The first four parts of this series covered Nottingham Forest, Notts County, Derby County/Leeds United, and Stoke City.
This episode in English Football Drama is going national. After four episodes diving into the circus that is Midlands football, we turn our attentions to England's national football team. This episode will focus on the 21st century up until about the 2010 World Cup, as the 'Golden Generation' of young English talent failed to achieve for country the glories that so many of them achieve for their club. There will be a bigger focus on off-the-pitch stories and tabloid scandals here, although obviously football is central. In the era before the general sanitisation of the English tabloid press, players on the national team were up there with the royal family for gossip reporting and red-top scandals. And scandals they provided.
Background
English football had something of a renaissance in the 1990s. It ceased to be a violence-ridden sport predominantly enjoy by white working class men and became a national game in the true sense.
After the glory of winning the World Cup for the home crowd at Wembley in 1966 against the Germans, England's footballing fortunes took a nosedive. England did not qualify for the 1974 or 1978 World Cups at a time when English clubs were dominant in Europe (all European Cups from 1977-82 were won by English sides, while Leeds would have likely won in 1975 if not for some incredibly suspect refereeing at a time when bribing the officials wasn't unheard of) and the national team stagnated. Performances picked up in the 1980s, but the most notable moment for the national side in this decade was a certain Diego Maradona cementing an increasingly bitter rivalry with the Hand of God in 1986.
Off the pitch, hooliganism, often called the 'English Disease', blighted the reputation of the game at home and abroad. The socially turbulent 1970s-80s created a violent subculture at the underbelly of the sport, arguably fuelled by media sensationalism, and earnt English football supporters a terrible reputation when they supported club or country in European football competitions. The darkest moment came at the 1985 European Cup final when 39 supporters of Italian side Juventus died in a stadium collapse caused when the crowd sought to flee rioting Liverpool supporters. While there were other factors at play, the horrific scene was too much and all English clubs were banned from European football until 1990. England's participation in the 1988 European Football Championship in West Germany was like a candle which drew in the unsavoury elements of English club support and there was widespread violence between English and German hooligans around the competition.
So, in 1990 a fairly unpleasant situation? What changed?
At home measures such as the general adoption of all-seater stadiums and corresponding ticket price increases killed off much of the old 'terrace culture' around clubs. The government had made it possible to ban convicted hooligans from attending any match and in some cases from leaving the country during a major international tournament. The issue was good optics for Margaret Thatcher and her general commitment in government to the traditional safe Tory issue of law and order.
The 1990 World Cup in Italy was the start of football's image change. England did better than usual, making it to the semi-final before what would become a highly common event. Losing penalties to Germany. Their success and the resonance of Paul Gascoigne crying on the pitch owing to his sheer pashun took football to a wider audience.
But in 1996, football came home. England hosted Euro 96 and a famous song by comedians David Baddiel and Frank Skinner epitomised the optimism of football's home nation during the festival-like competition with a sentimental nod to 'Thirty Years of Hurt' since 1966, the year that has weighed down England ever since. But above all, it showed supporting England as a thing that normal and respectable people did. The awarding of the competition to England was a major sign from world football that the 'bad old days' were over, and the competition went ahead in a positive atmosphere. There was just one small hitch, once again involving a penalty shoot out against Germany as England once again lost in that accursed format.
Things weren't perfect by the end of the decade and some of the old problems persisted. A 1995 friendly against Ireland in Dublin was abandoned owing to a riot by a group of English hooligans affiliated to various far-right organisations, and there was trouble at both the 1998 World Cup and Euro 00. But it was a far cry from previous decades, and there was a new optimism in the air about the English national team.
The Golden Generation
At the turn of the century, England had a huge amount of young and promising talent who would surely end the pain and bring the World Cup home once again. Players who would go onto become legends for their clubs such as right midfielder David Beckham, centre forward Michael Owen, central midfielder Steven Gerrard, and centre half Rio Ferdinand had all come of age. Manchester United's 1999 Champions League win was the first one for an English side since Liverpool in 1984, and it signalled a return to Europe's top competitions for English sides.
What better to signal a new era than by absolutely stuffing the main rivals? After the 1966 World Cup final, penalty agony in 1990 and 1996 and various other historical events, Germany became the team English fans wanted to beat the most. The tabloids revelled in the rivalry with their diplomatic headlines and terrace classics such as 'Ten German Bombers' and 'Two World Wars and One World Cup' remained commonplace. For their part, German fans tend to regard the Netherlands as their primary football rival.
And stuff them they did. In a World Cup qualifier in Munich in 2001, Germany suffered their first ever home loss in a World Cup qualifier as England beat them 5-1. This was a massive vindication of England's young side and the new manager Sven Goran Erikkson, who had received some scepticism as England's first foreign manager. German fans consoled themselves with the Dutch failure to qualify for the 2002 World Cup, but it marked a new era for England.
The side that humbled Germany included the core of the 'Golden Generation' with Gary Neville, Ashley Cole, Steven Gerrard, Rio Ferdinand, Sol Campbell, David Beckham, Paul Scholes, Emile Heskey, Michael Owen, and Jamie Carragher all appearing.
The CEO of the Football Association, Adam Crozier, publicly described this cohort as the 'Golden Generation' who would go on to do for England what many of them were already doing for their clubs.
Japan 2002
England qualified for the 2002 World Cup in Japan in true fashion, with a David Beckham free kick past Greece putting England through. David Beckham had rapidly become the poster boy for the Golden Generation owing to his skill with free kicks and his glamorous appearance and personal life, including being married to the famous pop star Victoria Beckham, known as 'Posh Spice' in the Spice Girls.
Quick stop off for the first affair to be dragged through the tabloids on the way to Japan. In the spring of 2002, it turned out that Anglo-Swedish TV presenter Ulrika Jonsson had been having an affair with Sven. The News of the World (Britain's premier sleaze rag/newspaper, known as the News of the Screws in satirical magazine Private Eye or if you're my great-granddad, as the Whore's Gazette) absolutely loved a bit of sleaze. So much, in fact, they gave Ulrika a column for the next four years so she could launch the odd diatribe at Sven and his unconventional personal life.
Sven was bullish about England's chances in 2002, in a formula which would become predictable, he expected that this year would be the year.
To be fair, it started well. A group stage tie with Argentina gave England a chance to give a bloody nose to another rival. Owing to the 1982 Falklands War and Diego Maradona's 'Hand of God' in 1986, Argentina normally sits 3rd in the list of international footballing rivalries after Germany and Scotland. I remember as a fairly young child my parents explaining to me that the Argies were a 'dirty' side after the Hand of God and an acrimonious game in the 1998 World Cup and that beating them was pretty much essential. So, it was to relief all round when David Beckham scored from a penalty.
Draws with Sweden and Nigeria took England into the Round of 16, amusingly at the expense of Argentina who were 2002's shock group stage exit. England brushed Denmark aside 3-0 to set themselves up for a quarter final tie with the favourites and eventual champions, Brazil.
Was this to be England's year? Victory against Brazil would likely assure playing Germany in the final given the probability of beating Senegal or Turkey in the semi-final (so goes the England fan's logic) and we would doubtless give Jerry another pasting to make it Two World Wars and Two World Cups. The nation was abuzz. I willingly went to school early to watch it on the projector, the only time I've ever turned up a minute early of my own free will.
And Brazil won. Well, they were going to, weren't they? A good Michael Owen goal set England up for a 1-0 win and dreams of glory in Tokyo, but Messrs Ronaldo and Rivaldo put two past David Seaman to knock England out.
The circular firing squad and media postmortem was mysteriously absent. Brazil had some truly sublime talents, and surely a young side would come good and win it next time, wouldn't they?
The Road to Euro 04
A few distractions to get through before we get into the meat of England's next footballing failure. In late 2003 England played a friendly against Australia which was notable for two things. England were roundly beaten by Australia, a nation with whom England has fierce cricket and rugby rivalries but doesn't generally register in football, to echoes of condemnation. And a young Wayne Rooney became the youngest player to ever play for England, aged just 17. More on him later.
England's qualification for Euro 04 was generally uneventful, except for a shock draw to Macedonia, involving this filthy banger direct from a corner.
Rio Ferdinand managed to get an eight month ban from football by missing a drugs test. This was a blow as it removed one of Sven's two preferred centre halves from the squad. It also removed one of the main drama generators, a man who had gained notoriety in 2000, when Channel 4 aired a sex tape featuring him and fellow internationals Kieron Dyer and Frank Lampard filmed on holiday in Cyprus. Fortunately for us, his ban allowed an even more sublime drama king to take centre stage. Enter John Terry. Another new appearance was Chelsea teammate Wayne Bridge. More on them later.
Penalties, again?
Euro 04 started strong with an early goal from Frank Lampard against France. France, despite being England's rivals of choice in most fields, is not an especially strong football rivalry. The idea of France as a major footballing nation is relatively new and as far as football goes there isn't any real bad blood. Even Zinedine Zidane's injury time equaliser and winner didn't create any lasting resentment.
Comfortable wins over Switzerland and Croatia put England 2nd in the group and set them up for a draw against host nation Portugal. Of course, Portugal had their own wonderkid as the world was introduced to a certain Cristiano Ronaldo. The two sides drew 2-2 after extra time and the dreaded penalties loomed. Fear not, for dead-ball specialist David Beckham stepped up to the mark for his country.
Well.
Beckham joined Stuart 'Psycho' Pearce and Gareth Southgate in the list of players who gained notoriety for fluffing a penalty.
Sven started to come under criticism in the tabloids for his failure to deliver the goods. He came under more criticism when he, once again, couldn't keep it in his trousers, starting a proud tradition for the Golden Generation to become better known for tabloid gossip.
