Answer · Bots · Architecture

Sports betting bot architecture on Pinnacle odds

Updated 9 November 2026 · PinnFeed · 2 min read

Answer: most betting or trading bots have five parts: (1) a data feed, (2) a local state of events and markets, (3) signal rules, (4) risk checks, and (5) execution on the account you bet with. PinnFeed fills part 1 and helps with 2–3: a WebSocket (Builder) that sends a full snapshot and then delta frames with every Pinnacle price change, including fair, margin and maxStake. It does not place bets; execution is yours, on a book that allows it, where betting is legal for you.

The pipeline#

Part	What it does	With PinnFeed
Ingest	Keep one connection open, answer `ping` with `pong`, reconnect with backoff	`wss://pinnfeed.io/v1/stream` (Python, Node)
State	Map of `market.id` → latest market; drop on `op: "remove"`	snapshot/delta
Signals	Price drop, offer above `fair`, line change	steam, EV screen
Risk	Stake caps, `maxStake >= 100`, max exposure per event, kill switch	Your code
Execution	Place the bet, log price and result	Your bookmaker or exchange

Minimal signal loop (Python)#

book = {}                                   # market id -> market

def on_frame(msg, alert):
    if msg["type"] == "snapshot":
        for m in msg["markets"]:
            book[m["id"]] = m
    elif msg["type"] == "delta":
        for ch in msg["changes"]:
            if ch["kind"] != "market":
                continue
            if ch["op"] == "remove":
                book.pop(ch["id"], None)
                continue
            new, old = ch["market"], book.get(ch["market"]["id"])
            book[new["id"]] = new
            if not old or (new["maxStake"] or 0) < 100:
                continue
            for o_new, o_old in zip(new["outcomes"], old["outcomes"]):
                if o_old["decimal"] and o_new["decimal"] / o_old["decimal"] - 1 <= -0.04:
                    alert(new["eventId"], o_new["name"], o_old["decimal"], o_new["decimal"])

Plug on_frame into any client from the language pages (Go, Rust, Kotlin). Log every signal and decision so you can measure it later against the close: bet tracker with CLV.

Design tips#

Run ingest on an always-on host, not a serverless function.
Keep a paper mode that logs what the bot would do before it bets anything.
Use one WebSocket per process and fan out internally; Builder allows 3 connections, Pro 10.
Handle 4008 slow_consumer by moving work off the receive loop.

Plans#

Builder (€99/mo): WebSocket, all 13 sports, all markets. Pro (€199/mo): 10 connections and raw frames. 3-day trial; a card is required and it is charged after day 3 unless you cancel. Prototype on Free REST first: signup.

Educational content for adults (18+). Automated betting can lose money quickly; no tips, no guaranteed profit.

Questions people ask#

Can PinnFeed place bets for my bot?#

No. It is read-only data. Execution happens on your own bookmaker or exchange account.

Is a bot allowed by bookmakers?#

Many books restrict automated betting in their terms. Check before you automate execution.

Can I build it with an AI coding tool?#

Yes: build a betting bot with Cursor and vibe coding.