Same match stats, but AI scores more?

[DeanMon]

Yeah, I read a lot of threads, and experienced myself that very often human dominates the game with a lot of shots and possessions but cannot score. Someone would come in and comment "It's your tactics!".

Okay, fine, maybe it is true. However, the argument now is that why, given the same match stats (shots, CCC, possessions,...), AI "seems" to score more than human?

Let's say I play MU, and I play cautiously against MC away, and positively against Leicester home. Then in respective matches, MC and MU have ~20 shots, ~10 on target, ~60% possessions, ~2-3 CCC, but MC won 4-1 but MU drew 0-0?

Please notice that I wrote "seems" because I might be biased here and sample size is too small. But if someone has some data to show me that it is not the case, then I will really appreciate. Otherwise, it is hard not to doubt that AI is "rigged" to be favored against human.

Thank you all.

Edited July 13, 2020 by DeanMon

[zyfon5]

12 minutes ago, DeanMon said:

Yeah, I read a lot of threads, and experienced myself that very often human dominates the game with a lot of shots and possessions but cannot score. Someone would come in and comment "It's your tactics!".

Okay, fine, maybe it is true. However, the argument now is that why, given the same match stats (shots, CCC, possessions,...), AI "seems" to score more than human?

Let's say I play MU, and I play cautiously against MC away, and positively against Leicester home. Then in respective matches, MC and MU have ~20 shots, ~10 on target, ~60% possessions, ~2-3 CCC, but MC won 4-1 but MU drew 0-0?

Please notice that I wrote "seems" because I might be biased here. But if someone has some data to show me that it is not the case, then I will really appreciate. Otherwise, it is hard not to doubt that AI is "rigged" to be favored against human.

Thank you all.

The AI is not rigged against human. Why you score more when you have better stats is simple. Because football is not played on a spreadsheet. What matters is what happen on the pitch. If we look at pure stats, one will assume that Manchester city is the best team in the premier league this season which we know is far from reality.

[DeanMon]

1 minute ago, zyfon5 said:

The AI is not rigged against human. Why you score more when you have better stats is simple. Because football is not played on a spreadsheet. What matters is what happen on the pitch. If we look at pure stats, one will assume that Manchester city is the best team in the premier league this season which we know is far from reality.

Hello, I think you misread my post. I played MU in 2 matches: MC-MU and MU-Leicester. MC and MU dominated 2 matches respectively with very similar match stats, but MC 4-1 MU and MU 0-0 Leicester.

It's fair to say AI is not rigged, but until some sort of data is provided, it is also fair to say AI is rigged.

[zyfon5]

5 minutes ago, DeanMon said:

Hello, I think you misread my post. I played MU in 2 matches: MC-MU and MU-Leicester. MC and MU dominated 2 matches respectively with very similar match stats, but MC 4-1 MU and MU 0-0 Leicester.

It's fair to say AI is not rigged, but until some sort of data is provided, it is also fair to say AI is rigged.

What proof do you like to see? Behind in stats but still win? Or leading in stats but still win? I can show you both.

[DeanMon]

11 minutes ago, zyfon5 said:

What proof do you like to see? Behind in stats but still win? Or leading in stats but still win? I can show you both.

I don't mean a sole example, of course I also won some matches by luck. I mean given the same situation (same domination, similar players' attributes,...), AI and human should have the same results in the long run.

It's hard to replicate the situation, maybe some developers from AI can do.

Let's settle this once and for all, then if someone accuses the game is broken, you guys can throw him the data that the game is fair, AI and human are treated the same. Therefore, human needs to adapt to win. Fair enough?  

Edited July 13, 2020 by DeanMon

[HUNT3R]

43 minutes ago, DeanMon said:

Otherwise, it is hard not to doubt that AI is "rigged" to be favored against human.

It isn't. The ME can't differentiate between human and AI. It doesn't need to. It's there to simulate a match.

These scenarios happen IRL and it happens in AI v AI matches too. In FM, I would say that stats look often more one-sided than it should, but still, teams IRL with more shots or possession fail to win fairly often. See here, for instance : 

https://old.reddit.com/r/footballmanagergames/comments/hpufdg/it_happened_again_ai_fming_ai/

 

[DeanMon]

2 minutes ago, HUNT3R said:

It isn't. The ME can't differentiate between human and AI. It doesn't need to. It's there to simulate a match.

These scenarios happen IRL and it happens in AI v AI matches too. See here, for instance : 

https://old.reddit.com/r/footballmanagergames/comments/hpufdg/it_happened_again_ai_fming_ai/

 

It seems not enough to simply say it isn't.

Don't you want to settle this one and for all to all the frustrated players that the game is fair? If so, give us a proof.

[HUNT3R]

1 minute ago, DeanMon said:

Don't you want to settle this one and for all to all the frustrated players that the game is fair? If so, give us a proof.

The only way to really prove it is if you see the ME code, which isn't going to happen.

 

The best I can do is provide you with a link to a post that goes into some detail on how the ME calculates events and matches:

Hopefully that helps.

[DeanMon]

8 minutes ago, HUNT3R said:

It isn't. The ME can't differentiate between human and AI. It doesn't need to. It's there to simulate a match.

These scenarios happen IRL and it happens in AI v AI matches too. See here, for instance : 

https://old.reddit.com/r/footballmanagergames/comments/hpufdg/it_happened_again_ai_fming_ai/

 

It's one more thing about your post is that I am questioning if AI is rigged against human. Between AI versus AI, I don't care. Of course it is fair AI vs AI, you don't need to prove that.

[zyfon5]

28 minutes ago, DeanMon said:

I don't mean a sole example, of course I also won some matches by luck. I mean given the same situation (same domination, similar players' attributes,...), AI and human should have the same results in the long run.

This would not happen even if the AI is not rigged. Football is a sport that has a lot of randomness more than any major sport that you can name. Only the game developers has the capability to test things like this on a large scale. So if you do not believe what the developers say then no one is able to convince you.