Young Rooney started strong as well, as it turned out that he had shagged a granny in a Liverpool brothel. Hopes of an incognito quickie with a lady of the night were dashed when a group of Everton fans (Rooney's then-club and boyhood team) found out and started chanting his name outside when he was doing the dirty.
At about this time, Ashley Cole began a relationship with the professional Geordie and pop star Cheryl Tweedy, a saga which would keep the News of the Screws engaged for some years.
Possibly the only bad thing about that sordid rag going under in 2011 is that I'm having a hell of a time finding the original news articles for this write up.
The Old Enemy Beckons
The 2006 World Cup is now on the horizon. And who is hosting it, other than Germany? What a coup it would be for the boys and England's national pride for, in a story lifted from a campy 1950s war film, our plucky boys to parachute in Berlin and win the World Cup from right under Fritz's nose?
Well, if they could get there. Northern Ireland, traditionally footballing minnows, beat England 1-0 in a qualifier. This was Sven's first defeat in a qualifier but the tabloids attacked his playing style and, being a Swede, the lack of pashun and dersiah that a redblooded Englishman would have. England did qualify with a match to spare, so surely it would be a smooth ride to Germany and for our plucky heroes to take on the Hun and win?
Like fuck it would. Sven managed to get into the tabloids again. At the time, the News of the Screws employed an undercover journalist who specialised in 'sting' stories aimed at prominent figures. Mazher Mahmood was popularly known as the 'Fake Sheikh' owing to his tendency to masquerade as an Arab oil sheikh. He approached Sven as an Arab businessman hoping to invest in English football, and Sven seemingly said told Mahmood to buy Aston Villa (a respectable Premier League side from Birmingham) as he hoped to take the Villa job after the 2006 World Cup and bring David Beckham in from Real Madrid to be his captain. The imbroglio fatally undermined Sven, who was forced to announce a resignation effective after the World Cup. Essentially working through his notice period, it isn't hard to see what impact on the squad that might have.
England turned up in Germany, and the tabloid media brought its legendary tact and diplomacy. England fans made no reference to past historic events before matches. Songs like this were condemned in the respectable press which saw them as crass chants from drunken shirtless louts, but many football fans see these chants as lighthearted banter.
On the pitch England beat Paraguay and Trinidad in the group stages before another turgid draw with Sweden, putting England through to the Round of 16 at the top of their group.
Sven began to deviate from the traditional 442 formation, using Rooney up front alone and playing a third man in central midfield, usually Michael Carrick. We'll talk a little more about the significance of 442 in a bit.
A 1-0 win over Ecuador set England up for another quarter-final tie, this time against Portugal, who had knocked them out on penalties in 2004.
Surely England wouldn't be knocked out on penalties by Portugal again would they? Curses aren't real surely?
A 0-0 snooze fest led to the inevitable. And the inevitable happened. England's players missed three penalties. This was now becoming a serious mental block, with England going out of the 1990, 1998, and 2006 World Cups on penalties as well as Euro 96 and 04.
It wasn't Sven's problem now. As agreed with the FA, he left the role thereafter. But who could replace him?
The Wally with a Brolly
After the pashunless Sven, it was time to get a proper Englishman back in charge. Enter Steve McClaren. McClaren had proper club pedigree, taking Middlesbrough to a UEFA Cup Final in 2006 after winning a League Cup in 2004.
His first media scandal came when the News of the Screws claimed Ashley Cole had taken part in a homosexual orgy. They settled out of court and retracted the stories but this was just another time in which the tabloid press pursued sensational stories about members of the England team.
McClaren gave John Terry the captaincy and dropped David Beckham from his line ups. There was a period of poor form in which England scored once in five matches, followed by when he told a press conference they could write what they wanted after a lacklustre win over Andorra because he wasn't going to say anything else. McClaren recalled David Beckham and form picked up once again. It wasn't enough though, as England lost against Russia before the final gameday of the qualification cycle for Euro 08. This meant that England had to avoid defeat against Croatia, who obligingly beat England 3-2 at Wembley while Russia took 2nd place with a win over minnows Andorra. This was the first time in 24 years that England had not qualified for a European Championship.
The Daily Mail's headline the next day became one of those legendary headlines that goes down in a country's collective pop-culture memory. The Wally with a Brolly epitomised what was then the shortest ever tenure in the England job.
Who could possibly prepare the club for the 2010 World Cup after a period sat on the sidelines of international football?
Enter Fabio
Fabio Cappello was another foreign hiring. He was widely welcomed in the English press as a notorious disciplinarian who could curb the excesses allowed under the relaxed Sven and the disastrous McClaren.
Fabio rigidly played 442 football. In England, 442 is more than a football formation, it is a way of life. 442 means standing up for traditional no-nonsense English values like work ethic, pluck, and willingness to die for the team over fancy foreigners with their poncy passing skills. This initially endeared him to the press, which demanded nothing more than the most patriotic form of football involved.
Capello saw Beckham dropped once again for his initial set of friendly matches. In goal he started to play David 'Calamity' James, who was perhaps unfairly scapegoated for being a somewhat eccentric keeper prone to errors.
In the 2010 qualification group, England did very well. They comfortably qualified, including with a win over Germany to keep the fans happy, and everyone looked forward to watching England bring it home from the first ever African World Cup.
But first, an interlude for the biggest of the tabloid shitstorms.
Viva John Terry
In January 2010 a certain individual successfully placed a 'Superinjunction' preventing details of an extra-marital affairs being published. The existence of the injunction was also not to be revealed. Another judge overruled this and the News of the Screws was free to report the information it had. Namely, John Terry had an affair with the partner of Wayne Bridge, former Chelsea teammate and current England teammate. Wayne Bridge in the end wasn't involved in the 2010 World Cup, but many were naturally shocked a captain would do something like this to a team member.
At the time there was a prominent scandal around the use of 'injunctions' by high profile people exploiting England's lax law on these issues, and John Terry was merely one of several.
In any case, the News of the Screws ended up apologising to the woman involved, in a fairly remarkable move which in hindsight was an early indication of the radical shifts in the tabloid press during the early 10s.
John Terry was stripped of the captaincy for the first time and replaced with Rio Ferdinand for the upcoming World Cup.
Germany, again?
It started so well. England was put into a relatively soft group on paper and were the clear favourites. The Sun's legendary tact once again.
Match number one of 'HMS Piss the Group' was against the USA. Again, not really a footballing nation, the expectation of an easy win was high. It started well with an early goal from Steven Gerrard before a goalkeeping error etched into my mind today. Rob Green mishandled a save, giving Clint Dempsey a goal. A frustrating 0-0 draw with Algeria indicated something was going wrong.
Were England about to be knocked out of an allegedly soft group? Fortunately a goal against Slovenia rescued it, but England finished 2nd in the group behind the USA, leading to a Round of 16 tie against Germany.
Once again, the red tops employed their legendary diplomacy
Did it go to penalties? No, but don't breathe a sigh of relief yet.
England got stuffed 4-1. Properly stuffed. The only consolation was that an obvious Frank Lampard goal was not awarded because the linesman didn't see it properly. In the days before VAR it really was that rudimentary. The reason this mattered was because England had been awarded a similarly controversial goal in the 1966 World Cup final, many Germans reassured themselves knowing that it could have gone so differently if Hurst's 2nd goal hadn't been allowed. At least they were now even on the controversial crossbar goal count.
Was this a consolation to the press?
Of course not
Once again, the red tops hounded a squad who had failed to live up to the increasingly lofty expectations.
But at least Wayne Rooney got one more romp in the tabloids out, as it turned out he had paid hush money to a high-class escort he had been 'seeing'.
Viva John Terry 2
Fabio quietly gave John Terry the captaincy back. He still had one shitstorm left in him. And it was ugly. He was accused of calling Anton Ferdinand, brother of Rio, a 'fucking black cunt' during a game between Chelsea and West London neighbours Queen's Park Rangers. After being charged by police for using racist language, Terry was once again stripped of the England captaincy, which went to Steven Gerrard.
Fabio, who had come under heavy fire after the 2010 World Cup, resigned over this. He didn't want to take the captaincy from John Terry but the FA forced his hand and he left the role in early 2012.
John Terry was actually acquitted of racially abusing Anton Ferdinand in a criminal trial, but an FA hearing found him guilty and punished him with a fairly short ban. This was controversial because of the different standards of proof. A conviction in English criminal courts needs proof 'beyond reasonable doubt', whereas the FA hearing used civil rules where guilt is based on the balance of probabilities. Terry apologised for his language, but he retained the Chelsea captaincy and never directly apologised to Anton.
Epilogue
With Fabio's departure the classic 'Golden Generation' era ended, even though players like Gerrard and Lampard still turned out for England. In the interests of time and length I shall give a summary until about 2016 before wrapping up.
Roy Hodgson managed England from 2012-6. England were eliminated in the group stage of the 2014 World Cup after losing to Uruguay and Italy and drawing to Costa Rica. Ironically Italy, the other team expected to go to the knockouts, was the other eliminated side. This was a bigger trauma than 2010, although massive changes in journalism (such as the News of the Screws closing down and far stricter regulation) meant that there was less of the constant tabloid coverage of the fallout.
2016 saw a decent showing in Euro 16, but once again the media got cocky before a tie against a theoretically weaker side only for England to get knocked out by them. Bravo Iceland.
Sam Allardyce replaced Roy Hodgson and lasted a grand total of two months before being caught telling undercover journalists how to get around various financial rules in football before criticising previous managers and some of his players.
Finally, Gareth Southgate came and has seemingly reinvented the side, making it to the semi-finals in 2018. But at least it wasn't penalties then.
Why did the Golden Generation Fail?
On paper they should have been at the top of the game. Pep Guardiola compared them to the Spanish flowering of talent which led to their 2010 World Cup win. There are a few factors here.

