Sigrex/telegram-signal-extract

Telegram Trading Signals

Telegram trading-channel messages paired with a machine-readable trade signal, or null when the message is not a signal.

The task this dataset encodes: given a raw channel message, decide whether it is an actionable open/close signal, and if so emit a JSON payload describing the trade.

{"symbol": "BTCUSDT", "side": "BUY"}

The payload follows the Sigrex signal payload format. Only symbol and side are emitted. size is deliberately omitted: no message in this corpus states a concrete trade size (leverage, margin and "% of deposit" are present in many messages, but those are not the same thing as size), so emitting one would be invention.

Files

file	rows	columns
data/signals.csv	35,145	telegram_signal_message, json_signal
data/signals-metadata.csv	35,145	+ channel, message_id, date, symbol, side
data/full.csv	195,847	telegram_signal_message, json_signal

• signals — only messages classified as real signals. Every json_signal is a JSON string.
• full — every extracted message. json_signal is a JSON string for the 35,145 signals and null for the other 160,702 rows. This is the detection/extraction training set.
• -metadata — the signal set with provenance columns, for analysis rather than training.

Columns

column	type	notes
telegram_signal_message	string	raw message text, verbatim (emoji and newlines preserved)
json_signal	string | null	{"symbol": ..., "side": "BUY"|"SELL"}, or null
channel	string	only in -metadata
message_id	int	only in -metadata
date	string	only in -metadata
symbol	string	normalized pair, e.g. BTCUSDT, XAUUSD (only in -metadata)
side	string	BUY or SELL (only in -metadata)

Composition

• 241,350 raw messages from 18 Telegram channel exports, deduplicated to 195,847 rows.
• 35,145 signals (~18% positive) / 160,702 non-signals.
• 986 distinct symbols. Side split: 23,428 BUY / 11,717 SELL.
• Positives are signals with an explicit side, a symbol and a concrete entry price; reports ("TP1 hit ✅"), profit recaps, position updates, promos, news and commentary are negatives.