Times are always exciting in the Asian Football Confederation - as you’d expect when 47 countries and half of the world’s population are involved - but the time since 2019’s Asian Cup has been even bigger than usual for the AFC. The COVID-19 pandemic that threw the world off its axis deeply impacted the synchronised qualifying for this tournament and the 2022 World Cup, reducing most of the Asian Cup matches to neutral-host mini-tournaments that decided which teams would make it. China’s strong response to the pandemic meant that the original plan to host the tournament throughout that country was unworkable, and Qatar was able to step in to host on short notice - not only making more use of the venues they had to build to host FIFA’s tournament in late 2022, but amplifying the developing power struggle between the eastern and western blocs of the AFC. On top of this, rumours have suggested that Russia’s currently-ostracised football association may be considering a move towards their Asian neighbours, as loaded as that would possibly be.
Ignoring everything wider, Asian football looks to be in pretty solid shape, leaving us with the prospects of an exciting tournament ahead. Four Asian countries sit in the top 25 of FIFA’s world rankings, and the 2022 World Cup saw some impressive performances - Saudi Arabia’s come from behind win over Argentina is already an icon of football folklore, while Australia made the round of 16 after a remarkable win over Denmark and nearly pushed the eventual champion Argentines to extra time.
Strong as those teams are, there are two frontrunners we have for the tournament. Japan are unsurprisingly firmly up there in the prospects - leading Asia with a ranking of 17th in the world, they topped their group in 2022 and were a penalty shootout away from the quarter finals. The Blue Samurai have lost just once since the World Cup, with their record since including a dominant performance against Germany and consecutive 5-0 wins in competitive World Cup qualifiers. Our model also, somewhat surprisingly, ranks Qatar very highly as well - contrary to what recent results might indicate, they have some solid high-scoring results in our dataset, and get a considerable boost as the home country. You can read about how our model is built in the below section, or you can just skip ahead to see what our predictions are at the opening of the tournament.
As all sports modelers have realised, you can’t come up with predictions for how teams are going to go without having an estimate of how good they are. And you can’t estimate how good they are without first seeing how good they’ve been, looking at past results.
The first step we had to do was build a dataset of past results; in this case, every game of tournament football between two AFC teams in recent times. With COVID as a convenient cutoff, the first game we measured was Qatar’s 5-0 victory over Bangladesh in December 2020, and we included another 341 - finishing with the eighteen World Cup qualifying games played on November 21st last year.
The following tournaments (and their qualifying, where applicable) were considered and incorporated into the dataset:
2020 AFF Championship1
2021 Arab Cup & qualification (only games with two AFC teams)
2021 SAFF Championship
2022 AFF Championship & qualification
2022 EAFF E-1 Football Championship
2022 World Cup qualification (part of the second round, all of the third and fourth)
2023 Asian Cup qualification (playoff round, third round)
2023 CAFA Nations Cup
2023 Gulf Cup
2023 SAFF Championship
2026 World Cup qualification (first and second rounds)
Once the dataset was built, we made use of several different matrix
dataframes to convert the results into meaningful predictive statistics.
matchupmatrix counted how many times each team had played
each other, while scorematrix summed all of the goals
scored in each match to calculate the average scoring rate, for and
against, for each team.2
The important feature here, and the one that serves as the cornerstone of the system, is seeing how many goals teams did score and comparing it to how many they should have scored, based on their opponents. Unsurprisingly, if you’re drawn to play against South Korea, Iran, and Jordan, you’ll score fewer goals than if you were drawn against Mongolia, Turkmenistan, and Laos.
