Inside the FM27 match engine: how it actually works

caiograco

Member
Joined
Nov 11, 2026
Messages
53
Reaction score
49
In my other thread (FM27 match engine: what we know so far) I've been measuring what things are worth by changing them and counting points. In this one I'm trying to dig into the actual code to answer the why. Like, why is pace and acceleration worth so much, how are they used in the ME, and why other attributes like passing don't seem to matter. I used the an agent to look inside FM27's match engine while it plays real saved matches. It recorded which parts of the engine read each attribute, how often, and at what moment, and then read the busiest parts closely enough to get the formulas they use.

The key learning here is that the engine doesn't play with the numbers on your player's screen. It turns them into match values, rolls dice from them, and spends most of its time on players deciding what to do. The way it uses the dice function is also weird and explains why so many attributes don't make a difference even when they are used by the ME a lot. In fact, it seems (from what I know so far) that pace and acceleration are not called inside the ME more than the other attributes, they are called for the functions that actually matter.

This is still work in progress and some parts I'm more confident than others because we have the actual formulas the game uses and the agent checked it against real matches. Others are our (me and the agent) best reading of what is going on.

I will try to organize everything in comments below so its easier to read and follow.
 
First, let me try to describe how the match engine actually runs a match, since everything else is based on this.

The ME is a step simulation. I think most of us suspected or knew that already. What it does exactly is that It moves the match on a quarter of a second at a time, about 24,000 steps in a match. Nothing happens between steps. Every player and the ball move a quarter of a second, and every decision is taken at some step. About three quarters of the steps are live play, the rest are stoppages (the clock keeps running while the ball is dead), plus half time and full time.

In each step of live play the engine does roughly this, in order:
  1. Moves everyone along their current path at their current speed, and updates how tired they are.
  2. Picks who gets to think this step (not everyone re-thinks every step).
  3. Looks at each team's situation, position/shape and where the ball is.
  4. Deals with a high ball if there is one. i.e. who can reach it, when, and whether to head it or control it.
  5. Lets the chosen players decide. The man on the ball weighs his options, the others choose where to stand and run.
  6. Moves the ball along its flight.
  7. Settles contact, in particular body contests, tackles, and loose balls.
Where the time goes. About 70% of the engine's work is step 5, players deciding what to do. Half of that is the one player on the ball (or about to receive it) weighing his options, the other half is the 21 others choosing where to be. The physics (moving players, the ball, aerial balls, body contests) is only about a tenth. The single busiest piece of the engine is a "how fast can he get there" estimate, which runs for lots of players every time anyone considers a pass. This is where Pace, Acceleration and Agility come in, and I think it's a big part of why they matter so much.

Who thinks when. About half the players re-think in a typical step (10-11 of 22 on average). The rest just carry on with their last decision. Most players re-think every step or every other step, but in some situations the engine holds a player back for half a second, or a full second if a roll on his Concentration comes up low. That's basically reaction speed.

To get a better sense of what is going one, I asked the agent to trace a move. He picked a defensive midfielder who had the ball. When the player had to make the decision it worked as follows:
  1. the engine worked out who he notices around him (his Vision die roll) and whether he reads their runs (his Anticipation die);
  2. it considered a pass to every teammate, working each one out as it would actually be kicked (his Passing, Technique, the foot he'd use);
  3. and for each pass, estimated if the receiver gets there first and which opponents could cut it out, using everyone's Pace, Acceleration and Agility;
  4. it then considered running with the ball (Dribbling, Pace, Acceleration) and shooting (Finishing or Long Shots, depending on the distance);
  5. it then picked an option. We saw the kick happening on the next step.
That one decision read over 13,000 attribute values in a quarter of a second, and most of them were other players' attributes. Every teammate he could pass to and every opponent who could intercept. His own attributes were barely touched.

The choice also isn't simply "take the best option". The engine rolls dice at several points inside the decision, which is why the same player in the same spot doesn't always do the same thing. I haven't figured out exactly how the final option is picked yet.
 
