A steam move<\/strong> is a sudden, sharp line movement at a professional bookmaker \u2014 usually Pinnacle \u2014 triggered by informed money. When Pinnacle shortens odds on an outcome, it means sharp bettors have hammered that side. Soft bookmakers (Bet365, DraftKings, FanDuel) take 30 seconds to several minutes to react. That lag window is where the edge lives.<\/p>\n Tools like OddsJam charge $199+\/month for their “Sharp Money” screen. Sports Insights keeps their detection logic proprietary. Both are black boxes \u2014 you pay, you trust, you hope the signals are real.<\/p>\n This tutorial builds an open-source steam move detector in Python<\/strong> that uses OddsPapi’s API to track Pinnacle line movements across 350+ bookmakers and alert you before soft books correct. Every line of code is tested against real data. You own the logic, you tune the parameters, and you can run it on the free tier.<\/p>\n Here’s a concrete example. Chelsea vs Manchester City, Full Time Result market:<\/p>\n At T = 30 seconds, Pinnacle dropped from 3.84 to 3.55 \u2014 a 0.29 move. Bet365 is still sitting at 3.60. That’s a steam move signal<\/strong>: the sharp book moved, the soft book hasn’t followed. You have a ~90-second window where Bet365’s price is stale relative to the true market price.<\/p>\n Two signals matter: (1) the magnitude<\/strong> of the Pinnacle move, and (2) the lag<\/strong> before soft books follow. Bigger moves with longer lags = stronger signals.<\/p>\n The key technical differentiator: OddsPapi returns a Only fixtures with This is the core data-capture function. For each fixture, it grabs the current price and<\/em> The This is the algorithm. Compare two consecutive snapshots for the sharp book. If Pinnacle’s price dropped by more than the threshold, check each soft book: if the soft book’s The Run it:<\/p>\n When the detector catches a steam move, the output looks like this:<\/p>\n Console output is fine for testing. For live monitoring, push signals to a Discord channel or Telegram chat.<\/p>\n To use it, create a Discord webhook in your channel settings and drop the URL into the polling loop:<\/p>\n For Telegram, swap in a call to the Bot API ( OddsPapi's free tier gives you 250 requests per month. Each call to To stretch the budget: reduce the fixture count, increase the poll interval, or poll only during kickoff windows (when lines move most). For all-day monitoring across multiple sports, upgrade to a paid plan.<\/p>\n Detecting steam moves is one thing. Proving they're profitable is another. OddsPapi's How Steam Moves Work (The 90-Second Explainer)<\/h2>\n
\n\n
\n \nTime<\/th>\n Event<\/th>\n Pinnacle (Chelsea Win)<\/th>\n Bet365 (Chelsea Win)<\/th>\n<\/tr>\n<\/thead>\n \n T = 0<\/td>\n Market open<\/td>\n 3.84<\/td>\n 3.60<\/td>\n<\/tr>\n \n T = 30s<\/td>\n Sharp money hits Pinnacle<\/td>\n 3.55<\/strong><\/td>\n 3.60 (stale)<\/td>\n<\/tr>\n \n T = 2min<\/td>\n Bet365 corrects<\/td>\n 3.55<\/td>\n 3.50<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n Old Way vs OddsPapi<\/h2>\n
\n\n
\n \nFeature<\/th>\n Scraping \/ Manual<\/th>\n OddsJam ($199+\/mo)<\/th>\n OddsPapi<\/th>\n<\/tr>\n<\/thead>\n \n Bookmakers<\/strong><\/td>\n 2-3 you scrape<\/td>\n ~100 (US-focused)<\/td>\n 350+<\/td>\n<\/tr>\n \n Pinnacle included<\/strong><\/td>\n No (blocked)<\/td>\n No (US books only)<\/td>\n Yes<\/td>\n<\/tr>\n \n Detection logic<\/strong><\/td>\n Your guess<\/td>\n Proprietary black box<\/td>\n Open source (this tutorial)<\/td>\n<\/tr>\n \n Alert delivery<\/strong><\/td>\n None<\/td>\n In-app only<\/td>\n You own it (Discord\/Telegram)<\/td>\n<\/tr>\n \n changedAt timestamp<\/strong><\/td>\n No<\/td>\n No<\/td>\n Every outcome, every book<\/td>\n<\/tr>\n \n Historical backtest<\/strong><\/td>\n No<\/td>\n Separate $79\/mo<\/td>\n Free tier included<\/td>\n<\/tr>\n \n Cost<\/strong><\/td>\n Your time + IP bans<\/td>\n $199-999\/mo<\/td>\n Free tier: 250 req\/mo<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n changedAt<\/code> timestamp on every outcome for every bookmaker. This tells you exactly when<\/em> a bookmaker last updated a price \u2014 without it, you’d need to poll continuously and diff prices yourself. With it, you can detect staleness in a single API call.