How are game results calculated?

[BLD14]

In my first season with Derby, and I've hit a mid season blip and am currently doing awful at the moment. My strikers arent scoring and my defenders arent defending. I was conceding a lot from corners. So changed my corner tactics to one I found online. I know I need to train it, but I conceded from a corner in the very next game....

 

I played what was my last game in charge before being sacked. I lost 4-1 at home to 9th Blackburn despite having more shots and shots on target, and more possession. Their striker got 4 goals, also despite me telling my vice captain central defender, who was strong in the air, to man mark him. Andre Wisdom my right full back has been awful for me and I cant remember the last good game he had for me (my first choice RB is injured) and he complains whenever I tell him he needs to improve.

 

After getting fed up, I reloaded before the match and skipped to the next day to sim the match. I kept the exact same tactics and team selection and won. I did this 3 more times and won twice and drew once. Andre Wisdom had a lowest match rating of 6.9 in this games and a highest of 7.9, scoring the winner in one of them.

 

My point is, why is the simmed results so drastically different from the time I actually played it? And why can Andre Wisdom suddenly be a competent footballer again? Is this game made to be harder than it is for the AI? Because if so, that's ridiculous and broken.

 

Thank you in advance for any replies

[FrazT]

I posted this in another thread and as you can see there are a lot of variables, starting immediately after the end of the previous match that can affect the result of the next match.

OK- this list may not be all the factors:

Final team talk after the last game- player's reaction to this talk
Training schedules and player performance from then till the next game
Injuries at training
Any player interactions/discussions
Pre-game odds
Team and formation selection
Team and individual player instructions
Opposition instructions
Pre-game team talk and expectations
Player reaction to team talk
As soon as the game starts, any change of any sort, however minor  from you or the AI
In game injuries
In game substitutions
In game shouts
In game tactical changes

All of these and probably some more that I have forgotten will affect the constant game outcome calculations that the code makes.

[Orikoru]

I've noticed exactly the same thing. In some of my saves where I've got frustrated at how rubbish my team is, I've reloaded the game, picked exactly the same team line-up, tactics etc, and Instant Resulted it, end up winning the game or draw instead of getting beat 4-0 when I played it myself. It doesn't make sense that it's that drastically different. Well the only way it might make sense is if you sim the game, the game AI radically changes (corrects?) your tactics within the opening minutes of the game. But surely it doesn't?

[zyfon5]

1 hour ago, Orikoru said:

I've noticed exactly the same thing. In some of my saves where I've got frustrated at how rubbish my team is, I've reloaded the game, picked exactly the same team line-up, tactics etc, and Instant Resulted it, end up winning the game or draw instead of getting beat 4-0 when I played it myself. It doesn't make sense that it's that drastically different. Well the only way it might make sense is if you sim the game, the game AI radically changes (corrects?) your tactics within the opening minutes of the game. But surely it doesn't?

Every time u reload the game the AI make changes to how they approach the game depending on what happened in game. So even with the exact same line up the results might not be the same. Regarding why some ppl get better results when they instant result, most likely they are just bad at match management and failed to respond correctly to what happened in game. I don't regard myself good at match management either but around 70% of the time I can get better results very often turning a loss from instant result to a win if I manage the match myself.

[yolixeya]

7 hours ago, BLD14 said:

My point is, why is the simmed results so drastically different from the time I actually played it? And why can Andre Wisdom suddenly be a competent footballer again? Is this game made to be harder than it is for the AI? Because if so, that's ridiculous and broken.

From what I could understand, when you simulate a match it doesn't use the same ME like when you play.

And just to add, this is the first time ever that I installed custom skin that has a instant replay button and of course I first tested it and was able to win games that I lost when simulating. So I don't think it's harder, and also I read people say that the game doesn't differentiate between human and AI.

[Britrock]

14 hours ago, yolixeya said:

From what I could understand, when you simulate a match it doesn't use the same ME like when you play.

And just to add, this is the first time ever that I installed custom skin that has a instant replay button and of course I first tested it and was able to win games that I lost when simulating. So I don't think it's harder, and also I read people say that the game doesn't differentiate between human and AI.

It uses the exact same ME, there's no reason that SI would double their workload and write a second one for simming that was different to their main one. Writing one ME is hard enough

21 hours ago, Orikoru said:

I've noticed exactly the same thing. In some of my saves where I've got frustrated at how rubbish my team is, I've reloaded the game, picked exactly the same team line-up, tactics etc, and Instant Resulted it, end up winning the game or draw instead of getting beat 4-0 when I played it myself. It doesn't make sense that it's that drastically different. Well the only way it might make sense is if you sim the game, the game AI radically changes (corrects?) your tactics within the opening minutes of the game. But surely it doesn't?