The match engine relies on effective values and die rolls. This was one of the most surprising things for me.

The engine almost never reads the numbers you see on the player's screen. There are two layers in between, and I'm quite confident on both (we (the agent and I) have the formulas and the agent checked them against real matches).

Layer 1: effective values. At kick-off and again at breaks in play (a few dozen times a match), the engine turns each attribute into an "effective" value: the attribute plus the player's match modifiers (mostly condition and sharpness), rescaled. The physical ones fade as a player gets tired. Over one match, a player's effective Agility, Balance and Acceleration dropped by about 10%. This is one of the ways Stamina pays: it keeps your Pace and Acceleration up late on.

Layer 2: dice. At the same moments, the engine rolls a die for 28 attributes, from 0 to the effective value, and until the next roll a lot of the play reads the die, not the attribute. Most mental attributes reach the pitch mainly this way. Passing, Finishing and the other ball skills don't have a die and are read directly. A better attribute still gives a better die on average, but a player's level swings from one spell to the next.

The weird part: most checks only ask if the die came up zero. This is what I mentioned in the first post. When the engine checks a die, most of the time it just asks "is it zero?" (for Decisions, about 255 of its 300 checks). A positive die lets the player consider something, like an option or a small adjustment to his position, and a zero means he doesn't, for the whole spell. How often you get a zero depends on the scale the attribute is kept on:

AttributeChance of a zero die at 5at 10at 15at 20
Decisions, Positioning, Marking, Composure, Concentration, Anticipation, Vision, Technique, Off the Ball4.0%2.9%2.3%1.9%
Work Rate, Bravery, Aggression13%8%6%4.5%
Teamwork8%5%3%2.4%
The skill and mental attributes are squeezed into the top part of their scale, so even a bad player rarely rolls a zero, and +5 moves that chance by about half a percentage point. So Decisions, Positioning or Marking can be read hundreds of thousands of times in a match and still barely move results. Work Rate, Bravery and Aggression are on a steeper scale, which I think is why Work Rate has a floor in the tests (bad at the bottom, flat from about 9 up).

What the engine looks at the most. Counting the dice, per outfield player per 90 minutes: the Vision die about 310,000 times, Anticipation 240,000, Decisions 210,000, Concentration 150,000. Then the physical attributes behind every "who gets there first" estimate: Agility about 90,000, Balance 70,000, Strength 60,000, Jumping Reach 40,000. Pace and Acceleration are only about 25,000 each, Passing 30,000, and Finishing about 100. So Pace and Acceleration aren't read more than the others, they are read in the places that decide what happens.

For me, the above is bonkers. We have all these attributes that actually do something in the ME but that at the end make no difference in the final result.
 
In terms of actual formulas for the physical attributes. It turns out to be pretty simple and mostly linear. These are for a fresh player, on the usual 1-20 scale. The agent checked all of the formulas in an actual running match. For Acceleration and Agility, exactly where they are used is partly my reading.

AttributeWhat it doesAt 10At 20
PaceTop speed: about 8.05 + 0.11 × Pace meters per second9.2 m/s10.3 m/s
AccelerationHow quickly he gets going, in every "who gets there first" estimate: about 2.2 + 0.09 × Acceleration3.14.0
AgilityHow fast he turns when he has to change direction: about 1.55 + 0.0225 × Agility, capped at 2.01.782.0
StaminaHow fast he gets tired when running above 75% of top speed, compared with a Stamina-20 player: 2.5 − 0.075 × Stamina1.75× as fast1×
StaminaHow fast he recovers when jogging: 0.5 + Stamina/400.75×1×
Jumping ReachHow high he jumps: (Jumping Reach + 5)/30 metres on top of his standing reach0.50 m0.83 m

Some notes on each:
Pace is a flat 6% faster everywhere for +5. Top speed is used in every run, and in every estimate of who gets to a ball first, which the engine does for several players every time anyone considers a pass. That's why it's worth so much. It also works from the effective value, so it drops as the player gets tired.

