Predicting UFC Outcomes

UPDATE (March 25, 2017)

The project outline below was origially drafted in September, 2016. Much has changed since then. Mabye I’ll get around to updating the description at some point. Here are a few of the changes:
- the model
- new features
- feature reduction of highly correlated features
- stepwise feature selection according to AIC
- set of scripts to streamline model testing and evaluation
- monte carlo simulations to explore most profitable betting strategies (developed by a friend and my brother)

Real World Results

Below are real world predictions made prior to each event. You’ll notice the performance between the model and Vegas is pretty similar below. A few things to remember about the model and the results listed below:
- the model doesn’t make prediction where one of the fighters is making a UFC debut
- these are all model predictions at the minimum confidence threshold of 50%
- in a betting situation, these predictions are filtered to a higher confidence threshold, as well as by other means.

Date	Bets	Moneyline	Probability	Model.Win	Vegas.Win	Thresh
2017-03-18	makwan amirkhani	-120	0.6285	0	0	60
2017-03-18	brad pickett	-138	0.5749	0	0	55
2017-03-18	scott askham	-142	0.5857	0	0	55
2017-03-18	daniel omielanczuk	153	0.6032	0	1	60
2017-03-18	darren stewart	-176	0.8005	0	0	75
2017-03-18	gunnar nelson	-334	0.6366	1	1	60
2017-03-18	corey anderson	125	0.5971	0	1	55
2017-03-18	joe duffy	-591	0.7547	1	1	75
2017-03-18	vicente luque	-128	0.5163	0	0	50
2017-03-18	marc diakiese	-202	0.6297	1	1	60
2017-03-11	beneil dariush	135	0.5341	0	1	50
2017-03-11	francisco trinaldo	160	0.5582	0	1	55
2017-03-11	ray borg	-124	0.5128	1	1	50
2017-03-11	mauricio rua	-150	0.7802	1	1	75
2017-03-11	michel prazeres	-241	0.7483	1	1	70
2017-03-11	rani yahya	-211	0.7123	0	0	70
2017-03-11	rony jason	105	0.5065	0	1	50
2017-03-11	tim means	-181	0.6000	0	0	60
2017-03-11	kelvin gastelum	-340	0.7107	1	1	70
2017-03-04	alistair overeem	-131	0.7104	1	1	70
2017-03-04	mirsad bektic	-675	0.6382	0	0	60
2017-03-04	luke sanders	-135	0.6983	0	0	65
2017-03-04	david teymur	305	0.5320	1	0	50
2017-03-04	marcin tybura	-156	0.5285	1	1	50
2017-03-04	paul craig	115	0.5811	0	1	55
2017-03-04	rashad evans	-220	0.5955	0	0	55
2017-03-04	stephen thompson	-134	0.6237	0	0	60
2017-02-19	elias theodorou	106	0.5708	1	0	55
2017-02-19	gerald meerschaert	-279	0.6066	1	1	60
2017-02-19	johny hendricks	120	0.5623	1	0	55
2017-02-19	paul felder	-346	0.7766	1	1	75
2017-02-19	santiago ponzinibbio	-330	0.6707	1	1	65
2017-02-19	thiago santos	-182	0.5703	1	1	55
2017-02-19	travis browne	101	0.5026	0	1	50
2017-02-11	randy brown	-143	0.7754	0	0	75
2017-02-11	derek brunson	-120	0.6034	0	0	60
2017-02-11	glover teixeira	-199	0.5439	1	1	50
2017-02-11	jacare souza	-555	0.8839	1	1	75
2017-02-11	dustin poirier	-431	0.6404	1	1	60
2017-02-11	islam makhachev	-327	0.5128	1	1	50
2017-02-11	rick glenn	-176	0.5366	1	1	50
2017-02-11	ryan laflare	-296	0.6748	1	1	65
2017-02-11	wilson reis	-591	0.8891	1	1	75
2017-02-04	abel trujillo	-110	0.5531	0	0	55
2017-02-04	curtis blaydes	-300	0.5549	1	1	55
2017-02-04	alex morono	-105	0.5609	0	0	55
2017-02-04	chris gruetzemacher	230	0.5427	0	1	50
2017-02-04	khalil rountree jr	-165	0.6463	1	1	60
2017-02-04	dennis bermudez	-200	0.7212	0	0	70
2017-01-28	raphael assuncao	-155	0.5353	1	1	50
2017-01-28	francis ngannou	-375	0.7663	1	1	75
2017-01-28	donald cerrone	-155	0.7404	0	0	70
2017-01-28	alessio di chirico	-110	0.7764	0	0	75
2017-01-28	jc cottrell	-145	0.5350	0	0	50
2017-01-28	jason knight	-150	0.6837	1	1	65
2017-01-28	sam alvey	-155	0.5182	1	1	50
2017-01-15	aleksei oleinik	-145	0.5814	1	1	55
2017-01-15	court mcgee	-140	0.5932	0	0	55
2017-01-15	joachim christensen	-225	0.6326	1	1	60
2017-01-15	dmitrii smoliakov	-110	0.6602	0	0	65
2017-01-15	frankie saenz	-165	0.5557	0	0	55
2017-01-15	marcin held	100	0.5154	0	1	50
2017-01-15	sergio pettis	-150	0.7080	1	1	70
2017-01-15	tony martin	-175	0.5132	1	1	50
2017-01-15	walt harris	-145	0.6143	1	1	60
2017-01-15	yair rodriguez	-445	0.7820	1	1	75
2016-12-30	antonio carlos junior	-140	0.5263	1	1	50
2016-12-30	cody garbrandt	160	0.5184	1	0	50
2016-12-30	tarec saffiedine	120	0.5112	0	1	50
2016-12-30	neil magny	-165	0.5515	1	1	55
2016-12-30	louis smolka	115	0.6123	0	1	60
2016-12-30	alex garcia	-190	0.5155	1	1	50
2016-12-30	tj dillashaw	-205	0.5100	1	1	50
2016-12-17	alex morono	-115	0.6271	1	1	60
2016-12-17	bojan velickovic	-155	0.7095	0	0	70
2016-12-17	colby covington	-360	0.5998	1	1	55
2016-12-17	cole miller	-110	0.5299	0	0	50
2016-12-17	eddie wineland	-225	0.5080	1	1	50
2016-12-17	hector sandoval	-130	0.6126	1	1	60
2016-12-17	scott holtzman	150	0.5767	0	1	55
2016-12-17	mike perry	-135	0.5634	0	0	55
2016-12-17	mickey gall	-120	0.5004	1	1	50
2016-12-17	urijah faber	-450	0.6846	1	1	65
2016-12-10	max holloway	-195	0.5800	1	1	55
2016-12-10	cub swanson	188	0.5041	1	0	50
2016-12-10	donald cerrone	-275	0.7071	1	1	70
2016-12-10	lando vannata	-170	0.5478	1	1	50
2016-12-10	kelvin gastelum	110	0.7367	1	0	70
2016-12-10	mitch gagnon	-156	0.5630	0	0	55
2016-12-10	nikita krylov	100	0.6733	0	1	65
2016-12-10	olivier aubin mercier	-165	0.5002	1	1	50
2016-12-10	rustam khabilov	-210	0.7076	1	1	70
2016-12-10	zach makovsky	-160	0.5500	0	0	55
2016-12-09	andrew sanchez	-205	0.7311	1	1	70
2016-12-09	francis ngannou	-550	0.5372	1	1	50
2016-12-09	corey anderson	-400	0.7424	1	1	70
2016-12-09	derrick lewis	-205	0.7983	1	1	75
2016-12-09	marc diakiese	-325	0.6540	1	1	65
2016-12-03	anthony smith	-115	0.6544	1	1	65
2016-12-03	brandon moreno	-125	0.7252	1	1	70
2016-12-03	brendan oreilly	100	0.5464	0	1	50
2016-12-03	ryan hall	-110	0.5732	1	1	55
2016-12-03	ion cutelaba	-215	0.6528	0	0	65
2016-12-03	jorge masvidal	-250	0.5396	1	1	50
2016-12-03	joseph benavidez	-200	0.5385	1	1	50
2016-12-03	devin clark	-105	0.5631	1	1	55
2016-11-26	jake matthews	-300	0.7287	0	0	70
2016-11-26	omari akhmedov	-180	0.5074	1	1	50
2016-11-26	damien brown	115	0.5549	1	0	55
2016-11-26	richard walsh	112	0.6033	0	1	60
2016-11-26	ben nguyen	110	0.5989	1	0	55
2016-11-26	marlon vera	125	0.5636	1	0	55
2016-11-26	derek brunson	-135	0.5036	0	0	50
2016-11-26	chris camozzi	-245	0.7258	0	0	70
2016-11-26	jason knight	131	0.6187	1	0	60
2016-11-19	gegard mousasi	-550	0.7373	1	1	70
2016-11-19	stevie ray	-105	0.5643	1	1	55
2016-11-19	teruto ishihara	-275	0.6556	0	0	65
2016-11-19	kyoji horiguchi	-199	0.7209	1	1	70
2016-11-19	kevin lee	-102	0.7378	1	1	70
2016-11-19	zak cummings	-170	0.5942	1	1	55
2016-11-19	ryan bader	-370	0.7399	1	1	70
2016-11-19	albert morales	235	0.5279	0	1	50
2016-11-19	krzysztof jotko	140	0.5699	1	0	55
2016-11-19	zak ottow	142	0.5038	0	1	50
2016-11-19	manvel gamburyan	140	0.6020	0	1	60
2016-11-19	luis henrique	-275	0.6747	1	1	65
2016-11-19	kamaru usman	-205	0.7164	1	1	70
2016-11-19	jack hermansson	-210	0.5305	0	0	50
2016-11-19	justin scoggins	-168	0.7722	0	0	75
2016-11-12	conor mcgregor	-137	0.5500	1	1	55
2016-11-12	stephen thompson	-191	0.5142	0	0	50
2016-11-12	chris weidman	-190	0.7122	0	0	70
2016-11-12	frankie edgar	-340	0.5929	1	1	55
2016-11-12	rafael natal	-165	0.6559	0	0	65
2016-11-12	vicente luque	-105	0.6086	1	1	60
2016-11-12	jim miller	130	0.7162	1	0	70
2016-11-12	khabib nurmagomedov	-305	0.6494	1	1	60
2016-11-05	rafael dos anjos	-145	0.5486	0	0	50
2016-11-05	charles oliveira	-105	0.5347	0	0	50
2016-11-05	beneil dariush	-115	0.6033	1	1	60
2016-11-05	marco beltran	-125	0.5642	0	0	55
2016-11-05	erick montano	-105	0.6659	0	0	65
2016-11-05	douglas silva de andrade	-125	0.5554	1	1	55
2016-11-05	alex nicholson	175	0.5269	0	1	50
2016-11-05	enrique barzola	-549	0.6046	1	1	60
2016-11-05	felipe arantes	180	0.5206	0	1	50
2016-10-08	Michael Bisping	-225	0.8037	1	1	75
2016-10-08	Gegard Mousasi	-300	0.9063	1	1	75
2016-10-08	Ovince Saint Preux	-155	0.7126	0	0	70
2016-10-08	Stefan Struve	-185	0.7629	1	1	75
2016-10-08	Mirsad Bektic	-700	0.5751	1	1	55
2016-10-08	Iuri Alcantara	-165	0.7111	1	1	70
2016-10-08	Damian Stasiak	145	0.6871	1	0	65
2016-10-08	Albert Tumenov	-260	0.5388	0	0	50
2016-10-08	Mike Perry	120	0.6247	1	0	60
2016-10-08	Leonardo Santos	175	0.5480	1	0	50
2016-10-01	John Lineker	105	0.7123	1	0	70
2016-10-01	Will Brooks	-270	0.5858	0	0	55
2016-10-01	Andre Fili	150	0.5666	1	0	55
2016-10-01	Shamil Abdurakhimov	-120	0.5583	1	1	55
2016-10-01	Elizeu Zaleski dos Santos	130	0.5261	1	0	50
2016-10-01	Nate Marquardt	180	0.5480	1	0	50
2016-10-01	Ion Cutelaba	-165	0.6151	1	1	60
2016-10-01	Curtis Blaydes	-210	0.5855	1	1	55
2016-09-24	Renan Barao	-500	0.7085	1	1	70
2016-09-24	Antonio Silva	350	0.5854	0	1	55
2016-09-24	Fancisco Trinaldo	-130	0.7309	1	1	70
2016-09-24	Thiago Santos	-620	0.6075	0	0	60
2016-09-24	Mike de la Torre	115	0.5159	0	1	50
2016-09-24	Gilbert Burns	-175	0.5683	0	0	55
2016-09-24	Rani Yahya	-140	0.6103	1	1	60
2016-09-24	Jussier Formiga	-200	0.5980	1	1	55
2016-09-24	Stevie Ray	-160	0.6967	0	0	65
2016-09-24	Vicente Luque	-500	0.6840	1	1	65
2016-09-17	Dustin Poirier	-170	0.6995	0	0	65
2016-09-17	Derek Brunson	-190	0.5975	1	1	55
2016-09-17	Roan Carneiro	105	0.5722	1	0	55
2016-09-17	Islam Makhachev	-135	0.5811	1	1	55
2016-09-17	Chas Skelly	-175	0.5081	1	1	50
2016-09-17	Gabriel Benitez	105	0.5951	1	0	55
2016-09-17	Belal Muhammad	-450	0.5708	1	1	55
2016-09-17	Antonio Carlos Junior	-255	0.6906	1	1	65
2016-09-17	Jose Quinonez	140	0.5547	1	0	55
2016-09-17	Randy Brown	-450	0.6144	1	1	60
2016-09-10	Jimmie Rivera	-170	0.5714	1	1	55
2016-09-10	Stipe Miocic	-110	0.6259	1	1	60
2016-09-10	Fabricio Werdum	-225	0.8458	1	1	75
2016-09-10	Brad Tavares	-200	0.6333	1	1	60
2016-09-10	Sean Spencer	105	0.5688	0	1	55
2016-09-03	Andrei Arlovski	120	0.6398	0	1	60
2016-09-03	Alexander Gustaffson	-550	0.7330	1	1	70
2016-09-03	Ryan Bader	-170	0.7937	1	1	75
2016-09-03	Nick Hein	-340	0.7009	1	1	70
2016-09-03	Nick Dalby	-190	0.5350	0	0	50
2016-09-03	Taylor Lapilus	-139	0.7598	1	1	75
2016-09-03	Rustam Khabilov	-560	0.7424	1	1	70

