This game is FAR FAR from realistic

[wwfan]

It does make sense to me and I do understand how it works in the game, but I still don't see how any researcher can break down and tease apart those factors, especially not when they are watching a proper game, and you are saying that training performance doesn't count.

How do you tell whether someone scored a goal because they had high composure or high finishing (and really think about this because I know what examples people are going to give, and they do not work, because you can argue all of them both ways)? In your examples, the first one sounded fine and reasonable, until I thought "why technique and not passing?" "why decision making and not creativity?" When is a player with good technique ever bad at passing?

Composure isn't just about finishing. It is about whether the player has a tendency to panic in high pressure situations, anywhere on the pitch. It can affect whether a defender gets rid of the ball first time, calmly plays it out of danger, or gets caught in possession. It can affect whether a forward hits it too early or straight at the keeper in big moments. The researcher will not be determining composure levels just based on finishing, but on the players overall tendency to remain composed or not during matches. It can modify pretty much all the technical and some mental attributes in certain match situations.

In the second example, things get even muddier. Accuracy of the shot is only down to the finishing attribute, right? I don't see how the researcher can ever get a good estimate of this, without being affected by composure, decision making etc. If training performance is not looked at, then when do you ever see a player shooting at an empty goal with no one marking him, which is what you seem to be suggesting is what the finishing stat describes. Once again, you mention technique. I can't tell whether you are referring to technique or finishing. If it is finishing, then what does finishing have to do with working for space with the ball, if it is the same thing as shooting? If it is technique, then why does technique allow a player to aim for a target better? And on another note, why is long shot different from finishing, rather than having a shooting stat and using creativity, composure, power and decision making as modifiers?

Technique relates to how much space and time a player needs for him to perform the other attributes well, whether he can perform them well when the ball isn't in the perfect position, plus how well he can create space for himself. A high technique player will only need to be in a tiny bit of space to hit his finishing or passing attributes, whereas a low technical attribute player needs more space and time. Thus, a Xavi type player will pass to 20 no matter how tightly marked he is (unless his composure fails), whereas a Leon Britton (who passes extremely well) probably needs more time and space to do so. As such, technique modifies other attributes as well.

As HUNT3R said, the "Finishing" attribute is "Accuracy" with shots inside the box. "Long shots" means "Accuracy" from outside the box. Of course a player still needs the technique, composure etc depending on the match and situation. In a lot of space, a low technique player will finish to his finishing attribute, providing he stays composed. However, in less space, he won't.

The way that it is possible to explain away everything makes it a perfect theory, and this is why it is flawed. It sounds scientific and logical but when you realize that this whole argument cannot be disproved, it gets really frustrating.

The examples given do make sense, but not if you look at it from a practical point of view. I get that this is just the way it works, but that's not the point. The point is that we have to look at the types of stats we have and really think about what they each mean. And what I see when I see a lot of the stats is "wow, these two attributes are not mutually exclusive, they can't predict anything with sufficient efficiency and accuracy". And this is why I question the way we look at the data we do now, though I realize that this is how the coding works.

I think the problem you are having is you are looking at every attribute as an absolute, and they simply aren't meant to work that way. All kind of things modify them. Mental stats such as composure, bravery, decision making, flair, teamwork etc, all modify the basic technical stats or how the player might perform in a certain situation. A player with poor decisions or teamwork might go alone when he has a teammate in a much better position to score, meaning his own finishing attribute / composure are the only things you are relying on. Low flair or creative freedom (which is merely a flair boost) might result in the player trying very safe shots and seeing them saved regularly, rather than trying to something a little different.

In addition, other elements, such as hidden stats, morale, team and media talk pressure, team discipline, team complacency, and team gelling can all affect a player's performance in any given match. You can also add in other aspects that are sometimes difficult to pick, such as whether the player is using his stronger or weaker foot, how fast he's being forced to move, the technical difficulty of the chance etc. His 17 finishing only relates to how well he is likely to finish when all other elements are in his favour. For finishing, a researcher is simply giving his opinion on how accurate the player is when shooting inside the box. That is being constantly modified by other game elements. It is never going to drop stupidly low, but it cannot be used as a constant for every situation the player is in. You have to look at other elements as well.

On these forums, in general, those who accept the relative nature of attributes don't get frustrated. Nearly every rant comes from people who read them as absolutes. If they were absolutes, the game would be far, far less sophisticated and the ME considerably worse. Their relative nature is necessary for the ME to work with any degree of believability.

[MSCCG]

I think the problem you are having is you are looking at every attribute as an absolute, and they simply aren't meant to work that way. All kind of things modify them. Mental stats such as composure, bravery, decision making, flair, teamwork etc, all modify the basic technical stats or how the player might perform in a certain situation.

I think you misunderstood me. I have a whole thing written and data collected but I think it would be better if I outlined my point of view better first in case we keep moving further and further apart.

Firstly, I agree that shooting aka finishing should be moderated by composure, decision making, bravery, etc (each to a different degree). In that sense, I am not disagreeing with you in any way.