Stamina works twice: it slows down how fast a player gets tired when he runs hard, and speeds up how fast he recovers when he jogs. Both feed condition, which lowers every effective physical attribute. So Stamina is really "keeping your Pace, Acceleration and Agility for 90 minutes", and a high press means more time running above 75% of top speed.

Strength and Balance count exactly the same in body contests, which is the single most frequent check in the engine. Each player gets half from his fixed Strength and Balance and half from his current Strength and Balance dice. Coming in from behind or the side gives a bonus, and the winner has to lead by a margin, otherwise neither wins.

Jumping Reach is about 3.3 cm per point, and if a ball is out of reach he just can't play it. Height counts too: the jump only adds anything once it beats his standing reach, which for a 1.83 m player means Jumping Reach above about 3-4. There's also a part of the engine that reads Strength, Jumping Reach and Heading together, which looks like the aerial duel (our best reading).

Dribbling decides how close to top speed a player can run with the ball. A better dribbler lets the ball run less far ahead and runs nearer his top speed, so Dribbling and Pace multiply each other.

Kick range for long passes and clearances: Strength counts three times as much as Long Shots (for keepers, Kicking and Strength equally), scaled by the foot he uses. Centre-backs use this the most.
 
On the mental side, most attributes work through their dice (see the post on dice above), but a few of them have a pretty clear job.

Vision is how far a player looks, Anticipation is how far he reads ahead. This is the most read part of the whole engine. Every quarter of a second, for every other player, the engine compares that player's distance (times 0.4) with the observer's Vision die. If the die is bigger, he notices that player. If his Anticipation die is bigger too, he reasons with where that player is going to be, not where he is now. On average a Vision-10 player takes in players up to about 40 m away and a Vision-20 player about 60 m, with big swings between rolls. We're confident on the check itself; that the distance is measured from the ball is our reading.

Concentration is basically reaction speed. Four times a second the engine rolls each player's Concentration. When it's limiting how often players re-think (which happens in some situations), a low roll makes the player wait a full second before he can respond instead of half a second. So a low Concentration player is a bit slower to react when the play changes. Confident on this one.

Determination picks the better or worse roll. Before rolling Concentration, the engine checks against a value built from Determination: pass and the player keeps the better of two rolls, fail and he gets the worse. It also changes at half time. Confident on this one too.

Decisions, Positioning, Marking and Composure mostly work as on/off switches: a zero die means the player doesn't consider an option or doesn't make an adjustment, a positive die means he does. Composure comes in for decisions on the ball under pressure (our reading). Like I said in the dice post, the chance of a zero barely moves with the attribute, which is why these show so little in the tests.

Work Rate sets a minimum effort when a player picks how fast to run. Below a certain effort level he may settle for a slower speed than the situation asks for (our best reading). Together with its steep die scale, that gives Work Rate a floor instead of a straight line.

Off the Ball and Concentration hold the line. When a player sets his position against what looks like the offside line, the engine uses his Off the Ball die when his team is attacking and his Concentration die when it's defending. A low die moves him off the line by a random amount: up to 2 m with a die of 0, 1 m with 10, nothing from 20 up. Small, but on an offside line it's the difference (our best reading of what the line is).

Teamwork comes in for awareness of teammates and positioning with the team, and Flair for the more ambitious options on the ball. Both are our best reading so far.
 
Now shooting and goalkeepers, which explains why Finishing and the GK attributes behave so differently in the tests.