[DeanMon]

@HUNT3R

I think me and many other frustrated players do not mind losing games. But at least, we need to be assured that we are losing fair and squared.

If we dominate and can't win a game, it's fine. However, there must be another game we are dominated but unbeaten. Given the same bus in the box, Greenwood CANNOT only hit 3/8 shots on target in there and still goalless, while Fernandinho scored a long shot, right? Soyuncu played like a superman, but Lindelof made a mistake leading to goal?

Again, the sample size is too small, and I might have confirmation bias. But if it is a complain from many players, then you guys should have a look.

Do you know how online casinos operate? They need to have their random number generator tested and certified. Are there any gamblers complaining about losing on online casinos, even it's transparent that casinos have a house edge? I don't think so.

I think we can learn a thing or two about that practice here.  

Edited July 13, 2020 by DeanMon

[Spanner]

Interesting that you make Casino comparisons, that is something that I can come back to later in another post.

Regarding the matches that you speak of and with game results versus the match stats, you are right that people are too quick to say ‘it’s your tactics’.
The ME is a complex yet basic concept which takes far more inputs than just your match tactics.

Tactics pre game, in-game and post-game can all have an effect on not just your current game but also future games.

The inputs you give the ME to work from effect the result it (churns-out) produces as its outcome.
For example:

1.       Training stats improve individual as well as team stats which help in that grouping of inputs.

2.       Coaches have an effect on team training, so get the best coaches you can for better results that also fit your tactics, players needs and overall club culture. This will give you much better cohesive inputs for the ME to work with.

3.       Interviews are a huge input for the ME. Pre-game and post-game, be very careful and please understand that the interviews are the scariest aspect of the ME. The are as important as tactics and in game decisions.

4.       Tactics are a factor which can not be ignored. Have the right players on the pitch so that you will win the individual as well as team based match-ups. For example, defenders are weak in the air. A very big opportunity there. Another example, the CB’s are very slow….set-up accordingly.

5.       Team talks and Team meetings also affect the way players can play.

Those are just a few examples.

Football is not about tactics, in real life. It’s a lot more and the ME does a good job replicating the randomness plus brutality of football. Only yesterday Chelsea lost 3-0 to Sheffield United. Did anyone see that happening?

Like with a Computerised Casino machine game, please remember that everything is connected and one thing can set off a domino effect of reactions.
(and yeah, I was banned from a casino because I played a computerised roulette game that basically became a free ATM for me).

Inputs – ME – Outputs
You have a lot more control over the outputs than you know/think.

[DeanMon]

4 hours ago, Spanner said:

Interesting that you make Casino comparisons, that is something that I can come back to later in another post.

Regarding the matches that you speak of and with game results versus the match stats, you are right that people are too quick to say ‘it’s your tactics’.
The ME is a complex yet basic concept which takes far more inputs than just your match tactics.

Tactics pre game, in-game and post-game can all have an effect on not just your current game but also future games.

The inputs you give the ME to work from effect the result it (churns-out) produces as its outcome.
For example:

1.       Training stats improve individual as well as team stats which help in that grouping of inputs.

2.       Coaches have an effect on team training, so get the best coaches you can for better results that also fit your tactics, players needs and overall club culture. This will give you much better cohesive inputs for the ME to work with.

3.       Interviews are a huge input for the ME. Pre-game and post-game, be very careful and please understand that the interviews are the scariest aspect of the ME. The are as important as tactics and in game decisions.

4.       Tactics are a factor which can not be ignored. Have the right players on the pitch so that you will win the individual as well as team based match-ups. For example, defenders are weak in the air. A very big opportunity there. Another example, the CB’s are very slow….set-up accordingly.

5.       Team talks and Team meetings also affect the way players can play.

Those are just a few examples.

Football is not about tactics, in real life. It’s a lot more and the ME does a good job replicating the randomness plus brutality of football. Only yesterday Chelsea lost 3-0 to Sheffield United. Did anyone see that happening?

Like with a Computerised Casino machine game, please remember that everything is connected and one thing can set off a domino effect of reactions.
(and yeah, I was banned from a casino because I played a computerised roulette game that basically became a free ATM for me).

Inputs – ME – Outputs
You have a lot more control over the outputs than you know/think.

Hi, let me express it mathematically like this:

ME(inputs) = ME(training, line-up, tactics, morale, interview, shouts, OI,...) = Outputs = (what happens in match) = (body language, match stats, results, ratings,...)

My question is that, how could I get many outputs right (body languages, match stats,...) but the results are wrong? How could ME explain this?

I cite casinos, because of the similarity of solution in dealing with frustrated customers. If the developers could design a test to prove the game fairness once and for all, there will be never a complaint about the ME again. I swear! SI just needs to write "our ME was tested for fairness by..." right on the package, the sales will go up 50% immediately!

 

@HUNT3R If the design of such a test is too difficult, I think developers should enlighten us by revealing more information about how ME works.

ME is like a black box here. If we players do not know about its mechanism, its state, its operation, whether it is accurate/fair, then what can stop us to doubt it is rigged against us?

SI developers could spend a few minutes to look at the reviews on Steam. Too much frustration and anger are there. If you guys are happy with your sales, then fine, just ignore me. If you guys love to retain your die-hard fans, it's something you guys can do for us.

 

@yolixeya Then just ignore me. (I reached the post limit)

It also comes down to what kind of business SI aims to develop with customers, right? Maybe ME is not rigged, but how about its skewness, or bias, or imperfection?

 

@Mrlee.1986 Maybe it is not rigged, but it could contain many flaws when interpreting the reality. If I am Pep, I can't blame reality, because reality is always fair and accurate. But ME "may" not be fair/accurate, because it represents its developers.

