Sports & Wagering — Cricket Intelligence for AI Agents

The lineup · ready now

Each product, in full.

Every one of these is a real, callable endpoint live today. Below: exactly what each delivers, the model behind it, who it's for, and the per-call price.

IPL · T20 · fantasy · real-time

Fantasy picks & captain

Ranked player picks and the optimal captain for any T20 match.

$0.10 / call

pay per call · deterministic

What's inside — 3 layers

Ranked picks. Every player in the squad scored and ranked by predicted fantasy value — an ensemble of expected points, probability of a top-6 finish, and a ceiling model. Separates safe floor picks from explosive ceiling picks.

Captain pick. The single highest-ceiling player, chosen by a 90th-percentile model — because captaincy doubles points and rewards upside, not average. Captain-hit rate 35.6%, more than double random.

Expected + ceiling per player. Each player carries both an expected-points figure and a ceiling, so a lineup builder can balance a safe core against high-variance differentials.

Use case

A fantasy contest bot calls this to build a Dream11-style lineup in one shot — it ranks all 22 players, names the captain that maximizes doubled points, and returns floor-vs-ceiling for differential picks. One call replaces an entire player-projection pipeline trained on 129,000 player-innings. For a fantasy app or tipster, it powers an AI picks feature or backs published tips with a validated model instead of gut feel.

For · fantasy bots · contest optimizers · fantasy apps · tipsters

tool · hsh-cricket-fantasy-picks

View full details →

ranked pickscaptainP(top-6)ceiling

T20 · live · per-ball

Live chase win-probability

Real-time win % for a T20 run chase, updated every ball.

$0.08 / call

pay per call · deterministic

What's inside — 3 layers

Calibrated win probability. A true, calibrated probability the chasing side wins from the exact match state — trained with log-loss, validated on Brier score. Beats the required-run-rate heuristic by 8.6 accuracy points.

Full situation read. Runs needed, balls remaining, current and required run rates, wickets in hand — the complete state returned alongside the probability.

Phase-aware confidence. Appropriately uncertain in the powerplay, confident by the death overs — accuracy climbs from 73% to 88% as the chase resolves.

Use case

An in-play betting bot prices a live chase market off a calibrated win probability instead of a crude run-rate rule — and acts on the gap between the model and the market line every ball. Trained on 475,000 ball-states from 4,406 real chases, AUC 0.88 on unseen matches. For a betting syndicate or broadcaster, it's a validated live win model or an on-air win-% graphic, without building 475K rows of training data first.

For · in-play betting bots · trading agents · broadcasters · syndicates

tool · hsh-cricket-chase-winprob

View full details →

win probabilitycalibratedper-ballAUC 0.88

T20 · in-play · live

Momentum & pressure index

Which side is gaining, and how much pressure is on the chase.

$0.07 / call

pay per call · deterministic

What's inside — 3 layers

Momentum delta. The swing in chasing-team win-probability between two match states — positive means the batting side is surging, negative means the bowling side is squeezing. Labelled into readable trends.

Pressure index. A 0-to-1 score combining how far the required rate exceeds a par chase rate with wickets lost — exactly how squeezed the chasing team is right now.

Before / after read. Both win-probability points that produced the swing, so the size and direction of the turn are explicit and auditable.

Use case

An in-play trader watches momentum for the exact moment a chase turns — acting the instant it flips to the bowling side, before the market fully reprices. Built on the same win-probability model behind the chase product, so momentum and pressure are consistent signals. For a broadcaster, it drives a live momentum graphic; for an analyst, it measures exactly when and how fast matches swung across a season.

For · in-play traders · betting bots · broadcasters · analysts

tool · hsh-cricket-momentum

View full details →

momentumpressure indexswingtrend

T20 · innings · live

Par-score / projected total

Projected first-innings total + calibrated P10-P90 confidence band.

$0.08 / call

pay per call · deterministic

What's inside — 3 layers

Projected total. The model's predicted final first-innings score from the current state — trained on the real scoring curve, so it doesn't blow up early the way run-rate extrapolation does.

Confidence band (P10–P90). Calibrated low–high bounds with 85%% coverage, returned alongside the projection.

Phase-routed accuracy. Powerplay / middle / death specialist models — death-overs median error 8.5 runs, band tightening as the innings resolves.

Use case

An over/under betting bot prices a first-innings total market off a trained projection with a calibrated confidence band — and in the powerplay, the model's restraint versus naive extrapolation flags mispriced totals. For a commentator or analyst, it answers 'is this a par score?' on air, or benchmarks actual totals against model-par to grade batting.

For · over/under bots · betting agents · commentators · analysts

tool · hsh-cricket-parscore

View full details →

projected totalpar scoreP10–P90 bandover/under

T20 · media · full-match

Over-by-over win-prob timeline

The full match win-probability curve — the broadcast worm.

$0.05 / call

pay per call · deterministic

What's inside — 3 layers

Full win-prob curve. Win-probability at every over of the chase, returned as an ordered series ready to plot — the broadcast 'worm' graph in data form.

Score-tagged points. Each point carries the score and wickets at that over, so the curve annotates itself for a graphic or a recap.

