Svenc
-
Posts
5,152 -
Joined
-
Last visited
-
Days Won
1
Content Type
Forums
Articles
FAQs
Online Manual
Support: Blocks
Support: Games
Bug Tracker
SIGames Manual (beta)
Profiles
Posts posted by Svenc
-
-
On 09/04/2023 at 16:29, enigmatic said:
Not sure what the moan is there... they won most of those games, and at least one of them they won by converting xG more efficiently than the opposition so it's the AI that should be moaning about it! The AI, of course, doesn't get to make a topic every time they create chances and yet the OP keeps a clean sheet...
Underlines my theory that the people moaning about this stuff have no understanding of football and just want the flukes to be in their favour every single time though
Historically it's also been the human manager being super successful (as usual...), thus his team rising in reputation to the point that every AI manager choses defensive tactics. As would happen to any manager, AI included, from the start were they managing Man City, Bayern, et all.
Thus not only the AI always having fewer shots -- but also the AI teams exclusively ever scoring from few/er shots. That's bound to happen even if the human manager weren't attacking 24/7. Toss a coin for long enough, you may even get a couple streaks... simple laws of probability. It seems some of the same familiar faces still popping in too.
SPOILER: If you can't do that to the AI in particular in matches where the AI is considered massive match favourite, chances are YOU SUCK AT THIS GAME. SPOILER ENDING. What I said about if the AI were one day actually to become "decent": RAGE QUIT CENTRAL. I don't mind anymore, mind. I actually compare "quitting" football to when I quit smoking. At first, it was pretty hard. Nowadays I'm wondering what the fuss is all about and I'm often actually pretty alienated not merely by players and federations/officials, but fans alike, acting as if a simple (and occasionally fun) game of kicking a ball around indeed WAS a matter of life and death.
0 -
I see these kind of topics have never "evolved". Despite many human managers evidently still outperforming the AI left, right front and center on all accounts on FM20, 22, 23 and 33. Except for one crucial one. Which is winning a few matches despite having fewer/ lesser chances -- which is not gonna happen to that super successful human manager, as it will be exclusively AI that will shut up shop / play on the counter, seeing the human manager team as the big dog to frustrate. You can only ever win with fewer shots if you aim to have fewer shots, Sherlock Obvious. If the human managers would do the same from kick-off in that scenario however, matches would play out like a Western movie shootout, except for nobody ever drawing the gun...
This is one of the reasons why I've stopped playing (the other is that I'm alienated by football these days). Because if the AI were to massively ever improve, the challenge would increase, and "it" would happen MORE oftently. Imagine an AI manager that could "read", as a good human manager can, where that space left to exploit would actually be... and dynamically in-match adjusting its formations/roles/duties and instructions accordingly.
Rage-quit central.
0 -
That said, no FM playing experience currently whatsoever (or football interest, for that matter). But the first post sounds like there MAY be something iffy about Conte's actual match management (which is non-existent on leagues not fully simulated, as matches aren't actually fuly simulated). Wouldn't be a first for AI managers on the series. In particular given the frequency of how often he's being sacked.
Still have vivid memories of AI Guardiola on about FM16ish, back when he was managing one of the most domestically dominating squads in the game at that time. Namely: Bayern. Which he semi-regularly got to scoring less than 60 goals per season, plus sometimes have them losing as many as 7/8 matches throughout. This was in parts connected to him frequently completely isolating Lewandwoski from all the rest, making him easy to mark to boot, as his "lone forward flat midfield" 4-1-4-1 primary "prefered formation" had that lone forward pretty much always on an attack duty regardless of the roles around that forward (and a couple other stuff).
If you're unsure what to look out for, you could "test" this by replacing Conte in the db with somebody else (e.g. another AI manager) doing better in this edition.
And of course, running a suitable amount of season simulations to rule out randomness (see sample size song).
edit: Some of the more recent experiments seem to go in that direction already, albeit just by altering some of Conte's traits.