For one more time, I do NOT say dominance = victory. What I am trying to say is that given the same dominance, AI and human MUST have similar results in the LONG run. I am questioning given the same dominance, human achieves less success than AI, and I seek proof to eliminate that thought.

 

@Spanner If you are trying to tell me that human's SoT could be worse than AI's SoT, then maybe we need to split stats into high quality SoT and low quality SoT, right?

Edited July 13, 2020 by DeanMon

[yolixeya]

25 minutes ago, DeanMon said:

ME is like a black box here. If we players do not know about its mechanism, its state, its operation, whether it is accurate/fair, then what can stop us to doubt it is rigged against us?

It really comes down to what kind of a man are you. There are people right now that think everything about COVID-19 is a conspiracy and that world leaders and even scientist are lying to us about it. For me, it's enough that SI says it's not rigged. I have no reason to think they are lying. If I would think that I would not be playing this game.

[Mrlee.1986]

I don’t understand why they would rig it? All that would achieve is having people moaning that “they can’t win every single game so it must be rigged” or “I lost a game that I dominated so it must be fixed”
(did anyone really expect bournemouth to thump Leicester last night? Did anyone expect Southampton to beat city last week even though the stats show city dominated (as shown in image)? 
 

(I wonder if Pep blamed the Match engine for that one)

Why would they rig the game against us, just to have people moan, so they stop playing/buying the game and the company lose money? 
 

im no expert but that seems a bit of a stupid way to run a business 😐

Edited July 13, 2020 by Mrlee.1986
Tidied up

[Spanner]

1 hour ago, DeanMon said:

Hi, let me express it mathematically like this:

ME(inputs) = ME(training, line-up, tactics, morale, interview, shouts, OI,...) = Outputs = (what happens in match) = (body language, match stats, results, ratings,...)

My question is that, how could I get many outputs right (body languages, match stats,...) but the results are wrong? How could ME explain this?

I cite casinos, because of the similarity of solution in dealing with frustrated customers. If the developers could design a test to prove the game fairness once and for all, there will be never a complaint about the ME again. I swear! SI just needs to write "our ME was tested for fairness by..." right on the package, the sales will go up 50% immediately!

 

@HUNT3R If the design of such a test is too difficult, I think developers should enlighten us by revealing more information about how ME works.

ME is like a black box here. If we players do not know about its mechanism, its state, its operation, whether it is accurate/fair, then what can stop us to doubt it is rigged against us?

SI developers could spend a few minutes to look at the reviews on Steam. Too much frustration and anger are there. If you guys are happy with your sales, then fine, just ignore me. If you guys love to retain your die-hard fans, it's something you guys can do for us.

 

@yolixeya Then just ignore me. (I reached the post limit)

It also comes down to what kind of business SI aims to develop with customers, right? Maybe ME is not rigged, but how about its skewness, or bias, or imperfection?

 

@Mrlee.1986 Maybe it is not rigged, but it could contain many flaws when interpreting the reality. If I am Pep, I can't blame reality, because reality is always fair and accurate. But ME "may" not be fair/accurate, because it represents its developers.

For one more time, I do NOT say dominance = victory. What I am trying to say is that given the same dominance, AI and human MUST have similar results in the LONG run. I am questioning given the same dominance, human achieves less success than AI, and I seek proof to eliminate that thought. 

With respect DeanMon I think your understanding of the Match Engine is slightly off, but that is just my opinion.
Understanding how you impact the Match Engine will help you better see that it is not rigged and you actually have a lot of control over the outputs it (the match engine) spits-out.

Everything you do on the game (and I mean everything) is connected and important.

With reference to what you have written a few things:

1.       Sales would go up 50% …. Sounds nice, but really not sure about that.

2.       The business model of Sports Interactive (SiGames) is not to frustrate its customers. Casinos have another agenda, which is written into their algorithms, Football Manager is not the same. I understand what you are trying to say, but Casinos are perhaps not the best comparison. 

3.       I have been playing for years, took a long break, came back and have recently played FM16 and FM20. You can take my word as gospel, there is no rigged system.

4.       Humans and AI are 2 different topics. Also the example you gave referring to Man United and Man City games / stats ….and the results it gave to you (human) and Man City (the AI) is not anything major.
…..stats are one thing but you must remember how to interpret the stats. This is very important to grasp. Please keep in mind that a shot on target can be a variety of ways (deflections, slow poor shots from long range, lucky rebounds, clear good chances, difficult headers etc…)

I wish you all the best in your quest to have FM success.

[Lebohang_Mokoena]

I think when people raise complaints such as these it's because they're putting too much emphasis on tactics. I'd argue that most players don't realise that tactics aren't that important in football manager, and are definitely much less important than they were ten to fifteen years ago.

When people have this issue, they're essentially saying 'I have a good tactical setup, which I know is working because in matches I have more shots & shots on target than the opposition, so the matches where this happens and I lose must be an error with the game rather than down to an external factor.'

I'd argue that this argument ignores the impact morale, team cohesion and dressing room atmosphere have on the game. In FM20, I've had about five or six seasons in the Dutch League with top six teams using a possession based tactic, and a similar pattern occurred in each; We'd lose a game against our rivals, or a big defeat against a top side, and that would cause morale to drop, which in turn would lead to matches where we'd play just as well (I'd still dominate possession and get a decent amount of shots on target) but we couldn't convert chances. These games I'd either draw or lose, which would affect morale, and so on. 

I'd try different tactics, which would occasionally work, though not as well compared to a gamble on a team meeting that paid off.   I've always wondered if morale affects players in a way as stamina or consistency do.

Generally speaking I wouldn't look at matches like this in isolation and also factor in morale, cohesion, dressing room atmosphere, and even press conferences and team talks.

[Nacaw]

2 hours ago, Mrlee.1986 said:

I don’t understand why they would rig it? All that would achieve is having people moaning that “they can’t win every single game so it must be rigged” or “I lost a game that I dominated so it must be fixed”
(did anyone really expect bournemouth to thump Leicester last night? Did anyone expect Southampton to beat city last week even though the stats show city dominated (as shown in image)? 
 

