complete_ufc_data.csv

Dataset Overview

File name: /data/complete_ufc_data.csv
Description: This dataset aggregates 30 years of UFC match history (since 1994), fighter statistics, and nine years of historical betting odds (since November 2014).

Column	Example	Description	Source
`event_date`	`2023-09-16`	UFC event date	Extracted from UFC match history
`event_name`	`UFC Fight Night: Grasso vs. Shevchenko 2`	UFC event name	Extracted from UFC match history
`weight_class`	`Womens Flyweight`	UFC weight class	Extracted from UFC match history
`fighter1`, `fighter2`	`Alexa Grasso`, `Valentina Shevchenko`	Fighter names	Extracted from UFC match history
`favourite`, `underdog`	`Valentina Shevchenko`, `Alexa Grasso`, `NaN`	Favourite and underdog fighters	Historical odds from betmma.tips
`favourite_odds`, `underdog_odds`	`1.67`, `2.88`, `NaN`	Betting odds (decimal)	Historical odds from betmma.tips
`betting_outcome`	`favourite`, `underdog`, `NaN`	Betting outcome	Historical odds from betmma.tips
`outcome`	`fighter1`, `fighter2`, `Draw`	Match result	Extracted from UFC match history
`method`	`S-DEC`, `U-DEC`, `KO/TKO Punches`	Victory method	Extracted from UFC match history
`round`	`5`	Winning round	Extracted from UFC match history
`fighter1_`, `fighter2_`		Fighter attributes	Extracted from UFC fighter statistics
`events_extract_ts`, `odds_extract_ts`, `fighter_extract_ts`	`2023-09-21 02:02:55.178363`	Data extraction timestamp

Code: Python scripts were used for web scraping and data preprocessing.
Functionality: Completed UFC data scraping (fighter stats and match results), historical betting odds scraping, and data cleaning.

Insight: Historical win probability shows a strong correlation between age and average strikes per minute with match success. Younger fighters or those with higher strike output have a statistical advantage, winning about 60% of matches.
Insight: The historical probability that the favourite wins rises from slightly above 50% to over 75% when the decimal odds difference exceeds 2.0. Moreover, as the odds gap widens, this likelihood increases, reaching about 90% when the odds difference exceeds 4.5.

Development status: Ongoing; machine‑learning models are being tested for predicting match outcomes based on fighter statistics.
Preliminary test: Initial models (GBM, logistic regression) achieve roughly 65% accuracy without betting odds features.
Future iterations: Planned testing of additional features such as win streaks, finish rates, derived attributes (endurance, wrestler/striker/slugger tags) and whether a fighter is a betting favourite.