<\/p>\nBuild the Steam Move Detector: Step by Step<\/h2>\n
Step 1: Setup and Configuration<\/h3>\n
import requests\nimport time\nfrom datetime import datetime, timedelta, timezone\n\nAPI_KEY = \"YOUR_API_KEY\"\nBASE_URL = \"https:\/\/api.oddspapi.io\/v4\"\n\n# --- Detection parameters ---\nSHARP_BOOK = \"pinnacle\"\nSOFT_BOOKS = [\"bet365\", \"1xbet\", \"draftkings\", \"fanduel\", \"betmgm\", \"betway\"]\nSPORT_ID = 10 # Soccer (change to 11 for basketball, 14 for NFL, etc.)\nMARKET_ID = 101 # Full Time Result (1X2)\nSTEAM_THRESHOLD = 0.05 # Minimum price drop to flag as steam\nPOLL_INTERVAL = 90 # Seconds between polling cycles\nOUTCOME_NAMES = {\"101\": \"Home\", \"102\": \"Draw\", \"103\": \"Away\"}\n\ndef api_get(endpoint, params=None):\n \"\"\"Authenticated API call with rate-limit safety.\"\"\"\n if params is None:\n params = {}\n params[\"apiKey\"] = API_KEY\n resp = requests.get(f\"{BASE_URL}\/{endpoint}\", params=params)\n resp.raise_for_status()\n time.sleep(0.2)\n return resp.json()<\/code><\/pre>\nSTEAM_THRESHOLD<\/code> is the minimum price drop at Pinnacle to count as a steam move. A drop from 3.84 to 3.55 = 0.29, well above the 0.05 default. Tune this based on your risk tolerance \u2014 tighter thresholds mean fewer but stronger signals.<\/p>\nStep 2: Fetch Upcoming Fixtures<\/h3>\n
def get_fixtures(sport_id, days_ahead=3):\n \"\"\"Fetch upcoming fixtures with odds available.\"\"\"\n today = datetime.now(timezone.utc).strftime(\"%Y-%m-%d\")\n to_date = (datetime.now(timezone.utc) + timedelta(days=days_ahead)).strftime(\"%Y-%m-%d\")\n\n fixtures = api_get(\"fixtures\", {\n \"sportId\": sport_id,\n \"from\": today,\n \"to\": to_date\n })\n\n with_odds = [f for f in fixtures if f.get(\"hasOdds\")]\n print(f\"Found {len(with_odds)} fixtures with odds (out of {len(fixtures)} total)\")\n return with_odds<\/code><\/pre>\nhasOdds: true<\/code> will return a bookmakerOdds<\/code> payload from the \/odds<\/code> endpoint. The \/fixtures<\/code> endpoint supports a max 10-day range, but 3 days is enough for most use cases.<\/p>\nStep 3: Snapshot Builder \u2014 Capture Current Odds State<\/h3>\n
changedAt<\/code> timestamp from Pinnacle plus your list of soft books.<\/p>\ndef snapshot_odds(fixture_id, bookmakers):\n \"\"\"\n Capture price + changedAt for every outcome across bookmakers.\n Returns: {slug: {outcome_id: {\"price\": float, \"changedAt\": datetime}}}\n \"\"\"\n snapshot = {}\n all_books = list(bookmakers)\n\n # Batch requests (API accepts comma-separated bookmaker slugs)\n data = api_get(\"odds\", {\n \"fixtureId\": fixture_id,\n \"bookmakers\": \",\".join(all_books)\n })\n\n for slug, book_data in data.get(\"bookmakerOdds\", {}).items():\n market = book_data.get(\"markets\", {}).get(str(MARKET_ID), {})\n snapshot[slug] = {}\n\n for oid, outcome in market.get(\"outcomes\", {}).items():\n player = outcome.get(\"players\", {}).get(\"0\", {})\n if not isinstance(player, dict) or \"price\" not in player:\n continue\n\n changed_raw = player.get(\"changedAt\", \"\")\n changed_dt = None\n if changed_raw:\n changed_dt = datetime.fromisoformat(changed_raw)\n\n snapshot[slug][oid] = {\n \"price\": player[\"price\"],\n \"changedAt\": changed_dt\n }\n\n return snapshot<\/code><\/pre>\nchangedAt<\/code> field is an ISO 8601 UTC timestamp on every outcome \u2014 it tells you the last time that specific bookmaker updated that specific price. This is what makes the entire detector work: you compare changedAt<\/code> across sharp and soft books to detect staleness without maintaining your own price history.<\/p>\nStep 4: Steam Detection Engine<\/h3>\n