0 -
Meh, such n00bs (all from within the space of about half a season back then). Unfortunately, AI mans on the last couple editions don't seem to use such ****** tactical decisions during matches anymore (including their own attacking set pieces, throwing everybody and the kitchen sink forward even early game as soon as they go a goal behind, lmao).
10 hours ago, Nacaw said:England should not be dominating France like that in a Euro semi final. I understand France had a man sent off, but they played 11vs11 for 56 min and still only managed a single shot. To make matters even worse, England are still dominating with "the old guard" in 2028. National teams go through so much change in 6 years. Here's England's team 6 years ago.
Yeah, that too. Without looking at what the France (presumably AI) did, it's hard to tell what specifically was going on though, whether it was tactical, match engine flaw or both (oft, it is both).
As for the real-life xG examples, I think this is still the record without a team scoring.
0 -
If one forward has basically almost double the amount of xG (league stats), this will be down to the team's and his behavior (tactics, PPMs, etc), and the patterns of play that continously emerge out of those. Basically, one of the forward will see much more of the ball in finishing situations, whereas the other will not. It is notable that the forward with lesser xG even had it boosted by 2 penaltys (whopping ~0.7xGs each), else it would look even "worse". The gulf is simply far too large to be simply down to attributes -- it's the repeat patterns of play that emerge. Plus it's not like both the forwards are leagues apparts in base attributes as well.
Re: Attributes. Debatable if the succceed, but SI have always tried to do somewhat of a simulation. Despite public perception being otherwise and fans as well as pundits arguing some players were absolute ***** and some godlike, players (and forwards) in particular on their level aren't as far apart as is perceived. (Which is also one of the reasons why on every release there's players to get barely average forwards scoring plenty-- whereas others struggle to get Messi going). It's not that Messi's that much more deadly (though he's one of the few elite players who outperform their xG every season.). It's like that he has multiple times the chances than your average forward per match (at least at Barca, his shot volume has come down significantly ever since his move to PSG). Which is party down to skill, but also the way how guys like these are continuously fed by their team mates and made the focal points of play almost every other attack (CR7's shot volume alone average throughout his career sits at like like ~6 per match, which is almost thrice as much as your average forward. It's probably no wonder that Benzema started to score more goals at Madrid when he left, obviously a huge part of Real's play consisted of feeding their starlet prior.)
Any competitive sports is about margins of edges on absolutely any level. If you compete with Messi, chances are you're already pretty damn good at what your doing, belonging to a seleect few of elite players of your generation.
On that note, even if you'd put a defender in front of the opposition goal as a lone forward and would feed the ball to him all match, eventually he's going to score some goals. In that I challenge the notion that scoring regularly does necessarily indicate that a player must be an above average forward. Any player put in front of the opposition goal and scoring some eventually only does his job in the team -- like a keeper would if he'd faced some shots or a play maker if he provided some key passes. No more, no less. What makes an above average forward is that he'd score more than you'd expect off him given his chances presented -- and/or carving out chances for himself (or others) that else may not have existed. And that over a longer period, as matches, even isolated fully league seasons, are always goingn to be influenced by large amounts of chance.3 -
On 07/03/2022 at 15:01, Weed07 said:
I've played every FM for easily 3000 hours and one of my tin foil hat theories is that the better your team is compared to AI clubs and the more you dominate, the more chances your players will miss to artificially keep scorelines from being unrealistically high and teams are more likely to score past you with less chances to make scorelines closer than they should be.
That even players pouring those absurdely many hours into this series actually believe this to be true may be the reason why there won#t ever be a truly competitive AI.
(And I've personally reported stuff that clearly hurt AI chance conversion in the past, mainly tacitcal).
Mind you, as the Gamasutra article outlines, this isn't anybody being "dumb". It's just human psychology. A lot of game devs are aware of this, which is because the Civ devs did as the article describes (artificially putting AI under a penalty) -- or why currently the X-Com devs at least on lower difficulty fudge the numbers in the players favor (so that upon a missed player shot, the chance of a hit artificially internally goes up with every miss). Even the devs of the upcoming D&D game Baldur's Gate 3 have subsequently patched an option for rigged dice into the game which the first two games never ever had. Even on those difficulties and options, there's player that are dead-pan convinced the game wouldN't be playing fairly towards them. Those are huge concerns for developers (far too huge if you ask me) -- if you've lost player trust once and alienated a player, he may be a lost costumer forever (unless hopelessly addicted).