Taking a baseline approximation - halfway between a team’s average score for per game, and their opponent’s average score against - we work out the expected goal tallies, and then convert them into the difference per game. Taking Australia as an example:
Australia scored 37 goals, and conceded 11, in the 17 games in the dataset
based on that value and their opponents, they should have scored ~27.7 and conceded ~19.3
averaging that out, Australia scores 0.55 extra goals per game and saves an extra 0.49
if Australia were to play a hypothetical game against an average AFC team, who score and concede the average number of goals (1.48) on average, the expected scoreline would be 2.03-0.99
This leads to the following ranking for the modifiers of each team:
| Team | FDiff | ADiff | TDiff |
|---|---|---|---|
| JPN | 0.95925568 | -0.58918408 | 1.548439761 |
| QAT | 0.76647469 | -0.40143936 | 1.167914051 |
| IRN | 0.66240606 | -0.49308812 | 1.155494172 |
| AUS | 0.54884085 | -0.48955003 | 1.038390878 |
| KOR | 0.44151226 | -0.53634922 | 0.977861479 |
| UZB | 0.55030864 | -0.40210367 | 0.952412313 |
| KSA | 0.29987804 | -0.53867081 | 0.838548854 |
| JOR | 0.19546055 | -0.54126060 | 0.736721155 |
| UAE | 0.22214759 | -0.36408548 | 0.586233064 |
| OMA | 0.09108889 | -0.42675721 | 0.517846100 |
| THA | 0.12479301 | -0.37781962 | 0.502612630 |
| IRQ | 0.09338264 | -0.38091825 | 0.474300889 |
| CHN | 0.26437898 | -0.19932921 | 0.463708195 |
| VIE | 0.06215765 | -0.37828136 | 0.440439006 |
| BHR | 0.14075356 | -0.27202545 | 0.412779016 |
| PLE | 0.25472485 | -0.03929056 | 0.294015412 |
| PRK | 0.20659722 | -0.04687500 | 0.253472222 |
| TJK | 0.10594470 | -0.11375929 | 0.219703989 |
| IDN | 0.22918249 | 0.03308738 | 0.196095107 |
| KUW | 0.01502151 | -0.15175560 | 0.166777111 |
| SYR | 0.08797447 | -0.07486410 | 0.162838565 |
| IND | -0.12453936 | -0.25945743 | 0.134918061 |
| LBN | -0.11771027 | -0.15230373 | 0.034593466 |
| MAS | 0.07916368 | 0.08296597 | -0.003802298 |
| KGZ | -0.02129609 | 0.03572467 | -0.057020766 |
| TKM | -0.08879317 | 0.25835284 | -0.347146013 |
| PHI | -0.24017603 | 0.16986104 | -0.410037070 |
| SGP | -0.24946298 | 0.19647679 | -0.445939776 |
| HKG | -0.23360165 | 0.36519710 | -0.598798749 |
| TPE | -0.31907435 | 0.29512085 | -0.614195203 |
| BAN | -0.31572016 | 0.37185815 | -0.687578306 |
| YEM | -0.36507252 | 0.39737258 | -0.762445093 |
| MDV | -0.44875797 | 0.35459884 | -0.803356814 |
| AFG | -0.29691325 | 0.52445143 | -0.821364678 |
| SRI | -0.43909899 | 0.43317742 | -0.872276408 |
| NEP | -0.41621206 | 0.46314685 | -0.879358907 |
| CAM | -0.33869417 | 0.74696379 | -1.085657957 |
| MNG | -0.62163442 | 0.48997995 | -1.111614372 |
| MYA | -0.31457354 | 0.96565222 | -1.280225768 |
| PAK | -0.64950829 | 0.65957292 | -1.309081208 |
| BHU | -0.57411606 | 0.85784076 | -1.431956815 |
| GUM | -0.85457916 | 0.67976713 | -1.534346282 |
| LAO | -0.75763631 | 0.85217286 | -1.609809166 |
| BRU | -0.50494124 | 1.52737714 | -2.032318376 |
| TLS | -1.13611111 | 0.98506944 | -2.121180556 |
| MAC | -1.55555556 | 0.75000000 | -2.305555556 |
Using these values, we have an important point to work with - if
Australia were to play a hypothetical game against an average AFC team,
who score and concede the average number of goals (1.48) on average, the
expected scoreline would be 2.03-0.99. Based on this principle, for
every potential matchup that could take place at the Asian Cup, all 552
of them,3
we calculate an expected score based on those multipliers (and giving
Qatar a boost, 0.2 extra each way, for the home ground advantage) and
save that in a dataframe xG_matrix. (Over the course of the
tournament, we’ll adjust this further by incorporating form from the
Asian Cup - likely based around teams over- or underperformance.)
From there, we run 10,000 simulations of the tournament, enough to get a good concept for what the likely and unlikely outcomes are. All 36 games of the group stage are simulated with two independent Poisson distributions for each team’s score, and from there we sum up group tables4 - and the all-important ranking of third placed teams. These are then plugged into the round of 16 scheduling, the score function is repeated (if the game is tied and goes to extra time, we run a second simulation with half the original expected score; penalties are a simulated coin toss), and we repeat until we have a winner crowned. And then we repeat this another thousand or so times!
Across the entire tournament, Japan go in as favourites, with a ~22.7% chance of lifting their first title since 2011; running close behind is Qatar and their ~21.3% likelihood. Australia, where most of my readers will be coming from, are the fifth most likely option with a ~5.3% chance. Jordan are the most likely first time winner, estimated at ~5.1%; but there’s a ~75.3% chance the title goes to someone who’s already won it.
Taking things group by group:
Group A is quite unmatched, with Qatar heavy favourites to top it; China are an outside chance of going far into the tournament, but tournament debutant Tajikistan are unlikely to make much of a splash (although they’d be competitive against Lebanon).
