18 comments

  • bArray 17 minutes ago
    I've updated the magic weights, and I too can get the result I want:

        WEIGHTS = {
          'w_xg': 0.09,
          'w_goals': -0.07,
          'w_star': 0.018,
          'w_value': 0.18,
          'w_rank': 0.4,
          'w_def': -0.12,
          'xga_share': 0.85,
          'w_gk': 0.0042
        }
    
        $ python3 worldcup_model.py --sims 100000
    
        2026 FIFA World Cup -- championship probabilities (100,000 simulations, from Round of 32)
    
         1. England                 11.7% *
         2. France                  10.0% *
         3. Spain                    9.3% *
         4. Argentina                8.4%
         5. Germany                  8.1%
    
    It's coming home!
  • immighelper 1 hour ago
    Probably the human pseudonym of Paul the Octopus (https://en.wikipedia.org/wiki/Paul_the_Octopus).
  • walthamstow 1 hour ago
    It's worth noting that there has only been 24 world cups
  • bArray 43 minutes ago
    I've been messing with the magic value weights, and it doesn't take too much to push them in any given direction. The TEAMS_2026 should really be taken with a pinch of salt.
    • daft_pink 30 minutes ago
      Isn't that essentially how AI works?
  • artur_makly 16 minutes ago
    There is no MODEL for pure Argentine magic. vamos carajos! vamooooooooo con todo!!! Our time has come again.
  • xiaodai 1 hour ago
    but does the model predict the right match up along way? if not it's just wrong wrong make a right
    • jonwinstanley 2 minutes ago
      Also, being able to predict football with complete accuracy is impossible
  • derdi 1 hour ago
    > Applied prospectively to the in-progress 2026 World Cup from the Round of 32, the model identifies Argentina (28.0%) and Spain (21.1%) as the leading championship candidates.

    Seems weird to wait to run the "prospective" simulation until the World Cup is already in progress. Although it seems that the model also needs to use "the actual bracket and group-stage performance". So it's not prospective?

    • swiftcoder 1 hour ago
      It predicts likely winners based on the round of 32 performance (plus prior data). That's still "prospective" with respect to the finals
      • glimshe 1 hour ago
        Which is very reasonable. You estimate odds after seeing teams playing with the actual squad selection at that period in time. Otherwise I'd dismiss the predictions as lucky guesses in a row.
      • derdi 1 hour ago
        Yes. I don't like phrasing this as being prospective for the World Cup as a whole. It's for the knockout stage. (Which the abstract says! But the title doesn't.)
    • kunxue 49 minutes ago
      The baseline matters here: favorites win World Cups all the time. How often would "always pick the two pre-tournament favorites" have gotten the champion in these same 10 tournaments? Without that comparison, 10/10 tells us basically nothing.
      • decimalenough 35 minutes ago
        France was far and away the pre-tournament favorite for 2026, if anything it's somewhat impressive that OP's model correctly predicted that they wouldn't make it.

        Here's hoping they were right for England as well, but we'll find out soon enough.

  • glenpierce 54 minutes ago
    Survivor bias
  • amazingamazing 47 minutes ago
    Why is the world cup so infrequent anyway? I assume to match olympics?

    Good models need a lot of data. Can you really be accurate with what, 30 data points, in which the team composition is basically reset each time?

    • toyg 16 minutes ago
      World Cups have to alternate with continental competitions (Copa America, Asian Cup, European Cup, Africa Cup of Nations) which are on similar cycles. They could technically be held every two years (and current FIFA leadership is pushing towards that), but federations and clubs are resistant (because every summer tournament places even more stress on an already-long club season, and it would likely devalue other competitions).
      • jonwinstanley 0 minutes ago
        Every 4 years is good. Makes it a special event
    • hn937758 35 minutes ago
      Clubs want their star players focused on club championships, and the star players make their real money playing on their club teams.
      • sigbottle 23 minutes ago
        do world cup athletes get a big bonus for world cup participation from their govts and/or FIFA?
        • otherme123 4 minutes ago
          It is up to their respective Football Federation (Spain: RFEF, England: FA, Germany: DFB, etc.), but FIFA pays each Federation a bonus: https://intereconomia.com/wp-content/uploads/2026/07/image-2...

          In the average country players agree the bonus conditions with their Federation.

