Strategy Lab · paper money, live prices · as of Nov 21, 2026
HourBasedStrategy
HourBasedStrategy is an open-source Freqtrade strategy from the freqtrade/freqtrade-strategies collection. We run its code unchanged with $1,000 of paper money on live Binance spot prices and publish what it is worth every day. Our rough label for it is “Indicator mix”, from its name; the label is ours, not the author's. It is one of 2 HourBasedStrategy versions in the lab, the same bot with different settings: see how the family is doing.
Share this record · Is this yours? Claim the page
What its code does. Reads 1-hour candles with its own maths, no standard indicator. Sells on its own sell signal and a profit target of 53% that falls to 0% the longer it holds. Stop-loss at a 10% loss. Read from the frozen file by a script, not by running it. A value the code works out while it runs is left out. How we know
Number 128 of 133 on today's board. The board ranks every bot that has traded by what its $1,000 has become since it joined.
Live record and backtest
| Live recordsince Nov 7 | Backtest45 days, not a record | |
|---|---|---|
| $1,000 became | $713.04 | $547.73 |
| Return | −28.7% | −45.2% |
| Holding the 20 coins | −28.0% | −42.9% |
| Trades closed | 31 | 67 |
| Won / lost | 10 / 21 | 22 / 45 |
| Worst drop, hourly | −29.1% | −45.5% |
| Worst drop, Freqtrade | −28.8% | −45.2% |
| Candles | 1h | 1h |
| Period | Nov 7 to Nov 21, 2026 | Oct 7 to Nov 21, 2026 |
Day by day
| Day | $1,000 was | Change on the day | Trades so far | Worst drop so far |
|---|---|---|---|---|
| Nov 21, 2026 | $713.04 | −9.5% | 31 | −29.1% |
| Nov 20, 2026 | $788.25 | −1.9% | 27 | −23.5% |
| Nov 19, 2026 | $803.40 | −1.1% | 25 | −20.1% |
| Nov 18, 2026 | $811.97 | −0.3% | 25 | −19.3% |
| Nov 17, 2026 | $814.53 | +1.0% | 22 | −19.0% |
| Nov 16, 2026 | $806.14 | +0.5% | 19 | −19.9% |
| Nov 15, 2026 | $802.26 | −6.5% | 19 | −20.2% |
| Nov 14, 2026 | $857.66 | −3.2% | 13 | −14.7% |
| Nov 13, 2026 | $885.67 | −5.6% | 11 | −12.0% |
| Nov 12, 2026 | $938.30 | −1.6% | 10 | −7.5% |
| Nov 11, 2026 | $953.97 | −1.0% | 9 | −7.5% |
| Nov 10, 2026 | $964.02 | −1.0% | 8 | −5.2% |
| Nov 9, 2026 | $973.62 | – | 8 | −2.8% |
How we know
- Source code: HourBasedStrategy.py in freqtrade/freqtrade-strategies, on its main branch. That file can change after we froze our copy; the hash below is the version we run.
- Code is frozen: fixed on Nov 7, 2026, the day it joined. SHA-256 of the file we run:
61755d49c46b3037c102f60de30841910472b6da06e9b578882b3dfc125b9ccc - Lookahead check: Passed. Freqtrade's lookahead analysis found no use of future prices.
- Same rules as every bot: $1,000 paper wallet, the same 20 coins against USDT on Binance spot, five open trades at most, the exchange's fee on every fill, no leverage and no shorting. Method and limits.
- Paper money: no real order is placed. Real fills can be worse than the candle's price in fast markets.
Run it
Run it for me (beta, not live yet)
We're building a runner that trades HourBasedStrategy on your own exchange account: trade-only access that can't withdraw, a budget and a loss limit you set, a kill switch, and a message for every trade. Join the list and we'll email you when the beta opens.
Run it on your own server, about 10 minutes
Free and open source: Freqtrade, the lab's 20 coins and rules, and this bot's frozen file. It starts on paper money, so it places no real orders until you say so. Exchanges: Bybit, OKX, Binance (Bybit: not for US residents; OKX: US residents use OKX US, which has its own rules; Binance: US residents use Binance.US, a separate exchange).
# 1. On a Linux server with Docker (or your own computer), in a new folder: mkdir -p tas-bot/user_data/strategies tas-bot/user_data/logs && cd tas-bot EXCHANGE=bybit # or okx, or binance curl -fsSL -o docker-compose.yml https://tradingagentstack.com/run/kit/docker-compose.yml curl -fsSL -o .env https://tradingagentstack.com/run/kit/env curl -fsSL -o user_data/config.json "https://tradingagentstack.com/run/kit/spot-$EXCHANGE.json?s=HourBasedStrategy" curl -fsSL -o 'user_data/strategies/HourBasedStrategy.py' 'https://raw.githubusercontent.com/freqtrade/freqtrade-strategies/main/user_data/strategies/HourBasedStrategy.py' # 2. Check it's the exact file the lab tests. The two hashes must match: sha256sum 'user_data/strategies/HourBasedStrategy.py' # on a Mac: shasum -a 256 # 61755d49c46b3037c102f60de30841910472b6da06e9b578882b3dfc125b9ccc # 3. Choose the bot and start it on paper money: echo 'STRATEGY=HourBasedStrategy' >> .env docker compose up -d docker compose logs -f # Ctrl+C stops watching, not the bot
Going live: the checklist
- Let it run on paper for at least a week and compare it with this page.
- Make an API key that can trade but can't withdraw, locked to your server's IP address (how).
- Keep only the money you can lose on that account, or use a sub-account, and set
FREQTRADE__AVAILABLE_CAPITALin.envto it. - In
.env: add the key and secret, setFREQTRADE__DRY_RUN=falseand pointDB_URLat a new file such aslive.sqlite. Then rundocker compose up -dagain. - Stop it any time with
docker compose stop. Stopping leaves open trades as they are; close them on the exchange if you want out completely.
Need an account or a server? Exchanges: Binance · Bybit · OKX · Bitget. A small Linux server: Vultr, or compare VPS hosts for trading bots.
Rather not run anything? These platforms run their own bots, no code: WunderTrading · Bitsgap · Cryptohopper · 3Commas. They are not this bot, and the lab hasn't tested them.
Check our number yourself: the commands the lab runs
These are the commands the lab runs, with the lab's own config. Freqtrade is free, and they need no exchange account and no API key.
# 1. Install Freqtrade (freqtrade.io), then in an empty folder: freqtrade create-userdir --userdir user_data curl -L -o config.json 'https://raw.githubusercontent.com/intoApex/tas-workers/main/strategy-lab/config.json' curl -L -o 'user_data/strategies/HourBasedStrategy.py' 'https://raw.githubusercontent.com/freqtrade/freqtrade-strategies/main/user_data/strategies/HourBasedStrategy.py' # 2. Get the candles: Binance spot, public data, no exchange keys freqtrade download-data -c config.json --userdir user_data --timeframes 5m 15m 1h 4h 1d --days 60 # 3. Repeat our live record: every candle since the day it joined the lab freqtrade backtesting -c config.json --userdir user_data --strategy HourBasedStrategy --timerange 20261107- --cache none # 4. Paper trade it yourself on live prices (dry run, no real orders) freqtrade trade -c config.json --userdir user_data --strategy HourBasedStrategy
Step 3 is the test the lab runs every day. It counts candles up to the moment you run it, so your number moves a little with the hour. If it is far from ours, tell us.