## 
## Attaching package: 'reshape2'

## The following object is masked from 'package:tidyr':
## 
##     smiths

***

Introduction

Ultimate Fighting Championship (UFC) is the world’s premier professional mixed martial arts organization. Fighters engage in one-on-one matches consisting of 3-5 rounds of 5 minutes each. Fighters are allowed to strike with punches, kicks, knees, elbows, or employ submission moves such as chokes or joint locks. Fighting occurs standing up, on the ground, or in a clinch position. Much like other professional sports, there are a variety of structured data gathered on individual fighters and on each fight match up. This project gathers much of that data, learns from it, and then makes predictions about future fight matchups. Specifically, this project seeks to answer the question: Given a specific mathcup, what is each fighter’s probability of winning?

---

Wrangling

The dataset ultimately developed by and for this project is not only an aggregation of data scraped from multiple sources, but is also comprised of newly generated data derived from the scraped data. Namely, the derivitive data refers to historic career statistics for each fighter at the time of each fight.

---

Aggregating The Data

This project intially gathered data from four sources:

• UFC Unofficial API: list of current UFC fighters
• Wikipedia: list of UFC events and event information
  
• FightMetric: detailed fight-by-fight statistics and fighter statistics for every UFC fight
  
• FightMatrix: proprietary current and historical fighter ranking system

However, in the end the only sources of data that were used in the final dataset were FightMetric and FightMatrix. The process of scraping data generally followed the pattern below using the Rvest package.

1. Collect and store relevant URLs in a list.

# Collect individual fighter urls.
for (letter in letters) {
  fightMetric_fighter_list_url <- paste("http://www.fightmetric.com/statistics/fighters?char=",
                                        letter, "&page=all", sep = "")
  
  fightMetric_fighter_list_page <- read_html(fightMetric_fighter_list_url)
  
  # Get inidividual FightMetric Fighter links for letter of alphabet from Fighter List page.
  fighter_links_per_letter <- fightMetric_fighter_list_page %>%
    html_nodes(".b-statistics__table-col:nth-child(1) .b-link_style_black") %>%
    html_attr("href")
  
  fightMetric_fighter_links <- c(fightMetric_fighter_links, fighter_links_per_letter)
}

2. Make copies of HTML from the URL list and store in a list.

#### Make copies of each FightMetric fighter page HTML to later scrape data we need. ####
fightMetric_fighter_html <- list()

for (i in 1:length(fightMetric_fighter_links)) {
  fightMetric_fighter_html[[i]] <- read_html(fightMetric_fighter_links[[i]])
}

3. Parse the HTML and put what we want in a dataframe.

#### Populate REVISED FIGHTER DATA ####
for (i in 1:length(fightMetric_fighter_html)) {
  
  # Get fighter name
  name <- fightMetric_fighter_html[[i]] %>%
    html_nodes(".b-content__title-highlight") %>%
    html_text(trim = TRUE)
  
  fightMetric_fighters_df[i, 1] <- name[[1]]
  
  # Get fighter pro record
  record <- fightMetric_fighter_html[[i]] %>%
    html_nodes(".b-content__title-record") %>%
    html_text(trim = TRUE) %>%
    strsplit(" ")
  
  record <- record[[1]][2] %>%
    strsplit("-")
  
  fightMetric_fighters_df[i, 2] <- record[[1]][1]
  
  fightMetric_fighters_df[i, 3] <- record[[1]][2]
  
  fightMetric_fighters_df[i, 4] <- record[[1]][3]
  
  # Get fighter stats
  stats <- fightMetric_fighter_html[[i]] %>%
    html_nodes(".b-list__box-list-item_type_block") %>%
    html_text(trim = TRUE) %>%
    gsub("^.*:(\\n *)*", "", .)
  
  stats <- stats[-10]
  
  for (s in 1:length(stats)) {
    if (s == "") {
      fightMetric_fighters_df[i, (4+s)] <- NA
    } else {
      fightMetric_fighters_df[i, (4+s)] <- stats[s]
    }
  }
}

This process was adapted and repeated to pull data for each of the three primary dataframes:

• ufc_fight_details_df: fight-by-fight statistics
  
• fightMetric_fighters_df: individual fighter statistics
  
• figher_ranking_df: current and historical fighter rankings by weightclass

---

Cleaning The Data

Much of the data that was scraped had to be cleaned to some extent. At the very least values were converted from character to numeric, and in other cases fixes such as removing ‘%’ signs, converting a fighter’s height into number of inches, converting Dates of birth into machine readable Date format, etc… were necessary. Here are a few examples:

Removing ‘%’ sign.

# Remove % sign in several of the columns and convert to numeric.
for (i in c(11,13,15,16)) {
  fightMetric_fighters_df[,i] <- fightMetric_fighters_df[,i] %>%
    gsub("%", "", .) %>%
    as.numeric()
  
  fightMetric_fighters_df[,i] <- fightMetric_fighters_df[,i]/100
}

Converting height into number of inches

# Convert Height to numeric inches.
fightMetric_fighters_df$Height <- fightMetric_fighters_df$Height %>%
  gsub("'|\"", "", .) %>%
  strsplit(split = " ")

for (h in 1:length(fightMetric_fighters_df$Height)) {
  if (fightMetric_fighters_df[h, 5][[1]] == "--") {
    fightMetric_fighters_df[h, 5][[1]] <- NA
  } else {
    fightMetric_fighters_df[h, 5][[1]] <- (as.numeric(fightMetric_fighters_df[h, 5][[1]][1])*12) +
                                     as.numeric(fightMetric_fighters_df[h, 5][[1]][2])
  }
}

fightMetric_fighters_df$Height <- unlist(fightMetric_fighters_df$Height)

There were also instances of UFC fighters with the same name, which was particulary troublesome when merging this fighter data with other dataframes on the fighter’s name. It gave the appearance of a dataset or a merge that was entirely accurate, but for a few inexplicable errors. So after further digging into the datasets and crosschecking with the original webpages, the issue was handled by identifying these fighters by their name in addition to a physical attribute, and then renaming one of them. Here’s an example.

