ONLINELEAGUE.co.uk

At the heart of ONLINELEAGUE.co.uk. Every clash of 2 teams is simulated in a complex process. All factors that make up real football are included.

Two teams compete against each other. That is the core, the heart of football and of ONLINELEAGUE.co.uk. Whether league or friendly match. Games are always simulated and there is no animated visual representation, no live games. The team managers cannot actively intervene in the action during this match calculation.

Of course, the team managers can determine exactly how their own team should react to certain situations, for instance, when which players should be substituted and when the tactics should be changed. This happens in the run-up to a match via the tactical settings, which are dynamically linked to certain conditions.

More information on tactical settings

Initial situation

In the lead-up to the actual match calculation, the two competing teams are matched against each other to determine where they currently stand. The following aspects play an important role.

PLAYERS

Each team is made up of its players. Each player has individual abilities.

To utilise these, a player must be fit. The higher his fitness rating, the better he can tap into his potential.

Form is also crucial. Is a player having a good day? Is he on a roll?

There are consistent players who can regularly play to their full potential, others fluctuate in their performances.

Young, highly talented players can excel in a single game and give an indication of the full potential they can realise.

The current, individual performance potential of each player for the upcoming fixture is determined on the basis of these parameters.

TEAM

Every style of play and tactical playing system requires certain key positions with different individual abilities. Or in other words: the right player for the right position in the system. Only then can a playing system be implemented successfully. If this is not the case, it has negative effects. A system with a designated playmaker, for example, is such a key position. If this playmaker lacks technical skill, has a bad shooting technique or has a poor tactical understanding of the game, his team will almost certainly have a hard time getting into the game.

So every position in the chosen playing system is subjected to a TARGET/ACTUAL analysis and evaluated. Taking the current performance level of each player into consideration, this results in an evaluation of each position, each part of the team and finally the whole team. This evaluation also takes the parts of the team and players of both teams that clash on the field into account.

IS THE TEAM HAVING A GOOD RUN; WHAT IS THE MORALE LIKE?

Team ratings are influenced by a supposed „run“/the team's morale: positive in the case of a winning streak - negative in the case of a winless streak.

HOME / AWAY

Every team has a slight home field advantage at the start of a new season. As the season progresses, the home field advantage changes with the achieved results. If a team wins frequently at home, this increases the home advantage. If there are a lot of defeats suffered at home, this can also turn into a home weakness, which then has a negative effect on the morale of the team.

The same is true for away games.

STADIUM / FANS

Having many spectators in your own stadium has a positive effect on the home team. A large, empty stadium, on the other hand, tends to negatively effect the home team, especially if a large portion of the terrace remains empty.

The neural network

ALL-KNOWING DATABASE

At the heart of the ONLINELEAGUE.co.uk match calculation is the neural network, a huge database in which millions of football data are stored. This data has been drawn up and constantly refined together with football experts. It contains information about how different playing systems in different set-ups react to each other. This means that there is not only information about how a 4-3-3 offensive system plays against a counter-attacking 4-2-3-1 system, but also how the exact behaviour changes with different player lineups. For example, the 4-3-3 with the best available players in key positions behaves very well. But if the players lack the skills crucial to this system, the result will be rather mediocre. Depending on how the opponent's system, in this case the 4-2-3-1, is set up, the resulting number and types of plays in a game will be completely different.

So the net is filled with every conceivable constellation of systems and player lineups. For example, there is data that predicts how everything will go down when the offensive 4-3-3 has rather slow and not very technically gifted players in the offensive lines, especially on the right side, and the 4-2-3-1 has extremely fast offensive players and well-trained midfielders in terms of shooting and tactical understanding. The data for this scenario will predict more offensive chances for the 4-2-3-1 system despite it generally being the less offensive formation. The reasoning being: certain special systems like the offensive 4-3-3 need absolute specialists in key positions. If these positions are not filled out adequately, this system

will perform far worse than a „standard“ system with exactly the same players. However, an offensive 4-3-3 with the proper specialists is of course a force to be reckoned with.