changedAt<\/code> is older<\/em> than Pinnacle’s changedAt<\/code>, it hasn’t reacted yet. That’s the steam signal.<\/p>\ndef detect_steam(prev_snapshot, curr_snapshot, fixture_meta):\n \"\"\"\n Compare two snapshots. Detect where Pinnacle moved but soft books haven't.\n Returns list of signal dicts.\n \"\"\"\n signals = []\n sharp = SHARP_BOOK\n\n if sharp not in prev_snapshot or sharp not in curr_snapshot:\n return signals\n\n for oid in curr_snapshot.get(sharp, {}):\n prev_sharp = prev_snapshot[sharp].get(oid)\n curr_sharp = curr_snapshot[sharp].get(oid)\n if not prev_sharp or not curr_sharp:\n continue\n\n # Price drop = shortening odds = sharp money on this side\n price_move = prev_sharp[\"price\"] - curr_sharp[\"price\"]\n if price_move < STEAM_THRESHOLD:\n continue\n\n # Pinnacle moved. Check each soft book for staleness.\n for soft_slug in SOFT_BOOKS:\n if soft_slug not in curr_snapshot:\n continue\n soft_data = curr_snapshot[soft_slug].get(oid)\n if not soft_data:\n continue\n\n sharp_moved_at = curr_sharp[\"changedAt\"]\n soft_moved_at = soft_data[\"changedAt\"]\n\n if not sharp_moved_at or not soft_moved_at:\n continue\n\n # Soft book's last update is OLDER than Pinnacle's = stale\n if soft_moved_at < sharp_moved_at:\n lag_seconds = (sharp_moved_at - soft_moved_at).total_seconds()\n edge_pct = ((soft_data[\"price\"] \/ curr_sharp[\"price\"]) - 1) * 100\n\n signals.append({\n \"fixture\": fixture_meta,\n \"outcome_id\": oid,\n \"sharp_prev\": prev_sharp[\"price\"],\n \"sharp_now\": curr_sharp[\"price\"],\n \"sharp_move\": round(price_move, 3),\n \"soft_book\": soft_slug,\n \"soft_price\": soft_data[\"price\"],\n \"soft_stale_sec\": int(lag_seconds),\n \"edge_vs_sharp\": round(edge_pct, 2),\n \"detected_at\": datetime.now(timezone.utc).isoformat()\n })\n\n return signals<\/code><\/pre>\nedge_vs_sharp<\/code> field tells you how much value remains. If Pinnacle moved Chelsea Win from 3.84 to 3.55 and Bet365 is still at 3.60, the edge is (3.60 \/ 3.55) - 1 = +1.41%<\/code>. That's the EV of betting Bet365's stale price against Pinnacle's new \"true\" price.<\/p>\nStep 5: The Polling Loop<\/h3>\n
def run_detector(fixtures):\n \"\"\"Main loop: snapshot -> compare -> alert -> repeat.\"\"\"\n all_books = [SHARP_BOOK] + SOFT_BOOKS\n prev_snapshots = {}\n\n print(f\"Steam Detector running | {len(fixtures)} fixtures | Poll every {POLL_INTERVAL}s\")\n print(f\"Sharp: {SHARP_BOOK} | Soft: {', '.join(SOFT_BOOKS)}\")\n print(f\"Threshold: {STEAM_THRESHOLD} price drop\")\n print(\"=\" * 60)\n\n while True:\n for f in fixtures:\n fid = f[\"fixtureId\"]\n match_name = f\"{f['participant1Name']} vs {f['participant2Name']}\"\n meta = {\"fixtureId\": fid, \"match\": match_name, \"kickoff\": f[\"startTime\"]}\n\n curr = snapshot_odds(fid, all_books)\n\n if fid in prev_snapshots:\n signals = detect_steam(prev_snapshots[fid], curr, meta)\n for s in signals:\n label = OUTCOME_NAMES.get(s[\"outcome_id\"], s[\"outcome_id\"])\n print(f\"\\n{'=' * 60}\")\n print(f\" STEAM MOVE DETECTED\")\n print(f\" {s['fixture']['match']}\")\n print(f\" Outcome: {label}\")\n print(f\" Pinnacle: {s['sharp_prev']} -> {s['sharp_now']} (moved {s['sharp_move']})\")\n print(f\" {s['soft_book']}: still at {s['soft_price']} (stale {s['soft_stale_sec']}s)\")\n print(f\" Edge vs sharp: +{s['edge_vs_sharp']}%\")\n print(f\" Time: {s['detected_at'][:19]} UTC\")\n print(f\"{'=' * 60}\")\n\n prev_snapshots[fid] = curr\n time.sleep(1) # Rate limit between fixtures\n\n ts = datetime.now(timezone.utc).strftime(\"%H:%M:%S\")\n print(f\"[{ts}] Poll complete. Next in {POLL_INTERVAL}s...