• The team's philosophy didn't really work. There was never a coherent playing style for the national team which players would be selected around or moulded into. Rather, managers tended to make systems work around players even if they didn't gel well.
  
• Rivalries in the squad. It probably reflects on English football well that there are several top end teams who compete for titles and European qualification, rather than the perhaps less exciting one club dominance of Germany and Italy or the duopolies of Scotland and Spain. But, Gerrard and Lampard seemed to struggle when paired and their club rivalries may have played a part. The bulk of the squad came from three clubs, Manchester United, Liverpool, and Chelsea who all won Champion's League honours at some point in the early 21st century and regularly competed for titles and domestic cups. Could they switch off the rivalries when on national duty?
  
• Finally, the press. We have gone through the tabloid obsession with these footballers and their personal lives. We have also discussed the cycle of bullishness followed by anger which follows the inevitable defeats. Is it unreasonable for a young adult to find these pressures too much? Probably not. Above all, it is hard to escape how much better England appear to play now that the press is less hostile towards the side, that their manager is a more nurturing figure, and that the sleazy red tops have either gone the way of the dodo (News of the Screws) or been forced to fundamentally change their approach in response to heightened regulation and changing social mores.
  

Manchester United fans cannot be satisfied with their team’s Premier League campaign, as they are currently 8th in the table. However, thanks to often hiccups of their rivals, the Red Devils are only five points behind Tottenham Hotspur, having played one game less. In the mid-week, Bruno Fernandes and the lads are a completely different team, and in a very tough group, they top the table. After a poor start to the Premier League season, nobody expected that Man Utd could win in Paris – again. Ole Gunnar Solskjaer’s side is three points ahead of their upcoming rivals and RB Leipzig, whom they trashed 5:0 at Old Trafford. If United haven’t lost away to Basaksehir – thanks to terrible defensive positioning – they wouldn’t worry about the place in the eight-finals. However, Manchester United needs a point from this one at the home ground to secure a place in the top two. Despite not so confident displays, the hosts tied four victories in a row, and the latest one came against Southampton last weekend. The Red Devils were 2:0 down at half time but managed to score a winner in the stoppage time. That one might boost their confidence ahead to this crucial game.
On the other hand, PSG has been in a variable form lately, as the visitors booked only two wins in the past five games. They lost to RB Leipzig 2:1 in Germany and suffered a surprising 3:2 defeat in Monaco, although they were 2:0 ahead. Last weekend, Thomas Tuhel’s side earned just a point against Bordeaux at Parc des Princes. Only six points in the current Champions League campaign cannot satisfy the last season’s finalists, and this is a must-win game for the visitors. Neymar and Kylian Mbappe are back in the squad, and the French champions will heavily rely on them in this clash. Nevertheless, other players can pose a real threat to the hosts, although Marquinhos, Julian Draxler, and Juan Bernat won’t be able to help their teammates.

Full-Time Winner Odds & Prediction

It is going to be an exciting clash and a very unpredictable one. Looking into each team’s strengths, PSG has more individual quality, but they don’t perform constantly this year. Man Utd can put masterclass display and yet lose to the bottom-half Premier League teams at Old Trafford. Nevertheless, we believe the hosts won’t suffer a defeat in this one.

Goals Market Prediction

Both teams have trouble keeping their nets intact lately. Their last two matches were pretty efficient ones, and we think both teams will score at least once in the game.
Manchester United AH 0 @ 2.10
BTTS Yes @ 1.55
Correct score 1:1 @ 7.00
Read the analysis here and share your opinion with us!

The hosts put a very good display in Munich last weekend in the Bundesliga, as they snatched a point against Bayern. Although they had a lead twice, the home side will definitely be satisfied with a draw. Emil Forsberg and the lads have been very productive in front of the oppositions’ goal lately. RB Leipzig scored nine times on three occasions, but during the same period, conceded seven goals. Julian Nagelsman’s side is still in the battle for a place in the knockout stage, as they are level with PSG and Manchester United. However, they need to think only about the victory in this game since it is very unlikely that PSG can lose against Istanbul Basaksehir at Parc des Princes. The German side has the weakest head-to-head record in this triangle, and three points for them is the only outcome that can guarantee them a place in the top two.
Manchester United had another poor first-half display, as they trailed 1:0 to West Ham last weekend. However, they needed 30 minutes to turn around the result and book a 3:1 victory. That one was their fourth in a row in the Premier League and lifted them to the 5th spot. The Red Devils are now only five points behind the Spurs, with one game played less. However, they missed the chance to secure a place in the Champions League knockout stage last Wednesday, as they lost to PSG 3:1 at Old Trafford. Although Ole Gunnar Solskjaer’s side had several excellent chances, they got punished. United’s crucial defeat occurred in Istanbul against Basaksehir, and now, they need to avoid the loss in order to remain in the top two.

Full-Time Winner Odds & Prediction

A great battle is ahead of us, and both sides will go for some points in this clash. We believe Manchester United is more experienced in games like this one, and they should pick up at least a point.

Goals Market Prediction

Both sides have been involved in efficient games lately. RB Leipzig will attack from the start, and the visitors will search for their chance to punish any home side’s mistake. Therefore, we don’t expect nets to remain intact at the end of this game.
Manchester United AH +0 @ 2.20
BTTS Yes @ 1.50
Correct score 2:2 @ 12.00
Read the analysis here and share your opinion with us!

After six straight victories, Ajax tied two defeats in a row, and that is not an encouraging moment for the hosts ahead to the most important match in this part of the season. They first lost at Anfield, while Twente booked a surprising win at Johan Cruijff Arena last weekend. Erik Ten Hag’s still tops the Eredivisie table, but they need to chase for a victory in this one to secure one of the top two spots in the group. A double triumph against Midtjylland brought them six points, while Ajax snatched the point in Bergamo after a 2:2 draw. They could have easily got back with a win, but they spilled a 2:0 lead. Dusan Tadic and the lads are very efficient this season, especially in the domestic championship. However, the hosts need to get back on the winning track if they want to book a ticket for the eight-finals.
Atalanta had more time to prepare for this clash since they didn’t play their Serie A match last weekend. The visitors are one point ahead of their next opponents, and a draw could see them progressing further in the competition. However, they also cannot be proud of their latest form since Atalanta has been pretty inconsistent lately. Gian Piero Gasperini’s side celebrated just twice on the past nine occasions, and interestingly, both victories came at the away grounds. Despite losing 5:0 against Liverpool at home, they managed to win all three points at Anfield just one round later. That’s how Atalanta is unpredictable. Luis Muriel and the lads are very productive in front of the oppositions’ goal this season, but they concede a lot as well. Last time, they were just one step away from the semi-finals, but can they get into the knockout stage again?