Now, if a game is to be computed, this neural net is called upon. It operates with the formations and player ratings of the two teams playing each other. Of course, the database doesn't have the exact same constellation as the one requested for the particular game. The neural network now makes a so-called approximation. Out of all the data close to the requested constellation, the data that best matches the request is evaluated. As a result, the neural network now provides us with basic information about the amount and types of attacking efforts of both teams, based on the entered „experience values“. This request process is done repeatedly throughout the game. Since the situation changes again and again throughout the game (players get tired, playing systems change, substitutions are made, injuries occur, morale rises or drops with goals/conceded goals), the net always delivers results in alignment with the current constellation. This means that team A can have the upper hand in the beginning of a game and later, because of tactical changes, substitutions, fatigued players and other influencing factors, team B can take control of the game.

Simulating of plays

The neural network determines not only the number of advances and noteworthy game occurances, but also the manner of the attacks. Does the game strongly feature long passes, short passes, counterattacks or is the game focused on the left or the right wing? Once the styles of play have been determined for team A or team B, this will be fully simulated. This means that the relevant players and opposing players who are directly or indirectly involved in a game situation are determined. This is done for each moment of a play. Considering the individual attributes of the players involved, the so-called „controlled randomness“ is used to decide how the game situation plays out. Depending on the type of attack, there are countless different possibilities of how a play can unfold.

For example, an attack starts with a player in possession of the ball in his own half. What happens/what does he do next?

He starts a dribble against an opponent.
He plays a short pass to the left / right / back / forward
He plays a long pass out to the wing
He is fouled
He is closely marked and commits a foul
He fumbles the ball
Etc.

What exactly happens depends, as described, on the skills of the players and their opponents. But also the playing system, the chosen tactics, the score and the current condition of the player control the random generator, which determines what happens next.

CONTROLLED RANDOMNESS

A simple example shows how controlled randomness works:

Player A is sent towards the opposing penalty area with a pass from his teammate; player B, from the opposing team, tries to beat him to the ball or take the ball from him to ward off the attack.

Player A has a technique rating of 65, player B has a duels rating of 34 (both scores are kept up to date in real time by global stats like fitness, form, etc). Since, in this case, defending the ball is generally a bit easier than securing the ball, the defender gets a bonus factor. His rating is increased by 20%. So the two are now facing each other with the values 65 to 41 (34 +20% = 41). The odds are 65 to 41 in favor of the striker securing the ball. Imagine that in this situation a number between 1 and 106 (the result of 65 + 41) is generated by chance. If the number is between 1 and 65, player A wins. If it is between 66 and 106, player B wins.

This is an oversimplification. In the actual implementation there are many more factors that influence the calculation. But the principle is the same, no matter if one or several skill ratings are taken into account.

Based on this process, all the other moments of a play with all possible outcomes are simulated. The result is a goal, a missed chance, a foul or something else.

Whether a player is injured, fouled or gets a yellow card or even a red card. All of this is determined by controlled randomness, using all the available information on players and game attributes.

This principle follows real footbal in so far that even a player vastly superior to his opponent can always loose a duel. In real life, the absolute underdog occasionally wins a duel or even scores a goal. Even if it happens very rarely, it happens!

BALANCING

And now we come to one of the most important aspects of game calculation: balancing. Meaning, the balancing of how much randomness takes effect in relation to player traits and how the influence of certain player traits is weighed. In our example, without chance, player A would ALWAYS get the ball. With controlled chance, he gets it more often but not always - which seems absolutely realistic. However, the long-term distribution of the two outcomes is checked, questioned and optimized over and over again by the fine tuning of the balancing. For this purpose, we regularly use evaluations and statistics and check for anything out of the ordinary. We deal with both purely statistical/mathematical findings as well as "perceived" anomalies, which are often brought to our attention by the community. Changes to this balancing are usually made gradually, meaning in very small steps. Most of the time, only one setscrew is adjusted in order to localize the effects in this highly complex structure and to be able to clearly assign these effects to this adjustment.

