Real Data, Scientific Models, Real Results

Real Data, Scientific Models, Real Results

The website has been operational for a month, and Monkey wants to share with everyone the methods and foundations of its operation.

Real Data:
Based on over 15,000 records of Hong Kong horse racing over the past two years, continuously accumulating new data, which forms the basis for various prediction models. This approach aims to completely avoid the influence of human factors.

Scientific Models:
For predictions, Monkey has utilized Machine Learning models like Naive Bayes, Logistic Regression, Random Forest, Neural Network, etc., exploring various prediction angles and continuously fine-tuning the models.

Clear Prediction Scope:
To avoid misleading complexity, Monkey deliberately focuses on the most basic element of horse betting, which is the placing (Place) of horses. This makes it easy to see whether the prediction is correct or not. Other types of bets (Q or T) are essentially combinations of these three horses, allowing users to freely interpret. Monkey also intentionally avoids using the traditional tip of four horses, as only the first three horses contribute to payouts. The return rate is also an important factor in calculating the success of horse betting.

Real Predictions:
Prediction results are usually updated one day before the race day, and members are allowed to submit their tips. Input or changes are automatically closed ten minutes before the start of the first race. Therefore, every tip on the site, whether from the computer model or human choice, is 100% made before the event and can be trusted.

Real Results:
Results and odds for each race are updated, and the ranking of tip providers is calculated in real-time. Users can see who the star performer of the day is and consider following their tips. For example, in every ten-race day, the model selects thirty horses, and hitting fourteen horses means a hit rate of 47%. It's a simple and straightforward calculation. Whether the tips are accurate or not, everyone uses the same standard to judge, making the results clear.

Regardless of good or bad results, they are recorded, planning the cumulative performance of each predictor. Because a single good or bad performance is not reliable, the aim is to find predictors or models that can consistently perform well in the long term.

In summary, while the website is constantly improving in hit rate, its integrity regarding data has always remained unchanged. Please continue to support us.