# Handle instances of same name in FIGHTMETRIC_FIGHTERS_DF
fightMetric_fighters_df[fightMetric_fighters_df$Name=="Dong Hyun Kim" & fightMetric_fighters_df$Reach==70, 1] <- "Dong Hyun Kim (Maestro)"
fightMetric_fighters_df[fightMetric_fighters_df$Name=="Michael McDonald" & fightMetric_fighters_df$Weight==205, 1] <- "Michael McDonald (The Black Sniper)"
fightMetric_fighters_df[fightMetric_fighters_df$Name=="Tony Johnson" & fightMetric_fighters_df$Weight==185, 1] <- "Tony Johnson (185lbs.)"

---

Shaping The Data

After all the data had been collected and stored in dataframes, there was some reshaping required in order to end up with a single dataframe suitable to subset from and present to a logistic regression or random forest function for modeling. At this point there were three dataframes - fight_details, fightMetric_fighters, and fighter_ranking - and the desired end state was to have a single fight_details that contained not only detailed fighter stats for each fight, but also each fighter’s attributes, and each fighter’s ranking. Furthermore, fight_details was currently in the form of each row representing one fighter’s stats for one fight, which meant there were two rows for each fight - one for each fighter’s stats. So, four things needed to happen…

1. Reshape fight_details to contain both fighter’s statistics for a single fight in a single row.

#### Reshape fight_details_df to accomodate logistical regression or random forest ####
colnames(ufc_fight_details_df)[2:3] <- c("Fighter1", "Fighter2")


# Grab even number rows from fight_details to later bind with odd number rows.
fighter2_df <- ufc_fight_details_df %>%
  slice(seq(2,nrow(ufc_fight_details_df),2))

# Rename columns in fighter2_df and original fight_details_df.
colnames(fighter2_df)[17:160] <- paste("F2", colnames(fighter2_df)[17:160], sep = "")
colnames(ufc_fight_details_df)[17:160] <- paste("F1", colnames(ufc_fight_details_df)[17:160], sep = "")

# Delete even number rows from fight_details.
ufc_fight_details_df <- ufc_fight_details_df[-seq(2,nrow(ufc_fight_details_df),2),]

# Combine columns from fight_details and fighter_2, so each row holds each fighter's data for
# a given fight.
ufc_fight_details_df <- bind_cols(ufc_fight_details_df, fighter2_df[17:160])

# Delete fighter_2 data frame.
rm(fighter2_df)

# Then, REMOVE rows with NA for FIGHTER.
na_row_index <- c()
for (n in 1:nrow(ufc_fight_details_df)) {
  if (is.na(ufc_fight_details_df[n,2]) | is.na(ufc_fight_details_df[n,3])) {
    na_row_index <- c(na_row_index, n)
  }
}

ufc_fight_details_df <- ufc_fight_details_df[-c(na_row_index),]
###################################

2. Include a new variable to indicate the winner - simply a 0 or 1 variable indicating if Fighter 1 wins or not.

# Mutate FIGHT_DETAILS to give logical form of result (Fighter1 or Fighter2) ready for logistic
# regression or random forest.
ufc_fight_details_df <- ufc_fight_details_df %>%
  mutate(F1Wins = as.character(Result) == as.character(Fighter1))

for (i in 1:nrow(ufc_fight_details_df)) {
  if (is.na(ufc_fight_details_df[i,305])) {
    ufc_fight_details_df[i,305] <- NA
  } else if (ufc_fight_details_df[i,305] == TRUE) {
    ufc_fight_details_df[i,305] <- 1
  } else {
    ufc_fight_details_df[i,305] <- 0
  }
}

3. Merge fightMetric_fighters with fight_details on each fighter’s name.

##### Merge FIGHTMETRIC_FIGHTERS_DF with FIGHT_DETAILS_DF ####

#### Merge FIGHTERS_DF to FIGHTER1 in FIGHT_DETAILS ####
fightMetric_fighters_df <- fightMetric_fighters_df %>%
  rename(Fighter1=Name)

ufc_fight_details_df <- ufc_fight_details_df %>%
  left_join(fightMetric_fighters_df, by = "Fighter1")

left_copy_fight_details <- ufc_fight_details_df[,1:306]
right_copy_fighter1_details <- ufc_fight_details_df[,307:323]

left_copy_double_rows <- as.numeric(rownames(left_copy_fight_details[duplicated(left_copy_fight_details),]))

left_copy_fight_details <- left_copy_fight_details[-c(left_copy_double_rows),]
right_copy_fighter1_details <- right_copy_fighter1_details[-c(left_copy_double_rows),]

ufc_fight_details_df <- left_copy_fight_details
########################

#### Merge FIGHTERS_DF to FIGHTER2 in FIGHT_DETAILS ####
fightMetric_fighters_df <- fightMetric_fighters_df %>%
  rename(Fighter2=Fighter1)

ufc_fight_details_df <- ufc_fight_details_df %>%
  left_join(fightMetric_fighters_df, by = "Fighter2")

left_copy_fight_details <- ufc_fight_details_df[,1:306]
right_copy_fighter2_details <- ufc_fight_details_df[,307:323]

left_copy_double_rows <- as.numeric(rownames(left_copy_fight_details[duplicated(left_copy_fight_details),]))

left_copy_fight_details <- left_copy_fight_details[-c(left_copy_double_rows),]
right_copy_fighter2_details <- right_copy_fighter2_details[-c(left_copy_double_rows),]

ufc_fight_details_df <- left_copy_fight_details %>%
  bind_cols(right_copy_fighter1_details) %>%
  bind_cols(right_copy_fighter2_details)

colnames(ufc_fight_details_df)[307:323] <- paste("F1", colnames(fightMetric_fighters_df)[2:18], sep = "")
colnames(ufc_fight_details_df)[324:340] <- paste("F2", colnames(fightMetric_fighters_df)[2:18], sep = "")

fightMetric_fighters_df <- fightMetric_fighters_df %>%
  rename(Name=Fighter2)
###################

4. Merge fighter_ranking with fight_details on each fighter’s name.

################### Assign FIGHTER_RANKING to fighters in NEW_FIGHTS #################

# Reset clean unique FightNumbers 
new_fights_df$FightNumber <- c(1:nrow(new_fights_df))

new_fights_df <- new_fights_df %>%
  mutate(F1Rank = 251, F2Rank = 251)

F1RankCol <- grep("F1Rank", colnames(new_fights_df))
F2RankCol <- grep("F2Rank", colnames(new_fights_df))

quarterly_date <- as.Date("2006-01-01", "%Y-%m-%d")
for (q in 1:(length(fighter_ranking_df)-9)) {
  temp_fighter_ranking <- fighter_ranking_df %>%
    select(c(q,353))
  
  quarterly_fight_details <- new_fights_df %>%
    filter(quarter(Date, with_year = TRUE) == quarter(quarterly_date, with_year = TRUE)) %>%
    select(FightNumber, Date, Fighter1, Fighter2, F1Rank, F2Rank)
  
  for (F1 in 1:nrow(quarterly_fight_details)) {
    if (tolower(quarterly_fight_details[F1, 3]) %in% tolower(na.omit(temp_fighter_ranking[,1]))) {
      quarterly_fight_details[F1,5] <- temp_fighter_ranking[tolower(na.omit(temp_fighter_ranking[,1])) == tolower(quarterly_fight_details[F1,3]), 2][1]
    }
    if (tolower(quarterly_fight_details[F1, 4]) %in% tolower(na.omit(temp_fighter_ranking[,1]))) {
      quarterly_fight_details[F1,6] <- temp_fighter_ranking[tolower(na.omit(temp_fighter_ranking[,1])) == tolower(quarterly_fight_details[F1,4]), 2][1]
    }
  }
  
  for (i in 1:nrow(quarterly_fight_details)) {
    new_fights_df[new_fights_df$FightNumber == quarterly_fight_details[i,1], c(F1RankCol, F2RankCol)] <- quarterly_fight_details[i, 5:6]
  }
  
  quarterly_date <- quarterly_date + months(3)
  if (quarter(quarterly_date, with_year = TRUE) > quarter(today(), with_year = TRUE)) {
    quarterly_date <- as.Date("2006-01-01", "%Y-%m-%d")
  }
}

new_fights_df$FightNumber <- c(1:nrow(new_fights_df))

new_fights_df$F1Rank <- as.numeric(new_fights_df$F1Rank)
new_fights_df$F2Rank <- as.numeric(new_fights_df$F2Rank)

#######################################################################################

---

Feature Engineering

As mentioned earlier, the dataset for this project is a combination of data scraped from several sources, as well as data derived (or newly created) from the scraped data. Specifically, the derived data includes career statistics for each fighter at the time of each fight. This data was not available anywhere online (to my knowledge), but was critical in developing an accurate description of each fighter at the time of each fight in which to build a model that predicts future outcomes. So, the basic approach to generating this new data was to sort through the entire fight dataset (ufc_fight_details_df) one fighter at a time in chronologically tally up their career stats. Below is an example of some of the code, which can almost certainly be refactored into something more economical, but for the time-being is made up of about 200 lines of nested for loops and if else statements.

