Fight Night

Helping MMA fans bet confidently using AI to determine the likelihood of the outcome; while formulating a more compelling approach to MMA rankings that disregard the hype and rewards objectively.

Comment

💸Inspiration

MMA has quickly become the fastest-growing sport in the world, yet the big war and one of the oldest organizations, the UFC, have been misusing the rankings and are unilaterally giving special consideration to more popular fighters; this becomes problematic when the company uses the list to determine fighter compensation and used by managers to secure individual sponsorships. This leads to fighter pay inequity, where results are a secondary consideration. Even top MMA commenters have spoken out about how the ranking seems to defy logic in the promotion's favor.

📊What it does

We trained a model using random foresting to return a likelihood of each fighter winning when the user enters two names. In addition, we can score each fighter's skills by applying an Elo rating system which is popular in zero sum games like chess. In Elo, each player starts at 1000 as they win or lose, gain or lose points, respectively, proportional to their opponent's skill. Post fight Elo for fighter A is equal to the inverse of one pulse ten to the power of the difference of current fighter B Elo and fighter A Elo divided by four hundred. Post fight Elo for fighter B is equal to the inverse of one pulse ten to the power of the difference of current fighter A Elo and fighter B Elo divided by four hundred. Over time this becomes an overall reflection of skill.

💻How we built it

We use two data sets containing fight statistics to predict the outcome. We used binary classifications. Build a graph using c born. Random foresting.

🧑‍🤝‍🧑Challenges we ran into

Deciding what kind of data to collect from UFC players and building an Elo prediction system was challenging. There was a wealth of information available, but determining which data was relevant and important for creating accurate predictions was difficult. Each player's physical condition and winning experiences in UFC are distinct and must be considered when building an Elo prediction system. These factors play a critical role in determining a player's performance and must be incorporated into the system to create accurate predictions. A comprehensive analysis of the unique characteristics of each player was essential to account for these differences and build a robust and reliable Elo prediction system.

✅Accomplishments that we're proud

Feb 11 UFC Saturday Night Card Vegas Odds:

  1. Volkanovski-Makhachev 34% +195 +130 -175
  2. Emmett-Rodriguez 46% +116 -135 +100
  3. Brown-Maddalena 33% +201 +150 -200
  4. Porter-Tafa 34% +194 +150 -200
  5. Menifield-Crute 22% +348 +330 -500
  6. Bukauskas-Pedro 33% +200 +165 -225
  7. Baghdasaryan-Culibao 63% -167 -150 +110
  8. Ross-Rodrigues 42% +136 +110 -150
  9. Prado-Mullarkey 35% +190 +165 -225
  10. Shainis-Jenkins 37% +168 +165 -225
  11. Reed-Lookboonmee 71% -247 -280 +200
  12. Bilder-Young 46% +119 +110 -150
  13. Brenner-Tukhugov 47% +112 +100 -135

Our Prediction:

  1. Makhachev
  2. Rodriguez
  3. Maddalena
  4. Tafa
  5. Crute
  6. Baghdasaryan
  7. Culibao
  8. Ross
  9. Prado
  10. Jenkins

Actual Result: 70%

  1. T
  2. T
  3. T
  4. T
  5. F
  6. T
  7. T
  8. F
  9. F
  10. T

🧠What we learned

We learned how to collect data through web scraping and data visualization and build an Elo prediction system using Python. The process involved using various tools and techniques to extract data from websites and visualize it meaningfully. We also gained an understanding of the Elo prediction system and its applications in sports betting and learned how to implement the system in Python. This involved developing algorithms and writing code to analyze the collected data and predict UFC wins based on the Elo system.

🥊What's next for Fight Night

We would like to extend the predictions to boxing, a notorious popular betting sport. To do this, we will need to gather data on boxers, analyze their performance, and incorporate this information into our prediction system. This will likely involve thoroughly analyzing boxers' different techniques and strategies and assessing their physical abilities and winning experiences. Additionally, we will need to consider the unique rules and scoring systems used in boxing and how these will impact the predictions. By extending our Elo prediction system to include boxing, we hope to provide a helpful tool for bettors and fans alike and increase the accuracy of our predictions in this popular sport.

Built With

Share this project:

Updates