(I wonder if Pep blamed the Match engine for that one)

Why would they rig the game against us, just to have people moan, so they stop playing/buying the game and the company lose money? 
 

im no expert but that seems a bit of a stupid way to run a business 😐

Oh, but he absolutely did. https://www.bbc.com/sport/football/53302856

Football is flawed like that, an inferior team can beat a superior team. Is the solution that the team with most shots of targets and most possession should get +1 goal at the end of the match? Never in a million years.

Obviously FM players feel frustration when you feel like you've done everything you could to optimise winning a match. That's GOOD, because that's the way coaches at all levels sometimes feel too. 

[Svenc]

Disclaimer: Reason for me posting this is that I want the AI of this game to become ever more intelligent. However, with more intelligent AI.... See, the reason this only ever goes in the AI's favor is nobody even bothers with similar ideas to this for long enough.

Back when the AI tactics barely protected ist own attacking Corners (good times, sigh!), with Leicester I had made half a season out of "rigging" the AI. One of the clues what was going on purely stats-wise is the oft much higher percentage of shots to shots on target for Leicester, but in terms of data the game offers, that's it.






Blame Match Of The Day and all that crap putting so much importance on shot Counts on a spread sheet (at least they have introduced xG, but too little too late).

That's all folks.

Edited July 13, 2020 by Svenc

[enigmatic]

30 minutes ago, Svenc said:

Dear Lord Have Mercy.

The reason this only ever goes in the AI's favor is nobody even bothers with somthing similar to this.

Blame Match Of The Day and all that crap putting so much importance on shot Counts on a spread sheet (at least they have introduced xG, but too little too late).

That's all folks.

The Burnley example is a worthy attempt. Lots of shots blocked, plenty of clean sheets despite usually having less than 40% possession, and had one of the best chance conversion ratios in the league despite mediocre finishers. All like defensive AI teams.

And also got very average results over the course of the season... just like AI defensive teams

 

[Zemahh]

1 hour ago, enigmatic said:

The Burnley example is a worthy attempt.

Appreciate the shout-out! 

I'd say part two was a clearer example of "FM'ing the FM" though (finished 6th as relegation favourites, with joint 1st place in Clean Sheets and 2nd fewest Goals Scored). I'm sure it'd be possible to create something even more extreme, but it does indeed show there's nothing "scripted" about defending well and scoring from the only shot you get. Thread linked by Svenc is a nice example, if you decide to do something like that at a top club that can afford real quality players, you can pretty much shut the AI down completely.

[enigmatic]

Think Part Two was too good! Whereas Burnley were exactly like the weaker sides in the AI; just sat very deep and frustrated teams but on a good day that was enough to pick up a point or three.

[ExeChris]

One whole season's worth of data; yes I did manage to score the second most goals in the league but I had to create by far the most chances to do so.

 

Sorry they are bad pics but I scored 77 goals but created 139 chances.  Nobody else created more than 95 chances.

Edited July 13, 2020 by ExeChris

[Svenc]

7 hours ago, Zemahh said:

Appreciate the shout-out! 

I'd say part two was a clearer example of "FM'ing the FM" though (finished 6th as relegation favourites, with joint 1st place in Clean Sheets and 2nd fewest Goals Scored). I'm sure it'd be possible to create something even more extreme, but it does indeed show there's nothing "scripted" about defending well and scoring from the only shot you get. 

The suspicious that there was is one of the reasons why SI are careful when improving their AI tactics and match management (the AI changes things around during a match according to scorelines, and may only attack for a couple minutes -- it REALLY doesn't care about who has the most shots come the end, and shouldn't). The main coder when he was still around on these boards is the most purist as you can get. And he took such accusations personal.

And yes, 9.9/10 FMers playing incredibly attacking tactics 24/7 oft not only exposing themselves but also compressing the space in the final third is the reason this never ever goes the other way (just take a look into the download sections to get a clue) . The rest is then being killed by the player overachievement and the AI reacting to changing Team reputations, e.g. the AI not playing exposing attacking football itself against the overachieving player anymore.

Interesting threads on playing defensive football, btw.

Edited July 14, 2020 by Svenc

[Zemahh]

11 hours ago, ExeChris said:

One whole season's worth of data; yes I did manage to score the second most goals in the league but I had to create by far the most chances to do so.

What that tells me, is that you are inefficient (whether tactically, or down to your personnel). Most of the tactics posted on this forum are overly-aggressive attempts to overwhelm the AI, using high-risking mentalities mixed with a bunch of Attack duties, which is how you get 30+ shots, 0 goals and a goal conceded from counter. Such tactics can also destroy the AI by a few goals margin if it lines up offensively, but as soon as your reputation improves and you start seeing defensive tactics, their success rate starts dropping. This is when you start seeing tons of low-quality shots from all kinds of angles, because the space is no longer there (and the more attacking the player's mentality, the less likely he is to remain patient).

Besides that, stats like "chances created" or "shots on target" hold very little value. For all you know, half of those could be obstructed set piece headers, which of course aren't going to be converted regularly. You're much better off analyzing your shots on match-to-match basis, clicking on all of them and excluding set pieces and headers, which is how you should find out how many quality chances you actually had (open-play, 6-yard box, by foot). It's tedious and I wish there was a faster way of doing it, but far more reliable than taking stats at face value.