Full-Time Winner Odds & Prediction

The stake in this clash is very high, as the winner will continue its Champions League campaign. We think Ajax is a bit closer to the victory, and they should be able to deliver it.

Goals Market Prediction

Atalanta won’t be focused just on defending its net and will search for the chance to score. Since both sides are very efficient in the final third, we believe that neither of the teams will be able to keep the clean sheet.
Ajax to win @ 2.25
BTTS Yes @ 1.45
Correct score 3:2 @ 19.00
Read the analysis here and share your opinion with us!

This will be a long read, the OP is pretty much me explaining how the prediction model works, if you just want to see the predictions for Best Girl 7 then skip to my first post in the thread.

Updates

(Skip to the introduction if this is your first time reading.)

• Update 1 - using a Logit-normal model instead of a Normal model

This is a minor fix in the grand-scheme of things. Instead of assuming the vote share follows a normal distribution we instead assume it follows a Logit-normal distribution. Random variables that follow a L-N distribution have their support bounded to the range [0,1] which means it will never consider impossible probabilities outside of this range which is what was happening before. This is a minor change because the probability of a character receiving a negative vote share or a vote share >100% in the old Normal model was negligible since almost every matchup is in the 10-90% vote share range and the standard deviation is ~5%.
In the calculations below everything stays the same except we are now modelling the Logit of the vote share i.e. we assume logit(V) is a Normally distributed random variable with mean logit(p) and variance σ² (different σ, estimated analytically using past contest data as before: 0.25 in early rounds, 0.30 in later rounds).

• Update 2 - improving same-show matchup predictions.

The problem: two characters from the same popular show can dominate opponents in the early rounds and appear to be roughly equal in strength but when they match against eachother one character is the clear favourite and wins by a landslide. This causes the winning character to have an artificially inflated score and results in them being predicted to do better than they should do in later rounds. To explain the fix for this I will use an example.
Example:
In round 4 of Best Girl 5 we had Megumin of Konosuba go up against Wiz, also from Konosuba. Megumin is one of the leads of the show whereas Wiz is a side-character so it's pretty obvious that Megumin should be the favourite here and will probably win by a large margin. Megumin's score going into round 4 was 5080 compared to 2808 for Wiz so the traditional model predicts a vote share of 5080/(5080+2808) = 64.40% for Megumin and a win probability of 99.11%. What actually happened was Megumin won by a scoreline of 12744-2316 and a 84.62% vote share, a full 20% higher and approximately 4 standard deviations away from expected! This would result in Megumin's score rising from 5080 to 6675 making her the overwhelming favourite to win the contest. This is problematic as Megumin likely would not have beaten an opponent from a different show by the same margin so Megumin is rated "too strong" at this point in the contest.
To attempt to fix this (I say attempt because nothing is perfect in statistics) I gathered 39 same-show matchups from seven different contests (Best Character 4, Best Guy 5/6, Best Girl 4,5,6,7) and plotted the expected vote share for the higher seed against the difference in the logit of the actual vote share and the logit of the expected vote share and then centred it on 0.5 (50% vote share). The idea is that two characters from the same show with the exact same score are still expected to have a 50-50 vote share but as the score of one character gets bigger than the other the vote shares become more and more lopsided than what the model predicts. This is what the plot looks like and we can see a general positive trend supporting the idea. A simple linear regression yields a gradient of approximately 5.25 for the line. I should mention that a linear regression may not be perfect since the data does not seem to be perfectly linear. However it is reasonably close to linear as 36/39 (~92%) of the residuals lie within 2 standard errors of the fitted line.
To calculate the new expected vote share for Megumin vs. Wiz we do the following:

Vote Share = logit^-1(5.25 * (0.6440 - 0.5) + logit(0.6440)) = 0.7939 

Which means instead of a >20% overperformance Megumin overperformed by just ~5% or roughly 1 standard deviation away from expected, her new win probability is effectively 100.00% (to 5 sf). This was the distribution of the difference in expected vote share from actual vote share before the adjustment (mean = 8.12% overperformance) and this is the distribution after the adjustment (mean = 1.56% overperformance). The mean being slightly above zero shows that it still isn't perfect but it is in line with the distribution of unique-show matchups (1.24% overperformance for the higher seed) which is a good thing as it means characters won't be punished or rewarded for being in a same-show matchup versus other characters in the bracket.
That's all for now, in the future when the dataset of same-show matchups get larger I hope to refine the regression coefficient to be a little more accurate. If more evidence emerges suggesting a linear regression is not suitable I may look into changing the adjustment.

Introduction

For a while now I’ve wondered how one could predict the winner of the contests of anime by using the numbers behind each character. What I would like to produce is a table for each character in the contest with the probabilities of them reaching a certain round such as the finals bracket, winning the whole thing or even just making it to the last 256 for a lower-seeded fan favourite. This would be a bit like what FiveThirtyEight have created for the UEFA Champions League, and ideally one could look at past forecasts to see how well the model forecasted the future.
But how exactly do you assign a probability for one character to receive more votes than their opponent? You could make some complex formula involving the seeds based on previous contest data – indeed statistically the higher seeded character wins around 90% of all matchups, but seeds don’t tell the full story. The seeding of a character is based on the number of votes they receive in the elimination round. In the elimination round voters will vote for any number of characters that they deem worthy of entering the bracket proper. The top 512 get in with the one that received the most votes seeded as #1, the second most voted as #2 etc. Often the top seed isn’t necessarily the most feared character in the contest. Best Guy 6 had Mumen Rider seeded at number 1, yes the side character from One Punch Man outseeded not only the protagonist of the show but a further 510 male characters who were in the running this year! Unsurprisingly Mumen Rider didn’t last as long as his top seed would suggest as he bowed out in Round 4 to 65th seed Jotaro Kujo!
Moreover the actual numbers of the seeds mean nothing in a statistical sense. If seed #1 had 2000 votes, seed #2 had 1500 and seed #10 had 1400 votes in the elimination stage respectively then in terms of raw popularity seed #2 is closer to seed #10 than seed #1 despite the numbers saying otherwise. Thus it is important to consider the elimination votes instead of the seedings.
So it’s clear that while a model must take seedings into account, they aren’t the be-all-end-all of the story and how a character performs against other characters once the main contest gets going is much more important. There are a couple key things you can look out for to identify which characters are overperforming or underperforming their seeds; firstly the vote share which is simply the number of votes a character receives in a matchup divided by the total number of votes for both characters.

E.g. If character 1 beat character 2 by a scoreline of 1500 to 500 then the vote share for character 1 is 1500/(1500+500) = 0.75, or 75% compared to 25% for character 2.

If a character has consistently had a higher vote share in previous rounds than the opponent they are going up against then that signals that there is a good chance they will win the matchup, irrespective of the seeding because they are beating opponents in a more convincing manner. Another key thing to look at is the strength of the opponents faced so far – this is a bit vaguer to explain in words but you can often tell when a character has made it far into the contest by beating bums versus an opponent who has had to knock out several protagonists and pulled off a couple upsets to get where they are.
In summary a good predictive model should take into account three things:

1. The seeding of the characters, based on the number of votes received in the elimination rounds.
2. The vote shares achieved in the contest so far.
3. The strength of the opponents faced so far.

The Model

(There is a little bit of mathematical/statistical knowledge required to understand in this section, you can skip to the example further down if you do not wish to read it and still get a good idea of how the model works.)
I propose the following model, for which we can make predictions from:

For any particular first round matchup M between character 1 and character 2 let X1 and X2 represent the number of votes character 1 and character 2 receive respectively.
Let N := X1+X2 be the total number of votes in M. Define V1 := X1/N and V2 := X2/N to be the vote shares of character 1 and character 2 respectively (note that V1 and V2 are random variables).
Let s1 be the number of votes character 1 received in the elimination round and let s2 be the number of votes character 2 received in the elimination round (note that these values are constants and not random). We shall call these values the score for the characters.
Finally define t := s1+s2 to be the total number of votes for either character in the elimination round and let p1 := s1/t and p2 := s2/t be the proportion of votes for character 1 and character 2 in the elimination round respectively.
Then under this model we make the assumption than V1 and V2 are Normally distributed random variables with means p1 and p2 respectively and have the same variance σ^2.

