---
title: "Crypto Hypothesis Tester — Automatically validate your market theories over time"
description: "You are my crypto hypothesis tester. I want to test this hypothesis:\n\"When ETH price rises more than 3% in 24h, NEAR price rises within the next 48 hours\"\n\nDo the following:\n\n1. Read memory at hypotheses/eth-near-correlation.md using memory_read.\nIf file does not exist, create it with:\n# Hypothesis: ETH pump → NEAR follows within 48h\n- Hypothesis: When ETH rises >3% in 24h, NEAR rises within 48h\n- Status: TESTING\n- Started: [today's date]\n- Observations: []\n- Confirmed: 0\n- Refuted: 0\n\n2. Fetch current prices from CoinGecko:\nhttps://api.coingecko.com/api/v3/simple/price?ids=ethereum,near&vs_currencies=usd&include_24hr_change=true\n\n3. Apply hypothesis logic:\n- If ETH 24h change > +3%: log as TRIGGER EVENT with today's date and ETH price\n- If a TRIGGER EVENT was logged 48h ago: check if NEAR price is now higher than at trigger time\n  - If YES: mark as CONFIRMED, increment confirmed counter\n  - If NO: mark as REFUTED, increment refuted counter\n\n4. Write updated log back to memory at hypotheses/eth-near-correlation.md\n\n5. Every Sunday send me a Telegram summary:\n\"🧪 Hypothesis Test Report\n\nTheory: ETH pump → NEAR follows within 48h\nStatus: TESTING (day [X] of 30)\n\nResults so far:\n✅ Confirmed: [X] times\n❌ Refuted: [X] times\n📊 Hit rate: [X]%\n\nLast trigger: [date] — ETH +[X]% → NEAR [outcome]\n\nVerdict: [PROMISING if >60% / INCONCLUSIVE if 40-60% / REJECTED if <40%]\"\n\nCreate a routine that runs every 6 hours to check prices and update the hypothesis log, and sends the weekly report every Sunday at 10:00 AM."
canonical: "https://hub.ironclaw.com/usecases/crypto-hypothesis-tester-automatically-validate-your-market-theories-over-time"
markdown: "https://hub.ironclaw.com/usecases/crypto-hypothesis-tester-automatically-validate-your-market-theories-over-time.md"
type: "use-case"
categories: ["Web 3 / Crypto","Research","Automation"]
author: "Evgeny"
---

# Crypto Hypothesis Tester — Automatically validate your market theories over time

## Example Prompt

> You are my crypto hypothesis tester. I want to test this hypothesis:
> "When ETH price rises more than 3% in 24h, NEAR price rises within the next 48 hours"
> 
> Do the following:
> 
> 1. Read memory at hypotheses/eth-near-correlation.md using memory_read.
> If file does not exist, create it with:
> # Hypothesis: ETH pump → NEAR follows within 48h
> - Hypothesis: When ETH rises >3% in 24h, NEAR rises within 48h
> - Status: TESTING
> - Started: [today's date]
> - Observations: []
> - Confirmed: 0
> - Refuted: 0
> 
> 2. Fetch current prices from CoinGecko:
> https://api.coingecko.com/api/v3/simple/price?ids=ethereum,near&vs_currencies=usd&include_24hr_change=true
> 
> 3. Apply hypothesis logic:
> - If ETH 24h change > +3%: log as TRIGGER EVENT with today's date and ETH price
> - If a TRIGGER EVENT was logged 48h ago: check if NEAR price is now higher than at trigger time
>   - If YES: mark as CONFIRMED, increment confirmed counter
>   - If NO: mark as REFUTED, increment refuted counter
> 
> 4. Write updated log back to memory at hypotheses/eth-near-correlation.md
> 
> 5. Every Sunday send me a Telegram summary:
> "🧪 Hypothesis Test Report
> 
> Theory: ETH pump → NEAR follows within 48h
> Status: TESTING (day [X] of 30)
> 
> Results so far:
> ✅ Confirmed: [X] times
> ❌ Refuted: [X] times
> 📊 Hit rate: [X]%
> 
> Last trigger: [date] — ETH +[X]% → NEAR [outcome]
> 
> Verdict: [PROMISING if >60% / INCONCLUSIVE if 40-60% / REJECTED if <40%]"
> 
> Create a routine that runs every 6 hours to check prices and update the hypothesis log, and sends the weekly report every Sunday at 10:00 AM.

## What the Agent Does

You define a market hypothesis once — for example "when ETH pumps 3%+, NEAR follows within 48 hours". The agent monitors prices every 6 hours, logs trigger events and their outcomes, and tracks confirmation rate over time.

Every Sunday it sends a Telegram report with how many times the hypothesis held vs failed, the current hit rate, and a verdict. After 30 days you have statistically grounded evidence whether your theory works or not — fully automated, no manual tracking needed.

## Skills and Tools

- http
- memory_read
- memory_write
- time
- message
- routine/cron

## Categories

- Web 3 / Crypto
- Research
- Automation

## Links

- [HTML page](https://hub.ironclaw.com/usecases/crypto-hypothesis-tester-automatically-validate-your-market-theories-over-time)