the chain.
If the second-stage read does not materially change the call, the desk says so plainly. We do not manufacture revisions for the sake of appearing responsive. A prediction that holds up across the toss window is a useful prediction — it tells the reader the conditions read was correct, and the captain-case did not need to change.
The conditions evidence we use
Conditions evidence on the desk is anchored to a defined sample and a date range. We do not use a single match's behaviour as a generic read. The conditions table on each match briefing lists the venue's last five completed fixtures in the same format, with the average powerplay run rate, the average death-over run rate, and the chase-versus-defending record over that sample. Small samples are labelled explicitly — a venue with fewer than three relevant fixtures is flagged as 'small sample' and the conditions read is downgraded accordingly.
We do not use ball-by-ball data as a primary conditions signal. The desk's working rule is that aggregated conditions data over a defined sample is more reliable than a single over's behaviour. The exceptions are dew windows and pitch cracks, which are time-sensitive and which we update when the ground report lands.
Why we do not publish a hit rate
The desk does not publish a hit rate over its prediction archive because the metric would mislead readers. Most of our predictions are conditional — a call that names a scenario under stated conditions is not the same as a binary call. The archive mixes fixture types (some fixtures have well-defined conditions reads; others do not), and the absolute sample is small in any given tournament window.
What the archive is for is auditing individual calls. A reader who wants to evaluate the desk's pattern can read the prediction-vs-outcome tables and see how often the conditions read held up, how often the captain-case landed, which risk labels we got right or wrong. That is a more honest use of the archive than a single percentage, and it does not collapse the desk's conditional calls into a metric that does not exist.
Working with the archive
The prediction archive is searchable by team, by venue, and by risk label. To use it well, pick a narrow slice — for example, all 'low risk' calls at one venue across the last ten fixtures — and read those entries before drawing conclusions. The desk's pattern is usually visible at that slice level; it is rarely visible at the aggregate level.
If you find a pattern the desk has not flagged, route it to the contact page. We treat reader-submitted findings as inputs to the methodology note on the predictions desk. Material corrections are added to the public corrections log.
Reading support on this page
Every section in this article is sourced. Where a figure is from a small sample we say so explicitly. The article is updated when fixtures confirm, when the toss lands, and when post-match review changes the read.
Frequently asked
What does 'most likely outcome' mean here?
It means the scenario with the highest probability under our stated assumptions. We do not assign a numeric probability to each scenario — that would imply a precision we cannot defend.
Why don't you publish a hit rate?
Hit rate over our archive is misleading because (a) most calls are conditional, (b) the archive mixes fixture types, and (c) the sample is small in absolute terms. The archive is for auditing individual calls, not for aggregating a metric.
What happens if a player pulls out after the prediction is published?
The prediction page appends a 'late-news' note, the captain-case is revised if the withdrawn player was named, and the post-match review explains how the late change affected the call.
Can I submit my own prediction for the desk to review?
Reader predictions are not published, but readers can route tips and methodology questions via the contact page.