Which attribute a shot uses depends on the kind of shot. One part of the engine gives a player his shooting skill for a shot (confident on this one):

ShotAttribute used
Inside 18 yards (the depth of the box)Finishing
Between 18 and 22 yardsa straight blend of Finishing and Long Shots
Beyond 22 yardsLong Shots
HeaderHeading
Direct free kickFree Kick Taking (from long range, mixed with Long Shots)
PenaltyPenalty Taking
That skill is then set against how hard the shot is, sometimes with part of it rolled as a die. This only runs a handful of times per chance, while the player is weighing up the shot. So Finishing is one of the least read attributes in the whole engine (about 100 times a match for an outfield player), and it still decides most shots from inside the box. That's why Finishing converts chances but doesn't create them, and why tables based on xG miss it.

Goalkeepers (our best reading, since these only run when a shot or cross comes in and are hard to catch):
  • Reflexes sets how fast the keeper covers the distance to the ball, and a good roll shortens his reaction further as long as his Concentration die isn't zero.
  • Agility can cap that reaction.
  • Aerial Reach does for him what Jumping Reach does for outfield players: how high he gets.
  • Handling goes into a random check on how much of the shot's pace he takes off (catch or parry, most likely).
  • One on Ones and Command of Area work through their dice when deciding whether he comes off his line.
All of this happens at the save, not before, so it changes goals and not xG. That fits the other thread, where Agility and Reflexes save about 0.8-0.9 goals a season per level while the xG the keeper faces doesn't move.
 
Next, everything else that feeds into how a player plays in a match: condition, sharpness, morale, personality, CA, familiarity, instructions and traits.

Condition, sharpness, morale and cohesion all go into one modifier that's added to every attribute when the effective values are worked out. The agent measured how big each one is by changing one at a time for the whole squad and converting back into attribute points (1-20):

ChangePace, AccelerationDecisions, Positioning, MarkingVision, Anticipation, Composure, ConcentrationTechnical (Passing, Finishing...)
Condition 90%−0.8−0.9 to −1.1−0.4 to −0.6about 0
Condition 70%−2.7−2.8 to −3.3−1.8 to −2.3−1.0
Sharpness 80%−1.3−0.7 to −0.9−0.4 to −0.6−0.2 to −0.5
Sharpness 60%−2.8 to −2.9−1.7 to −2.1−1.1 to −1.5−0.7
Morale 1 (from 13)0−0.6 to −0.9−0.3 to −0.6−0.4 to −0.6 (Flair −1.6)
Team cohesion 30 (from 76)−0.4−1.1 to −1.2−0.5 to −0.7−0.5
So condition and sharpness hit the body first, and condition also hits the reading of the game: a player at 70% condition plays with almost 3 points less Pace, Acceleration, Decisions and Positioning. Low sharpness also cuts effective Stamina, so an unsharp player gets tired faster too. Morale moves most things by less than a point (mostly Flair), which is why it doesn't show in the results. Club cohesion isn't used at all. This was one match and one squad, so take the exact sizes as rough.

Pressure and Professionalism. Pressure feeds a "nerves" term into the same modifier: going from Pressure 1 to 20 lifts almost every mental and technical value by about a third to half a point. Professionalism adds a smaller term. Both also have their own dice. This is why Pressure was the one hidden attribute with a clear effect in the other thread. Adaptability isn't used in the match at all.

Current Ability is read, but only as a limit inside that modifier, through the gap between Potential and Current Ability. For established players the gap is small and the limit sits at its floor, which fits with CA doing nothing in my tests.

Tactical familiarity scales the dice. Right after the dice are rolled, each of the eight tactical areas scales two dice:

AreaDice it scales
In possession shapeFirst Touch, and a team cohesion die
Passing styleDecisions, Composure
Build upTechnique, Vision
Chance creationAnticipation, Off the Ball
Out of possession shapeConcentration, Marking
Line of engagementWork Rate, Aggression
Pressing intensityAnticipation, Tackling
Pressing trapDecisions, Positioning
The multiplier goes from 0.5 (area value below 16) through 1.0 (41-50) up to 1.6 (91-100), and each die moves at most 5 either way. So familiarity works on the mental side of the game, never on the body. It doesn't touch the effective values themselves, which is why attributes and familiarity just add up in the tests. We're confident on this one.

