You are browsing as a guest. Sign up (or log in) to start making projects!

FIFA 26

  • 5 Devlogs
  • 9 Total hours

model which predicts all matchups for this year's world cup

Ship #2

This is the second ship on my project, if you didn’t review the first one, it’s a world cup simulator made using a Random Forest model and Poisson distribution model to generate scores. All matches were simulated locally to generate a consensus result base viewable across the site. The new feature, reason for reship, was the knockout matchups starting and now there is an interactive bracket so you can simulate your own games but also keep real results once they occur. The site is updated daily, either at the end of the day when matches are complete, or the next morning so expect changes depending on when you view the site. A challenging aspect this time around was organizing the bracket structure such that the teams flow downward when they win. Two things to note, the site does take a while to load the first time as it needs to gather all the data and simulation function, but there should be no other lag or problems once active. Lastly, though the third place matchup doesn’t show up at first, it appears after both semifinal matchups are simulated. Enjoy!

  • 2 devlogs
  • 4h
  • 14.80x multiplier
  • 65 Stardust
Try project → See source code →
Open comments for this post

39m 16s logged

First: Thank you everyone for their reviews on my first ship, this will likely be the last devlog for this project as it is a time-based project(can’t really update after the world cup ends) but I appreciate all the feedback. As for this current devlog, there was a small fix which needed to be done, I originally said on my last post that after a matchup occurs in real-time, it is locked and can’t be simulated anymore. This is not the case anymore and you have the option of locking/unlocking games that have occurred in real time already.

0
0
4
Open comments for this post

3h 42m 54s logged

The group stage just completed today, but luckily the knockout functionality was essentially all set to go. After inputting the matchups, the site now has new knockout features equipped. There is an interactive bracket, to simulate your own matchups that includes penalty shootouts, along with a simple page to display advancement probabilities to each round. Another ship should be out shortly to reflect these new changes.

0
0
3
Ship #1

I made a project in honor of the FIFA 2026 World Cup. It simulates all 72 group stage matches, consolidates them into easy to read data, viewable on the website link, and allows you to simulate your own matchups as well. The link is live at: https://fifa-26.streamlit.app/. Challenges occurred in this project in how the model was trained to simulate games with accuracy, converting the python script into a viewable website, and ensuring results were realistic and presentable. I am most proud of how users are able to make their own simulations within the website, but also be able to view a general view of data that was done through numerous simulations. For example, advancement probabilities for each team at each position in their group. Note, when opening the link for the first time, it may take a while to load as it needs to grab the data through the GitHub repo, but after loading, it shouldn’t have many delays.

  • 3 devlogs
  • 4h
  • 15.16x multiplier
  • 67 Stardust
Try project → See source code →
Open comments for this post

1h 44m 9s logged

This might be the last devlog for a bit, but the project is officially live with the help of streamlit as this is a python project. The link is already live, so you can simulate your own matches and see what the consensus result was through multiple simulations. This project will be shipped soon but updated again for when the knockout stage matchups are set in stone. Attached is how it looks on the main page.

0
0
4
Open comments for this post

1h 35m 39s logged

Update: I added the FIFA rankings as another metric to simulate the games. I also strayed away from using the poisson model to predict outcome and now only use it for scoring purposes as the random forest model is much more accurate. Attached are the updated probabilities.

0
0
5
Open comments for this post

1h 5m 49s logged

So I recently decided to train a model to help predict the group stage matches for the FIFA world cup. It’s a combination of a Random Forest Model and Poisson Distribution to help predict the games. Currently I’m only training the model based off of a teams ELO and recent form(performance in past 5 games), but plan on adding FIFA’s point system as another metric. I attached the most common match result of each game after running the simulation 1000 times and the probability of where each team finishes this year(first, second, third qual., third elim., and fourth)

0
0
3

Delete project?

Are you sure you want to permanently delete this project? This action cannot be undone.

All devlogs, followers, and associated data will be removed.

Followers

Loading…