TACTICAL SETTINGS

A team's tactical adjustments, even those made to counteract the playing system and tactical settings of one's opponent, influence both the outcome of the neural network as well as the outcome of the situations that make up simulated game plays. Thus, decisions made by controlled chance are also dependent on the tactical settings.

CONDITIONAL ACTIONS

Actions taken by the team manager due to certain occurances are immediately taken up and implemented in the calculation. If a player is injured and it was determined which player should replace him, the conditional action takes effect and the player is replaced according to the specifications. If there are several conditional actions that can take effect, the action higher up on the list is always used. The higher an action is on the list, the higher its priority.

Of course, such a substitution changes the given situation and the neural net is „consulted“ to get the results that come from this new situation.

CHANGES IN PLAYER RATINGS

The longer a player is on the pitch, the more exhausted he becomes. Of course, his fitness plays a big role in this. For example, if a player has a 95 in fitness, he is well prepared for the match and will usually be well able to play until the final whistle. How fast his fitness will decrease during a match and whether this will lead to a light or heavy exhaustion depends on the player's stamina. The worse a player's stamina is, the faster he will enter a state of exhaustion, which will in turn make the player slower and less concentrated. In the end, almost all of his abilities suffer from this. Even a technically good player will likely not go into dribblings, let alone successfully complete them, if he is very fatigued. Therefore, just like in real football, it is important to quickly substitute fatigued players and bring in fresh players. Since only a maximum of 3 substitutions are possible, careful squad planning is key. What counts are not only the obviously important skills like technique, speed, shooting technique etc. In order to indeed finish a game victoriously, a good portion of stamina is vital. At least for a large part of the players. Players with poor stamina but extraordinary talent in one or few areas are more suitable as super-subs for 20 minutes game time. For such a role, insufficient stamina is forgivable.

CYCLE

The simulation was explained using a single, very small component. Within an entire game calculation, a very large number of these small building blocks are lined up one after the other, interdependently called upon and result in a complete game play.

Each play that is simulated to its conclusion can effect the result of the neural network, which is reconsulted and therefore always reflects the exact state of the game. The next play is determined based on the result of the previous one and again begins with the first building block. This cycle repeats itself again and again until the game is over.

(IM)PROBABILITY

Team A has 20 offensive plays in the game, many good scoring opportunities and far more ball possession. Team B makes it to the opponent's goal 2 or 3 times and wins 1-0. A scenario that is rare but can occur every now and then. Why can it happen? Because it is within the realm of possibility, just like in real football. And because all the probabilities of the simulated game build on individual skill and a pinch of luck (or bad luck). There can be many reasons why a team can create a multitude of scoring opportunities without any player ever being able to find the back of the net: there is simply no prolific goalscorer or the top goalscorer(s) are having a bad day. Or the opponent's goalkeeper is extremely good - in general or just on that day. Or it is simply not meant to be, according to the football gods. We try to take all these possibilities and factors into account in the manner we have described.

It follows, that for each and every match, 90 minutes plus stoppage time (possible extra time and penalty shootout) are simulated from the first to the last minute. There are no predefined scenarios or predetermined game outcomes because one team is supposedly stronger on paper than the other.

It is therefore by design that there will always be results that nobody can really explain. This is the only way a football simulation can approximate all the facets that make up real football.

The performance of a team over the entirety of a whole league season can however always be explained and understood. After all, every team is subject to these circumstances.

For a team manager to be upset over an "unfairly" lost game is not only understandable but to a certain degree also an important aspect of the game, without which football, more than any other sport, would not be as popular as as it is. In real football and likewise in ONLINELEAGUE.co.uk.

Back to the table of contents