\r \r

Instructions for Greyhound Predictor Skill\r

\r When activated (e.g., user says "predict Monmore R5 greyhounds" or "upcoming greyhound predictions"), follow these steps:\r \r

1. Parse user input: Extract race details like track (e.g., Monmore, Towcester), race number/date, or "upcoming" for today's races.\r \r
2. Fetch data: \r
   • Use web_search or browse_page to get upcoming race cards from sites like gbgb.org.uk, sportinglife.com/greyhounds, or timeform.com/greyhounds.\r
   • Example: Browse "https://www.gbgb.org.uk/racing/todays-trials-meetings/" for today's races, or API like "https://api.gbgb.org.uk/api/results?page=1&date={today}&track={track}" for form.\r
   • Gather: Dog names, traps, recent form (e.g., 12131 = positions in last 5 races), best times, odds, trainer form.\r \r
3. Analyze data: \r
   • Calculate basic metrics: Win rate (wins/races), average position, recent speed (time/distance), trap bias (e.g., inside traps win more in sprints).\r
   • Use rules: Favor dogs with form like 111 (recent wins), low traps in short races (270-480m), wide traps in stayers (650m+).\r
   • If code_execution available, run a simple Python model (see script below) on fetched data to score probabilities.\r \r
4. Predict:\r
   • Winner: Dog with highest score (e.g., best form + trap advantage).\r
   • Second: Strong chaser (good recent places, stalking trap).\r
   • Output: "Winner: [Dog Name] (Trap X) - Reasons: Recent wins, trap bias. Second: [Dog Name] (Trap Y) - Consistent placer."\r \r
5. Handle errors: If no data, say "Couldn't fetch race info—try specifying track/date."\r \r

Sample Python Code for Prediction (if code_execution tool enabled)\r

If your OpenClaw supports code_execution, include this in instructions to run a basic model. Paste data into a DataFrame.\r \r import pandas as pd\r from sklearn.linear_model import LogisticRegression\r from sklearn.model_selection import train_test_split\r from sklearn.preprocessing import StandardScaler\r \r

Sample data (replace with fetched race data)\r

data = pd.DataFrame({\r 'trap': [1, 2, 3, 4, 5, 6],\r 'win_rate': [0.4, 0.3, 0.35, 0.2, 0.45, 0.25], # Wins/races\r 'avg_position': [2.1, 3.0, 2.5, 4.0, 1.8, 3.5],\r 'recent_form_score': [0.8, 0.6, 0.7, 0.4, 0.9, 0.5] # Custom score from form\r })\r data['winner'] = [1, 0, 0, 0, 0, 0] # Dummy target for training (use historic data)\r \r

Train simple model\r

X = data[['trap', 'win_rate', 'avg_position', 'recent_form_score']]\r y = data['winner']\r X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)\r scaler = StandardScaler().fit(X_train)\r X_train = scaler.transform(X_train)\r model = LogisticRegression().fit(X_train, y_train)\r \r

Predict for new race\r

new_data = pd.DataFrame(...) # Fill with fetched data\r preds = model.predict_proba(scaler.transform(new_data))[:, 1]\r top_dog = new_data.iloc[preds.argmax()]['dog_name']\r print(f"Predicted winner: {top_dog}")