HSH·Intelligence
HomeSports & wageringPar-score / projected total
T20 · innings · live

Par-score / projected total.

The projected final first-innings total from any mid-innings state — now with a calibrated P10–P90 confidence band. A phase-routed quantile model (powerplay / middle / death specialists) that returns not just a projected total but honest low–high bounds, with 85% band coverage. In the death overs it tightens to a median error of 8.5 runs. The "is this a par score?" answer every over/under bettor and trading agent wants.

$0.08 / call
deterministic price · pay per call
toolhsh-cricket-parscore
How to buy

Two ways in — pick yours.

For people & teams
  1. Buy credits with a card — no wallet, no crypto.
  2. Get an API key emailed to you instantly.
  3. Call this product with your key — credits debit per call.
  4. Check your balance anytime via Check balance.
Buy credits with card →
For AI agents
  1. Point your agent at the x402 endpoint, or describe the need in plain language.
  2. Pay per call in USDC (Base or Solana) — no signup.
  3. Receive the signal inline, settled on-chain.
View agent catalog →
What you get · 3 layers

The signal, in full.

Pass the match state; receive a complete, model-backed read.

1

Projected total

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

projected_final_total
2

Naive baseline

A naive run-rate extrapolation is returned alongside the model projection, so the model's edge over the method everyone else uses is explicit.

naive_extrapolation
3

Confidence band

Every projection ships with calibrated P10-P90 bounds (85% coverage) that widen early and tighten through the innings — an explicit read on how much to trust the number.

low_p10 high_p90 phase band_width
Where it comes from

Fit to the real scoring curve.

Trained on 4,683 real T20 matches from Cricsheet — free, legal, ball-by-ball — then validated on unseen matches. No fragile live-feed dependency; the intelligence is in the model.

Real scoring curves
Trained on the actual run-accumulation shape of thousands of T20 innings, which is non-linear.
trained
Gradient-boosted regressor
XGBoost over state features — Phase-routed quantile ensemble — death-overs median error 8.5 runs, calibrated 85%% P10-P90 band.
proprietary methodology
Venue context
Optional venue input adjusts for high-scoring vs bowler-friendly grounds.
feature-engineered
Out-of-sample MAE
Death-overs median error 8.5 runs; 85%% calibrated P10-P90 band coverage, validated on unseen matches.
out-of-sample
Sample output

What comes back.

{
  "projected_final_total": 179.5,
  "low_p10": 164.6,
  "high_p90": 201.2,
  "phase": "death",
  "band_width": 36.6,
  "naive_extrapolation": 170.6
}
Who calls it

Use cases.

For autonomous agents
Over/under bot
Price the total
Price a first-innings total market off a trained projection, not a naive rate.
Betting agent
Spot early value
In the powerplay, the model's restraint vs naive extrapolation flags mispriced totals.
Dashboard bot
Live par read
Show a projected total that updates sensibly ball by ball.
For people & teams
Bettor
Beat the naive line
Get a projection with a calibrated confidence band, model-projected not run-rate math.
Commentator
Is it par?
Answer 'what's a good score here' on air with a real model.
Analyst
Score benchmarking
Compare actual totals to model-par to grade batting performances.
How to call it

Discover, pay, receive.

01 · Discover
Find the tool
Via MCP at dod.hshintelligence.com/mcp or the /x402/catalog.
02 · Pay
Pay per call
Send the x402 payment in USDC on Base. Fixed deterministic price.
03 · Receive
Get the signal
The signal returns inline — ready for your model, lineup, or strategy.
# call the parscore tool
curl -H "X-Payment: <x402-usdc-base-payload>" \
"https://dod.hshintelligence.com/mcp" \
-d '{"tool":"hsh-cricket-parscore","input":{...}}'
Questions

FAQ.

How much better than run-rate extrapolation?
Projected from the real T20 scoring curve, not linear run-rate — so it stays sane early and ships a calibrated P10-P90 confidence band (85% coverage) at every state.
How accurate by the death overs?
Death-overs median error of 8.5 runs, with a calibrated 85%% P10-P90 confidence band.
What does it need?
Balls bowled, wickets, and runs so far. Venue is optional context.
Is it only for the first innings?
It's designed for projecting a first-innings total; for a chase, use the chase win-probability or difficulty products.
More in this category

Sports & wagering.