Course for beginners

Lesson 1. Why this project exists

The project answers one question: can a simple rule earn more than buying a coin and holding it? An example rule: "hold SOL only while its price is above its 50-day average".

The internet is full of rules like this, all with beautiful charts. Usually it is self-deception: the rule was fitted to the past, fees were forgotten, or it was tested on one lucky period. This project protects itself in three ways:

1. Write it down first, test second. The rule and the bar for success are written down and sealed with a fingerprint (a hash). The test cannot be adjusted after seeing the result.
2. Test on years the rule has not seen. A rule is chosen on 2018–2021 and judged from 2022 onwards.
3. Compare with luck. A statistical test answers whether such a result could appear by chance.

Today's result: 2 of 33 hypotheses pass their pre-registered test. That is a result too: it saves the money that would go into trading "beautiful" rules.

Lesson 2. How it works

Everything runs by itself on GitHub's servers. Nobody needs to keep a computer switched on.

• Every day at 03:00–07:00 UTC fresh prices are downloaded and checked, market conditions are recalculated and this website is rebuilt.
• Every Monday old hypotheses are re-tested on the new data. If a verdict changes, an alert is sent.
• Once a month an AI model may suggest new ideas. A person checks and locks each one before it is tested.

Lesson 3. The site's pages

Open the home page: it answers two questions — what has been found and what the market looks like today. The menu at the top leads to five pages; the language list is on the right.

How to read a market card. The big number is today's value. The framed word under it is the zone (for example "greed" or "above"). The line at the bottom shows the last 12 months. "What is this?" opens a short explanation. What matters is not the number itself but a change of zone: that is when an alert is sent (lesson 7).

Lesson 4. How to read a coin page — SOL as an example

Every coin has its own page. Market → "Coins" → click SOL Solana, or open Solana directly. For other coins only the ticker in the address changes: ETHUSDT, BTCUSDT, EVERUSDT…

A coin page is read in four steps:

1. 12-month chart. Blue line = price, yellow dashes = 50-day average, white dots = 200-day average. Price above both averages = uptrend, below both = downtrend. The scale is logarithmic: a rise from $10 to $20 looks the same as from $100 to $200.
2. Numbers. Change over 7 days, 30 days and a year; distance from the averages; volatility. Volatility of 70% means a typical yearly move of about 70% up or down. Today: BTC 25%, SOL 35%, EVER 262%. Three more lines put the price in context: how far it is below its highest close (SOL: -61%), what $100 invested a year ago is worth now (SOL: $53), and how closely the coin moves with Bitcoin, from −1 to 1 (SOL: 0.83).
3. What the tested rules say today. Four rules from the hypotheses and what each would do today: "in the market", "out of the market" or a share of a position. Next to each is the verdict of its test. None has passed, so this is an observation, not a signal.
4. Studies that include this coin. For SOL the most interesting is H-014 (1-week momentum): on SOL the rule reached a Sharpe of 0.79 against 0.36 for holding, but it failed on BTC, so the verdict is "weak".

Three questions to ask about any coin: is the price above or below its 200-day average? Is volatility higher or lower than usual? Is Bitcoin rising or falling (the last row of the rules table)?

Lesson 5. How to read a hypothesis

Open any hypothesis from the table on the Research page, for example H-003. At the top is a coloured verdict and a sentence "in plain words".

Then a results table, one row per coin and period. The columns are always compared in pairs, rule against "hold":

• Sharpe: return per unit of risk. Higher is better. Holding Bitcoin gives about 0.5.
• CAGR: average yearly growth.
• Max drawdown: the worst fall from a peak. −50% means the money halved at some point.
• p-value: the chance the result is luck. The smaller the better; it must be below the bar written in the hypothesis.

Under the table is a chart of how 1 dollar would have grown with the rule (solid line) and with holding (dashed). The list below lets you switch coin and period.

Lesson 6. Watching Everscale (EVER)

EVER is not on Binance, so its prices come from Gate.io. You can follow it in three places:

1. The EVER card on the Market page ("Coins" section): how far the price is from its 50-day average. Plus means the H-018 rule is "in the market", minus means "out".
2. The EVER page: Everscale, with a chart, numbers and the state of each rule.
3. Hypotheses H-018 and H-019, tested on EVER. Since mid-2022 the SMA50 rule returned -9% a year against -69% for holding, with a worst drawdown of -84% against -100%.

Important. The rule "beats" holding mainly because holding lost almost everything. EVER's daily trading volume is small, so the price can jump even on small trades. This is information to watch, not a recommendation.

Lesson 7. Alerts in Telegram

On this website nothing has to be configured: just open the page once a day. Alerts are for people who run their own copy of the project (it is free):

1. Create a bot with @BotFather and copy its token.
2. Send your bot any message, then find your chat id (for example with @userinfobot).
3. Fork the repository, then in the fork: Settings → Secrets and variables → Actions → add TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID.
4. Settings → Pages → Source: GitHub Actions, then Actions → research-agents → Run workflow.

An alert arrives only when something changes: a market card moves to another zone, or a hypothesis gets a new verdict. On quiet days there are no messages.

Lesson 8. How to test your own idea

The easiest way does not need any code:

1. Open a new issue "Hypothesis idea".
2. Describe the rule in plain words: when you are in the market, when you are out, which coins.
3. Explain why it should work, and what counts as success. Write this before looking at any charts.

The idea is turned into a hypothesis file, locked and tested. The result appears on this website in all languages, with a page of its own.

If you know a little Python, read CONTRIBUTING.md: a hypothesis is a short YAML file.

Lesson 9. What the project does not do

• It does not give advice. No verdict here tells you to buy or sell. Even a "pass" would only be a reason for further checks.
• It does not trade and does not manage anyone's money.
• It does not promise the future. All results describe the past; markets change.
• Not everything is included. Taxes, slippage on large orders and the risk of an exchange failing are not in the numbers. Fees are.
• Small coins are risky in a special way. With little trading volume, the price you see may not be the price you get.

If you want to act on something here, first ask: what happens to me if the rule stops working tomorrow?

Lesson 10. Practice: tasks for one week

• Open the Market page and find today's Fear & Greed zone.
• Find the coin with the highest volatility in the "Coins" table.
• Open the SOL page: is the price above or below its 200-day average?
• Open H-005 and explain in one sentence why it failed.
• Open the EVER page and compare its volatility with Bitcoin's.
• Come back in a week: did any market card change zone?
• Write down one idea of your own: the rule, why it should work, what counts as success.

Glossary for beginners