4. Example block of code to generate derivitive data.

# Add columns to be populated to FIGHT_DETAILS
ufc_fight_details_df <- ufc_fight_details_df %>%
  mutate(F1CarFightTime = NA, F2CarFightTime = NA, F1UFCWins = NA, F2UFCWins = NA, F1UFCLosses = NA,
         F2UFCLosses = NA, F1UFCNC = NA, F2UFCNC = NA, F1CarSigStr = NA, F2CarSigStr = NA,
         F1CarSigStrAtt = NA, F2CarSigStrAtt = NA, F1CarSigStrAbs = NA, F2CarSigStrAbs = NA,
         F1CarTD = NA, F2CarTD = NA, F1CarTDAtt = NA, F2CarTDAtt = NA, F1CarTDAbs = NA, F2CarTDAbs = NA,
         F1CarOppSigStrAtt = NA, F2CarOppSigStrAtt = NA, F1CarOppTDAtt = NA, F2CarOppTDAtt = NA)

#### Begin loop to generate additional fighter data ####
for (fighter in unique_fighters$.) {
  fighter_career_data <- ufc_fight_details_df %>%
    filter((Fighter1 == as.character(fighter)) | (Fighter2 == as.character(fighter))) %>%
    arrange(Date)
  
  for (i in 1:nrow(fighter_career_data)) {
    
  #### Adding up total career time in ring for each fighter going into each fight - Fighter1 spot. ####
    if (fighter == fighter_career_data[i,2]) {
      
      if (i == 1) {
        # F1CarFightTime when F1
        fighter_career_data[i, 342] <- 0
        
        # F1CarSigStr when F1
        fighter_career_data[i, 350] <- 0
        
        # F1CarSigStrAtt when F1
        fighter_career_data[i, 352] <- 0
        
        # F1CarSigStrAbs when F1
        fighter_career_data[i, 354] <- 0
        
        # F1CarTD when F1
        fighter_career_data[i, 356] <- 0
        
        # F1CarTDAtt when F1
        fighter_career_data[i, 358] <- 0
        
        # F1CarTDAbs when F1
        fighter_career_data[i, 360] <- 0
        
        # F1CarOppSigStrAtt when F1
        fighter_career_data[i, 362] <- 0
        
        # F1CarOppTDAtt when F1
        fighter_career_data[i, 364] <- 0
      } else {
        # F1 was F1 last fight
        if (fighter == fighter_career_data[(i-1),2]) {
         
          # F1CarFightTime when F1 this fight was F1 last fight
          fighter_career_data[i, 342] <- fighter_career_data[(i-1), 342] + fighter_career_data[(i-1), 341]
          
          # Get fight data from last fight when F1 this fight was F1 last fight.
          temp_fight_data <- c(fighter_career_data[(i-1), 18], fighter_career_data[(i-1), 19],
                               fighter_career_data[(i-1), 162], fighter_career_data[(i-1), 23],
                               fighter_career_data[(i-1), 24], fighter_career_data[(i-1), 167],
                               fighter_career_data[(i-1), 163], fighter_career_data[(i-1), 168])
          
          # F1CarSigStr through F1CarOppTDAtt when F1 this fight was F1 last fight
          for (n in 1:8) {
            if (is.na(temp_fight_data[n])) {
              fighter_career_data[i, ((174+n)*2)] <- 0
            } else {
              fighter_career_data[i, ((174+n)*2)] <- fighter_career_data[(i-1), ((174+n)*2)] + temp_fight_data[n]
            }
          }
        }

Some widdling down of the features occured in going from the 382 features of ufc_fight_details_df to the subset of 48 features ultimately used in the logisitic regression model. This process was part logic, part domain expertise. Perhaps more on the logic side were features of physical description, like age, height, weight, reach, weight class, and stance. Career statistics features are the product of both logic and domain expertise. You don’t need to know anything about Mixed Martial Arts or UFC to think it’s a good idea to consider each fighter’s career statistics when making a prediction, but, for example, data for round-by-round statistics were not included because not all fights have the same number of rounds, and there would have been a considerable amount of additional data wrangling to present this data to the model in a useful way. The decision to include each fighter’s rank might also seem natural, but on the other hand an excellent data scientist without as much domain expertise might think that statistical descriptions of each fighter’s performance should be enough, or even superior than a somewhat subjective or circumstantial indicator like rank. But the ranking variable takes into consideration factors such as: the quality of a fighter’s past opponenents, and the chronology of past performance (e.g. a fighter with a long successful career that is now fighting past his prime).

---

Prediction Methods

After collecting and wrangling the data, we have a dataset of over 5,000 MMA fights ready to train a statistical model to make a predictions. Given we’re trying to predict a binary outcome - Win or Lose - for each fighter in a given fight, there are a few methods that would be suitable. For this project I tested two different methods - Logistic Regression and Random Forest, and ended up going with Logistic Regression because it performed better on test data.

---

Logistic Regression

---

Testing

# Variable for the date ten years ago today.
ten_years_ago <- ymd(today()) - years(10)
# This gives us the subset of fights (rows) we'll use for the model.
row_subset_fights <- ufc_fight_details_df %>%
  filter(Date > ten_years_ago) %>%
  filter(Gender == "Men") %>%
  filter(WeightClass != "Strawweight") %>%
  filter(WeightClass != "Atomweight") %>%
  filter(WeightClass != "Catch Weight") %>%
  filter(WeightClass != "Open Weight") %>%
  filter(WeightClass != "Other") %>%
  filter(WeightClass != "Super Heavyweight") %>%
  filter((F1UFCWins>0 | F1UFCLosses>0 | F1UFCNC>0) & (F2UFCWins>0 | F2UFCLosses>0 | F2UFCNC>0)) %>%
  arrange(Date)
# Subset of columns to be used in the model subset.
model_dataset_columns <- read.csv(paste(getwd(), "/FreshModelColumns.csv", sep = ""))
model_dataset_columns <- as.character(model_dataset_columns$ColumnNames)
# Get subset fight_details dataset - all rows; only columns from selected_features
fresh_model_dataset <- row_subset_fights %>%
  subset(select = model_dataset_columns)
# Factor Weightclass.
fresh_model_dataset$WeightClass <- as.factor(fresh_model_dataset$WeightClass)
# Factor F1Stance
fresh_model_dataset$F1Stance <- fresh_model_dataset$F1Stance %>%
  as.factor()
# Factor F2Stance
fresh_model_dataset$F2Stance <- fresh_model_dataset$F2Stance %>%
  as.factor()
##########!!!!!!!!! SUBSETTING !!!!!!!!!!###########
# Split fight_details_df into training and testing sets.
set.seed(109) 
split_ufc_data <- fresh_model_dataset$F1Wins %>%
  sample.split(SplitRatio = 0.70)
train_ufc_subset <- fresh_model_dataset %>%
  subset(split_ufc_data == TRUE)
test_ufc_subset <- fresh_model_dataset %>%
  subset(split_ufc_data == FALSE)
# Remove and save columns we don't need, but may need later... just in case.
saved_column_names <- c("FightNumber","Fighter1","Fighter2","Date")
column_pos <- c()
for (column in saved_column_names) {
  column_pos <- c(column_pos, grep(column, colnames(fresh_model_dataset)))
}
saved_columns <- fresh_model_dataset %>%
  select(column_pos)
train_ufc_subset <- train_ufc_subset %>% select(-column_pos)
test_ufc_subset <- test_ufc_subset %>% select(-column_pos)
####################################
######### LOGIT MODEL ##############
ufc_fresh_logit_model <- glm(F1Wins ~ ., data = na.omit(train_ufc_subset), family = binomial)
# Prediction - probability that F1 wins
test_ufc_subset$F1WinsProb <- round(predict(ufc_fresh_logit_model, newdata = test_ufc_subset, type = "response"), 4)
#########################################
########## THRESHOLD MODELS #############
WinThreshold <- 0.5
# Filter out fights to get only those where F1WinsProb > 0.6 or < 0.4
test_threshold_subset <- test_ufc_subset %>%
  filter(F1WinsProb > WinThreshold | F1WinsProb < (1-WinThreshold))
test_threshold_subset$PredF1Wins <- as.numeric(test_threshold_subset$F1WinsProb > WinThreshold)
#######################################
######### LOGIT CONFUSION MATRIX #######
logitPerformance <- confusionMatrix(as.factor(test_threshold_subset$PredF1Wins), as.factor(test_threshold_subset$F1Wins),
                                    dnn = c("PredF1Wins", "F1Wins"))
logitPerformance

Confusion Matrix and Statistics

          F1Wins
PredF1Wins   0   1
         0 119  99
         1 237 479
                                          
               Accuracy : 0.6403          
                 95% CI : (0.6085, 0.6711)
    No Information Rate : 0.6188          
    P-Value [Acc > NIR] : 0.09412         
                                          
                  Kappa : 0.1761          
 Mcnemar's Test P-Value : 7.782e-14       
                                          
            Sensitivity : 0.3343          
            Specificity : 0.8287          
         Pos Pred Value : 0.5459          
         Neg Pred Value : 0.6690          
             Prevalence : 0.3812          
         Detection Rate : 0.1274          
   Detection Prevalence : 0.2334          
      Balanced Accuracy : 0.5815          
                                          
       'Positive' Class : 0               
                                          

roc(F1Wins ~ F1WinsProb, test_threshold_subset, auc = TRUE, plot = TRUE)

Call:
roc.formula(formula = F1Wins ~ F1WinsProb, data = test_threshold_subset,     auc = TRUE, plot = TRUE)

Data: F1WinsProb in 356 controls (F1Wins 0) < 578 cases (F1Wins 1).
Area under the curve: 0.6701

---

Random Forest

---

Testing

############################ RANDOM FOREST MODEL ###################################
#set.seed(388)
ufc_fresh_randomForest_model <- randomForest(as.factor(F1Wins) ~ ., data = na.omit(train_ufc_subset),
                                            importance=TRUE)
test_ufc_subset$F1WinsProb <- NULL
test_ufc_subset$RFF1WinsProb <- predict(ufc_fresh_randomForest_model, test_ufc_subset, type = "prob")[,2]
test_ufc_subset$RFPredF1Wins <- as.numeric(test_ufc_subset$RFF1WinsProb > 0.5)
randomForestPerformance <- confusionMatrix(as.factor(test_ufc_subset$RFPredF1Wins), as.factor(test_ufc_subset$F1Wins), dnn = c("F1Wins", "PredF1Wins"))
randomForestPerformance

Confusion Matrix and Statistics

      PredF1Wins
F1Wins   0   1
     0  91  77
     1 265 501
                                         
               Accuracy : 0.6338         
                 95% CI : (0.602, 0.6648)
    No Information Rate : 0.6188         
    P-Value [Acc > NIR] : 0.1817         
                                         
                  Kappa : 0.1362         
 Mcnemar's Test P-Value : <2e-16         
                                         
            Sensitivity : 0.25562        
            Specificity : 0.86678        
         Pos Pred Value : 0.54167        
         Neg Pred Value : 0.65405        
             Prevalence : 0.38116        
         Detection Rate : 0.09743        
   Detection Prevalence : 0.17987        
      Balanced Accuracy : 0.56120        
                                         
       'Positive' Class : 0              
                                         

roc(F1Wins ~ RFF1WinsProb, test_ufc_subset, auc = TRUE, plot = TRUE)

Call:
roc.formula(formula = F1Wins ~ RFF1WinsProb, data = test_ufc_subset,     auc = TRUE, plot = TRUE)

Data: RFF1WinsProb in 356 controls (F1Wins 0) < 578 cases (F1Wins 1).
Area under the curve: 0.6282

varImpPlot(ufc_fresh_randomForest_model, n.var = 10)

---

UFC Project Workflow

The practical application of this project is to generate predictions for UFC fights on a weekly basis in order to wager on the fights. There are four main steps in this workflow.

