Excel LADZ - Weekly Update: Monday 24 July

G’day lads, Part 1 of the NBA In-Season Tournament model has been released onto YouTube. Part 2, which includes simulating the knockout phase of the tournament, will be published within the next couple days. If you have any questions about the model (already available on the Patreon page), don’t hesitate to post them on this post or on Discord!

2023 Women’s World Cup

I’ve created a model to predict/simulate the knockout stages of the 2023 FIFA Women’s World Cup. I’m adding some aesthetic “final touches” to the model, which will be uploaded to Patreon to view in the days before the knockout phase begins. Below was my process lads:

1. From the Official FIFA Website, I recorded each nation’s Ranking Score. As a sort of ELO Rating, this score is a measure of how strong each national side is; e.g. the United States has a score of 2090.03, while Jamaica has a rating of 1536.81.
2. Using these scores it is possible to obtain a win percentage for a team in a matchup. Taking the USA vs Jamaica example, according to the formula below, the USA has a 96.03% chance of winning. That is, progressing to the next stage.

1. I then made a 16 team knockout bracket. Beginning with the Round of 16, successful teams that “won” their simulated matchup (a random decimal decided the winner of each match) progressed to the Quarter Finals, then the Semi Finals, and then a winner was determined in the World Cup Final.
2. This single simulation was then run 1,000 times using a Data Table with Excel’s What-If Analysis within the ‘Data’ tab.
3. The simulated results were then compiled and analysed to check each team’s chance of progressing past each stage and ultimately winning the tournament.

Note: Exactly the same process can be done with a Men’s competition, however it involves a slightly different formula which can be found on the Fifa website or even on Wikipedia.

Australian Football League (AFL) Model.

As you may have noticed, an AFL model has been published on the Patreon page. While not totally completed (sports betting features, venue options, and player values have yet to be added), this model demonstrates an important concept in sports modelling.

In an AFL game there are two ways to score points:

• Goal: worth 6 points
• Behind: worth 1 point

As a result, an Expected Score of 97 cannot be treated as 97 “individual scores”. Rather, it must be simulated as a combination of both goals and behinds: the two scoring options. This is to avoid inflating a team’s probability of winning.

Consider a match where Team 1 is expected to score 105 points, to Team 2’s 76. Taking this to be 105 scores vs 76 scores would give Team 2 virtually no chance to win the match. This is obviously untrue.

Rather, the 105 points may be composed of 16 goals and 9 points (25 scores), versus 12 goals and 4 points (16 scores). Simulating 16 vs 12 expected goals (as well as the behinds separately) will result in a more accurate reflection of a team’s probability of winning.

This is similar to the NBA model, whereby the final score is calculated by summing the simulated score from Free Throws, 2 Pointers and 3 Pointers.

Short Term Plans

• EPL Model
  • With the English Premier League Season starting in early August, I am planning to release a string of models. These include predicting the whole season, halftime markets, corner markets, and much more.

If you have any questions lads, leave a comment here or on Discord! I look forward to hearing about your ideas and/or questions!