Team instructions reach the pitch in three ways (confident on where they land, some of the uses are my reading):
  1. Shape, set once at kick-off. Width, defensive line and mentality move the formation's positions. For example, "much deeper" puts the line 1.5 m deeper at kick-off.
  2. Switches on each player. Passing length, dribbling, shooting from distance, tackling, creative freedom, overlaps and so on become switches the decisions check all the time. Passing directness becomes a number per player, and it's one of the busiest values in on-ball decisions.
  3. Checked during play. The rest (pressing trap, counter-press, crossing type, bypass press) is checked at the moment it applies.
Traits are switches inside the same decisions (confident on where they're checked). For example, Tries Killer Balls Often changes which teammates a player looks for, Knocks Ball Past Opponent and Runs With Ball Often are checked when he decides to dribble, and Tries To Break Offside Trap when a forward picks where to stand. So a trait changes what a player tries, not how well he does it.
 
Here is my attempt at summarizing everything above and also from what we know from previous knowledge and test.
  1. pace and acceleration matters because they influence "who gets there first", and is used for every player any time someone is deciding on a pass.
  2. Jumping reach matters because it decides the whole areial game on its own. It carries 3.3cm per point and a ball is only in play if it is on reach.
  3. Antecipation and concentration also rank high because they are also used for every player, every tick, by the player reading where others are going and how quick to react.
  4. Work Rate has a floor because of the die roll. It has a steep scale and its checks are always against zero.
  5. Balance matters as much as Strength in duels as they count exactly the same in the body contest.
  6. Finishing and the goalkeeper attributes change goals but not xG, because they're only read at the shot.
  7. Passing, Technique, Vision and Decisions show little even though the engine checks them all the time. This is because most checks only ask if the die is zero, which +5 barely changes, and pass choices depend mostly on the physical estimates of the receiver and the opponents.
  8. Familiarity adds to attributes instead of multiplying them, because it scales the dice, not the effective values.
  9. Sharpness and condition matter more than anything else on the start screen, because they take several points off the physical attributes, while morale moves things by less than a point.
  10. CA does nothing in the match: it only sets a limit that sits at its floor for established players.
There are still many open questions, and I will keep digging. Some things that are top of mind. How are final choice made, how traits enter the equations. There are many other parts of the engine that we haven't checked.

Also, if anyone is doing something similar, let me know, so we can cross check the work and speed the process.
 
@caiograco So can you suggest the best changes we make if we are losing? What would be the most effective changes, like I’ve noticed changing wingers always does the trick when losing.
 
@caiograco So can you suggest the best changes we make if we are losing? What would be the most effective changes, like I’ve noticed changing wingers always does the trick when losing.
I don't think this answers any of that. It is just about understanding why we see the results in-game and how things work under the hood.
 
This is incredible work, thanks a lot for sharing.

1) irl the attributes of different players + the tactic combine together. For example a guy who can cross, crossing to a guy with high jumping reach in a tactic that asks for a lot of crossing. Does this sort of behavior happen in the engine as well? Are those an emergent property of the engine?

2) Maybe this is common knowledge but those calculations that you mention seem to make sense in the context of a full match but what happens when the match is compressed like if you use the Quick Sim button or when the engine is calculating results for hundreds of matches at a time in a few seconds?
 
Some more results on consistency, home advantage and traits.

Consistency and Important Matches are one roll before kick-off. Before the match starts, the engine rates how big the match is (competition, stage, importance, rivalry) against the player's reputation, and blends the two hidden attributes (consistency for an ordinary match, and important matches for a big one) based on that. That blend, together with the player's condition and fatigue, sets the odds of a random "bad day" that's subtracted from all his attributes for the whole match. If it hits, it's subtracted from all his effective values for the whole match, through the same modifier as condition and sharpness. Neither attribute is read during play after that. I need to better understand how big the effect is and how likely a "bad day" is.

