Skip to main content
PYON

Official Competition Rubric

Hackathon Marking Scheme

Build the best trading strategy on Pyon and get scored across 7 dimensions. Total: 100 points.

100
Total Points
7
Categories
21
Criteria
4
Score Levels

Weight Distribution

Research & Thesis
15
Strategy Design & Complexity
15
AI Utilisation
15
Backtest Performance
20
Risk Management & Volatility
15
Portfolio & Exposure Management
15
Presentation & Documentation
5

Scoring Categories

Each criterion uses a multi-level scale. Judges assign a score within each band. Tap a category to expand.

Depth of fundamental and technical research conducted before building the strategy.

AI-Powered Research used (factor scores, company analysis, peer comparison)

/5
0 = No research2 = Basic single-ticker lookup4 = Multi-factor analysis with comparisons5 = Full thesis with bull/bear/base cases

Investment thesis quality (clear rationale, scenario analysis, risk assessment)

/5
0 = No thesis2 = Vague rationale4 = Structured thesis with catalysts5 = Institutional-grade thesis with scenarios

Research-to-Strategy pipeline (Convert to Strategy used effectively)

/5
0 = Not used2 = Converted but thesis loosely connected4 = Strategy clearly derived from thesis5 = Seamless research-to-strategy with AI context

Sophistication and logical coherence of the strategy graph in the Builder.

Node variety (triggers, conditions, actions, routers used appropriately)

/5
0 = Single node2 = Basic trigger-action pair4 = Multi-condition branching5 = Rich graph with routers, multiple triggers, layered conditions

Logical flow & connectivity (no disconnected nodes, sensible edge routing)

/5
0 = Broken graph2 = Functional but messy4 = Clean and logical5 = Elegant, well-organised with annotation notes

Advanced nodes (LLM Request, API Request, Persistent Variables, Scheduling)

/5
0 = None used2 = One advanced feature4 = Multiple advanced features5 = Creative integration of LLM/API nodes with stateful logic

Grade Boundaries

Distinction85, 100

Exceptional across all dimensions. Institutional-quality, strong risk-adjusted returns.

Merit70, 84

Strong AI usage, solid backtest, meaningful risk management.

Pass50, 69

Functional strategy across core platform features.

Developing30, 49

Gaps in testing, risk management, or AI usage.

Incomplete0, 29

Strategy not functional or untested.

Judging Notes

How to Judge

  1. Each team presents their strategy live on Pyon. Judges observe the Builder canvas, backtest results, and research reports.
  2. Score each of the 21 criteria independently. Sum across all 7 categories for the final score out of 100.
  3. Ties are broken by (a) Backtest Performance, then (b) Risk Management, then (c) AI Utilisation, in that order.

Important

Backtests use historical data only. Judges should reward strategies that demonstrate awareness of overfitting, out-of-sample validation, and realistic assumptions over raw return numbers.

Pyon Hackathon Marking Scheme v1.1, April 2026