World Cup 2018 start soon and you might have heard about @blocktrades’ prize of 2000 SBD if you guess the most right results. I don’t know much about soccer but I know quite a lot about data analysis and if you want to win, this analysis might help you. If you didn't join so far, here is article describing how to do that.
Wisdom of the crowd
Collective opinion of a lot of people will often produce better result than an opinion of a single expert because false expectations of a single person are filtered by voice of others that have the right information. Let’s look at some examples: most people agree that Spain, France and Brazil will win all their matches, most people also agree that Saudi Arabia and Iran will lose all. 96 % of people think that Spain will beat Morocco and 95 % think that Spain wins over Iran. Only 2 % of people think that Costa Rica will beat Brazil which is extremely unlikely and since you get only 1 point if that happens, you should prefer to bet on more likely results where you can get the same 1 point much more likely. The hardest think to find in the data is separating teams that are on similar level like Serbia and Switzerland, Poland and Colombia or Japan with 39 % chance of win and Senegal with 33 % chance of win while 28 % likelihood of tie. We can do some tricks like removing people what believe that Costa Rica wins over Brazil but even that moves the percentages only slightly.
Looking at the data
We analyzed 954 valid entries and most people (913) agreed that Spain beats Morocco (24 believe in tie), 906 people think Spain will win over Iran while 24 people believe in Spain's loss and 24 in tie), 892 trusts Brazil to win over Switzerland while 25 think Brazil will lose and 37 think it will end up with a tie. Win and loss in graph is from a view point of the first mentioned team.
| Spain vs Morroco | Iran vs Spain | Brazil vs Switzerland |
|---|---|---|
| 913 - 17 (24 ties) | 24 - 906 (24 ties) | 892 - 25 (37 ties) |
 |
 |
 |
| Mexico vs Sweden | Poland vs Colombia | Serbia vs Switzerland |
|---|---|---|
| 396 - 243 (315 ties) | 182 - 438 (334 ties) | 291 - 353 (310 ties) |
 |
 |
 |
Change when we eliminate noise
When we eliminate people that believe Brazil will lose to Costa Rica, Spain will lose to Morocco and similar matches that most people agree on with each iteration we get only slight change in percentage distribution of wins and loses as you can see down below.
Sweden's chances agains South Korea slightly rises when we eliminate users with marginal views. Sweden's chance rises from 59% to 63% and South Korea's chance decreases from 16% to 13%.
| Denmark vs Australia | Japan vs Senegal | Russia vs Saudi Arabia |
|---|---|---|
| 50% -> 54% vs 33% -> 31% | 39% -> 37% vs 28% -> 29% | 88% -> 91% vs 7% -> 6% |
 |
 |
 |
| Win probability | Team | Team | Win probability | Tie probability | |
|---|---|---|---|---|---|
| 88% | Russia | vs | Saudi Arabia | 5% | 7% |
| 10% | Egypt | vs | Uruguay | 74% | 15% |
| 47% | Morocco | vs | Iran | 23% | 30% |
| 20% | Portugal | vs | Spain | 51% | 29% |
| 94% | France | vs | Australia | 3% | 3% |
| 90% | Argentina | vs | Iceland | 4% | 6% |
| 21% | Peru | vs | Denmark | 49% | 31% |
| 57% | Croatia | vs | Nigeria | 22% | 21% |
| 36% | Costa Rica | vs | Serbia | 37% | 27% |
| 91% | Germany | vs | Mexico | 4% | 5% |
| 94% | Brazil | vs | Switzerland | 3% | 4% |
| 59% | Sweden | vs | South Korea | 16% | 24% |
| 90% | Belgium | vs | Panama | 4% | 6% |
| 4% | Tunisia | vs | England | 90% | 6% |
| 58% | Poland | vs | Senegal | 18% | 24% |
| 68% | Colombia | vs | Japan | 16% | 16% |
| 55% | Russia | vs | Egypt | 19% | 27% |
| 92% | Portugal | vs | Morocco | 3% | 5% |
| 90% | Uruguay | vs | Saudi Arabia | 4% | 6% |
| 3% | Iran | vs | Spain | 95% | 3% |
| 87% | France | vs | Peru | 4% | 9% |
| 50% | Denmark | vs | Australia | 17% | 33% |
| 72% | Argentina | vs | Croatia | 8% | 20% |
| 92% | Brazil | vs | Costa Rica | 2% | 7% |
| 45% | Nigeria | vs | Iceland | 29% | 26% |
| 19% | Serbia | vs | Switzerland | 46% | 35% |
| 86% | Belgium | vs | Tunisia | 6% | 8% |
| 90% | Germany | vs | Sweden | 3% | 7% |
| 15% | South Korea | vs | Mexico | 65% | 20% |
| 91% | England | vs | Panama | 4% | 5% |
| 39% | Japan | vs | Senegal | 33% | 28% |
| 31% | Poland | vs | Colombia | 37% | 32% |
| 52% | Uruguay | vs | Russia | 20% | 28% |
| 14% | Saudi Arabia | vs | Egypt | 65% | 21% |
| 96% | Spain | vs | Morocco | 2% | 3% |
| 3% | Iran | vs | Portugal | 93% | 4% |
| 8% | Denmark | vs | France | 73% | 19% |
| 34% | Australia | vs | Peru | 42% | 24% |
| 6% | Nigeria | vs | Argentina | 80% | 14% |
| 17% | Iceland | vs | Croatia | 60% | 23% |
| 5% | South Korea | vs | Germany | 92% | 3% |
| 42% | Mexico | vs | Sweden | 25% | 33% |
| 2% | Sebria | vs | Brazil | 92% | 6% |
| 55% | Switzerland | vs | Costa Rica | 20% | 25% |
| 23% | Japan | vs | Poland | 56% | 21% |
| 18% | Senegal | vs | Colombia | 60% | 23% |
| 39% | England | vs | Belgium | 30% | 31% |
| 28% | Panama | vs | Tunisia | 43% | 29% |
My recommendation
Choose winning team according to table above that is highlighted i.e. more than 40% of people agree that the team wins (W) and if no team has above 40%, I recommend to choose a tie (T).
About problems with the data
We didn’t analyze data in other languages or with typos. We ignored if someone specified that Japan vs Poland will end up with win of Japan and tie of Poland. If you missed some matches or duplicated matches, your entry was ignored as well.
Coming soon
After the first stage, expect analysis of posts for the second stage of the World Cup.