Home advantage doesn't come from the engine knowing who's at home (as far as I can tell so far). The engine marks the home team with a flag, but it's used mostly before kick-off. What it does is kind of interesting. It seems it makes the home manager expect about a goal more, which feeds his mentality choices.


The engine's home flag is mostly used before kick-off. The engine marks the home team with one flag (set for the home team and never for the away team in 40 saved matches). There's a second home flag that's never read at all. What the first one does is kind of interesting. It seems it makes each manager's pre-match expectation about a goal better for the home team, and that expectation feeds his mentality choices and a pre-match nerves term. During play it's only used to tell the two sides apart in the match statistics. So the engine itself doesn't give the home team a direct boost during the match, and whatever home advantage there is has to come from somewhere else in the match (maybe the crowd, like @Purity suggested in another thread). But I/the agent haven't found where yet.

There is also a sportsmanship check that looks like going down easily (our best reading, not traced in a match yet). One of the on-ball routines rolls against a player's sportsmanship cubed, depending on the distance and angle to the nearest opponent, and it's more likely to fire when his team needs a result. That is kind of cool but it needs more testing.
 
OK. This is the single most important thread I've seen in years. And my reading is that the game engine itself is as we would want it: an actual "real-time" (as much as a computer can emulate decisions from 22 tired brains) simulation.

Our grievances with the engine are thus more about the number of levers and the weight/scales of decisions.

Amazingn findings caiograco. This thread, for once, actually makes me love the game more (unlike all the meta attribute testings we do all the time hehe). It seems to EXPLAIN why some tactics work or no, and why some attributes are more important. No use in trying to make an amazing total football tactic if even great players with perfect fit have low percentage chance of going at the right place.

Meta isn't because of large-scale precomputations, but because of small-scale insiduous adjustments.

Thank you mate for sharing it with us!
 
This is incredible work, thanks a lot for sharing.

1) irl the attributes of different players + the tactic combine together. For example a guy who can cross, crossing to a guy with high jumping reach in a tactic that asks for a lot of crossing. Does this sort of behavior happen in the engine as well? Are those an emergent property of the engine?

2) Maybe this is common knowledge but those calculations that you mention seem to make sense in the context of a full match but what happens when the match is compressed like if you use the Quick Sim button or when the engine is calculating results for hundreds of matches at a time in a few seconds?
Thanks!
1) Yes, and it seems to be emergent rather than a special "combo" rule. When a player considers crossing, the engine weighs each teammate as a target, and that check reads the target's Heading, Jumping Reach, First Touch and Finishing together with the crosser's Crossing. Then the cross itself uses Crossing and Technique, the crossing instructions are checked when the ball is kicked, and the header is decided by reach and the aerial duel. So a good crosser, a big target man and a crossing tactic make each step more likely to happen and to work, without a formula for the combination. I haven't measured how much that's worth compared to the parts on their own, though.

2) Good question, and I don't know yet. My tool only runs the full engine, step by step, about 3 seconds a match. What I can say is that matches between other clubs that I have looked at (and these are a selected sample that the game saves in a temporary folder somehow), replay exactly (same stats, same score, etc...) than on the full engine. (i've re-simulated those using the real engine). So at least those use it. Whether Quick Sim or leagues you don't follow use a simpler, faster engine I don't know. Someone on FM26 found that with match detail set to none the game used a quicker engine where attributes mattered less, so it's worth checking for FM27.
 
OK. This is the single most important thread I've seen in years. And my reading is that the game engine itself is as we would want it: an actual "real-time" (as much as a computer can emulate decisions from 22 tired brains) simulation.

Our grievances with the engine are thus more about the number of levers and the weight/scales of decisions.

Amazingn findings caiograco. This thread, for once, actually makes me love the game more (unlike all the meta attribute testings we do all the time hehe). It seems to EXPLAIN why some tactics work or no, and why some attributes are more important. No use in trying to make an amazing total football tactic if even great players with perfect fit have low percentage chance of going at the right place.