These assumptions aren’t going to be 100% true for each matchup, to see why note that a voter can vote for both characters in the elimination round so s1 and s2 may contain the same voter whereas X1 and X2 cannot since a person can only vote for one of them in the contest proper. This is exacerbated when two characters from a very popular show that have been dominating opponents meet up in a later round – on paper it looks like it should be close to a 50-50 split but more often than not it is a very one-sided affair because the voter pool is virtually identical for both. The proportions observed in previous rounds are irrelevant because one character may be a more established fan-favourite than the other. In other words the more distinct the voter pool of the two characters is the stronger the assumption that the expected vote share follows the proportions from previous rounds.
The second assumption is that the vote shares follow a normal distribution with identical variance for each character. I will address this assumption later, though do note that empirical evidence suggests that the standard error (used to estimate the standard deviation) is approximately 0.05 in the early rounds and jumps up to 0.10 in round 6 and the finals bracket.

Computing Probabilities with this model

What we would like to predict is the probability that (w.l.o.g.) character 1 receives more votes than character 2 given the observed elimination round votes, that is to find Pr ( X1 > X2 | s1, s2 ). Then by using the model assumptions and the properties of the Normal distribution,

Pr( X1 > X2 | s1, s2 ) = Pr( X1 > X2 | p1, p2 ) = Pr( X1/N > X2/N | p1, p2 ) = Pr( V1 > V2 | p1, p2 ) = Pr( V1 – V2 > 0 | p1, p2 ) = Pr( 2V1 – 1 > 0 | p1, p2 ), since V2 = 1 – V1 = Pr( D > 0 ), where D := 2V1 - 1 ~ Normal(p1-p2, 4σ^2)) = Pr { [D – (p1 – p2)] / 2σ > [0 + (p2 – p1)] / 2σ } = 1 – Φ((p2 – p1)] / 2σ) 

Where Φ: ℝ → [0,1] is the Cumulative Distribution Function of a Standard Normal random variable.

Updating the score

So we have found the estimated probabilities that a character wins a particular matchup. Now suppose we observe what actually happened in round 1 and the winners progress to the next round, how do we make predictions for the future rounds? This is done by updating the score to match what we have observed.
Let x1 and x2 be the observed number of votes for characters 1 and 2 respectively and suppose (w.l.o.g.) that character 1 is the winner (so x1 > x2). We compute the observed score, s1*, for character 1 as s1* := t * x1/(x1+x2) and redefine the score of character 1 to be the observed score, that is set s1 <- s1*.
The above process can now be repeated in round 2 and beyond.

Justifying The Normal Assumption

For any particular character we assumed that V ~ Normal( p, σ² ), to test this assumption we can look at a sample distribution of (V – p) which should follow a Normal distribution with zero mean and variance σ² . I looked at data for two different contests: Best Guy 6 and Best Girl 6, both of which took place in the last year and aggregated the differences by round. I wanted to look at four things to test the assumption:

• The mean should be approximately zero.
  
• A histogram and a Normal quantile-quantile plot for a visual check to see if the data matches a Normal distribution. The histograms should follow a bell-shape curve and the Q-Q plots should follow a straight line if the data is Normally distributed.
  
• A Shapiro-Wilk test for normality. If the S-W test gives a p-value smaller than 0.05 then there is significant evidence that the data is not normally distributed.
  
• The standard error should be roughly the same in the early rounds and rise in the later rounds as the contest attracts more attention, introducing newer voters and making the finals bracket more volatile.
  

Sample distribution of (V-p) in Best Guy 6 by round
On visual inspection it seems that the data does follow a Normal distribution in each round and the S-W test agrees with this conclusion with the exception of round 2 when there was a big outlier in the matchup between Ainz ooal Gown and Cocytus. Based on the scores for both characters Ainz was expected to win with a vote share of ~65% but instead won with a massive 88% share for a difference of 23%. This is the downside of the model I was speaking about earlier, since both Ainz and Cocytus are in the same show the pool of voters voting for both characters is virtually identical and so we cannot make the normal assumption for this matchup. If you remove this matchup from the data then the S-W test gives a non-significant p-value of 0.3122.
Sample distribution of (V-p) in Best Girl 6 by round
Similarly the data from Best Girl 6 also seems to follow the Normal distribution with the exception of Round 1 which saw a massive upset between 406th seed Himari Takanashi and 107th seed Yui. This upset seems to be some form of SAO spite-voting (which is funny since Asuna would go on to win the contest) and highlights a second flaw of the model in that it can’t really predict spite-voters or strategic voters since they represent a different population to those that have voted in a characters’ matchups so far. Removing this outlier gives a non-significant S-W test p-value of 0.07224.
In both contests we see that the standard error stays relatively constant at around 0.05 until you reach round 6 (last 16) when it seems to double to around 0.10. The model will incorporate this by having the standard deviation be 0.05 until round 6 when it will change to 0.10. One reason for this increase in variance could be the large jump in people voting in later rounds as the contest gets bigger exposure. Finally the means for each round are slightly above zero suggesting that characters with higher scores (usually higher seeds) typically overperform relative to their expected vote share, this is because the differences for each round are taken with respect to the higher seed. There are a number of possible reasons for this, one being that the population of voters who have seen both characters may aggressively favour the higher seed over the lower seed, skewing their result. Still the mean is close enough to zero that the assumption seems valid.

Example

That all might seem like a lot to take in so I think an example will make things clearer. Let’s suppose we’re in a simple 4-girl contest and the matchups are Holo vs. Megumin and Kaguya vs. Mai with the winners facing off in the final. Each girl received the following number of votes in the elimination round to determine their seedings #1-#4:

Seed	Girl	Elimination Round Votes (score)
1	Kaguya	2600
2	Megumin	2400
3	Holo	2400
4	Mai	2400

By just eyeballing the numbers you can tell that Kaguya should be the favourite over Mai while Megumin and Holo should each have a 50% chance of advancing but what does the model say?
Kaguya vs. Mai
Consider Kaguya as character 1 and Mai as character 2 then p1 = 2600/(2600+2400) = 0.52 and p2 = 2400/(2600+2400) = 0.48. These are not the probabilities for each character to advance to the next round but instead are the expected vote shares for each character (52% for Kaguya and 48% for Mai). To find the probabilities that either character advances we use the equation derived above based on the model,

Pr(Kaguya wins) = 1 - Φ((0.48 – 0.52)/(2 * 0.1)) = 1 – Φ(-0.2) ≈ 0.579. 

Which implies

Pr(Mai wins) ≈ 0.421. 

So Kaguya is the clear favourite and is expected to win around 58% of the time. Now suppose the actual results come in and big shock! Kaguya loses by a scoreline of 4500-5500, or a 45-55 voter share ratio. Since Mai has won and moved on to the next round we need to update her score, her new score is the value her score should have been to minimize the difference which is

Mai's new score = (2400 + 2600) * 0.55 = 2750. Kaguya's new score = (2600 + 2400) * 0.45 = 2250. 

Note that Kaguya’s observed score falls down to 2250 so that their new scores perfectly reflect the 55-45 observed ratio.
Holo vs. Megumin
With the same setup as above we have that p1=0.50 and p2=0.50 and the probability that Holo advances to the next round is:

Pr(Holo wins) = 1 - Φ((0.50 – 0.50) / (2 * 0.1)) = 1 – Φ(0) = 0.500 => Pr(Megumin wins) = 0.500. 

So there is a 50% chance that Holo wins and a 50% chance that Megumin wins. Now suppose the results come in and in classic anime fashion Holo also wins by a scoreline of 5500-4500. Note that based on the seeds this would be classed as a big upset since Holo is seeded lower than Megumin but in reality because their votes in the elimination round were identical it isn’t. Holo’s updated score is

Holo's new score = (2400 + 2400) * 0.55 = 2640. Megumin's new score = (2400 + 2400) * 0.45 = 2160. 

and we move on to the final!
Mai vs. Holo
Going into the final Mai (2750) has a higher score than Holo (2640) despite winning by the same victory margin in the previous round. This is because Mai defeated a stronger opponent than Holo did, which was the third thing we wanted our model to incorporate. With the same setup as above we have that p1 = 2750/5390 ≈ 0.51 and p2 = 2640/5390 ≈ 0.49 and so

Pr(Mai wins) = 1 - Φ((0.49 – 0.51) / (2 * 0.1)) = 1 – Φ(-0.1) ≈ 0.540. 

Which implies

Pr(Holo wins) ≈ 0.460. 