\")\n time.sleep(POLL_INTERVAL)<\/code><\/pre>\n# Fetch fixtures and start detecting\nfixtures = get_fixtures(SPORT_ID, days_ahead=1)\nrun_detector(fixtures[:5]) # Monitor 5 fixtures to stay within rate limits<\/code><\/pre>\n============================================================\n STEAM MOVE DETECTED\n Chelsea FC vs Manchester City\n Outcome: Home\n Pinnacle: 3.84 -> 3.55 (moved 0.29)\n bet365: still at 3.60 (stale 87s)\n Edge vs sharp: +1.41%\n Time: 2026-04-12T16:32:05 UTC\n============================================================<\/code><\/pre>\nStep 6: Add Discord or Telegram Alerts<\/h3>\n
def send_discord_alert(signal, webhook_url):\n \"\"\"Push a steam move signal to Discord via webhook.\"\"\"\n label = OUTCOME_NAMES.get(signal[\"outcome_id\"], signal[\"outcome_id\"])\n msg = {\n \"content\": (\n f\"**STEAM MOVE** | {signal['fixture']['match']}\\n\"\n f\"Outcome: {label}\\n\"\n f\"Pinnacle: {signal['sharp_prev']} -> {signal['sharp_now']} \"\n f\"(moved {signal['sharp_move']})\\n\"\n f\"{signal['soft_book']}: {signal['soft_price']} \"\n f\"(stale {signal['soft_stale_sec']}s)\\n\"\n f\"Edge: +{signal['edge_vs_sharp']}%\"\n )\n }\n requests.post(webhook_url, json=msg)<\/code><\/pre>\nDISCORD_WEBHOOK = \"https:\/\/discord.com\/api\/webhooks\/YOUR_WEBHOOK_URL\"\n\n# Inside the polling loop, after detecting signals:\nfor s in signals:\n send_discord_alert(s, DISCORD_WEBHOOK)<\/code><\/pre>\nsendMessage<\/code> to your chat ID). The detection logic stays the same \u2014 only the delivery changes.<\/p>\nHow to Tune the Detector<\/h2>\n
\n\n
\n \nParameter<\/th>\n Default<\/th>\n Tighter<\/th>\n Looser<\/th>\n<\/tr>\n<\/thead>\n \n STEAM_THRESHOLD<\/code><\/td>\n0.05<\/td>\n Fewer signals, higher confidence<\/td>\n More signals, more noise<\/td>\n<\/tr>\n \n POLL_INTERVAL<\/code><\/td>\n90s<\/td>\n Faster detection, more API calls<\/td>\n Fewer calls, may miss fast windows<\/td>\n<\/tr>\n \n SOFT_BOOKS<\/code><\/td>\n6 books<\/td>\n Focus on slowest movers<\/td>\n More edge opportunities<\/td>\n<\/tr>\n \n MARKET_ID<\/code><\/td>\n101 (1X2)<\/td>\n Single market focus<\/td>\n Scan multiple markets (loop IDs)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n Free Tier Budget Math<\/h3>\n
\/odds<\/code> is 1 request. If you monitor 5 fixtures every 90 seconds, that's ~5 requests per poll cycle. At 250 req\/month, you get roughly 50 poll cycles<\/strong> \u2014 enough to cover a Saturday afternoon of matches or a midweek Champions League slate.<\/p>\nBacktest It: Did the Steam Moves Actually Win?<\/h2>\n
\/historical-odds<\/code> endpoint \u2014 free on every tier \u2014 lets you pull the full price history for any past fixture and check whether Pinnacle's move predicted the result.<\/p>\ndef check_closing_line(fixture_id, bookmaker=\"pinnacle\"):\n \"\"\"Pull historical prices and compare opening vs closing line.\"\"\"\n data = api_get(\"historical-odds\", {\n \"fixtureId\": fixture_id,\n \"bookmakers\": bookmaker\n })\n\n book = data.get(\"bookmakers\", {}).get(bookmaker, {})\n market = book.get(\"markets\", {}).get(str(MARKET_ID), {})\n\n for oid, outcome in market.get(\"outcomes\", {}).items():\n snaps = outcome.get(\"players\", {}).get(\"0\", [])\n if not snaps:\n continue\n opening = snaps[0][\"price\"]\n closing = snaps[-1][\"price\"]\n move = opening - closing\n label = OUTCOME_NAMES.get(oid, oid)\n print(f\" {label}: opened {opening} -> closed {closing} (moved {move:+.3f})\")<\/code><\/pre>\n