1. Update FightMetric_fighters_df: After each fight event, existing fighter data changes and sometimes new fighters are introduced to UFC. So, we want to pull this new information into our dataset of fighters (FightMetric_fighters_df). Specifically, we do this by running two blocks of code labeled for FightMetric_fighters_df that can be found in the UFC.R script beginning at line ~250 and ~1022, respectively.

2. Update fighter_ranking_df: Of course, after each event fighter’s win and lose, and so the fighter rankings change. There are actually two compenents to our fighter_ranking_df - we have historical rankings and current rankings. After each event we need to update the current rankings of each fighter. But the historical rankings will not change. These are recorded in the dataset on a quarterly basis. In other words, we have historical rankings for each weightclass for each quarter going back to Q1 of 2006. So updating of the historical rankings only occurs once a quarter when they are released. Theoretically this means our model could have a few cases of rankings that are slightly off, but as a practical matter quarterly rankings are probably pretty accurate as the vast majority of fighters do not fight more than once per quarter. We update the current rankings by running the block of code beginning at line ~1079 of UFCUpdater.R, and we update the historical rankings once a quarter by running the block of code at line ~1030 of UFCUpdater.R.

3. Update ufc_fight_details_df: After each event there are new fights and detailed statistics to be added to our ufc_fight_details_df. While this dataset was originally built using the UFC.R script, it can be a lengthy and somewhat hands-on process to scrape nearly 10,000 web pages once, but to do so every week would be overly burdonsome. So, there is another script, UFCUpdater.R, that scrapes only the recently completed fights and data, computes their derivitive data, and then merges them back into ufc_fight_details_df. We simply run the UFCUpdator.R script to accomplish this.

4. Generate predictions: Once we have updated our dataset, we are ready to generate predictions for the set of upcoming fights. To accomplish this we run the UFCPredictor.R script. This script does a few things. First, it scrapes the upcoming fight matchups. Then it computes and gathers the relevant variables for each fighter from our dataset. Then it refits the model with the most recent dataset. Finally, it applies our model to the set of data for the upcoming fights to produce a probability that Fighter 1 will win.

---

Roadmap For Improvement

The status of this project is both functional and practical, and it accomplishes what it sets out to do. Nonetheless, there is room for improvement. The following is a working list of items to explore for improving the performance of the model.

Feature Engineering
1. A feature that incorporates how a fight ends: For example, does a fighter win by KO/TKO, Submission, Decision, etc…
2. A feature that incorporates whether or not this is a fighter’s first fight in UFC.
3. A feature to incorporate a fighter’s current winning or losing streak.
4. Further testing of different variable combinations (removing variables that appear insignificant, for example)

Predictive Modeling
1. Explore implementing and Ensemble model of both Logistic Regression and Random Forest.
2. Explore clustering on sets of fights predicted incorrectly to see if we can develop rules to improve overall predictive accuracy with a multi-stage predictive approach… or a multi-faceted model (i.e. generate multiple fits for subsets of fights that meet certain criteria).
3. Explore Specificity, Precision, and other model performance measures as a way to weight or hone model predictions.

Usability
1. Automate dataset updates and prediction generation.
2. Build a user-friendly dashboard or interactive interface with Flexdashboard or Shiny, respectively.

Code Base
1. Implement testing suite to help avoid “breaking” an increasingly complex code base during future code maintenance and feature additions.
2. Plenty of room for refactoring, which will simplfy readability and collaboration, and may present some gains in execution speed.

---

Appendix

Data

# Creating table to explain the data.
data_table <- data.frame(colnames(ufc_fight_details_df),
                         rep(NA,length(ufc_fight_details_df)),
                         as.character(ufc_fight_details_df[3449,]),
                         rep("FightMetric",length(ufc_fight_details_df)),
                         stringsAsFactors = FALSE)
# Assign column names.
colnames(data_table) <- c("Variables", "Unabreviated", "Example", "Source")
# Generate unabreviated variable names.
data_table$Unabreviated <- data_table$Variables %>%
  gsub("F1", "Fighter1 ", .) %>%
  gsub("F2", "Fighter2 ", .) %>%
  gsub("Str", "Strike ", .) %>%
  gsub("Att", "Attempt ", .) %>%
  gsub("Car", "Career ", .) %>%
  gsub("Sig", "Significant ", .) %>%
  gsub("Opp", "Opponent ", .) %>%
  gsub("TD", "Take Down ", .) %>%
  gsub("Abs", "Absorbed ", .) %>%
  gsub("pM", "per Minute ", .) %>%
  gsub("Acc", "Accuracy ", .) %>%
  gsub("Def", "Defense ", .) %>%
  gsub("Avg", "Average ", .) %>%
  gsub("Dist", "Distance ", .) %>%
  gsub("KD", "Knock Down ", .) %>%
  gsub("Pass", "Guard Pass ", .) %>%
  gsub("NC", "No Contest ", .) %>%
  gsub("R1", "Round 1 ", .) %>%
  gsub("R2", "Round 2 ", .) %>%
  gsub("R3", "Round 3 ", .) %>%
  gsub("R4", "Round 4 ", .) %>%
  gsub("R5", "Round 5 ", .) %>%
  gsub("Tot", "Total ", .) %>%
  gsub("Rev", "Reversal", .) %>%
  gsub("SA", "Strikes Absorbed ", .) %>%
  gsub("Sub", "Submission ", .) %>%
  gsub("Ref", "Referee ", .) %>%
  gsub("SL", "Strikes Landed ", .)
# Insert data sources
self_derived <- grep(".*Car.*", data_table$Variables)
fightmatrix <- grep("F1Rank|F2Rank", data_table$Variables)
data_table$Source[self_derived] <- "Derived"
data_table$Source[fightmatrix] <- "FightMatrix"
kable(data_table, caption = "Composition of the primary datset, [ufc_fight_details_df]", align = 'l')