Because there's an element of randomness involved. How boring would it be if the same game always produced the same result? One minor change in a game can have huge knock on effects that can lead to massively different results by the end - i.e. Something as simple as your striker scoring early, now the opposition has to chase the game and they're more attacking so there's more space behind them. If you replay that game but this time he misses instead, the opposition can now sit deep and defend and a 4-0 win originally might now be a 0-0 draw

[Orikoru]

24 minutes ago, Britrock said:

It uses the exact same ME, there's no reason that SI would double their workload and write a second one for simming that was different to their main one. Writing one ME is hard enough

Because there's an element of randomness involved. How boring would it be if the same game always produced the same result? One minor change in a game can have huge knock on effects that can lead to massively different results by the end - i.e. Something as simple as your striker scoring early, now the opposition has to chase the game and they're more attacking so there's more space behind them. If you replay that game but this time he misses instead, the opposition can now sit deep and defend and a 4-0 win originally might now be a 0-0 draw

I would say an element of chance yes, but not complete randomness. Randomness implies that what we do is largely irrelevant. 

[yolixeya]

9 hours ago, Britrock said:

t uses the exact same ME, there's no reason that SI would double their workload and write a second one for simming that was different to their main one. Writing one ME is hard enough

I didn't think that thera are 2 ME but that it simulates it just like it simulates it when the league loaded as view only. I didn't know that it actually uses ME. Then it's definitely easier to win when you play because you are able to spot things and make corrections. That is if you know what you are doing.

[HUNT3R]

There's the 'proper' full ME and then there's the quick sim ME. Leagues not loaded and leagues not in full detail mode will use the quick sim ME. Your own league (by default) will be full detail, so the matches in that uses the full ME.

That said, I don't know how instant result works. An easy way to check would be to IR the match and see if you can go and watch it back. If you can't view how the match played out, it used the quick sim ME.

[Svenc]

13 minutes ago, HUNT3R said:

That said, I don't know how instant result works. An easy way to check would be to IR the match and see if you can go and watch it back. If you can't view how the match played out, it used the quick sim ME.

If you get better results at instant results, it's likely that the assistant then taking over match management making better decisions.

I think the title of this thread is misleading.

The engine doesn't calculate a result and works from that. The ME calculates a virtual match of football -- pass by pass, positioning by positioning, bug, er goal by goal -- and whatever is the result at the end of that is the result.

Therefore, that's I've always viewed it, the element of chance comes in in the way whether the chances are converted or aren't. (If SI would have it correct, even the best ones would be roughly 50/50 chances most of the time, and players on their level not make a gigantic difference.) See also Rashidi advocating his SIBOT (?) Chance Analysis, which no doubt gives you another edge over the game's AI managers -- including your assistant.  

 

That is in the long run. In individual matches, all kinds of crap should go. These chances then come about by individual quality (say tons of dribbles by a player who averages like half a dozen per match), tactics, errors, or a mix. Up to you to analyze what's what.

Edited May 19, 2020 by Svenc

[HUNT3R]

1 minute ago, Svenc said:

If you get better results at instant results, it's likely that the assistant then taking over match management making better decisions.

Yeah, I left this part out on purpose as I had no idea.

I don't use Instant Result. I know the Assistant is in charge when you Go On Holiday, but wasn't sure about IR.

[yolixeya]

18 minutes ago, HUNT3R said:

That said, I don't know how instant result works. An easy way to check would be to IR the match and see if you can go and watch it back. If you can't view how the match played out, it used the quick sim ME.

I don't know either as I don't use it that often but I will check that out when I get home.

[JordanMilly]

@HUNT3R @yolixeya - Just checked it (as the skin I use has instant results, but I never use it), and it uses the full ME as any other match played in a league your team is participating in. The assistant manager is the one making the decisions though, unless you've put in place a match plan.

[enigmatic]

People consider the match engine to be at fault when they repeat the game with the same tactics and still lose, and they consider the match engine to be at fault when they repeat the game with the same tactics and win. And also at fault when they change the tactics and don't win.

 

A strange game. The only way to win is not to replay....

[sporadicsmiles]

The instant result button uses the ME, same as if you would play it. This is clear since you can go back and rewatch the match whenever you use it. So if you are seeing better results it means your AssMan is better at reacting to what is going on in the game and getting performances from your players than you are. You should go and watch the matches you sim to see what changes are made, what your team does differently when you are not in charge. This will give you an idea of what you can do to improve your own performance. 