Model-consistent. Every point comes from the same validated AUC-0.88 chase model, so the curve matches the live win-probability product exactly.

Use case

A match-report bot generates the win-probability worm for every match with no manual charting — feed the over-by-over states, receive the full curve. For a broadcaster or sports-media app, it's the classic on-air win-probability graphic sourced from one API call; for a recap bot, the curve's peaks and troughs become an automatic match narrative.

For · match-report bots · broadcasters · sports media · recap agents

tool · hsh-cricket-timeline

View full details →

win-prob curvethe wormover-by-overmedia-grade

T20 · matchup · historical

Batter-vs-bowler matchup

Head-to-head record and dominance for any batter-bowler pair.

$0.05 / call

pay per call · deterministic

What's inside — 3 layers

Head-to-head record. Balls faced, runs scored, and dismissals for the exact pair — the raw history of how this specific matchup has played out.

Bayesian-adjusted strike rate. Strike rate in the matchup, shrunk toward the batter's overall rate so a 4-ball sample doesn't produce a wild number.

Dominance score. A single number above or below 1.0 — above means the batter dominates this bowler, below means the bowler has the edge.

Use case

A fantasy bot weights picks by who's likely to bowl to whom — boosting batters who historically dominate the expected bowlers. An in-play agent pulls the matchup the moment a bowling change happens and adjusts its read. Built from 111,565 real batter-bowler pairs with Bayesian shrinkage. For a preview writer or analyst, it grounds 'key battles' content in real head-to-head data.

For · fantasy bots · in-play agents · preview writers · analysts

tool · hsh-cricket-matchup

View full details →

head-to-head111K pairsdominanceBayesian

T20 · first-innings · live

First-innings win-probability

Win % while a team is still setting a total.

$0.04 / call

pay per call · deterministic

What's inside — 3 layers

Setting-team win probability. The chance the team batting first goes on to win, from their current first-innings state and match context — a harder prediction than the chase, honestly scored at AUC 0.76.

Venue-aware. A venue batting-index feature tells the model whether this is a high-scoring ground or a bowler's pitch — worth six AUC points over match-state alone.

Projected-total context. A naive projected total is returned alongside as a reference point for the setting side.

Use case

A betting bot prices the innings break — taking a model-backed view on the match before the chase even starts, while the setting team is still batting. Lifted from AUC 0.70 to 0.76 by adding venue batting-index and team-strength Elo. For a broadcaster, it's a first-innings win graphic; for an analyst, it quantifies how score, wickets, and venue shape first-innings win odds.

For · betting bots · trading agents · broadcasters · analysts

tool · hsh-cricket-first-winprob

View full details →

first-inningsvenue-awareEloAUC 0.76

T20 · player · form

Player form & fatigue

Hot/cold form rating and bowling-workload fatigue flag.

$0.04 / call

pay per call · deterministic

What's inside — 3 layers

Form rating. Recent scoring (exponentially weighted) against the player's own career baseline, labelled red-hot, in-form, steady, cold, or out-of-form.

Fatigue flag. Recent bowling workload over the last few appearances, flagged low, moderate, or high — a heavily-worked bowler is a fade signal.

Gated to real players. Restricted to 502 established IPL players with stable histories, so ratings surface genuine stars rather than small-sample noise.

Use case

A fantasy bot nudges picks toward in-form players and away from cold ones, and flags heavily-worked bowlers as fades. Form is recent fantasy output versus a player's own baseline; fatigue is recent bowling load — both gated to 502 established IPL players so the signal is stars, not minnow noise. For a tipster or analyst, it backs 'in form' claims with a real ratio and tracks trajectories across a season.

For · fantasy bots · preview agents · tipsters · analysts

tool · hsh-cricket-form

View full details →

form ratingfatigue502 playersgated

T20 · innings-break

Chase difficulty rating

How hard a target is to chase, graded from history.

$0.03 / call

pay per call · deterministic

What's inside — 3 layers

Historical success rate. The real fraction of chases that succeeded at this target level, from 4,406 completed chases — not a guess, the actual record.

Difficulty label. A readable grade — Easy, Gettable, Tough, Hard, or Very Hard — mapped directly from the historical success rate.

Transparent by design. Grounded directly in outcomes, no black box — which makes it the most defensible signal in the line.

Use case

A betting bot gets an instant difficulty grade the moment the first innings ends — anchoring an innings-break view in the real success rate of every historical target band (targets under 120 were chased 88% of the time; 200+ only 17%). For a commentator, it answers 'is this gettable?' at the halfway point with a data-backed grade instead of a hunch.

For · betting bots · preview agents · commentators · analysts

tool · hsh-cricket-chase-difficulty

View full details →

difficulty4,406 chasessuccess rateinnings-break

Nine cricket-intelligence products, built for agents.

Each product, in full.

What's inside — 3 layers

Use case

What's inside — 3 layers

Use case

What's inside — 3 layers

Use case

What's inside — 3 layers

Use case

What's inside — 3 layers

Use case

What's inside — 3 layers

Use case

What's inside — 3 layers

Use case

What's inside — 3 layers

Use case

What's inside — 3 layers

Use case

Discover, pay, receive — autonomously.

Trained models, not raw scores.