Autonity Historical Series · E-04 · Competition Design

Forecastathon Season Scoring:
S1–S4 Composite

Autonity's Forecastathon ran four seasons of forecasting competitions on AFP markets. Seasons 1–3 used a four-component composite that rewarded performance, liquidity provision, accuracy, and community growth. Season 4 simplified to raw PnL.

Mechanics documented for educational record. NTN was the native token at the time of Forecastathon; it had no guaranteed liquidity and scoring marked NTN_ILLIQUID on every output. Not financial advice.

Historical · Autonity S1–S4 Educational Interactive calculator
01 · Why four components?

What the composite was trying to measure

A simple PnL leaderboard favours participants who get lucky on one big position, not participants who help the market function. Autonity designed the S1–S3 composite to reward four distinct contributions to market quality:

Performance
Percentage return on starting capital. The basic measure of whether you made money over the season.
(end − start) / start × 100
Volume
Logarithmic scaling of total trading volume. Rewards liquidity provision without allowing wash-traders to dominate.
log10(total + 1) × scale
Accuracy
How often your directional bets were correct, weighted by position size. Distinguishes lucky from skilled forecasters.
custom per-season formula
Referrals
New participants brought to the platform, each worth 10 composite points, capped at 50 (5 referrals).
min(referrals × 10, 50)
02 · Formula reference

Exact scoring formulae by season

// S1, S2, S3 - Four-component composite
composite = performance + volume_score + accuracy_score + referral_score

performance = (balance_end − balance_start) / balance_start × 100
volume_score = log10(volume_total + 1) × volume_scaling
accuracy_score = weighted_directional_accuracy × accuracy_weight
referral_score = min(referrals_count × 10, 50)

// S4 - Simplified (PnL only, no referral component)
composite = (balance_end − balance_start) / balance_start × 100
NTN_ILLIQUID flag

Every Forecastathon output carried the NTN_ILLIQUID flag. NTN (Autonity's native token) had no established external market at the time of the competition. Reported NTN balances and rewards could not be reliably converted to any reference currency. This flag is preserved in every historical scoring record to prevent misrepresentation of results.

03 · Interactive calculator

Score a participant across seasons

Enter a participant's season statistics. The calculator produces the composite score and component breakdown for the selected season format.

🔒 All inputs are processed locally in your browser. No data is transmitted. Do not enter real personal data — use synthetic or anonymised inputs only.
Season Format
Preset
Sum of all position notionals opened during the season
Season-specific multiplier (e.g. 2.0 for S1)
Pre-computed directional accuracy from season records
Season-specific multiplier for the accuracy component
New users you brought to the platform (cap 5 = 50 pts)
--
Composite Score
04 · Season evolution

Why S4 dropped the composite

The S1–S3 composite created measurable market quality benefits: participants competed on volume and accuracy, not just luck. But it also created complexity that made results harder for participants to predict and verify. By Season 4, Autonity simplified to pure PnL percentage: easier to audit, easier to communicate.

The tradeoff: S4's simple measure re-exposed the platform to luck-based strategies. A participant who opened one large position on the right binary contract could outrank every consistent performer. This tension between composite quality scores and simple transparent metrics is unsolved across prediction market design.

Live equivalent

Polymarket and Kalshi use leaderboards sorted by PnL percentage (similar to S4). Metaculus uses calibration scoring (similar to S1–S3 accuracy component). No major prediction market currently uses all four components. The Forecastathon composite design remains a reference point in academic market microstructure literature on decentralised forecasting incentives.