← Home

Benchmark Bots

Automated players that set the skill bar for rewards eligibility.

📊 The Data Chain

BeatTheBots creates market intelligence from player predictions. Each prediction contributes to the weighted consensus — a skill-adjusted view of what the market believes will happen.

Our benchmark bots use this market data to set performance thresholds. To earn rewards, you must demonstrate skill by beating these automated baselines.

Each bot has different data access. Level 1 sees only raw predictions. Level 2 adds player skill ratings. Level 3 includes external event data. Level 4 uses AI reasoning with full database access.

⚡ To earn rewards:Your series score must beat the Level 2 Gatekeeper (MarketBot).

🤖 Meet the Bots

MathBot
Level 1Pure Mathematician
1499
R-Score

Uses unweighted average of all predictions. Optimises for accuracy while avoiding the consensus trap (predicting at consensus = zero score). Balances timing and conviction.

Data Access
Scoring rules and formulas
All player predictions (values only)
Current event score
Ability to calculate projected scores
No Access
Player R-Scores (skill ratings)
Event participant data (teams, form)
MarketBot
Level 2The Gatekeeper
1530
R-Score

Uses R-Score prediction data to detect which direction skilled players are betting. Reads market momentum rather than simply following consensus. This is the eligibility bar for rewards.

Data Access
Everything MathBot can see
Player R-Scores for momentum analysis
No Access
External league standings/form data
Live match scores from APIs
OracleBot
Level 3External Intelligence
1501
R-Score

Uses opponent-adjusted team strength from league standings and live scores. Calculates form by weighting results against opponent quality. Applies conviction-based response to live score deviations: holds position on small swings, revises on large ones.

Data Access
Everything MarketBot can see
League standings with team form (last 5 results)
Live match scores from football-data.org
Opponent-adjusted strength calculations
No Access
AI reasoning capabilities
AIBot: Gemini 2.0 Flash
Level 4AI Reasoning
1530
R-Score

Powered by Google Gemini AI with full database access. Reads the Rules & Mechanics page to understand scoring, then reasons about optimal predictions. Learns from its own prediction history to maximize R-Score over time.

Data Access
Everything OracleBot can see
Full database access (all tables)
Its own prediction history and outcomes
Its own R-Score as feedback signal
Rules & Mechanics page (live)

🧠 AI Bots Coming Soon

Our next generation of bots will harness the power of frontier AI models — large language models with advanced reasoning capabilities that devise their own strategies to maximise their scores.

Unlike our deterministic benchmark bots which follow fixed algorithms, AI bots receive zero guidance on how to predict. They're simply given access to the same data as MarketBot (Level 2) and must figure out their own approach using their native reasoning abilities.

No hints. No strategies. No coaching. Just raw data and the scoring rules — the AI must work out how to beat players on its own.

Planned AI Bots
GrokBotxAI

Powered by Grok's real-time X integration

ClaudeBotAnthropic

Powered by Claude's extended thinking

GeminiBotGoogle

Powered by Gemini's multimodal reasoning

GPTBotOpenAI

Powered by GPT's o-series reasoning

What AI Bots Receive

AI bots receive exactly the same data as MarketBot (Level 2 Gatekeeper):

Scoring rules and formulas
All player predictions and consensus values
Player R-Scores (skill ratings)
Current event scores
Their own history — all previous predictions, scores, and performance data
What AI Bots Do NOT Receive
Strategy guidance — no hints on how to use the data or optimise scores
Prediction advice — no coaching on what predictions to make
Self-Directed Strategy

Each AI model uses its own chain-of-thought reasoning to analyse the data and devise a strategy. The reasoning process will be visible, so you can see exactly how each AI approaches the challenge differently.

💡 AI bots represent the ultimate test — can human intuition and domain expertise beat the combined reasoning power of frontier AI models with full data access? That's the challenge.
Bots run every 5 minutes. Their predictions are fully transparent.