However, I think that the current system is not sophisticated and streamlined enough and what we think it does right now might be a superstition.

For example, shooting and long shots, if you just look at the physical act, are not so far apart given their definitions that one is inside the box and the other is outside. The difference between a shot and a long shot in real life seems to not be the aim/technique (i.e. finishing is purely aim, long shot is... also aim?), but other factors such as power, creativity (which I have been led to believe is qualitatively equivalent to vision?), decision making etc just like you said in comparing different levels of success in finishing. Why can't we combine the two stats (i.e. functions like passing) and then use mental and physical attributes to modify the final outcome? This would be truly relative and more sophisticated because it matches real life better. The technique doesn't change one foot outside or inside the box. Think about how Robben is rated in the game, with 17 finishing and 17 long shots. If you look at his goals, http://www.youtube.com/watch?v=htKJp5_PvaI, you see that the types of shots that he makes are not qualitatively different inside and outside the box (same technique, as in he pulls back his foot quite far and strikes with the side of his foot - minus the volleys, which I can ignore because many people on the forum says that finishing is no longer the important factor in volleys, but technique and first touch). But he often tries to come closer into the box to make the same shot, which means that it is easier to shoot the closer you are to the goal (of course, but my point is that it is an aiming issue, not a different technique). In my opinion, he was rated by the researcher twice for doing the same thing.

Likewise corners and free kicks. The technique used for a corner or a free kick is not far apart, I believe that there is an argument to combine the two into a "dead ball" stat and then use other technical and mental attributes to rate the two in the match engine. This is relative, and this is more sophisticated and true to real life.

I did a bit of research in the stat distributions in the players and I realize now that the system is either not as sophisticated as we believe, or the calculations are very wrong. Either way, it is not ideal in matching reality. For example, free kick is related to passing, and so is corners, and crossing. This is true in real life. However, when people complain about free kicks, many people reply that free kicks are moderated by passing, long shots, and corners (when the free kick is at the side). Looking over free kick rating - free kicks, passing, corners and crossing in the starting database co-vary quite a lot. If I had the data in spreadsheet form I could put all these numbers into the computer and run a statistical test, but just looking over the data, it looks like these stats might be at least partly measuring the same thing. This is troublesome, because there are issues of multicollinearity that would cause the model to predict values in the match engine poorly. For example, if the numbers are additive, and there is some factor within both free kicks and passing that measure the same thing, then you just doubled that underlying factor. If it is a percentage, you have just overestimated the probability.

This problem is not just restricted to within technical stats, but also technical and mental. For example the collinearity issue also exists between passing and creativity. If you look at the top 700 passers in the game, only around 90 of them have creativity at or below 10. You would expect that if passing was mutually exclusive to creativity, which is what would be required to make a powerful prediction, then creativity would have a much more even spread than it does now. You could argue that it is not that the players being good at passing and creativity are caused by one thing, but that they are excellent at both. This would make me feel better (though the issue of the researcher's ability to tease apart the two is worrisome), but I have a piece of evidence that makes me sceptical. When you look at free kick attributes, the total distribution of free kicks does not seem to be related to long shots. If passing is related to free kicks, and so are corners and crossing, this is a discontinuity and must be investigated. It may be that the researchers are not being consistent with their estimates, or that the game doesn't actually use long shots as a predictor for shooting from free kicks, both worrying. If it can be justified that a researcher looks at a player making a pass and rates both passing and creativity, then why is it not justified to look at a free kick and rate free kicks and long shots together? I understand free kicks are not all long shots, but right now they look like they are not related at all. I would expect it to be related a little bit, given the other attributes that are related.

On another note, I just have to object to you just tacking every single factor onto the final product of finishing. I do tend to agree with you, but I would never make that argument because it is too easy. I'm thinking about what would disprove this argument, as in what I would have to see in order for me to be proven wrong? In the argument that I make in this post, what I would have to see for me to lose my worries about my perceived problem with multicollinearity is that all technical attributes are either entirely independent of each other, or that only a single technical attribute is used per calculation (+ mental and physical attributes). If it is the former I have to object to this being the case looking at the data, and if it is the latter, then the researchers are unable to do their job correctly because they have to derive a huge number of attributes with a single action which is almost impossible. Furthermore, it cannot be an accurate representation in the game because stats are relative to each other, given that they are bounded by the CA which is also eaten up by footedness as well as other stats. The end result is a single template for certain types of player, which results in a static game with lowered uniqueness between individuals (or that this game is carried by PPMs, another worry).

To be honest, it's a little useless debating theory when we have no experimental evidence. I'd love to run some experiments by modifying attributes of a player one by one and either sim one match or a season and see what the end results are. This will probably end the debate once and for all. As of right now, I believe that attributes should be shrunken, or if the same number is kept, to be derived from more general observations from the researcher.