Meta isn't because of large-scale precomputations, but because of small-scale insiduous adjustments.

Thank you mate for sharing it with us!
Thanks, and yes, I agree. I'm surprised how detailed it is. And a bit annoyed that so many attributes matter for how the game is played but don't actually influence the results.
 
Club cohesion isn't used at all. This was one match and one squad, so take the exact sizes as rough.
In that post, you say that Club cohesion isn't used, but in the table, you showed that modifying it did affect effective attributes (if I read your table correctly). Do you remember how and why your interpretation was that it wasn't used at all?

Again, amazing work. Hope you're having fun!
 
In that post, you say that Club cohesion isn't used, but in the table, you showed that modifying it did affect effective attributes (if I read your table correctly). Do you remember how and why your interpretation was that it wasn't used at all?

Again, amazing work. Hope you're having fun!
Thanks, and yes, digging into the code is fun.

On your question. There are two different cohesion values in the player's data. The one in the table is team cohesion, and lowering it from 76 to 30 does lower the effective attributes (about a point off Decisions, Positioning and Marking). Club cohesion is a separate value, and when we changed it no effective value moved at all, which matches my earlier test where club cohesion 25 vs 100 made no difference to results.
 
The agent dug into two things people asked about: what traits actually do, and how the engine treats the home team.

Traits. They're switches inside about 20 of the decision routines. Most of them loosen a threshold or act like an extra player instruction. Some examples (confident on these, read from the code):

TraitWhat it changes
Shoots From Distanceconsiders shooting from up to 30 m instead of 20 m, but only if his instructions allow long shots and his Flair roll beats his Decisions roll
Avoids Long Shotsno shots from 20-30 m at all
Tries To Break Offside Trapstands 1.5 m off the last defender instead of 3.0 m
Tries Killer Balls Oftena pass towards goal no longer needs a Vision roll of at least 16
Plays Short Simple Passesthe safe short pass becomes the choice whenever his passing directness is below 12 instead of 7
Tries Long Range Passeslong passes are penalised from 15 m instead of 12 m
Possesses Long Flat Throw+8 m on long throws
There are about 25 in total. So @DRAGN, to your question: traits don't change how much an attribute is worth, they change what the player tries. Killer Balls lets a low-Vision player attempt passes he otherwise wouldn't, and Shoots From Distance opens up shots from further out. When the agent tested them, they changed the style of play but not the chances created. Shoots From Distance adds about 2.5 shots a match, but all from 20-30 m and worth very little.

The home side's expected margin. Before kick-off each manager works out an expected goal margin from the two team strengths, with a home bonus added for the home side and taken off the away side, plus a flat +0.5 when the teams don't share a ground. For two equal teams the home side expects about +1.26 goals and the away side −0.76.

That expectation is then used in three places:
  • The manager's confidence: +5 for a home side expected to win, −5 for an away side expected to lose.
  • The AI manager's expectation level: 13 at home against 10 away, which feeds its tactical choices.
  • Discipline: a team whose goal difference falls short of its expected margin loses discipline more and more as the match goes on.
During play the home flag is only used to tell the sides apart in the statistics. Swapping it between the teams accounts for at most a fifth of the home advantage, so the rest has to come in with the saved match itself. We haven't found which value that is yet: there's no attendance-like number in the match, team or manager data..
 
Something I've always wondered (if it is available in your testing tool): position proficiency.

Does it affect rolls? Apply multipliers on attributes? all of them or only mental? Does it affect mainly rolls ont positionning or also on passes/shots? That info would be great to understand how important it is to have a player both "playing in his natural position" and "playing a known role". Specifically on how their play is affected.

IRL, vision, positionning, off-the-ball should be drastically lower when playing out-of-position as all automatisms are broken. But long shots and passing, for instance, shouldn't be affected. I'd like to know how (and if) they considered it this way!
 
Back
Top