[ExeChris]

That's true and I don't doubt that half my chances could be obstructed set piece headers etc, but surely that should be the same for the other teams as well?   Surely in the game a clear cut chance is a clear cut chance and a half chance is a half chance whatever the team?  I have good players, international strikers worth tens of millions of pounds so the personnel is ok.  No doubt my tactics could be better, but I find it annoying that every club bar one scores more goals per chance than me over the season (and this isn't unique to this season)

[Nacaw]

45 minutes ago, ExeChris said:

That's true and I don't doubt that half my chances could be obstructed set piece headers etc, but surely that should be the same for the other teams as well?   Surely in the game a clear cut chance is a clear cut chance and a half chance is a half chance whatever the team?  I have good players, international strikers worth tens of millions of pounds so the personnel is ok.  No doubt my tactics could be better, but I find it annoying that every club bar one scores more goals per chance than me over the season (and this isn't unique to this season)

Sorry, but no. I don't know exactly what arbitrary information is used to give a chance the label "CCC", but it's very clear that they are not all created equal. Sometimes my team will get what looks to me like a great chance, cross into the 6 yard box with no marking, and it's not counted as even a half chance. Then my winger stands on the byline and hits the side netting, clear cut chance counted!

I wouldn't go as far as saying it's a meaningless stat, but it's not be-all and end-all for measuring your teams performance.

if we had Expected goals (xG) in the game, then I'd trust it a lot more.

Edited July 14, 2020 by Nacaw

[HUNT3R]

52 minutes ago, ExeChris said:

Surely in the game a clear cut chance is a clear cut chance and a half chance is a half chance whatever the team?

This is the last match I played, so it was just a random match and not picked because of anything special.

Half Chances. Which of these, would you say, qualifies as a half chance, if any:

 

And then which of these, if any, are Clear Cut chances? If you decide any of them aren't clear cut, would you say they're half chances at least or not even that?

[Svenc]

31 minutes ago, Nacaw said:

Sorry, but no. I don't know exactly what arbitrary information is used to give a chance the label "CCC", but it's very clear that they are not all created equal. Sometimes my team will get what looks to me like a great chance, cross into the 6 yard box with no marking, and it's not counted as even a half chance. 

In theory, over a fully Season and a breadth of different chances, this should even out some. There's a reason why the tactical forums (NOT the download section) has oft promoted the idea of "multi-dimensional" tactics, which is tactics that ideally see a variety of assists and finishers. This either way will be a case of the dynamic AI match management being more efficient as far as those flawed stats are concerned, and little more.

As argued, if the AI can do something the player cannot, the suspicious is immediately that there must be something fishy going on...

Edited July 14, 2020 by Svenc

[ExeChris]

3 minutes ago, Svenc said:

In theory, over a fully Season and a breadth of different chances, this should even out some. There's a reason why the tactical forums (NOT the download one) has oft promoted the idea of "multi-dimensional" tactics, which is tactics that see a variety of assists and finishers. This either way will be a case of AI match management being more efficient as far as those flawed stats are concerned, and little more.

As argued, if the AI can do something the player cannot, the suspicious is immediately that there must be something fishy going on...

This is what I mean.  No I don't know what always quantifies a clear cut chance; often I watch my game on comprehensive highlights and what looks to me a CCC doesn't register as even a HC and vice versa, but surely over a season, the AI teams would also have these chances that are classified the same way, which is why I get frustrated that I seem to need many more of them to score a similar amount of goals.

[HUNT3R]

On 13/07/2020 at 21:37, ExeChris said:

One whole season's worth of data; yes I did manage to score the second most goals in the league but I had to create by far the most chances to do so.

 

Sorry they are bad pics but I scored 77 goals but created 139 chances.  Nobody else created more than 95 chances.

You're only scoring half your 'chances'?

I'd have a look at those chances. See if there's any pattern(s) that stand out or any specific players who are to blame, etc.

I'm not sure that's normal, though I only have my own save to compare with. I've had similar rates with all the teams I've managed. Here's the latest:

So I've had only 64 chances created, which is only 6th best. I scored 55 goals. Whether they were 'chances' I can't tell you from the stats.

Looking at another screen, I see this: 

So conversion rates for what the game thinks a CCC is, is very similar for both me and my opposition.

[HUNT3R]

9 minutes ago, ExeChris said:

This is what I mean.  No I don't know what always quantifies a clear cut chance; often I watch my game on comprehensive highlights and what looks to me a CCC doesn't register as even a HC and vice versa, but surely over a season, the AI teams would also have these chances that are classified the same way, which is why I get frustrated that I seem to need many more of them to score a similar amount of goals.

Depends. If your tactic, player selection or in-match decisions aren't good enough, you may be giving up easier chances than you're creating.

If you're insanely attacking and counter pressing etc, you may be squeezing play in that final this so much that the chances you create aren't as good, as often, as you'd hope. On the other hand, being constantly overly aggressive, may lead to you being very open and giving easier chances away.

If you are a top team and you have a lot of smaller, stylish players in a team who is just plain defensive and picked a team to suit, you may struggle to hold onto the ball and they may just bully you in both penalty boxes.

There are a lot of different possible scenarios and sometimes/often? we do things different to how other teams do, so your team stands out.

To me, my stats almost always look okay (I prefer watching matches over watching stats so I tend not to care too much about them) but then I also don't do silly things like counter pressing on Attack duty with a relegation favourite etc.

[ExeChris]

That's true if the stats only related to my games, but the stats for the whole league include the games between AI clubs, so my tactics don't make any difference to how many chances they create and take.

And I checked the CCC page as above, for my last 29 league games apparently I've created 57 (scored 44) and had 29 created against me (scored 19).  I can't believe this - there is no way I've only missed 13 clear cut chances in the last 29 league game.

Anyway, perhaps I am moaning too much, this is my last season as I'm approaching 75 years old (in the game!) so am planning on quitting at the end of this season.  If I could just finish ahead of Bayern before I finish...

[HUNT3R]

Just now, ExeChris said:

That's true if the stats only related to my games, but the stats for the whole league include the games between AI clubs, so my tactics don't make any difference to how many chances they create and take.

I don't think you understand. If you do something like what I've listed as examples - YOU are in 100% of YOUR matches. So YOUR stats may look different to the other teams who only play you twice in 38 matches.

