How the Kairon Score works
Every market on Kairon carries a 0–100 score and a tier letter (D-, C, B, B+, A-, A, A+, S, S+, SS, SS+). The score answers one question: is this market worth your attention right now? It does NOT predict outcomes — that's what AI Strategy and the score-bot do. The score ranks opportunity quality.
The five buckets
Score = volume + liquidity + momentum + excitement + urgency. Maximum 100 by construction. Source is lib/scoring/index.ts. Identical for every surface — Dashboard, Search, market detail, lab page.
log10(volume24hr + 1) × 10Activity > size. A small market that just woke up beats a frozen whale.
log10(liquidityNum + 1) × 8Depth of the order book. Filters out questions nobody can actually trade.
volume24hr / volumeNumAcceleration. % of all-time volume that printed in last 24h.
Binary: (1 − abs(yes − 50)/50). Multi-outcome (3+): 1 − normalised HHI50/50 markets carry more information than 1% certainties. Multi-outcome reads the price flatness (HHI) so a 4-way 25/25/25/25 split scores the same as a binary 50/50.
(1 − daysLeft/30) × 10 if < 30 daysReward markets near resolution where the probability is sharp.
The exact coefficients (10, 8, 20, 10, 10) stay our IP — same trade-off Bloomberg makes with its index methodology. The structure + inputs are public so any score can be sanity-checked against the maths.
The 4-bucket breakdown reorganises the 5 formula inputs for UX (urgency + excitement fold into Momentum). Whale + News default to 0 unless the caller passes a tally — Free users see those greyed; Pro unlocks live values.
Tier ladder
- The score ranks attention-worthiness, not P&L expectancy. A high-tier market can still settle against you — that's why the score-bot at /lab/score-performance exists.
- Markets under $5,000 24h volume get capped at tier A and a LOW CONFIDENCE chip so quiet markets can't ride excitement+urgency stacking into a fake high tier. Raw uncapped tier is still surfaced in the chip tooltip so you can see what the formula thought.
- Whale + News bars need server-side tallies which most callsites don't pass yet. Backfill in progress per §1.3 ship list.
- Score weights are not auto-tuned. Bumping them is a deliberate version change — when v1.1 ships, every callsite + the bot version on /lab/score-performance bumps together so historical comparisons stay clean.