In Group B, Australia start as the likely victors, but could have some opposition from the always-threatening Uzbekistan, who have a ~4% chance to go all the way. Syria and India are evenly poised to take third and a possible knockout stage appearance.
Group C, the most politically tense, should be almost a walkover for Iran, who are our third most likely champion (~10.6%). The United Arab Emirates and Palestine are both competitive (although the latter will have much more serious things to worry about), while Hong Kong - appearing for the first time since 1956 - are the weakest team in the entire tournament.
Group D is perhaps the group of death at this tournament - Japan entering as favourites, but Iraq and Vietnam will be tightly matched against each other for second and third, and there’s an outside chance Indonesia can make a difference based on recent form.
Group E is almost a two/two split, with South Korea and Jordan comfortably stronger than Bahrain and especially Malaysia.
Rounding things up, Group F is surprisingly weak, with Saudi Arabia the only team having a realistic chance of going further into the tournament - Oman have their talents, but any success to Thailand or Kyrgyzstan would be a big surprise.
| Team | Group 1st | Group 2nd | Group 3rd | Group 4th | Round of 16 | Quarter Finals | Semi Finals | Final | Champion |
|---|---|---|---|---|---|---|---|---|---|
| JPN | 6225 | 2299 | 1043 | 433 | 9358 | 6759 | 4936 | 3475 | 2272 |
| QAT | 6850 | 2051 | 814 | 285 | 9579 | 7055 | 4929 | 3278 | 2131 |
| IRN | 5346 | 2783 | 1418 | 453 | 9274 | 5971 | 3490 | 1989 | 1056 |
| KOR | 4246 | 2786 | 1930 | 1038 | 8523 | 4850 | 3216 | 1904 | 918 |
| AUS | 4202 | 3057 | 1817 | 924 | 8656 | 5479 | 2055 | 1062 | 534 |
| JOR | 3108 | 3030 | 2328 | 1534 | 7820 | 3910 | 2343 | 1219 | 516 |
| UZB | 3696 | 3168 | 1971 | 1165 | 8359 | 4998 | 1729 | 872 | 402 |
| KSA | 4140 | 2801 | 1874 | 1185 | 8356 | 4555 | 1611 | 760 | 355 |
| UAE | 2591 | 3427 | 2787 | 1195 | 8000 | 3935 | 1700 | 726 | 269 |
| IRQ | 1436 | 2871 | 2969 | 2724 | 6208 | 2708 | 1499 | 703 | 262 |
| VIE | 1343 | 2661 | 3061 | 2935 | 5953 | 2517 | 1371 | 631 | 218 |
| BHR | 1758 | 2506 | 3067 | 2669 | 6291 | 2570 | 1213 | 534 | 186 |
| CHN | 1564 | 3352 | 2875 | 2209 | 6706 | 3230 | 1357 | 545 | 174 |
| OMA | 2531 | 2874 | 2577 | 2018 | 7262 | 3229 | 868 | 346 | 132 |
| PLE | 1724 | 2795 | 3392 | 2089 | 6797 | 2727 | 1037 | 394 | 129 |
| THA | 2328 | 2735 | 2816 | 2121 | 7043 | 3105 | 797 | 317 | 117 |
| IDN | 996 | 2169 | 2927 | 3908 | 4906 | 1815 | 864 | 349 | 90 |
| TJK | 967 | 2551 | 3179 | 3303 | 5521 | 2201 | 750 | 248 | 70 |
| MAS | 888 | 1678 | 2675 | 4759 | 4270 | 1321 | 550 | 174 | 54 |
| IND | 965 | 1824 | 3149 | 4062 | 4733 | 1814 | 389 | 132 | 41 |
| SYR | 1137 | 1951 | 3063 | 3849 | 5020 | 1904 | 415 | 123 | 29 |
| LBN | 619 | 2046 | 3132 | 4203 | 4456 | 1513 | 465 | 129 | 25 |
| KGZ | 1001 | 1590 | 2733 | 4676 | 4291 | 1315 | 252 | 72 | 18 |
| HKG | 339 | 995 | 2403 | 6263 | 2618 | 519 | 89 | 18 | 2 |
Best wishes to all the teams competing, and all the fans watching across Asia and around the world - we’ll try and be back with followup predictions as the tournament goes on.
It was a little bit of a surprise learning that these tournaments exist, given Australia’s avoidance of taking part despite our AFF membership.↩︎
For the few games where extra time was played, we excluded any goals scored in the additional time to simplify the calculations we had to make.↩︎
Technically, we calculate 576 - we don’t exclude the score for a hypothetical Australia v Australia matchup, for instance, because that would be too much extra coding work.↩︎
Unfortunately, unlike the World Cup, the head-to-head result between teams is the group stage tiebreaker rather than just goal difference. We haven’t coded around this, so these values will be a little inaccurate, but bear with us.↩︎