Variables	Unabreviated	Example	Source
Date	Date	2009-03-07	FightMetric
Detail	Detail	Kick to Head At Distance	FightMetric
F1Age	Fighter1 Age	32	FightMetric
F1BodyStr	Fighter1 BodyStrike	0	FightMetric
F1BodyStrAtt	Fighter1 BodyStrike Attempt	0	FightMetric
F1CarFightTime	Fighter1 Career FightTime	3790	Derived
F1CarOppSigStrAtt	Fighter1 Career Opponent Significant Strike Attempt	540	Derived
F1CarOppTDAtt	Fighter1 Career Opponent Take Down Attempt	5	Derived
F1CarSigStr	Fighter1 Career Significant Strike	216	Derived
F1CarSigStrAbs	Fighter1 Career Significant Strike Absorbed	215	Derived
F1CarSigStrAtt	Fighter1 Career Significant Strike Attempt	505	Derived
F1CarStrAbspM	Fighter1 Career Strike Absorbed per Minute	3.4	Derived
F1CarStrAcc	Fighter1 Career Strike Accuracy	0.43	Derived
F1CarStrDef	Fighter1 Career Strike Defense	0.4	Derived
F1CarStrLpM	Fighter1 Career Strike Lper Minute	3.42	Derived
F1CarTD	Fighter1 Career Take Down	17	Derived
F1CarTDAbs	Fighter1 Career Take Down Absorbed	0	Derived
F1CarTDAcc	Fighter1 Career Take Down Accuracy	0.52	Derived
F1CarTDAtt	Fighter1 Career Take Down Attempt	33	Derived
F1CarTDAvg	Fighter1 Career Take Down Average	2.43	Derived
F1CarTDDef	Fighter1 Career Take Down Defense	0	Derived
F1ClinchStr	Fighter1 ClinchStrike	2	FightMetric
F1ClinchStrAtt	Fighter1 ClinchStrike Attempt	4	FightMetric
F1DistStr	Fighter1 Distance Strike	20	FightMetric
F1DistStrAtt	Fighter1 Distance Strike Attempt	42	FightMetric
F1DOB	Fighter1 DOB	1976-10-05	FightMetric
F1GroundStr	Fighter1 GroundStrike	1	FightMetric
F1GroundStrAtt	Fighter1 GroundStrike Attempt	1	FightMetric
F1HeadStr	Fighter1 HeadStrike	18	FightMetric
F1HeadStrAtt	Fighter1 HeadStrike Attempt	42	FightMetric
F1Height	Fighter1 Height	73	FightMetric
F1KD	Fighter1 Knock Down	1	FightMetric
F1LegStr	Fighter1 LegStrike	5	FightMetric
F1LegStrAtt	Fighter1 LegStrike Attempt	5	FightMetric
F1Pass	Fighter1 Guard Pass	0	FightMetric
F1ProLosses	Fighter1 ProLosses	5	FightMetric
F1ProNC	Fighter1 ProNo Contest	0	FightMetric
F1ProWins	Fighter1 ProWins	12	FightMetric
F1R1BodyStr	Fighter1 Round 1 BodyStrike	0	FightMetric
F1R1BodyStrAtt	Fighter1 Round 1 BodyStrike Attempt	0	FightMetric
F1R1ClinchStr	Fighter1 Round 1 ClinchStrike	2	FightMetric
F1R1ClinchStrAtt	Fighter1 Round 1 ClinchStrike Attempt	4	FightMetric
F1R1DistStr	Fighter1 Round 1 Distance Strike	20	FightMetric
F1R1DistStrAtt	Fighter1 Round 1 Distance Strike Attempt	42	FightMetric
F1R1GroundStr	Fighter1 Round 1 GroundStrike	1	FightMetric
F1R1GroundStrAtt	Fighter1 Round 1 GroundStrike Attempt	1	FightMetric
F1R1HeadStr	Fighter1 Round 1 HeadStrike	18	FightMetric
F1R1HeadStrAtt	Fighter1 Round 1 HeadStrike Attempt	42	FightMetric
F1R1KD	Fighter1 Round 1 Knock Down	1	FightMetric
F1R1LegStr	Fighter1 Round 1 LegStrike	5	FightMetric
F1R1LegStrAtt	Fighter1 Round 1 LegStrike Attempt	5	FightMetric
F1R1Pass	Fighter1 Round 1 Guard Pass	0	FightMetric
F1R1Rev	Fighter1 Round 1 Reversal	0	FightMetric
F1R1SigStr	Fighter1 Round 1 Significant Strike	23	FightMetric
F1R1SigStrAtt	Fighter1 Round 1 Significant Strike Attempt	47	FightMetric
F1R1SigStrPercent	Fighter1 Round 1 Significant Strike Percent	0.48	FightMetric
F1R1SubAtt	Fighter1 Round 1 Submission Attempt	0	FightMetric
F1R1TD	Fighter1 Round 1 Take Down	0	FightMetric
F1R1TDAtt	Fighter1 Round 1 Take Down Attempt	0	FightMetric
F1R1TDPercent	Fighter1 Round 1 Take Down Percent	0	FightMetric
F1R1TotStr	Fighter1 Round 1 Total Strike	40	FightMetric
F1R1TotStrAtt	Fighter1 Round 1 Total Strike Attempt	64	FightMetric
F1R2BodyStr	Fighter1 Round 2 BodyStrike	NA	FightMetric
F1R2BodyStrAtt	Fighter1 Round 2 BodyStrike Attempt	NA	FightMetric
F1R2ClinchStr	Fighter1 Round 2 ClinchStrike	NA	FightMetric
F1R2ClinchStrAtt	Fighter1 Round 2 ClinchStrike Attempt	NA	FightMetric
F1R2DistStr	Fighter1 Round 2 Distance Strike	NA	FightMetric
F1R2DistStrAtt	Fighter1 Round 2 Distance Strike Attempt	NA	FightMetric
F1R2GroundStr	Fighter1 Round 2 GroundStrike	NA	FightMetric
F1R2GroundStrAtt	Fighter1 Round 2 GroundStrike Attempt	NA	FightMetric
F1R2HeadStr	Fighter1 Round 2 HeadStrike	NA	FightMetric
F1R2HeadStrAtt	Fighter1 Round 2 HeadStrike Attempt	NA	FightMetric
F1R2KD	Fighter1 Round 2 Knock Down	NA	FightMetric
F1R2LegStr	Fighter1 Round 2 LegStrike	NA	FightMetric
F1R2LegStrAtt	Fighter1 Round 2 LegStrike Attempt	NA	FightMetric
F1R2Pass	Fighter1 Round 2 Guard Pass	NA	FightMetric
F1R2Rev	Fighter1 Round 2 Reversal	NA	FightMetric
F1R2SigStr	Fighter1 Round 2 Significant Strike	NA	FightMetric
F1R2SigStrAtt	Fighter1 Round 2 Significant Strike Attempt	NA	FightMetric
F1R2SigStrPercent	Fighter1 Round 2 Significant Strike Percent	NA	FightMetric
F1R2SubAtt	Fighter1 Round 2 Submission Attempt	NA	FightMetric
F1R2TD	Fighter1 Round 2 Take Down	NA	FightMetric
F1R2TDAtt	Fighter1 Round 2 Take Down Attempt	NA	FightMetric
F1R2TDPercent	Fighter1 Round 2 Take Down Percent	NA	FightMetric
F1R2TotStr	Fighter1 Round 2 Total Strike	NA	FightMetric
F1R2TotStrAtt	Fighter1 Round 2 Total Strike Attempt	NA	FightMetric
F1R3BodyStr	Fighter1 Round 3 BodyStrike	NA	FightMetric
F1R3BodyStrAtt	Fighter1 Round 3 BodyStrike Attempt	NA	FightMetric
F1R3ClinchStr	Fighter1 Round 3 ClinchStrike	NA	FightMetric
F1R3ClinchStrAtt	Fighter1 Round 3 ClinchStrike Attempt	NA	FightMetric
F1R3DistStr	Fighter1 Round 3 Distance Strike	NA	FightMetric
F1R3DistStrAtt	Fighter1 Round 3 Distance Strike Attempt	NA	FightMetric
F1R3GroundStr	Fighter1 Round 3 GroundStrike	NA	FightMetric
F1R3GroundStrAtt	Fighter1 Round 3 GroundStrike Attempt	NA	FightMetric
F1R3HeadStr	Fighter1 Round 3 HeadStrike	NA	FightMetric
F1R3HeadStrAtt	Fighter1 Round 3 HeadStrike Attempt	NA	FightMetric
F1R3KD	Fighter1 Round 3 Knock Down	NA	FightMetric
F1R3LegStr	Fighter1 Round 3 LegStrike	NA	FightMetric
F1R3LegStrAtt	Fighter1 Round 3 LegStrike Attempt	NA	FightMetric
F1R3Pass	Fighter1 Round 3 Guard Pass	NA	FightMetric
F1R3Rev	Fighter1 Round 3 Reversal	NA	FightMetric
F1R3SigStr	Fighter1 Round 3 Significant Strike	NA	FightMetric
F1R3SigStrAtt	Fighter1 Round 3 Significant Strike Attempt	NA	FightMetric
F1R3SigStrPercent	Fighter1 Round 3 Significant Strike Percent	NA	FightMetric
F1R3SubAtt	Fighter1 Round 3 Submission Attempt	NA	FightMetric
F1R3TD	Fighter1 Round 3 Take Down	NA	FightMetric
F1R3TDAtt	Fighter1 Round 3 Take Down Attempt	NA	FightMetric
F1R3TDPercent	Fighter1 Round 3 Take Down Percent	NA	FightMetric
F1R3TotStr	Fighter1 Round 3 Total Strike	NA	FightMetric
F1R3TotStrAtt	Fighter1 Round 3 Total Strike Attempt	NA	FightMetric
F1R4BodyStr	Fighter1 Round 4 BodyStrike	NA	FightMetric
F1R4BodyStrAtt	Fighter1 Round 4 BodyStrike Attempt	NA	FightMetric
F1R4ClinchStr	Fighter1 Round 4 ClinchStrike	NA	FightMetric
F1R4ClinchStrAtt	Fighter1 Round 4 ClinchStrike Attempt	NA	FightMetric
F1R4DistStr	Fighter1 Round 4 Distance Strike	NA	FightMetric
F1R4DistStrAtt	Fighter1 Round 4 Distance Strike Attempt	NA	FightMetric
F1R4GroundStr	Fighter1 Round 4 GroundStrike	NA	FightMetric
F1R4GroundStrAtt	Fighter1 Round 4 GroundStrike Attempt	NA	FightMetric
F1R4HeadStr	Fighter1 Round 4 HeadStrike	NA	FightMetric
F1R4HeadStrAtt	Fighter1 Round 4 HeadStrike Attempt	NA	FightMetric
F1R4KD	Fighter1 Round 4 Knock Down	NA	FightMetric
F1R4LegStr	Fighter1 Round 4 LegStrike	NA	FightMetric
F1R4LegStrAtt	Fighter1 Round 4 LegStrike Attempt	NA	FightMetric
F1R4Pass	Fighter1 Round 4 Guard Pass	NA	FightMetric
F1R4Rev	Fighter1 Round 4 Reversal	NA	FightMetric
F1R4SigStr	Fighter1 Round 4 Significant Strike	NA	FightMetric
F1R4SigStrAtt	Fighter1 Round 4 Significant Strike Attempt	NA	FightMetric
F1R4SigStrPercent	Fighter1 Round 4 Significant Strike Percent	NA	FightMetric
F1R4SubAtt	Fighter1 Round 4 Submission Attempt	NA	FightMetric
F1R4TD	Fighter1 Round 4 Take Down	NA	FightMetric
F1R4TDAtt	Fighter1 Round 4 Take Down Attempt	NA	FightMetric
F1R4TDPercent	Fighter1 Round 4 Take Down Percent	NA	FightMetric
F1R4TotStr	Fighter1 Round 4 Total Strike	NA	FightMetric
F1R4TotStrAtt	Fighter1 Round 4 Total Strike Attempt	NA	FightMetric
F1R5BodyStr	Fighter1 Round 5 BodyStrike	NA	FightMetric
F1R5BodyStrAtt	Fighter1 Round 5 BodyStrike Attempt	NA	FightMetric
F1R5ClinchStr	Fighter1 Round 5 ClinchStrike	NA	FightMetric
F1R5ClinchStrAtt	Fighter1 Round 5 ClinchStrike Attempt	NA	FightMetric
F1R5DistStr	Fighter1 Round 5 Distance Strike	NA	FightMetric
F1R5DistStrAtt	Fighter1 Round 5 Distance Strike Attempt	NA	FightMetric
F1R5GroundStr	Fighter1 Round 5 GroundStrike	NA	FightMetric
F1R5GroundStrAtt	Fighter1 Round 5 GroundStrike Attempt	NA	FightMetric
F1R5HeadStr	Fighter1 Round 5 HeadStrike	NA	FightMetric
F1R5HeadStrAtt	Fighter1 Round 5 HeadStrike Attempt	NA	FightMetric
F1R5KD	Fighter1 Round 5 Knock Down	NA	FightMetric
F1R5LegStr	Fighter1 Round 5 LegStrike	NA	FightMetric
F1R5LegStrAtt	Fighter1 Round 5 LegStrike Attempt	NA	FightMetric
F1R5Pass	Fighter1 Round 5 Guard Pass	NA	FightMetric
F1R5Rev	Fighter1 Round 5 Reversal	NA	FightMetric
F1R5SigStr	Fighter1 Round 5 Significant Strike	NA	FightMetric
F1R5SigStrAtt	Fighter1 Round 5 Significant Strike Attempt	NA	FightMetric
F1R5SigStrPercent	Fighter1 Round 5 Significant Strike Percent	NA	FightMetric
F1R5SubAtt	Fighter1 Round 5 Submission Attempt	NA	FightMetric
F1R5TD	Fighter1 Round 5 Take Down	NA	FightMetric
F1R5TDAtt	Fighter1 Round 5 Take Down Attempt	NA	FightMetric
F1R5TDPercent	Fighter1 Round 5 Take Down Percent	NA	FightMetric
F1R5TotStr	Fighter1 Round 5 Total Strike	NA	FightMetric
F1R5TotStrAtt	Fighter1 Round 5 Total Strike Attempt	NA	FightMetric
F1Rank	Fighter1 Rank	20	FightMatrix
F1Reach	Fighter1 Reach	76	FightMetric
F1Rev	Fighter1 Reversal	0	FightMetric
F1SApM	Fighter1 Strikes Absorbed per Minute	3.8	FightMetric
F1SigStr	Fighter1 Significant Strike	23	FightMetric
F1SigStrAtt	Fighter1 Significant Strike Attempt	47	FightMetric
F1SigStrPercent	Fighter1 Significant Strike Percent	0.48	FightMetric
F1SLpM	Fighter1 Strikes Landed per Minute	3.69	FightMetric
F1Stance	Fighter1 Stance	Orthodox	FightMetric
F1StrAcc	Fighter1 Strike Accuracy	0.42	FightMetric
F1StrDef	Fighter1 Strike Defense	0.61	FightMetric
F1SubAtt	Fighter1 Submission Attempt	0	FightMetric
F1SubAttAvg	Fighter1 Submission Attempt Average	0	FightMetric
F1TD	Fighter1 Take Down	0	FightMetric
F1TDAcc	Fighter1 Take Down Accuracy	0.34	FightMetric
F1TDAtt	Fighter1 Take Down Attempt	0	FightMetric
F1TDAvg	Fighter1 Take Down Average	3	FightMetric
F1TDDef	Fighter1 Take Down Defense	0.91	FightMetric
F1TDPercent	Fighter1 Take Down Percent	0	FightMetric
F1TotStr	Fighter1 Total Strike	40	FightMetric
F1TotStrAtt	Fighter1 Total Strike Attempt	64	FightMetric
F1UFCLosses	Fighter1 UFCLosses	2	FightMetric
F1UFCNC	Fighter1 UFCNo Contest	0	FightMetric
F1UFCWins	Fighter1 UFCWins	5	FightMetric
F1Weight	Fighter1 Weight	205	FightMetric
F1Wins	Fighter1 Wins	1	FightMetric
F2Age	Fighter2 Age	31	FightMetric
F2BodyStr	Fighter2 BodyStrike	3	FightMetric
F2BodyStrAtt	Fighter2 BodyStrike Attempt	5	FightMetric
F2CarFightTime	Fighter2 Career FightTime	401	Derived
F2CarOppSigStrAtt	Fighter2 Career Opponent Significant Strike Attempt	20	Derived