In addition why are people surprised when a game is different if they replay it? There is an inherent bias in this view, since how many people replay games they won the first time they played it? I'd hazard very few. The only time I replay games is if I have to go back over part of a save because I forgot to save or had a crash (a rare occurrence, but it happens). And then I will try to get similar results, which can be surprisingly hard. Football is highly non-linear. It is not like an equation where if you feed it the same inputs you get the same outputs. It is more chaotic, similar inputs can lead to vastly different outcomes. And what happens in the game directly feeds back into the outcome of the game. It is a very dynamic process. It would be significantly more worrying if you always got the same result every time you replayed a game, because that would indicate the simulation is not dependent enough on what is actually happening. 

[Svenc]

6 hours ago, sporadicsmiles said:

It would be significantly more worrying if you always got the same result every time you replayed a game, because that would indicate the simulation is not dependent enough on what is actually happening. 

Equally worrying was it if there wouldn't be a general trend -- not over isolated "reloads", but replaying the same match at least a dozen times. Then you should see a trend. One "frustrating" scenario for those one-off reloads is the AI sitting Deep and sometimes deservedly, sometimes luckily, getting a win/draw. IIRC for a much older iteration there was a thread demonstrating that better players dropped those points less so (also having more goals from open play as opposed to a set piece, which was counted over the course of the matches), whilst lesser players in this case it was downloading tactics (and thus never learning how to play) less so. 

Edited May 20, 2020 by Svenc

[fmFutbolManager]

On 18/05/2020 at 00:59, BLD14 said:

In my first season with Derby, and I've hit a mid season blip and am currently doing awful at the moment. My strikers arent scoring and my defenders arent defending. I was conceding a lot from corners. So changed my corner tactics to one I found online. I know I need to train it, but I conceded from a corner in the very next game....

Conceding from corners is normal.

On 18/05/2020 at 00:59, BLD14 said:

I played what was my last game in charge before being sacked. I lost 4-1 at home to 9th Blackburn despite having more shots and shots on target, and more possession. Their striker got 4 goals, also despite me telling my vice captain central defender, who was strong in the air, to man mark him. Andre Wisdom my right full back has been awful for me and I cant remember the last good game he had for me (my first choice RB is injured) and he complains whenever I tell him he needs to improve

More shots and more shots on target don't really tell us much. The quality of shots is better. Nevertheless, have you ever watched Man City play? They almost always have more possession, more shots and more shots on target and still get beat. It happens. I know you never want to hear this but wouldn't football and Football Manager be really rubbish if you got points awarded for total number of shots rather than goals? It adds predictability to it. Makes it a lot more exciting. Yep, I also moan all the time when I lose despite dominating, but I keep coming back because it is unpredictable.

As for man marking it sounds like that was the wrong thing to do against this player if he scored four goals against you. I assume he has a lot better movement to be able to find space against your defender. I seldom get a central defender to man mark anyone, a DM is usually better for this. You risk your central defender being pulled out of position and if he's not as mobile getting beaten too often.

Finally, the match ratings are not always a reflection on individual performance. If your central midfielder just sits in his position pinging passes to team mates he won't be getting 9.0 ratings every match, but if his job is just to keep possession and keep the ball moving then he's doing what you've asked of him. Identify what you want your full-back to be doing and check his stats. If they're good enough then great. I play with Complete Wing-Backs and primarily I want them overlapping and pinging balls into the box so things like their pass completion and headers won are not really important. Chances created, successful dribbles, tackles lost and cross accuracy (although not so much the latter) are more important. 

On 18/05/2020 at 00:59, BLD14 said:

After getting fed up, I reloaded before the match and skipped to the next day to sim the match. I kept the exact same tactics and team selection and won. I did this 3 more times and won twice and drew once. Andre Wisdom had a lowest match rating of 6.9 in this games and a highest of 7.9, scoring the winner in one of them.

My point is, why is the simmed results so drastically different from the time I actually played it? And why can Andre Wisdom suddenly be a competent footballer again? Is this game made to be harder than it is for the AI? Because if so, that's ridiculous and broken.

This is such a small sample size. My guess is that when you played the matches yourself you didn't just hit play and leave things for 90 minutes. You would have made changes through the match which affected the outcome. Match results are dependant on so many variables that you will always get a random set of events which lead to the outcome.

The game isn't made to be harder for you than the AI. My guess is that the user makes the game harder than it needs to be. 

[Federico]

6 ore fa, sporadicsmiles ha scritto:

Football is highly non-linear. It is not like an equation where if you feed it the same inputs you get the same outputs.

But if it's true to say that Football is non-linear, don't you think that this cannot apply to Football Manager since, being it a graphic representation of mathematical laws, should actually give same outputs if fed by same inputs?

Again, to me any comparison with real football is inconsistent.