So we expect Mai to win the final against Holo approximately 54% of the time. This is nice to compute but we had to wait and see who would be in the final to find out what their chances of winning the contest was, how can we find out the probability that one of the girls would win the whole thing back in round 1? Let’s use Holo as an example.
Finding Holo’s chances of winning in round 1
The probability Holo wins the contest is the same as the probability of Holo reaching the final multiplied by the probability Holo wins in the final conditioned on her getting there. We already computed the first probability to be 0.500 and by the Law of total probability the second probability is

Pr(Holo wins the final | Holo reaches final) = Pr(Holo beats Mai) * Pr(Mai reaches final) + Pr(Holo beats Kaguya) * Pr(Kaguya reaches final) = (0.50 * 0.421) + (0.421 * 0.579) ≈ 0.454 

since her only possible opponents are Mai or Kaguya and we don’t yet know which one will reach the final. Thus the probability Holo wins the contest when all four girls are remaining is 0.500 * 0.454 ≈ 0.227. Note that this is not exactly one in four because Kaguya’s high score weighs the chances more in her favour. If we compute the probabilities for the other three girls we find that:

Girl	Win prob in round 1	Percentage
Kaguya	0.335	33.5%
Megumin	0.227	22.7%
Holo	0.227	22.7%
Mai	0.211	21.1%

So you would expect Kaguya to be a big favourite to win the whole thing out of the four, but more often than not someone other than her will win.

Generalising to bigger contests

If you’re savvy to how the above computations work, you’ll notice that as the number of rounds in the knockout contest increases (resulting in the number of participants increasing by a factor of 2 raised to the power of the number of rounds) the number of computations required to compute the overall win probabilities drastically increases. Finding the win probabilities of a 512-man contest in round 1 can only ever be done by a computer and so that’s what I set out to do. You can find my script (written in R) used to generate the output files in a folder in the Outputs section. I won't claim it’s optimised, indeed forecasting the winner from round 1 takes several minutes to compute on my old laptop but it gets the job done and later rounds fly by almost instantly. If you want a fun challenge try and write a script that computes the probabilities in a faster than exponential order of time.

Outputs

Below is a download link to a folder containing the probability forecasts for the recent Best Girl 6 and Best Guy 6 contests, which I used as a template to write my script. The script is also included along with a readme file to help you recreate the outputs. Please let me know if the link is broken!

Google Drive download link

Best Girl 6

Megumin was the clear favourite going into round 1 as she boasted a massive 3028 adjusted votes in the elimination rounds, which was significantly higher than second seed Aqua (2880) and third seed Holo (2663). This is reflected in the pre-contest probabilities as Megumin was given a 36% chance to win compared to 25% for Aqua and 15% for Holo. This probability increased further in round 2 after she won her round 1 matchup with a 91% vote share – the highest of the entire contest.
Megumin remained the strong favourite until round 4 at which point cracks began to show in her dominance – she was still doing well but so was Holo, who also had an easier ride to the finals as Mikasa, Mayuri and Saber were all still alive on Megumin’s side of the bracket. By the end of round 5 Holo took the lead as Jibril and Hachikuji had suddenly emerged as strong candidates in Megumin's half of the bracket. Mayuri was no longer looking like a pushover for Megumin and indeed Megumin would bow out in arguably the biggest upset of the contest to her in the next round leaving Holo as the clear favourite… Or so you would think, but Holo herself had a relatively poor round 6 as well, defeating the weak Yunyun by a smaller margin than expecting whilst Aqua and Mikasa posted dominant victories against tougher opposition. Mikasa would crush Mayuri in the quarter-finals to become the new favourite after Holo bowed out in a very surprising loss to Winry.
Also flying under the radar this whole time was Yuuki Asuna who in round 5 had under a 1% chance to win the title. Her stock had risen though after knocking out the dangerous Jibril in convincing fashion in round 6. With Mikasa as her quarter-final opponent she was given a 32% chance of winning, but she defied the odds and won in a dominant fashion to set up an unlikely final with Winry, who similarly defeated Aqua in equally convincing style!
The estimated probability that Asuna would make the final was 17% in the pre-contest and only 9% for Winry, at this point Asuna was deemed the favourite by the model, and was given a 59% chance of defeating Winry. The predicted vote share was 52-48 in Asuna’s favourite which she demolished by taking home the sixth crown with a whopping 63% of the vote!
Model Accuracy in Best Girl 6
Overall the model correctly favoured the winner in 466/511 matchups (91.2%) which was higher than the 460/511 matchups (90.0%) matchups won by the higher seed, suggesting evidence that the model predicts as good or better than just predicting the higher seed to advance. The success rate by round is broken down below:

Round	Correct Predictions (Model)	Correct Predictions (Seeds)
1	235/256 (91.8%)	234/256 (91.4%)
2	122/128 (95.3%)	120/128 (93.8%)
3	57/64 (89.1%)	55/64 (85.9%)
4	29/32 (90.7%)	27/32 (84.4%)
5	14/16 (87.5%)	14/16 (87.5%)
6	5/8 (62.5%)	5/8 (62.5%)
Finals	4/7 (57.1%)	5/7 (71.4%)
Overall	466/511 (91.2%)	460/511 (90.0%)

Best Guy 6

Best Guy 6 was a much more different affair to Best Girl 6 in that the elimination round votes for the top seeds were a lot closer together. This is reflected in the probabilities as seven characters were given a 5% probability or greater of winning the whole thing in the pre-contest (as opposed to four in Best Girl 6). Note that the number one seed, Mumen Rider is quickly identified as being seeded too high and is actually considered the underdog in his round 4 matchup against 65th seed Jotaro Kujo, who he lost to.
I remember in the early rounds the perceived “big three” were Reigen Arataka, Satou Kazuma and Edward Elric and indeed after round 2 these were the three favorites according to the model, though Shirogane Miyuki and Levi Ackerman were also identified as strong candidates.
Kazuma became the outright favourite next after crushing his round 3 opponent with a 86% vote share – a dominant showing that none of the other favourites could reply to. Second-favourite Edward Elric went a bit off the boil in rounds 4 and 5 – he still won handily but not by enough to keep pace with Reigen and Kazuma who shared the title of favourite for those rounds.
Everything changed in round 6 though – Kazuma survived a scare against Killua Zoldyck, winning by just a single vote whilst Reigen saw opponents in his half of the bracket grow stronger. Levi became the second favourite at this point whilst Edward Elric emerged as the most likely character to win after crushing Alphonse, though admittedly his stock may have rose a little too high since Alphonse is from the same show after all.
Levi proved his superiority over Reigen in the quarter-finals as he beat him by a margin pretty similar to what the model predicted. Interestingly Saitama beating Kazuma wasn’t so out of left field as I thought at the time; according to the data he had a 45% chance of making it to the semi-finals. It was in the semi-finals that one of the biggest upset of the contest occurred when Saitama defeated Edward Elric to book his place in the final against Levi who at this point was crushing opponents left and right. Saitama was given only a one in five shot of beating the titan-killing prodigy and he did not take it as Levi won by an even more comfortable margin than he was already predicted.
Model Accuracy in Best Guy 6
For Best Guy 6 the model correctly favoured the winner in 470/511 matchups (92.0%) which was higher than the 460/511 matchups (90.0%) matchups won by the higher seed, suggesting further evidence that the model predicts as good or better than just predicting the higher seed to advance. The success rate by round is broken down below:

Round	Correct Predictions (Model)	Correct Predictions (Seeds)
1	241/256 (94.1%)	241/256 (94.1%)
2	121/128 (94.5%)	118/128 (92.2%)
3	57/64 (89.1%)	56/64 (87.5%)
4	26/32 (81.3%)	23/32 (71.9%)
5	14/16 (87.5%)	12/16 (75.0%)
6	7/8 (87.5%)	6/8 (75.0%)
Finals	5/8 (62.5%)	5/8 (62.5%)
Overall	470/511 (92.0%)	460/511 (90.0%)

In summary over the two sample contests the model correctly favoured the winner in 936/1022 matchups compared to 920/1022 if you used a simple model that just favoured the higher seeds. This corresponds to an error rate of 8.4% for the Normal model versus an error rate (AKA the upset rate) of 10.0% for the simple model.