3 minutes ago, ExeChris said:

I can't believe this - there is no way I've only missed 13 clear cut chances in the last 29 league game.

That's also a point that was made. CCCs (or what counts as a CCC) aren't always accurate.

[DeanMon]

1 hour ago, HUNT3R said:

I don't think you understand. If you do something like what I've listed as examples - YOU are in 100% of YOUR matches. So YOUR stats may look different to the other teams who only play you twice in 38 matches.

That's also a point that was made. CCCs (or what counts as a CCC) aren't always accurate.

@HUNT3R

First, a chance is different from a shot, as a shot from the circle is not a chance.

Whatever definition SI uses for chance, it has be the the same for both AI and human. A chance is a chance, regardless who is in control, which league it happens,...

Some comments are like "human plays too overwhelming against AI so chances are rushed and missed...". So if AI Liverpool plays against AI Aston Villa, they don't play overwhelmingly?

The factor I can think of is the quality of strikers. Do you think Hoffenheim strikers are that low comparing to the average?

Thank you @ExeChris for providing the data, my point is clear here, human's conversion rate (of chances) is extremely lower than AI's average conversion rate, despite it is a closed pool of teams.

If chance is not a well-defined statistics, please get rid of it. Otherwise, it misleads our interpretation about the gameplay. 

Edited July 14, 2020 by DeanMon

[Nacaw]

1 hour ago, ExeChris said:

That's true if the stats only related to my games, but the stats for the whole league include the games between AI clubs, so my tactics don't make any difference to how many chances they create and take.

And I checked the CCC page as above, for my last 29 league games apparently I've created 57 (scored 44) and had 29 created against me (scored 19).  I can't believe this - there is no way I've only missed 13 clear cut chances in the last 29 league game.

Anyway, perhaps I am moaning too much, this is my last season as I'm approaching 75 years old (in the game!) so am planning on quitting at the end of this season.  If I could just finish ahead of Bayern before I finish...

The point is, some tactics can create better quality CCC's than others. If the AI has a tactic that does this, they should score more than half. 

I'm not even sure what counts as a chance create under team stats? Just CCC? Both half chances and CCC? Some other definition? 

I went and checked my current game 30 league games into the season, and according to team overview I've created 138 chances and scored 100 goals. I play counter attack football, and would expect to get good quality chances. But then I look at Man Utd, and they've score 94 from 111 chances, and play possession based.. I think my conclusion is that chances created isn't a very helpful for analysing how your team is doing.

If I look at CCC on the tactics page for the last 30 league games, I'm converting 43 out of 60. My opponents are 2 out of 3. The same page is also telling me that I've had 14 assists from the halfway line or my own half, but it shows 0 long balls under assist type 

I wish we had better statistics. 

 

 

[Zemahh]

1 hour ago, DeanMon said:

So if AI Liverpool plays against AI Aston Villa, they don't play overwhelmingly?

"Play overwhelmingly"? AI's tactics don't even come close to what human players use most of the time, you will never see it throw everyone forward like some tactics on this forum do. It also isn't capable of setting up complex attacking patterns with hours of thought put into them, like human players can, which of course results in worse tactics and less chances created.

The most adventurous you will see it be in most games it's expected to win, is a Positive 4-2-3-1 or something similar, it doesn't even switch to Attacking unless desperately chasing a goal in the final minutes (in most games, anyway). Compare that to your usual download-section tactic with Attacking mentality, 8 Attack duties and a plethora of instructions, and it becomes obvious why the AI wastes less chances; it also creates less of them, but of higher quality, due to more conservative tactics.

Another SI Staff quote on "cheating AI":

On 12/09/2019 at 14:39, Jack Joyce said:

The AI don't suss out your tactic and create counters, tactically they're just seeing a team win the league the previous season and thinking "we should defend more against this team" because they expect to get pumped. Most of the time where people have this kind of issue, it's because they used a very direct/counter attacking tactic to win the league which doesn't work as well against packed defences. You need to adapt to the opposition and analyse your team's performance and spot flaws with your play. "The AI are cheating" is too often a convenient target when people can't work out the actual reason why their team is losing.

What would be the point of rigging things? We're all active players of the game too. If you lose more we don't benefit from it, making users angry doesn't benefit us either. There'd be no reason to do anything like this.

Imagine if the AI was actually smart enough to actively adapt to your tactics like human managers would. It's accused of cheating now and it barely knows when to line-up offensively and when not to, imagine the outcry if it could shut down specific threats identified during matches. Half of the player base would never win another game and the forum would be on fire. 

Edited July 14, 2020 by Zemahh

[DeanMon]

20 minutes ago, Zemahh said:

"Play overwhelmingly"? AI's tactics don't even come close to what human players use most of the time, you will never see it throw everyone forward like some tactics on this forum do. It also isn't capable of setting up complex attacking patterns with hours of thought put into them, like human players can, which of course results in worse tactics and less chances created.

The most adventurous you will see it be in most games it's expected to win, is a Positive 4-2-3-1 or something similar, it doesn't even switch to Attacking unless desperately chasing a goal in the final minutes (in most games, anyway). Compare that to your usual download-section tactic with Attacking mentality, 8 Attack duties and a plethora of instructions, and it becomes obvious why the AI wastes less chances; it also creates less of them, but of higher quality, due to more conservative tactics.

Another SI Staff quote on "cheating AI":

Imagine if the AI was actually smart enough to actively adapt to your tactics like human managers would. It's accused of cheating now and it barely knows when to line-up offensively and when not to, imagine the outcry if it could shut down specific threats identified during matches. Half of the player base would never win another game and the forum would be on fire. 

You are making a lot of assumptions here, my friend. You don't see my tactics, you think I play Attacking/Very Attacking very game?

