Skip to main content

​
Overview

All data is stored in a single flat CSV file (conventionally named sumo.csv). Each row represents one match between two wrestlers. Rows with a known outcome are used as training and evaluation data; rows with a blank result column are the undecided matches the models are asked to predict.

​
Column Schema

ColumnTypeDescriptionExample
weight1numericWeight of wrestler 1 (lbs)401
weight2numericWeight of wrestler 2 (lbs)379
wins1numericNumber of wins for wrestler 16
wins2numericNumber of wins for wrestler 25
age1numericAge of wrestler 1 (years)35
age2numericAge of wrestler 2 (years)31
height1numericHeight of wrestler 1 (cm)190
height2numericHeight of wrestler 2 (cm)184
resultlogical/numericMatch outcome: 1 = wrestler 1 wins, 0 = wrestler 2 wins, blank = undecided1
There are no additional columns. The result column is always the last (rightmost) column.

​
Example Data

The following is a representative excerpt from sumo.csv: 
weight1,weight2,wins1,wins2,age1,age2,height1,height2,result
401,379,6,5,35,31,190,184,1
344,370,8,11,36,34,189,192,1
351,245,3,2,22,27,189,169,0
335,291,3,12,32,30,177,176,0
375,306,0,0,35,28,186,185,0
326,295,1,2,29,30,173,174,0
355,375,4,7,25,27,176,185,1
311,328,2,2,22,31,185,183,0
386,333,2,5,37,28,187,187,1
289,366,2,1,27,29,180,184,1
To represent an undecided match (one you want to predict), leave the result field blank: 
weight1,weight2,wins1,wins2,age1,age2,height1,height2,result
370,355,4,3,28,26,183,180,

​
Encoding Rules

​
Decided matches

result must be either 1 or 0:
ValueMeaning
1Wrestler 1 wins
0Wrestler 2 wins

​
Undecided matches

Leave the result field empty (trailing comma, no value). read.csv() will import this as NA. After sumo.read() processes the file, these rows have result = NA and are selected with: 
undecided <- filter(data, is.na(result))
A blank result field and a missing result field are equivalent in CSV — both are imported as NA by R’s read.csv().

​
Validation Rules

Before loading the file with sumo.read(), confirm the following:
  • All eight feature columns (weight1, weight2, wins1, wins2, age1, age2, height1, height2) must contain numeric values — no text, no empty cells.
  • result must be 1, 0, or blank. Any other value (e.g. "win", 2) will be coerced to NA by as.numeric() and the row will be treated as undecided.
  • The header row must be present and column names must match exactly (case-sensitive).
  • The file must use standard CSV encoding (comma-delimited, UTF-8 or ASCII).
If a feature column contains a non-numeric value, mutate_all(as.numeric) will silently coerce it to NA, which causes glm() and neuralnet() to drop that row or error. Inspect your data with summary(data) after loading to catch unexpected NA values.

​
How sumo.read() Processes the File

sumo.read() performs three sequential transformations: 
sumo.read <- function(csv) {
  data <- tibble(read.csv(csv)) %>%   # 1. Read CSV into a tibble
    mutate_all(as.numeric) %>%         # 2. Coerce every column to numeric
    mutate(result = as.logical(result)) # 3. Cast result: 1→TRUE, 0→FALSE, NA→NA
  return(data)
}
StepOperationEffect on result
1read.csv()"1" → 1, "" → NA (character)
2mutate_all(as.numeric)"1" → 1.0, blank already NA
3mutate(result = as.logical(result))1 → TRUE, 0 → FALSE, NA → NA
After loading, the caller is expected to split the tibble into decided and undecided subsets: 
data      <- sumo.read("sumo.csv")
history   <- na.omit(data)                  # decided matches  → train/evaluate models
undecided <- filter(data, is.na(result))    # undecided matches → run inference
Keep decided and undecided matches in the same file. The split is done in code, so you only need to maintain one CSV. Append a new row with a blank result whenever you want a new prediction.

Build docs developers (and LLMs) love

Get started for freeTalk to us