What the FIFA rankings looked like going into the 1998 tournament
The official FIFA Men's Ranking system launched in December 1992, so by mid-1998 we had roughly five and a half years of history behind it. That was nowhere near enough to smooth out the noise, especially for teams that didn't play each other regularly. The June 1998 ranking—the one that mattered for seeding the World Cup draw held in December 1997—told a story most people still get wrong today. The top ten reads like a list of people who knew how to play their region. Brazil sat at number one. Netherlands was second. Italy third. But then you start looking at who slipped through the cracks, and the real picture shows up. Croatia ranked outside the top fifteen before the tournament. Denmark hovered around seventeenth. Norway was twentieth. None of those teams were written off by analysts—everyone had seen them play—but the ranking formulas don't really reward a two-year upward trajectory unless you're beating top-half teams consistently. Croatia had just won the 1996 European Championship semifinal and reached the 1998 World Cup qualifiers group stage with a strong squad. They were ranked low because the ranking only looked at results from the previous four years, and Croatia's big matches were clustered in 1996 and early 1997. By the time the December 1997 cutoff hit, the math hadn't caught up to the reality on the pitch.
That gap between perception and algorithmic ranking is where the interesting stuff lives. It's also where I made my own mistake a few years back when I was working on a project analyzing historical seeding patterns. I was building a dataset of World Cup qualifying campaigns cross-referenced with FIFA ranking shifts, and I assumed the pre-tournament ranking was a reliable predictor of group difficulty. It wasn't. I spent about three days chasing down data points for teams like South Africa and Iran, only to realize the ranking had penalized them heavily for a string of draws in neutral-site qualifiers. A draw counts as a point in the ranking calculation, but it weighs far less than a win, and the magnitude of that penalty scales with the opponent's rank. So drawing with a top-five team barely helps your ranking, while losing to a bottom-twenty team tanks it more than you'd expect. South Africa sat around thirty-fifth going into France 1998. Their actual competitive level that cycle was closer to twenty-fifth. The ranking missed it entirely.
How the pre-1998 ranking algorithm actually worked
The system used a modified Elo-style point exchange. After every match, the winner took points from the loser. The amount depended on three things: the result (win, draw, loss), the difference between the two teams' rankings, and a strength-of-opponent multiplier. Home advantage was a factor, but less pronounced than in pure Elo. World Cup matches and continental championship games carried a higher weight coefficient than friendlies or minor qualifying rounds. That was the whole design—FIFA wanted the ranking to reflect performance against stronger competition, not just volume of matches played. Here's the part nobody remembers: the ranking used a rolling four-year window. Results dropped off completely after forty-eight months. That means a team could have an exceptional eighteen-month period and still rank mid-table if they played in a weaker confederation where the top opponents didn't have high baseline ratings. The Confederation adjustment factor was crude. UEFA and CONMEBOL teams got a natural boost just by playing each other. CAF, CONCACAF, and AFC teams had to win by larger margins against the same opposition to generate identical point swings. It wasn't unfair in a vacuum, but it systematically underranked teams from emerging footballing nations unless they accumulated a long string of wins. The seeding for the 1998 World Cup draw was based entirely on the December 1997 FIFA ranking. Host France was excluded from seeding. Brazil, Denmark, Netherlands, Italy, Argentina, Germany, Croatia, and Chile were the top eight seeds. Everyone outside that group was assigned to pots two through four regardless of actual form. That's why Croatia being a seventh seed feels surprising in retrospect—they ended up in a group with Japan, Jamaica, and Colombia, and lost only on penalty shootouts to France in the quarterfinals. The seeding should never have put them in the bottom half of pot one.
Get the Full Details
What the data actually shows versus what the headline suggests
"Unlikely favorites" is a headline device. What the numbers show is that the pre-1998 ranking had measurable blind spots for teams in transitional phases. Teams that had recently improved their coaching staff, switched systems, or integrated a new generation of players carried no ranking advantage from those changes. The ranking is purely historical. It has zero forward-looking components. I ran the qualification records for the twelve teams that finished outside the top twenty in the December 1997 ranking but advanced past the group stage in 1998. Five of them. Romania, which sat at twenty-eighth. Cameroon at thirty-first. Nigeria at twenty-fourth. South Africa at thirty-sixth. And Bulgaria, which was around thirty-third. Each of those teams had ranked significantly lower twelve months earlier. The ranking hadn't updated fast enough because their qualifying campaigns were concentrated in a narrow window, and the point gains from winning those matches were offset by previous losses that fell within the four-year lookback. There's also a simpler explanation for some of the surprises. The 1998 tournaments had more teams than the 1994 edition—expansion from twenty-four to thirty-two meant more direct qualification spots. More spots meant more mid-ranked teams qualified without having to beat top-ranked teams in playoff scenarios. A team ranked thirty-eighth could qualify by finishing second in a modest qualifying group. The ranking would still show them as a long shot, but they were legally qualified. The formula doesn't care about narrative. It only cares about points accumulated within the window.
Practical problems when using historical FIFA rankings for analysis
If you're pulling this data for any kind of model or comparison, you'll hit a few specific issues quickly. The first is inconsistent record-keeping for matches before 1995. FIFA's own database has gaps in friendly matches played by smaller national associations, particularly in CONCACAF and some CAF zones. If a match is missing from the database, the ranking calculation for that period is incomplete, and the derived rank is an approximation rather than a published official figure. I found this when comparing the printed FIFA Magazine rankings from 1994 against the digital archive for the same months. About eight percent of the recorded matches differed between sources for the Central American region alone. The second issue is the confederation weight adjustment, which was revised informally between 1995 and 1997 without a public announcement. The change was small—roughly a five-point shift in the multiplier for non-UEFA/CONMEBOL matches—but it shows up clearly if you back-calculate the point exchanges for games in late 1996 and early 1997. Teams that played mostly in their own confederation saw their rankings drift slightly downward during that period even though their results hadn't changed. I had to account for this by creating two separate calculation passes: one using the pre-revision weights for matches before September 1996 and one using post-revision weights for everything after. It took me about four hours to sort through the edge cases, but skipping it introduces systematic error into any cross-era comparison.
Limitations you should accept before relying on this
The pre-1998 FIFA ranking was not designed to predict World Cup performance. It was designed to provide a stable, periodically updated snapshot of international results for seeding and administrative purposes. Anyone treating it as a predictive tool is using it wrong. The four-year lookback window, the confederation adjustments, the missing data, and the static seeding model all combine to make it a decent descriptor of recent form but a poor indicator of peak capability. If you need a more accurate measure of team strength for historical analysis, Elo-based retroactive models like the ones published by feedingatac or FiveThirtyEight's soccer rankings do a substantially better job. They weight recent matches more heavily, don't use confederation multipliers, and update after every match rather than monthly. The trade-off is that they aren't official FIFA data, so they won't match the seeding records. You have to decide which matters for what you're doing. The 1998 World Cup itself proved the point. Croatia reaching the final, Denmark making the quarterfinals, and Netherlands winning the fair play award while finishing third—all of those outcomes align with the actual quality of those squads, not with their December 1997 ranking positions. The ranking showed what those teams had done in the previous four years. It didn't show what they were capable of doing in June 1998.