"AI's tactics don't come close to what human uses?" Take any team in PL, play against AI Liverpool, post the screenshot of final match stats to see how overwhelmingly AI Liverpool plays, then we talk again.

I am not afraid of a smart AI, and as I said I am not afraid of losing to AI, but I have no tolerance of "magic" AI..... Look at the screenshots above of @ExeChris, explain it to me.

[zyfon5]

41 minutes ago, DeanMon said:

You are making a lot of assumptions here, my friend. You don't see my tactics, you think I play Attacking/Very Attacking very game?

"AI's tactics don't come close to what human uses?" Take any team in PL, play against AI Liverpool, post the screenshot of final match stats to see how overwhelmingly AI Liverpool plays, then we talk again.

I am not afraid of a smart AI, and as I said I am not afraid of losing to AI, but I have no tolerance of "magic" AI..... Look at the screenshots above of @ExeChris, explain it to me.

Playing as a top Japan side against 2025 Liverpool in club world cup. Liverpool got completely shut out by a Japan side that is at best an upper mid table championship club level.

 

Edited July 15, 2020 by zyfon5

[DeanMon]

6 minutes ago, zyfon5 said:

Playing as a top Japan side against 2025 Liverpool in club world cup. Liverpool got completely shut out by a Japan side that is at best an upper mid table championship club level.

 

You understand I mean AI Klopp Liverpool in the context, right? Your 2025 Liverpool could be managed by ultra-defensive Simeone.

[zyfon5]

3 hours ago, DeanMon said:

You understand I mean AI Klopp Liverpool in the context, right? Your 2025 Liverpool could be managed by ultra-defensive Simeone.

How about this? This Chelsea side is not defensive by any means (still managed by lampard). And before you use the same argument again that they are not Klopp's Liverpool, the fact that this Chelsea side makes it into the club world cup is already proof that it is one of the best European sides so I'm not sure why does it matter to you. And if you are still not convinced that you do not need more chances than the AI to score when I get back to my laptop I can post screenshots of matches from the West Brom side I managed last season which have stats similar to this.

Edited July 15, 2020 by zyfon5

[HUNT3R]

7 hours ago, DeanMon said:

Whatever definition SI uses for chance, it has be the the same for both AI and human. A chance is a chance, regardless who is in control, which league it happens,...

Some comments are like "human plays too overwhelming against AI so chances are rushed and missed...". So if AI Liverpool plays against AI Aston Villa, they don't play overwhelmingly?

The factor I can think of is the quality of strikers. Do you think Hoffenheim strikers are that low comparing to the average?

No. I think your strikers are probably just fine.

The tools are the same. The ME treats everyone the same. So then the obvious question is - if you are the only team with different stats - what are you doing and, more importantly, what are you doing different?

It may be your tactic. It may be how you select players or just the quality of players in certain positions. It may be poor in-match decisions/reactions.

Look at Bordeaux in my save. They're 18th in the league. 63 chances created. The scored only 27. That's worse than even your own stats.

AI Liverpool might. I don't know exactly how they play in FM. Most favourites start on Positive, even if big favourites. Liverpool in FM may be as aggressive as PSG in my save. PSG currently start matches on Attacking. That said, even then, users' attacking tactics are usually far more attacking than how the AI sets up. I made up examples of the sort of reasons why you may see something different. None of us know how you set up, your player attributes or what you do in matches.

There's another thing - I saw someone post recently about the last 5 matches conceding a late goal every time. Now, obviously we know (if trailing) the AI will try and come back into the match. We'd do the same. The tactic was (though "only" Positive Mentality) very attacking and that was probably most, if not all of the reason why. When asked if any changes are made toward the end of a match as there's a clear pattern, the only answer was to drop Mentality from Positive to cautious. Still keeping the aggressive TIs. Still keeping the top heavy formation with both fullbacks as WB/A. It's a change that makes very little sense and it doesn't even necessarily tighten up anything. It's more disjointed which may cause even more issues. Even the AI wouldn't do something as silly as that. If it drops to Cautious, roles, duties and TIs will follow. It might not be drastic, but it certainly wouldn't still have both fullbacks on attack and extreme pressing settings with a much higher LoE.

7 hours ago, DeanMon said:

Thank you @ExeChris for providing the data, my point is clear here, human's conversion rate (of chances) is extremely lower than AI's average conversion rate, despite it is a closed pool of teams.

So I must be an AI bot? Because I can do it. I've posted my screenshots. I've posted some thoughts around why it might be different as well.

[Bruce Drundrige]

Anyone who thinks the AI is rigged because of unfair results or 'unscored goals' can't have watched much football, or perhaps isn't familiar with the agony of supporting a team whose results are definitely not deserved despite how well we played! Unfair, unexplainable or shock results and scenarios are part of the joy of football (and sport in general).

Interesting that it's always the team who dominate but don't win who complain?

Also, be careful what you wish for in terms of the intelligence of the AI. If it was as smart as us, and could react like us to in-game events, we'd never win.

 

[DeanMon]

2 hours ago, zyfon5 said:

How about this? This Chelsea side is not defensive by any means (still managed by lampard). And before you use the same argument again that they are not Klopp's Liverpool, the fact that this Chelsea side makes it into the club world cup is already proof that it is one of the best European sides so I'm not sure why does it matter to you. And if you are still not convinced that you do not need more chances than the AI to score when I get back to my laptop I can post screenshots of matches from the West Brom side I managed last season which have stats similar to this.

Hello, there might be a misunderstanding here. Let me clarify:

1) Many complain that why human does not often win despite dominating the match. Many "excuse" that human can't score because human's tactics are "too overwhelming" AI's tactics, which may be true.

2) Now my argument is that: when AI's tactics is "too overwhelming" human's tactics (AI Liverpool), it seems to score just fine.

The poster said AI's tactics are not as overwhelming as human's tactics, that is why AI can score. I said it is wrong, please check the current AI Liverpool to see how overwhelming it is.