Final Words

The Normal Model seems to achieve the three things we set out to do and based on data from recent contests while also having good predictive power. With that said there are some improvements and adjustments that could be made to make it even better. The first thing would be to deal differently with matchups between characters from the same show; these are normally one-sided and can result in artificially inflated score values for the winner. A good example of this was Ainz ooal Gown’s dominant win over Cocytus in round 2 of Best Guy 6 giving him a much higher score than he should have had at that stage. He would lose to Gilgamesh (who was higher seeded) in the next round despite being predicted to be the strong favourite because of this higher score, whilst simultaneously passing on some of the inflated score points to Gilgamesh, creating a knock-on effect. One solution would be to freeze the scores for characters in same-show matchups. Secondly you could experiment with the value of the standard deviation and possibly vary it depending on the seed of the character. You could also introduce a weighting parameter to the score updating function so that earlier rounds are weighted a little heavier than they currently are. In the end I decided to stick with the vanilla model because the simplest is usually the best (and I didn’t fancy testing stuff for another couple days haha!)
I hope you found this to be an interesting read; I will be posting the updated probability forecasts for each girl every day in the Best Girl 7 contest threads along with the daily results post. If you have any feedback on the model please let me know, this was a very fun project to take on!

Although Atletico Madrid already sits in 2nd place, they are only two points clear from Lokomotiv Moscow. The Spanish side has boked just one victory so far in the competition, the one in the second round after being 2:1 down to Salzburg. Diego Simeone’s side has troubles in finding the back of the oppositions’ net in this competition, as they scored only four times. In the opening match of the Champions League campaign, they lost to Bayern Munich 4:0. However, Joao Felix and the lads haven’t lost eight times in a row in all competitions, and they want to extend that unbeatable streak. Jan Oblak managed to keep the clean sheet four times in a row, but he’s going to have a tough challenge in this one. Diego Costa, Luis Suarez, and Lucas Torreira are going to be sidelined for this clash.
Bayern Munich travels to Madrid without any pressure, as they have already secured their place in the eight-finals. With four wins in the current Champions League campaign, they tied 15 consecutive victories in this competition. The defending champions missed the chance to win only once in the last 13 matches, although they conceded seven times in a row. Robert Lewandowski and the lads are still lethal in front of the oppositions’ goal, as they netted 15 times in this competition, in addition to 31 goals in the Bundesliga. Several players will not be at Hans Flick’s disposal for this clash, including Alphonso Davies, Joshua Kimmich, Javi Martinez, Corentin Tolisso, and probably Lucas Hernandez.

Full-Time Winner Odds & Prediction

Atletico Madrid needs a victory in this clash, but Bayern Munich will do everything to keep their unbeatable run going, despite several setbacks with injured players. We won’t be surprised if this game ends in a draw.

Goals Market Prediction

Although the hosts haven’t conceded in the last four games, Bayern Munich is capable of scoring against any team. Also, the visitors didn’t manage to keep their net intact in the past seven matches, so we think both teams will score in this one.
Draw @ 4.00
BTTS Yes @ 1.60
Correct score 1:1 @ 7.50
Read the full analysis here and share your opinion with us!

[Official||LIVESTREAM] 2020"Atalanta vs Liverpool" liVe STrEaMs-reddit Liverpool and Atalanta clash at Anfield Road in a very interesting game for Matchday 4 of 2020/21 UEFA Champions League. Check out the match preview, information, storylines, predictions and odds, and how to watch or live stream the game.

[![official@LiVeStrEAM @]]Bayern Munich vs RB Salzburg LiVe StreaMs@rEddIT

📲💻➲➧►watch live Bayern Munich vs RB Salzburg➲➧►

📲💻➲➧►watch Bayern Munich vs RB Salzburg➲➧►

Bayern Munich receive RB Salzburg this Wednesday trying to secure their spot in the next round of the UEFA Champions League 2020/21. Check out the match preview, information, storylines, predictions and odds, and how to watch or

[Official||LIVESTREAM] 2020"Atalanta vs Liverpool" liVe STrEaMs-reddit

📲💻➲➧►watch live Liverpool vs Atalanta➲➧►

📲💻➲➧►watch Liverpool vs Atalanta➲➧►

Liverpool and Atalanta clash at Anfield Road in a very interesting game for Matchday 4 of 2020/21 UEFA Champions League. Check out the match preview, information, storylines, predictions and odds, and how to watch or live stream the game.

!2020![Official!Reddit]"Liverpool vs Atalanta" liVe STrEaMs-reddit Liverpool and Atalanta clash at Anfield Road in a very interesting game for Matchday 4 of 2020/21 UEFA Champions League. Check out the match preview, information, storylines, predictions and odds, and how to watch or live stream the game.

oFFICIAL@]"Atalanta vs Liverpool" LiVe StreaMs@rEddIT

📲💻➲➧►watch live Liverpool vs Atalanta➲➧►

📲💻➲➧►watch Liverpool vs Atalanta➲➧►

Liverpool and Atalanta clash at Anfield Road in a very interesting game for Matchday 4 of 2020/21 UEFA Champions League. Check out the match preview, information, storylines, predictions and odds, and how to watch or live stream the game.

[![official@LiVeStrEAM @]]Bayern Munich vs RB Salzburg LiVe StreaMs@rEddIT Bayern Munich receive RB Salzburg this Wednesday trying to secure their spot in the next round of the UEFA Champions League 2020/21. Check out the match preview, information, storylines, predictions and odds, and how to watch or

After an abysmal start of the season, the home side managed to stabilize its form and enter a winning streak. PSG opened the domestic campaign with two straight narrow defeats against Lens and fierce rivals Marseille. It seemed that they simply didn’t have enough time to prepare for new challenges after losing in the Champions League finals against Bayern Munich. Also, the Saints couldn’t count on their star players in that period. However, Thomas Tuchel’s side managed to get back on the winning track with five straight wins, and PSG climbed to the 2nd spot. Kylian Mbappe and the lads have scored ten times in the last two games, while in the previous four occasions, the hosts conceded just once. PSG is full of confidence, and they want to book their first three points in the continental competition.
On the other hand, Manchester United fans cannot be satisfied with displays of their lads. The Red Devils opened the Premier League campaign with a shocking 3:1 defeat against Crystal Palace at Old Trafford. Although Ole Gunnar Solskjaer’s team struggled in the next clashes, they tied three victories. However, they fell to a heavy 6:1 defeat against Tottenham Hotspur before the international break. Bruno Fernandes and his teammates managed to improve the game somehow last weekend, as they were confident against Newcastle United with a 4:1 victory. Manchester United needs to stabilize its form urgently if they want to be competitive this season.

Full-Time Winner Odds & Prediction

Even though many vital players will be sidelined for this match for the home side, PSG is still the favorite in this clash. We believe the hosts will keep up where they left off and celebrate a win in this clash.

Goals Market Prediction

Manchester United is capable of scoring against any side, while the defensive work is their primary issue. We don’t need to mention that PSG is a goal machine, and they were incredibly efficient lately. We expect to see an exciting clash and goals in both nets.
PSG to win @ 1.55
BTTS Yes @ 1.60
Correct score 3:1 @ 11.00
Read the analysis here and share your opinion with us!

Alright, yesterday I created a quick match thread to calm down the increasing flow of new redditors on the sub, we went up to 3,598 online users on the subreddit at some point...
So today, let's have a serious, interesting and calm discussion about what happened during the game yesterday.

Atalanta 1 — 2 Paris Saint-Germain

Quarterfinals

Venue: Estádio da Luz
Attendance: Unkown
Referee: Anthony Taylor (ENG)
Kick-off: 21:00 local.

Lineups

Paris Saint-Germain lineup 4-4-2: Navas; Bernat, Kimpembe, T. Silva (Captain), Kehrer; Gueye, Marquinhos, Herrera; Neymar Jr., Icardi, Sarabia
Atalanta lineup 3-4-2-1: Sportiello; Toloi, Caldara, Djimsiti; Hateboer, De Roon, Freuler, Gosens; Gomez, Pasalic; Duvan Zapata

L'EQUIPE RATINGS

MATCH REPORT

GOALS:

• [1]-0 - Mario Pasalic (27')
• 1-[1] - Marquinhos (90')
• 1-[2] - Choupo Moting (90+3')

MATCH HIGHLIGHTS

• CBS Extended Highlights
• UEFA Highlights
• Optus Sports Highlights

STATISTICS:

	Atalanta	PSG
Shots (on target)	9 (4)	16 (6)
Ball Possession	39%	61%
Passes	367	593
Distance Covered	108.2 kms	104.9 kms
Fouls	29	13
Corners	4	3
Offsides	2	0
xG	0.58	3.29

FULL STATISTICS
FULL MATCH REPORTS:

UEFA: "Tuchel's side pull off stunning late comeback."