[Red_Arrow]

To give the researchers perspective I will use an unnamed Aldershot player. He is a good finisher, except when he is under pressure. Thus a decent finishing attribute and a low composure. He has a moderate technique when striking the ball, but an awful first touch. Thus a moderate technique but a low first touch. His decision making is generally dreadful. These combine to give a player who will take his chances when not under pressure but when he doesn't have time to think about what he is going to do. It helps this is a player who I have watched come through the youth and reserve teams and into the first team, and in nothing pre season games. It is however quite easy to break down a players ability into the determining factors.

It does (to me) work this way in the game. In my current save as Delemont in the Swiss leagues I have a striker with 13 finishing and 4 composure. He bangs the ball in on a regular basis but don't ask him to score a one on one because the ball will go wide or into the keepers midriff.

[wwfan]

@ MSCCG: I have given the model you are talking about plenty of thought over the years. However, I always revert back to the belief that to model the complexities of human behaviour accurately, more rather than less attributes are always going to be better.

[MSCCG]

@ MSCCG: I have given the model you are talking about plenty of thought over the years. However, I always revert back to the belief that to model the complexities of human behaviour accurately, more rather than less attributes are always going to be better.

If only that were the case. In the mathematical modelling of human behaviour, we rely heavily on the Akaike criterion. Basically, this measures up the information lost when we take out a parameter against how well the model predicts behaviour. If two models predict a similar pattern of behaviour, then the model with the lower number of parameters is the more powerful one.

The variability in human behaviour is not actually related so much to a complex myriad of mental and technical attributes but rather a simple choice procedure, where each player optimises their behaviour in a way that matches failure and success rate. I think that all behaviour in the game can be modelled by using a simple equation, taking technical attributes as success/failure rates and mental attributes as buffers to the sensitivity of a player's ability to discern those success/failure rates.

If you're interested there are articles that show how basketball shots are selected and modelled almost to perfect by the generalised matching law used in behaviour such as - Vollmer and Bouret, An Application of the Matching Law to Evaluate the Allocation of Two and Three Point Shots by College basketball player, Journal of Applied Behavior Analysis, 2000

[timmytiger69]

so, I'm still confused..... :o

[HUNT3R]

so, I'm still confused..... :o

With? Ask away. There are plenty of people on this forum that wouldn't hesitate to explain anything you're unsure about or need clarification on.

[Karnage94]

I might be a bit off the mark but this is how I think the whole finishing/composure/technique thing works. I see finishing as how accurate a player is when it comes to putting the ball into the corner of the net. Putting it in the areas that the keeper can't reach. A player with a high finishing rating will be able to do this more often than not.

Adding composure to this, if a player is more composed, he is more likely to hold himself together and take the more high pressure chances such as having a defender or two swarmed around him with an advancing keeper. A player with lower finishing but higher composure will still be able to take his chances I feel. He might not place the ball into the corner as well as somebody with higher finishing and lower composure but he will take the higher pressure chances more often. A player with high finishing but low composure in a high pressure situation may still hit the target but he could scuff his shot leading to an easy save as an example.

On the subject of technique and finishing I find that a player with high technique can bend and curl the ball out of reach of the keeper, which will aid his finishing. A player with low technique in contrast will be less able to bend the ball to his pleasing. Sometimes you need to bend the ball slightly away from the keeper or to add dip so he can't reach it rather than hitting it fairly straight. I might be wrong but that's how I see it.

[Bababui]

@ MSCCG: I have given the model you are talking about plenty of thought over the years. However, I always revert back to the belief that to model the complexities of human behaviour accurately, more rather than less attributes are always going to be better.

However, if the additional attributes are poorly defined, not modeled well and not properly balanced they can make a poor situation even worse.

[Dekker2]

They should really remove or modify the FM tip that mentions composure: because it only mentions it in the context of one-on-ones, people seem to ingore all the other places in the game that composure has an effect, and that a player's composure on the ball could easily be measured by the scouts.

[wwfan]

If only that were the case. In the mathematical modelling of human behaviour, we rely heavily on the Akaike criterion. Basically, this measures up the information lost when we take out a parameter against how well the model predicts behaviour. If two models predict a similar pattern of behaviour, then the model with the lower number of parameters is the more powerful one.

The variability in human behaviour is not actually related so much to a complex myriad of mental and technical attributes but rather a simple choice procedure, where each player optimises their behaviour in a way that matches failure and success rate. I think that all behaviour in the game can be modelled by using a simple equation, taking technical attributes as success/failure rates and mental attributes as buffers to the sensitivity of a player's ability to discern those success/failure rates.

If you're interested there are articles that show how basketball shots are selected and modelled almost to perfect by the generalised matching law used in behaviour such as - Vollmer and Bouret, An Application of the Matching Law to Evaluate the Allocation of Two and Three Point Shots by College basketball player, Journal of Applied Behavior Analysis, 2000

Valid only if you think behaviourism and quantitative analysis is an accurate or effective way of understanding human action. We'd be here forever and a day debating that.