Your example Chelsea is the best team of Europe, but its tactics may not be the most overwhelming.

Edited July 15, 2020 by DeanMon

[DeanMon]

53 minutes ago, HUNT3R said:

No. I think your strikers are probably just fine.

The tools are the same. The ME treats everyone the same. So then the obvious question is - if you are the only team with different stats - what are you doing and, more importantly, what are you doing different?

It may be your tactic. It may be how you select players or just the quality of players in certain positions. It may be poor in-match decisions/reactions.

Look at Bordeaux in my save. They're 18th in the league. 63 chances created. The scored only 27. That's worse than even your own stats.

AI Liverpool might. I don't know exactly how they play in FM. Most favourites start on Positive, even if big favourites. Liverpool in FM may be as aggressive as PSG in my save. PSG currently start matches on Attacking. That said, even then, users' attacking tactics are usually far more attacking than how the AI sets up. I made up examples of the sort of reasons why you may see something different. None of us know how you set up, your player attributes or what you do in matches.

There's another thing - I saw someone post recently about the last 5 matches conceding a late goal every time. Now, obviously we know (if trailing) the AI will try and come back into the match. We'd do the same. The tactic was (though "only" Positive Mentality) very attacking and that was probably most, if not all of the reason why. When asked if any changes are made toward the end of a match as there's a clear pattern, the only answer was to drop Mentality from Positive to cautious. Still keeping the aggressive TIs. Still keeping the top heavy formation with both fullbacks as WB/A. It's a change that makes very little sense and it doesn't even necessarily tighten up anything. It's more disjointed which may cause even more issues. Even the AI wouldn't do something as silly as that. If it drops to Cautious, roles, duties and TIs will follow. It might not be drastic, but it certainly wouldn't still have both fullbacks on attack and extreme pressing settings with a much higher LoE.

So I must be an AI bot? Because I can do it. I've posted my screenshots. I've posted some thoughts around why it might be different as well.

@HUNT3R

Let me restate my points:

1) The definition of chance is the same for both human and AI.

2) The chance conversion rate of human is lower than the AVERAGE chance conversion rate of ALL AI.

3) The quality of human's strikers is no less than the quality of ALL AI's average striker.

4) You cannot say the quality of human's chance is THAT MUCH lower over the large sample, because the quite-average Hoffenheim also plays defend/attack against certain teams just as average AI.

So what is the reason here?

Of course, 1 season might still be not large enough, but I cannot rest my case until I see human's conversion rate is close to ALL AI's AVERAGE conversion rate over a large sample.

Edited July 15, 2020 by DeanMon

[HUNT3R]

26 minutes ago, DeanMon said:

@HUNT3R

Let me restate my points:

1) The definition of chance is the same for both human and AI.

2) The chance conversion rate of human is lower than the AVERAGE chance conversion rate of ALL AI.

3) The quality of human's strikers is no less than the quality of ALL AI's average striker.

 

I posted screenshots of my season. What do you say about that? I am able to compete and show similar stats to that of the AI. How do your stats compare to it? Can you post screenshots of the same screens as me and @ExeChris?

You're making it a general thing (AI vs human) where it could just be AI vs YOU specifically and then there would be ways to improve the conversion rates.

[santy001]

One of the biggest issues by far is that the entire premise on which to argue it is either:

a) The in-game AI is in some way self-aware, sentient and will move parameters in games to penalise you. 

b) SI a game developer, who have a business model dependant upon which is solely selling you their game, are putting this priority second to some kind of arbitrary desire to cheat players. 

 

Conflating 1 shot for one team with 1 shot for another team is completely meaningless as well. Expand this out to your clear cut chances = all other clear cut chances and so on. It's like trying to say water from a polluted, toxic river is the equivalent of tap water because water=water. 

Most games that tend to implement AI cheating to provide a challenge are open about it. It also tends to serve as the games difficulty slider in games such as an RTS/Strategy game giving the AI money/strategical resources it doesn't have directly to provide a greater challenge. If a game does that, there's no reason to keep it a secret.

[DeanMon]

4 minutes ago, santy001 said:

One of the biggest issues by far is that the entire premise on which to argue it is either:

a) The in-game AI is in some way self-aware, sentient and will move parameters in games to penalise you.

b) SI a game developer, who have a business model dependant upon which is solely selling you their game, are putting this priority second to some kind of arbitrary desire to cheat players. 

 

Conflating 1 shot for one team with 1 shot for another team is completely meaningless as well. Expand this out to your clear cut chances = all other clear cut chances and so on. It's like trying to say water from a polluted, toxic river is the equivalent of tap water because water=water. 

How about this?

Your ME is not perfect in interpretating real football. It contains flaws. It penalizes human unfairly in terms of "incorrect" pep talk, training, tactical instructions, shouts... which lead to discrepant match results? 

[santy001]

It's not my ME but you're trying to create a strawman of sorts because I never claimed the ME is perfect.

Also, if you mismanage your team... then yes there are going to be consequences to mismanagement. Making a mistake, or rather a series of mistakes, and getting an adverse result isn't a discrepancy. Bears no relevance to the topic at hand, but you might as well complain that sticking your hand in a fire burns. The AI too makes mistakes and it gets punished for them. 

---

There's only two ways in which it can happen, either the AI has gone rogue and is far more sophisticated than any other video game AI in existence or you're saying SI have counter-intuitively and counter-productively designed the game against all their aims as a business and are doing it to cheat you. 

[DeanMon]

24 minutes ago, HUNT3R said:

I posted screenshots of my season. What do you say about that? I am able to compete and show similar stats to that of the AI.

@HUNT3R

Yeah, such data speaks more than thousands words. May you post data of all teams?

There is one more thing is that you games must not be reloaded for any, because if they were, then the data is cherry-picked which will be skewed in your favoured. Hofenheim's data must be verify for that fact as well.