F2CarOppTDAtt	Fighter2 Career Opponent Take Down Attempt	1	Derived
F2CarSigStr	Fighter2 Career Significant Strike	34	Derived
F2CarSigStrAbs	Fighter2 Career Significant Strike Absorbed	3	Derived
F2CarSigStrAtt	Fighter2 Career Significant Strike Attempt	55	Derived
F2CarStrAbspM	Fighter2 Career Strike Absorbed per Minute	0.45	Derived
F2CarStrAcc	Fighter2 Career Strike Accuracy	0.62	Derived
F2CarStrDef	Fighter2 Career Strike Defense	0.15	Derived
F2CarStrLpM	Fighter2 Career Strike Lper Minute	5.09	Derived
F2CarTD	Fighter2 Career Take Down	1	Derived
F2CarTDAbs	Fighter2 Career Take Down Absorbed	0	Derived
F2CarTDAcc	Fighter2 Career Take Down Accuracy	0.2	Derived
F2CarTDAtt	Fighter2 Career Take Down Attempt	5	Derived
F2CarTDAvg	Fighter2 Career Take Down Average	0.5	Derived
F2CarTDDef	Fighter2 Career Take Down Defense	0	Derived
F2ClinchStr	Fighter2 ClinchStrike	4	FightMetric
F2ClinchStrAtt	Fighter2 ClinchStrike Attempt	6	FightMetric
F2DistStr	Fighter2 Distance Strike	10	FightMetric
F2DistStrAtt	Fighter2 Distance Strike Attempt	31	FightMetric
F2DOB	Fighter2 DOB	1978-02-09	FightMetric
F2GroundStr	Fighter2 GroundStrike	0	FightMetric
F2GroundStrAtt	Fighter2 GroundStrike Attempt	0	FightMetric
F2HeadStr	Fighter2 HeadStrike	11	FightMetric
F2HeadStrAtt	Fighter2 HeadStrike Attempt	32	FightMetric
F2Height	Fighter2 Height	72	FightMetric
F2KD	Fighter2 Knock Down	0	FightMetric
F2LegStr	Fighter2 LegStrike	0	FightMetric
F2LegStrAtt	Fighter2 LegStrike Attempt	0	FightMetric
F2Pass	Fighter2 Guard Pass	0	FightMetric
F2ProLosses	Fighter2 ProLosses	6	FightMetric
F2ProNC	Fighter2 ProNo Contest	0	FightMetric
F2ProWins	Fighter2 ProWins	14	FightMetric
F2R1BodyStr	Fighter2 Round 1 BodyStrike	3	FightMetric
F2R1BodyStrAtt	Fighter2 Round 1 BodyStrike Attempt	5	FightMetric
F2R1ClinchStr	Fighter2 Round 1 ClinchStrike	4	FightMetric
F2R1ClinchStrAtt	Fighter2 Round 1 ClinchStrike Attempt	6	FightMetric
F2R1DistStr	Fighter2 Round 1 Distance Strike	10	FightMetric
F2R1DistStrAtt	Fighter2 Round 1 Distance Strike Attempt	31	FightMetric
F2R1GroundStr	Fighter2 Round 1 GroundStrike	0	FightMetric
F2R1GroundStrAtt	Fighter2 Round 1 GroundStrike Attempt	0	FightMetric
F2R1HeadStr	Fighter2 Round 1 HeadStrike	11	FightMetric
F2R1HeadStrAtt	Fighter2 Round 1 HeadStrike Attempt	32	FightMetric
F2R1KD	Fighter2 Round 1 Knock Down	0	FightMetric
F2R1LegStr	Fighter2 Round 1 LegStrike	0	FightMetric
F2R1LegStrAtt	Fighter2 Round 1 LegStrike Attempt	0	FightMetric
F2R1Pass	Fighter2 Round 1 Guard Pass	0	FightMetric
F2R1Rev	Fighter2 Round 1 Reversal	0	FightMetric
F2R1SigStr	Fighter2 Round 1 Significant Strike	14	FightMetric
F2R1SigStrAtt	Fighter2 Round 1 Significant Strike Attempt	37	FightMetric
F2R1SigStrPercent	Fighter2 Round 1 Significant Strike Percent	0.37	FightMetric
F2R1SubAtt	Fighter2 Round 1 Submission Attempt	0	FightMetric
F2R1TD	Fighter2 Round 1 Take Down	0	FightMetric
F2R1TDAtt	Fighter2 Round 1 Take Down Attempt	4	FightMetric
F2R1TDPercent	Fighter2 Round 1 Take Down Percent	0	FightMetric
F2R1TotStr	Fighter2 Round 1 Total Strike	14	FightMetric
F2R1TotStrAtt	Fighter2 Round 1 Total Strike Attempt	37	FightMetric
F2R2BodyStr	Fighter2 Round 2 BodyStrike	NA	FightMetric
F2R2BodyStrAtt	Fighter2 Round 2 BodyStrike Attempt	NA	FightMetric
F2R2ClinchStr	Fighter2 Round 2 ClinchStrike	NA	FightMetric
F2R2ClinchStrAtt	Fighter2 Round 2 ClinchStrike Attempt	NA	FightMetric
F2R2DistStr	Fighter2 Round 2 Distance Strike	NA	FightMetric
F2R2DistStrAtt	Fighter2 Round 2 Distance Strike Attempt	NA	FightMetric
F2R2GroundStr	Fighter2 Round 2 GroundStrike	NA	FightMetric
F2R2GroundStrAtt	Fighter2 Round 2 GroundStrike Attempt	NA	FightMetric
F2R2HeadStr	Fighter2 Round 2 HeadStrike	NA	FightMetric
F2R2HeadStrAtt	Fighter2 Round 2 HeadStrike Attempt	NA	FightMetric
F2R2KD	Fighter2 Round 2 Knock Down	NA	FightMetric
F2R2LegStr	Fighter2 Round 2 LegStrike	NA	FightMetric
F2R2LegStrAtt	Fighter2 Round 2 LegStrike Attempt	NA	FightMetric
F2R2Pass	Fighter2 Round 2 Guard Pass	NA	FightMetric
F2R2Rev	Fighter2 Round 2 Reversal	NA	FightMetric
F2R2SigStr	Fighter2 Round 2 Significant Strike	NA	FightMetric
F2R2SigStrAtt	Fighter2 Round 2 Significant Strike Attempt	NA	FightMetric
F2R2SigStrPercent	Fighter2 Round 2 Significant Strike Percent	NA	FightMetric
F2R2SubAtt	Fighter2 Round 2 Submission Attempt	NA	FightMetric
F2R2TD	Fighter2 Round 2 Take Down	NA	FightMetric
F2R2TDAtt	Fighter2 Round 2 Take Down Attempt	NA	FightMetric
F2R2TDPercent	Fighter2 Round 2 Take Down Percent	NA	FightMetric
F2R2TotStr	Fighter2 Round 2 Total Strike	NA	FightMetric
F2R2TotStrAtt	Fighter2 Round 2 Total Strike Attempt	NA	FightMetric
F2R3BodyStr	Fighter2 Round 3 BodyStrike	NA	FightMetric
F2R3BodyStrAtt	Fighter2 Round 3 BodyStrike Attempt	NA	FightMetric
F2R3ClinchStr	Fighter2 Round 3 ClinchStrike	NA	FightMetric
F2R3ClinchStrAtt	Fighter2 Round 3 ClinchStrike Attempt	NA	FightMetric
F2R3DistStr	Fighter2 Round 3 Distance Strike	NA	FightMetric
F2R3DistStrAtt	Fighter2 Round 3 Distance Strike Attempt	NA	FightMetric
F2R3GroundStr	Fighter2 Round 3 GroundStrike	NA	FightMetric
F2R3GroundStrAtt	Fighter2 Round 3 GroundStrike Attempt	NA	FightMetric
F2R3HeadStr	Fighter2 Round 3 HeadStrike	NA	FightMetric
F2R3HeadStrAtt	Fighter2 Round 3 HeadStrike Attempt	NA	FightMetric
F2R3KD	Fighter2 Round 3 Knock Down	NA	FightMetric
F2R3LegStr	Fighter2 Round 3 LegStrike	NA	FightMetric
F2R3LegStrAtt	Fighter2 Round 3 LegStrike Attempt	NA	FightMetric
F2R3Pass	Fighter2 Round 3 Guard Pass	NA	FightMetric
F2R3Rev	Fighter2 Round 3 Reversal	NA	FightMetric
F2R3SigStr	Fighter2 Round 3 Significant Strike	NA	FightMetric
F2R3SigStrAtt	Fighter2 Round 3 Significant Strike Attempt	NA	FightMetric
F2R3SigStrPercent	Fighter2 Round 3 Significant Strike Percent	NA	FightMetric
F2R3SubAtt	Fighter2 Round 3 Submission Attempt	NA	FightMetric
F2R3TD	Fighter2 Round 3 Take Down	NA	FightMetric
F2R3TDAtt	Fighter2 Round 3 Take Down Attempt	NA	FightMetric
F2R3TDPercent	Fighter2 Round 3 Take Down Percent	NA	FightMetric
F2R3TotStr	Fighter2 Round 3 Total Strike	NA	FightMetric
F2R3TotStrAtt	Fighter2 Round 3 Total Strike Attempt	NA	FightMetric
F2R4BodyStr	Fighter2 Round 4 BodyStrike	NA	FightMetric
F2R4BodyStrAtt	Fighter2 Round 4 BodyStrike Attempt	NA	FightMetric
F2R4ClinchStr	Fighter2 Round 4 ClinchStrike	NA	FightMetric
F2R4ClinchStrAtt	Fighter2 Round 4 ClinchStrike Attempt	NA	FightMetric
F2R4DistStr	Fighter2 Round 4 Distance Strike	NA	FightMetric
F2R4DistStrAtt	Fighter2 Round 4 Distance Strike Attempt	NA	FightMetric
F2R4GroundStr	Fighter2 Round 4 GroundStrike	NA	FightMetric
F2R4GroundStrAtt	Fighter2 Round 4 GroundStrike Attempt	NA	FightMetric
F2R4HeadStr	Fighter2 Round 4 HeadStrike	NA	FightMetric
F2R4HeadStrAtt	Fighter2 Round 4 HeadStrike Attempt	NA	FightMetric
F2R4KD	Fighter2 Round 4 Knock Down	NA	FightMetric
F2R4LegStr	Fighter2 Round 4 LegStrike	NA	FightMetric
F2R4LegStrAtt	Fighter2 Round 4 LegStrike Attempt	NA	FightMetric
F2R4Pass	Fighter2 Round 4 Guard Pass	NA	FightMetric
F2R4Rev	Fighter2 Round 4 Reversal	NA	FightMetric
F2R4SigStr	Fighter2 Round 4 Significant Strike	NA	FightMetric
F2R4SigStrAtt	Fighter2 Round 4 Significant Strike Attempt	NA	FightMetric
F2R4SigStrPercent	Fighter2 Round 4 Significant Strike Percent	NA	FightMetric
F2R4SubAtt	Fighter2 Round 4 Submission Attempt	NA	FightMetric
F2R4TD	Fighter2 Round 4 Take Down	NA	FightMetric
F2R4TDAtt	Fighter2 Round 4 Take Down Attempt	NA	FightMetric
F2R4TDPercent	Fighter2 Round 4 Take Down Percent	NA	FightMetric
F2R4TotStr	Fighter2 Round 4 Total Strike	NA	FightMetric
F2R4TotStrAtt	Fighter2 Round 4 Total Strike Attempt	NA	FightMetric
F2R5BodyStr	Fighter2 Round 5 BodyStrike	NA	FightMetric
F2R5BodyStrAtt	Fighter2 Round 5 BodyStrike Attempt	NA	FightMetric
F2R5ClinchStr	Fighter2 Round 5 ClinchStrike	NA	FightMetric
F2R5ClinchStrAtt	Fighter2 Round 5 ClinchStrike Attempt	NA	FightMetric
F2R5DistStr	Fighter2 Round 5 Distance Strike	NA	FightMetric
F2R5DistStrAtt	Fighter2 Round 5 Distance Strike Attempt	NA	FightMetric
F2R5GroundStr	Fighter2 Round 5 GroundStrike	NA	FightMetric
F2R5GroundStrAtt	Fighter2 Round 5 GroundStrike Attempt	NA	FightMetric
F2R5HeadStr	Fighter2 Round 5 HeadStrike	NA	FightMetric
F2R5HeadStrAtt	Fighter2 Round 5 HeadStrike Attempt	NA	FightMetric
F2R5KD	Fighter2 Round 5 Knock Down	NA	FightMetric
F2R5LegStr	Fighter2 Round 5 LegStrike	NA	FightMetric
F2R5LegStrAtt	Fighter2 Round 5 LegStrike Attempt	NA	FightMetric
F2R5Pass	Fighter2 Round 5 Guard Pass	NA	FightMetric
F2R5Rev	Fighter2 Round 5 Reversal	NA	FightMetric
F2R5SigStr	Fighter2 Round 5 Significant Strike	NA	FightMetric
F2R5SigStrAtt	Fighter2 Round 5 Significant Strike Attempt	NA	FightMetric
F2R5SigStrPercent	Fighter2 Round 5 Significant Strike Percent	NA	FightMetric
F2R5SubAtt	Fighter2 Round 5 Submission Attempt	NA	FightMetric
F2R5TD	Fighter2 Round 5 Take Down	NA	FightMetric
F2R5TDAtt	Fighter2 Round 5 Take Down Attempt	NA	FightMetric
F2R5TDPercent	Fighter2 Round 5 Take Down Percent	NA	FightMetric
F2R5TotStr	Fighter2 Round 5 Total Strike	NA	FightMetric
F2R5TotStrAtt	Fighter2 Round 5 Total Strike Attempt	NA	FightMetric
F2Rank	Fighter2 Rank	51	FightMatrix
F2Reach	Fighter2 Reach	72	FightMetric
F2Rev	Fighter2 Reversal	0	FightMetric
F2SApM	Fighter2 Strikes Absorbed per Minute	2.26	FightMetric
F2SigStr	Fighter2 Significant Strike	14	FightMetric
F2SigStrAtt	Fighter2 Significant Strike Attempt	37	FightMetric
F2SigStrPercent	Fighter2 Significant Strike Percent	0.37	FightMetric
F2SLpM	Fighter2 Strikes Landed per Minute	3.12	FightMetric
F2Stance	Fighter2 Stance	Orthodox	FightMetric
F2StrAcc	Fighter2 Strike Accuracy	0.51	FightMetric
F2StrDef	Fighter2 Strike Defense	0.57	FightMetric
F2SubAtt	Fighter2 Submission Attempt	0	FightMetric
F2SubAttAvg	Fighter2 Submission Attempt Average	0.6	FightMetric
F2TD	Fighter2 Take Down	0	FightMetric
F2TDAcc	Fighter2 Take Down Accuracy	0.29	FightMetric
F2TDAtt	Fighter2 Take Down Attempt	4	FightMetric
F2TDAvg	Fighter2 Take Down Average	3.17	FightMetric
F2TDDef	Fighter2 Take Down Defense	0.57	FightMetric
F2TDPercent	Fighter2 Take Down Percent	0	FightMetric
F2TotStr	Fighter2 Total Strike	14	FightMetric
F2TotStrAtt	Fighter2 Total Strike Attempt	37	FightMetric
F2UFCLosses	Fighter2 UFCLosses	0	FightMetric
F2UFCNC	Fighter2 UFCNo Contest	0	FightMetric
F2UFCWins	Fighter2 UFCWins	2	FightMetric
F2Weight	Fighter2 Weight	185	FightMetric
Fighter1	Fighter1	Matt Hamill	FightMetric
Fighter2	Fighter2	Mark Munoz	FightMetric
FightNumber	FightNumber	3449	FightMetric
Gender	Gender	Men	FightMetric
Judge1	Judge1	NA	FightMetric
Judge1Score	Judge1Score	NA	FightMetric
Judge2	Judge2	NA	FightMetric
Judge2Score	Judge2Score	NA	FightMetric
Judge3	Judge3	NA	FightMetric
Judge3Score	Judge3Score	NA	FightMetric
Method	Method	KO/TKO	FightMetric
Ref	Referee	Dan Miragliotta	FightMetric
Result	Result	Matt Hamill	FightMetric
Round	Round	1	FightMetric
Time	Time	3:53	FightMetric
TotFightTime	Total FightTime	233	FightMetric
WeightClass	WeightClass	Light Heavyweight	FightMetric