Paris were so nearly the first to land a blow, Neymar racing clear in the third minute only to slot wide with just Marco Sportiello to beat. Atalanta soon found their feet, with Hans Hateboer’s back-post header – smartly saved by Keylor Navas – serving notice of their growing intent.
La Dea’s ascendancy was rewarded before the half-hour when Duván Zapata played the ball neatly into the path of Mario Pašalić, whose first-time effort from just inside the penalty area arced perfectly past Navas.
Paris carved relatively little out of their superior possession for much of the second half, at least until the 90th minute when Neymar squared for Marquinhos to tap in from close range.
Crestfallen, Gian Piero Gasperini's side had barely caught their breath when substitute Kylian Mbappé raced down the left and crossed for fellow replacement Eric Maxim Choupo-Moting to help himself to an added-time winner.

The Guardian: "PSG's late, late double ends Atalanta's Champions League dream" by Barney Ronay

In the end it took £400m of attacking talent, 93 minutes of increasingly frantic football, and a man who mustered up five goals in a single year at Stoke to wrench a thrilling Champions League quarter-final the way of Paris Saint-Germain.
It had to be Neymar, one way or another. The world’s most expensive footballer had played for much of the game like a man trapped inside another kind of storyline, finding space fluently but shooting at goal like a man wearing wooden clogs.
Gian Piero Gasperini’s Atalanta underdogs had played with familiar verve to lead 1-0 with 90 minutes already up. At which point PSG’s labouring star machine finally found some relief.
Eric Maxim Choupo-Moting, on as a 79th-minute substitute, swung a fine cross into the box, and the ball was deflected in off a combination of an Atalanta defender and Marquinhos.
Three minutes later, with extra time looming, Neymar played an extraordinary, high-pressure pass, taking the ball in a small space outside the area, freezing the moment, then sliding the ball inside José Palomino to meet the run of Kylian Mbappé. His low cross was equally precise. Choupo-Moting turned the ball into the net to set up a semi-final with either RB Leipzig or Atlético Madrid next Tuesday.
Atalanta will remain the wider story and not just because of another stirring show from this high-intensity team of low-cost parts. A Champions League quarter-final represents an all-time high, albeit one that will be coloured by other emotions. The horrors of March in Bergamo are well-documented. This is a region still processing its grief. Freewheeling success on the football field is not a fix or a balm for this kind of real-life pain. But all things considered, it is not a bad place to start.
The other story is, of course, Neymar, and a step into fresh territory for Thomas Tuchel’s version of that oddly brittle PSG star vehicle. Neymar remains, for all the outstanding moments, an object of some frustration among those baffled by the choice to move at his peak to an environment as lukewarm as Ligue 1.
And yet, Neymar did deliver and did so when his team needed it most. There are stronger opponents in the draw but not many better attacks, with Mbappé likely also to be fit for the next round.
The Estádio da Luz had provided an agreeably haunting stage for the latest instalment of midsummer Covid-bal. The late evening kick-off meant Lisbon was at least mercifully cool, a step down from the predicted mid-August firepit.
Atalanta had Marco Sportiello in goal, with Josip Ilicic absent for domestic reasons. PSG were without Ángel Di María and Marco Verratti, with Mbappé on the bench.
The action was frantic from the start. It took Neymar two minutes to produce an astonishing miss, running in on goal unimpeded from the centre circle, then shanking the ball five yards wide of the right-hand post.
Unconcerned by their own suicidally high defensive line, Atalanta set about besieging the PSG end. With 10 minutes gone Hans Hateboer drew a clawing save from Keylor Navas and the game settled into a series of crisp, direct Atalanta attacks, broken up by the occasional white-shirted counter.
For a while that powerful PSG midfield began to dictate its more stately rhythms but it was Atalanta who took the lead. Duván Zapata found space just outside the box, tumbling over as the ball ran on to Mario Pasalic. His first-time shot zinged into the top corner.
The response was familiarly staccato. Neymar produced a nutmeg on Pasalic then pinged a shot just wide. Otherwise Atalanta held their high line and swarmed around the ball, their combinations zippy and urgent where PSG seemed to be playing through a midsummer haze.
Neymar had time before the break to miss another presentable chance, blasting miles over the bar after seizing on a terrible pass from Hateboer and sending Tuchel into mild frenzy in his dugout.
Tuchel has something agreeably pale and gaunt about him at the best of times, striding about the touchline looking like a sad-eyed victorian wraith. In Lisbon, left leg encased in a surgical boot, he could be seen shielding his brow and flinching in horror at times in that first half.
Mbappé, his remaining ace, was sent out to warm up at the interval. And PSG did start with more purpose after the break. Neymar continued to find space but still lacked any sense of edge. A low free-kick on 50 minutes was his fourth shot at goal, his third off target.
With 31 minutes to play Mbappé was on, replacing Pablo Sarabia. His impact was instant, a series of surges down the left offering a threat where PSG had been reliant on Neymar’s twists and spins.
Atalanta flooded that side, surrounding Mbappé with five defenders at times. It looked like being enough, right up until that extraordinary ending.

La Gazzetta dello Sport: "Atalanta, going out like this hurts! Pasalic deludes, but PSG overturns everything in the recovery minutes" by Francesco Fontana

Nerazzurri ahead 1-0 until the 90', then the mockery: Marquinhos' draw and Choupo-Moting's overtaking goal. The Parisians went to the semi-finals and Mbappé's entry was decisive.
Players in tears over an incredible result, unbelievable stuff. Atalanta lost 2-1 on PSG, killing the minutes from 90 minutes (Marquinhos' goal) to 93 minutes, when Choupo-Moting signed the overtaking after Pasalic's great goal at 27 minutes of the first half. The Champions League quarter-finals speak French, in the semi-finals Tuchel's team will go there and now await the winner of Leipzig-Atletico Madrid, on the pitch tomorrow night. In any case, despite the defeat, the Goddess can return to Bergamo with a very high head.
Compared to the hypothesis of the eve there is no news in Atalanta's house: Gasperini bets on Caldara, Hateboer and Pasalic, preferred to Palomino, Castagne and Malinovskyi. So, in his 3-4-2-1, space to Sportiello in goal. In defence also Toloi and Djimsiti with De Roon-Freuler on the middle and Gosens on the left. In front, of course, Gomez and Zapata. On the other side, without Verratti and Di Maria (out for injury and disqualification respectively), Tuchel chooses the 4-3-3 with Keylor Navas between the posts, behind Kehrer, Thiago Silva, Kimpempe and Bernat. In midfield Herrera, Marquinhos and Gueye with Sarabia, Icardi and Neymar in attack (Mbappé starts on the bench). English referee Taylor.
First ring at 3': left Papu from inside the area, para Navas on the ground. The real danger comes 1' later with Neymar who, incredibly, makes a one-on-one mistake with Sportiello kicking out. Super start in this first quarter final. The Goddess is in the game, PSG concedes something. On the 11th assist top of Gomez, who from a distance fishes Hateboer on the second post: dunked by the Dutchman's head, Navas is ok in answering (ditto a few moments later on Caldara, albeit offside). Until the 27th minute, the French push to corner the opponent, then the result changes: a ball to the limit for Zapata who, in a daring way, serves Pasalic, amazing with a left-footed shot to beat Navas on the second post. Goddess ahead. And he'll stay there until halftime, also because Neymar, in the 41st minute, didn't take advantage of Hateboer's mistake (Sportiello's back pass was too short) and sent him out. And by quite a lot. Thrill. We go into the locker room after 1' of recovery.
Tuchel (on crutches due to a sprained left ankle with a fractured fifth metatarsus) is not happy and sends Mbappé to warm up immediately, on the field at 60'. At the same time Gasperini puts Malinovskyi and Palomino in place of Papu and Djimsiti, who just before (57') makes a mistake on the fly: what a chance for the Goddess. In the central part of the shot Atalanta shortens, PSG struggles to create. And it's no coincidence that, at this stage, Sportiello is still watching: it will seem absurd, considering adversary and level of competition, but that's how it is. The changes continue: Gasp throws in the fray Muriel (out Pasalic), Tuchel risks more with Paredes and Draxler for Gueye and Herrera. We're at 74', when Sportiello puts his big foot on Mbappé's right. Navas doesn't make it because of an injury accused just before, on 79' space for Sergio Rico, the reserve goalkeeper. Eleven minutes to go, Mbappé made another mistake (very good Palomino in closing). Gasp called for Castagne and Da Riva, midfielder of the 2000 class at his debut in the Champions League. PSG tries, tries and tries. And at the end, in the 90th minute, he finds a draw with Marquinhos, who puts it in the fray from a few steps. And in the 93rd minute he even overtakes Choupo-Moting, who receives a "chocolate" from Mbappé. Unbelievable, unbelievable stuff. Although a round of applause for this Atalanta is certainly not to be missed.
POST-MATCH INTERVIEWS &amp; QUOTES

• Thomas Tuchel Post-Match Press Conference.
• Eric Choupo-Moting Post-Match Press Conference.