Most "predictions" published on the internet are feelings expressed as numbers. The desk treats predictions differently: every published call is the output of four signals combined. The desk walks through each signal, the weighting, and the back-testing record so you can calibrate the value of a call.
Signal 1: Pitch read.
The pitch read is the single most-predictive signal. The desk's pitch read protocol is short — four observations, 90 seconds — but the observations are weighted heavily in the final call.
Pitch read weighting in the final call: 35%.
Specifically, the desk's pitch read translates into a captain pick type (top-order / spinner / pace) and a contest format signal (small-league vs. mega contest leaning).
Signal 2: Dew factor.
The dew factor is the second most-predictive signal. On dew nights, the chase becomes the easier side; the second-innings total ceiling rises; spinner captains underperform; pace-bowling all-rounders gain value.
Dew factor weighting in the final call: 25%.
Signal 3: Head-to-head.
Head-to-head is the third signal. The desk weights head-to-head by player continuity — a player who's played in three of the last five match-ups has a meaningful data point. Franchise "history" without roster overlap is not signal.
Head-to-head weighting in the final call: 20%.
Signal 4: Recent form.
Recent form is the last signal, weighted at 20%. The desk weights the last five innings at 70%, last season at 20%, career-aggregate at 10%. Form is signal, but it's the last filter — not the first.
Combining the signals.
The four signals are combined into a call:
- All four signals agree → confidence is High, call is published as the desk's pick of the night.
- Three of four signals agree → confidence is Medium, call is published with caveats naming the dissenter.
- Two-of-four or worse → confidence is Low, the desk publishes the disagreement and asks readers to pick for themselves.
What the desk's predictions aren't.
The desk's predictions are not:
- Sportsbook odds — the desk does not publish or imply wagering advice.
- Guarantees — T20 matches have a high-variance outcome. Even the High-confidence call is a probability statement, not a promise.
- Subscription-locked — every prediction is freely readable. The desk may publish a brief newsletter in future; the desk itself will remain free.
Contest vs. prediction.
A prediction is a match-level call — which side wins. A contest call is a player-level pick — which XI scores most fantasy points. The two are related but different. A batter can score top fantasy points on the losing side, and a captain pick can carry an XI to top-half finish on a contest where the desk's match-level call was wrong.
The desk publishes predictions (match-level) and captain picks (player-level). For your XI, the captain pick is what matters most.
Back-testing the signals.
The desk publishes a track record for the four signals, weighted against the back-test. Recent numbers:
- Pitch read — accurate surface-type identification in 87% of matches.
- Dew factor — accurate chase-vs-defend call in 71% of night matches.
- Head-to-head — useful in 62% of matches; rarely a strong signal on its own.
- Recent form — useful in 78% of matches for captain picks specifically.
High-confidence picks specifically.
When all four signals agree, the desk's published pick has cleared top-3 in fantasy points in 62% of matches over the last three IPL seasons. The hit-rate is meaningful — well above the random baseline of ~33% (three players in the top three is meaningless).
The hit-rate in mega contests is lower — 14% top-0.1% finishes on differential picks. Above the random baseline of ~1%, but not a guarantee.
What the reader does with the call.
The desk's calls are frames, not certainties. Read the reasoning, apply it to your own XI, and pick for yourself. The picks are reasoning examples — useful as a comparison, not as a buy-instruction.
Further reading.
The desk's match predictions lane applies the four signals to every published call. The tips lane shows how to apply them to your own XI selection.