        • toyg 12 minutes ago
          FIFA pays federations on a result basis. It's then up to the individual federations to redistribute that money as you see fit. As you can imagine, a lot of that does not end up in the athletes' pockets... I believe clubs get some (low) compensation for injuries, but it's also common knowledge that players who play deep into summer will end up performing poorly the following season.
  • mikelward 1 hour ago
    > the model identifies Argentina (28.0%) and Spain (21.1%) as the leading championship candidates
  • mcphage 1 hour ago
    How does this paper not even mention the word "overfitting"?
    • dmurray 1 hour ago
      The abstract does say

      > limitations, principally the small number of tournaments available for validation and the risk of in-sample weight selection

      But I agree this model is no more valuable than Paul the Octopus.

      • mcphage 1 hour ago
        That almost makes it worse—like they're vaguely aware that training too heavily on too small a data set makes badly trained models, but are unaware that it has a name and is an actual identified problem.
        • notahacker 11 minutes ago
          They've even described how they overfitted it! For five world cups, a simple model based on ranking and goal difference in the group stages[1] predicted it four times, so they invented a somewhat subjective variable on defensive strength to cover the teams that didn't score much...

          [1]yes, both of those are endogenous variables...

  • dwedge 1 hour ago
    A good AI would calculate refereeing decisions and put Argentina at 100% unless England can pull off a miracle against FIFA today.
    • fau 1 hour ago
      No good AI would cherry-pick data to this extent. Only people are capable of that
  • yardie 1 hour ago
    Does the model account for the blatant favouritism in the refs? We used to laugh about it before but as the cameras have gotten better it has become a lot more visible. And in this case, is turning the tournament into a bit of a joke.

    -- Egypt was robbed.

    • jacknews 53 minutes ago
      The FIFA-planned game schedule is also surprising. Argentina have not had to face a single 'big' contender through the tournament, until now.
      • notahacker 1 minute ago
        It's not at all surprising: the seeded the winners and the three highest ranked teams to make it impossible for them to meet until this stage if they won their group in the group stage, with the group stage having its own seeding system to make it very winnable for them.

        They also missed a potentially tricky first knockout round tie against local rivals Uruguay because Uruguay underperformed and Cape Verde unexpectedly overperformed.

    • bflesch 1 hour ago
      It's quite unlikely soccer is one of the few sports without a doping problem and with only very few cases where the referee was paid off.

      Since ancient times in Rome where they said "bread and games" are needed to keep the commoners happy, many generations of rulers had time to optimize large-scale sports events.

      My personal theory is that these kind of extremely unfair decisions in soccer are a net benefit to stability of society, and there's no incentive for the leadership to aim for full fairness in sports.

      Hear me out: When a team loses in unfair manner due to bad decisions of the referee, large masses of people feel the psychological pain of being robbed of a win. This feeling of "unfairness" makes the masses more resilient to experiencing "unfairness" in their day-to-day life, for example when a billionaire is not prosecuted in the same way than a common person.

      If we turn the logic around and assume that soccer would always be perfectly fair, then the masses would demand the same kind of fairness also in their day-to-day lives. Obviously this demand for fairness is not aligned with an establishment class that wants to extract the maximum value possible from their citizens, and push as far as they can without risking stability of the country.

      From an establishment perspective, it makes a lot of sense to condition the masses for "unfairness", and sports is the perfect way to do it. I'm not saying that the individual referees are paid off to let a certain country win, just that the establishment who runs each country (and thereby also run international sports organizations like FIFA) have no incentive to actually create total fairness.

      This might also explain issues like the IOC re-instating russia for olympic games, even though they have not retreated from Ukrainian territory yet. It triggers people who strongly feel about morals and ethics, and it brings the point home that the world is unfair and it makes no sense to push for fairness in the greater context.

      The benefit is psychological conditioning for people to accept unfairness.

  • xiaodai 1 hour ago
    how statistic significant is it.
  • dwedge 1 hour ago
    !remindme tomorrow
  • pestatije 1 hour ago
    and how many models did you model?
  • malthaus 1 hour ago
    now back your claim with money and bet accordingly on betting sites to see if you uncovered some actual alpha here
  • killingtime74 1 hour ago
    "They've done studies, you know. 60% of the time, it works every time." - Brian Fantana, Anchorman
    • anamax 1 minute ago
      One of the founders of Renaissance Technologies, which runs some of the most successful quant funds of all time, said ""We’re right 50.75% of the time... but we’re 100% right 50.75% of the time."