# Creating table to explain the data.
data_table_fighters <- data.frame(colnames(fightMetric_fighters_df),
                         rep(NA,length(fightMetric_fighters_df)),
                         as.character(fightMetric_fighters_df[fightMetric_fighters_df$Name=="Gegard Mousasi",]),
                         rep("FightMetric",length(fightMetric_fighters_df)),
                         stringsAsFactors = FALSE)
# Assign column names.
colnames(data_table_fighters) <- c("Variables", "Unabreviated", "Example", "Source")
# Generate unabreviated variable names.
data_table_fighters$Unabreviated <- data_table_fighters$Variables %>%
  gsub("Str", "Strike ", .) %>%
  gsub("SL", "Strikes Landed ", .) %>%
  gsub("pM", "per Minute ", .) %>%
  gsub("Acc", "Accuracy ", .) %>%
  gsub("Def", "Defense ", .) %>%
  gsub("Avg", "Average ", .) %>%
  gsub("SA", "Strikes Absorbed", .) %>%
  gsub("Sub", "Submission ", .) %>%
  gsub("Att", "Attempt ", .) %>%
  gsub("TD", "Take Down ", .)
kable(data_table_fighters, caption = "Composition of the primary datset, [ufc_fight_details_df]", align = 'l')

Variables	Unabreviated	Example	Source
Name	Name	Gegard Mousasi	FightMetric
ProWins	ProWins	39	FightMetric
ProLosses	ProLosses	6	FightMetric
ProNC	ProNC	2	FightMetric
Height	Height	74	FightMetric
Weight	Weight	185	FightMetric
Reach	Reach	76	FightMetric
Stance	Stance	Orthodox	FightMetric
DOB	DOB	1985-08-01	FightMetric
SLpM	Strikes Landed per Minute	3.56	FightMetric
StrAcc	Strike Accuracy	0.5	FightMetric
SApM	Strikes Absorbedper Minute	1.18	FightMetric
StrDef	Strike Defense	0.69	FightMetric
TDAvg	Take Down Average	1.6	FightMetric
TDAcc	Take Down Accuracy	0.65	FightMetric
TDDef	Take Down Defense	0.61	FightMetric
SubAttAvg	Submission Attempt Average	1.2	FightMetric
Age	Age	31	FightMetric