But it's correct to say that inputs are different.

[fmFutbolManager]

7 minutes ago, Federico said:

But if it's true to say that Football is non-linear, don't you think that this cannot apply to Football Manager since, being it a graphic representation of mathematical laws, should actually give same outputs if fed by same inputs?

Again, to me any comparison with real football is inconsistent.

But it's correct to say that inputs are different.

It is not given the same inputs though. When you save and reload the calculation isn't the same so the result isn't the same.

[enigmatic]

And football's a pretty damn non-linear game. Pretty hard to imagine Palace scoring twice against both Manchester clubs if you asked them to repeat the same feat using the same everyone behind the ball tactics again. At the same time, hard to imagine PSG, Barcelona or even Ajax throwing away three goal leads in the Champions League if they got to save and reload.

Very different things can happen with very similar sides. Both the Premier League teams that have won games by nine goals lost the away game...

[Svenc]

1 hour ago, Federico said:

But if it's true to say that Football is non-linear, don't you think that this cannot apply to Football Manager since, being it a graphic representation of mathematical laws, should actually give same outputs if fed by same inputs?

Given that every action in a match/reload must be some weighted "dice roll", so to speak -- be it a pass, a tackle, whatever -- that sounds somewhat illogical. Sometimes that tackle is a success, sometimes it isn't, same as the pass, starting from the very first one… and that's when the butterfly starts to spread its wings. That's how code eventually works, usually. Even actions with a high probability of succeeding can result in bad "rolls" in a row. 

Players don't usually like that, however, or respond badly to such, as entertainingly outlined here by a RPG developer (games that internally also work with a load of stats and Maths). His 95% chance of a hit would translate to bookies odds of 1.05 or 50/1000. As said though, over a dozen or so reloads one should see a trend in the results. That's rarely ever done though.

Edited May 20, 2020 by Svenc

[sporadicsmiles]

3 hours ago, Federico said:

But if it's true to say that Football is non-linear, don't you think that this cannot apply to Football Manager since, being it a graphic representation of mathematical laws, should actually give same outputs if fed by same inputs?

 

You could probably map the phase space of FM should you desire, and try to track all the variables. But you would have to repeat a match a huge number of times. I have no idea the details of the ME programming, so it is difficult to comment. I would rather imagine there are a lot of if loops whose path is determined by a weighted RNG. This is pure speculation mind, so nobody go around quoting this on other threads. I will be unhappy if I see that!! But yeah, which would be a way to model the chaotic system using programming (since I do not know how to do it otherwise). In that case you can expect different outcomes from similar inputs (and remember your inputs are never ever the same for an FM match even if you repeat to keep them close). And then the trick is to balancing the thing so that your method of replicating the non-linear aspects of football are accurate.

Just again to reiterate, I am purely throwing speculation out there because I find this an interesting topic in general (I do a lot of simulations and computational studies in my job), and it is a fun thing to speculate on. Please nobody take this as facts. 

[Federico]

So the argument "playing with Instant result  gives me good results back, while if I play that game by myself I lose" is, generically, non-valid. I mean there are too many factors that could affect the game, in good or bad, made both by AI and the user.

[Svenc]

2 hours ago, Federico said:

So the argument "playing with Instant result  gives me good results back, while if I play that game by myself I lose" is, generically, non-valid. I mean there are too many factors that could affect the game, in good or bad, made both by AI and the user.

I'd argue it is not. Over the longer term (in this case a dozen+ reloads of the same match), the AI Manager if he's superior would yield better overall results as his match management would be superior. Similar to how better players of this game fare better than worse ones. Some of which outperforming any AI of the game by ridiculous margins. Some proclaiming they get the sack early in every other save. There's only one difference in either case: The user providing the input.

That was my argument anyway since my first post. If an AI Manager gets consistently better results, it is because he makes better decisions. SI program that AI, and they are/have always been receiving feedback on where to better it, which is kind of the true level of difficulty of the game. I've reported shenanigans in the past, and that evidently cost the AI points on that release.

It also made for horrible football, and like a dozen+ corners alone average in every match. Whilst horrible football is subjective, I even tried to copy similar tactics to check for raw results on another release. The same team that would yield an almost 80% win percentage for a half of the season tested dropped that by a good 20% plus for the other half tested if provided with tactics such as the AI's in my link above. Wins turned into added draws. Draws turned into the odd additionally loss. Shot conversions dropped from ~16% to below average. And I'd argue back then that was realistic, and deserved. Perhaps it should have been even more severe.

edit: That's purely "tactics" as one factor, mind. Naturally, there could be (and have been) also issues with AI Team and player selections, substitutions, and more.

Edited May 20, 2020 by Svenc