There are many ways to parametrise the Bradley-Terry model. This blog uses the odds parametrisation, which is perhaps the easiest to interpret.
More specifically, each team is assigned a “skill level” that is estimated from head-to-head results. These skill levels are reported on a relative odds scale – that is, the bottom placed team gets a default rating of 1.00 and every other team is assigned a relative rating that determines their odds of winning.
For example, at the end of the 2018 regular season the Eels were the bottom placed team with a skill level of 1.00. The top skilled team were the Sharks with an estimated skill level of 7.39. If the Sharks were to play the Eels in the upcoming week, then the odds of the Sharks defeating the Eels would be 7.39 to 1, or 88% confidence.
Similarly, the Broncos finished the 2018 regular season with an estimated skill level of 6.69, while the Dragons finished with a skill level of 6.06. Thus, if the Broncos were to play the Dragons in the upcoming week, then the odds of the Broncos winning would be 6.69 to 6.06, which is the same as 1.10 to 1, or 52% confidence.
Finally, homeground advantage for the 2018 season was estimated to increase the home team’s skill level by 1.34 times. Thus, if the Broncos were to play at home then their skill level would increase to 8.97 (1.34 times 6.69). The odds of defeating the Dragons would then be 8.97 to 6.06, which is the same as 1.48 to 1, or 59% confidence.
cross ✛ validated 