You'll see this kind of stuff all the time with every game that has random elements and RNG attached to it (which, according to somebody, is always being fudged in disfavor of the player, even if long-term research over huge samples shows it's "just"). However, it's particularly true for "symmetrical" games, e.g. a player having the same goals, input and options as an AI. Such as FM: AI mangers aim for the same stuff a player does (winning games), and they do the same things a player does to increas the chance: (Hopefully) developing a sound squad, making tactical adjustments, engaging in sound man management, etc. The logics are simple: 1) An AI will never be as good/creative with those same input as a player (totally true for any game out there). 2) Therefore, in order for that AI to stay somewhat competitive, devs have to rig their games (untrue, as you could as well code it so that player creativity is limited. SI have subsequently already done so in the past in particular regarding their tactical UI BTW which in very early years in the sliders had a lot more stuff to "fiddle" with -- a subsequent and then much discussed streamlining that also made it "easier" to code better tactical AI decision in later games).. There's a third: 3) There's multiple games that DO rig AI (rubber banding in various racing games, etc.) -- so devs not only can do this, but they have done so in the past, so it may well be the case for game X also.I don't know how many threads these boards had during the years (surely not every release, but this remains a classic), but even at a basic glance and results it's always the players who dominate the game left right front and center. On every release somebody breaking all-time historical league records for goals scored in succession all the same.
This goes for FM2022 also, which I'm not playing (big parts due to my almost zero interest in football these days). As you can probably guess, this topic has been one of my bug-bears for very long. If the AI is better at something, it's always been the case that the AI has beeen coded to be better at something -- as on occasion, SI are taking such AI criticism on board.0 -
Up to debate what 3d itself has added. However Being able to "read" the match play either 2d/3d (really doesn't matter), even at a basic level, is one of the the biggest advantages you could have over AI managers. And still have. AI tactics may or may not improve (as SI evidently try to struck a balance in whom they want to challenge) -- however they won't ever be able to "read" the second by second play. That second to second play is not an interpretation". That's what's "going on", possible bugs to report included.
And it's not that hard. Never been. A couple dots kicking and running around based on instructions.... spotting whether say full backs are too adventurous and being constantly caught ouf of position due to either instructions, roles/duties, PPMs (or a combo) and stuff like that -- if anybody has base struggles with that, they'd have problems "analyzing" actual football matches in real life.
'0 -
On 20/02/2022 at 03:33, _mxrky said:
2.3 xg (0.7 of which was a pen) vs 1.7 is hardly the fm games that people complain about. You thought you did something though
NOt sure if that's what he meant, that isolated game. Spurs did this to Pep a couple seasons in a row now -- with a couple more interesting stats to boot (IIRC there was even a Cup tie in between two seasons ago, not sure). And I'd haphazard the guess that Pep is not that kind of manager who gawks at a couple of numbers and concludes: "Everything must have been going fine".
Manchester City 2 - 3 Tottenham (February 19 2022) | EPL | 2021/2022 | xG | Understat.com
Tottenham 1 - 0 Manchester City (August 15 2021) | EPL | 2021/2022 | xG | Understat.com
Tottenham 2 - 0 Manchester City (November 21 2020) | EPL | 2020/2021 | xG | Understat.com
Tottenham 2 - 0 Manchester City (February 02 2020) | EPL | 2019/2020 | xG | Understat.com
Manchester City 2 - 2 Tottenham (August 17 2019) | EPL | 2019/2020 | xG | Understat.comUnsure about the current patch iteration, not played much since Nov (and then mostly Pathfinder WOTR). But in actual football, there is copious evidence beyond merely xG stats that good and bad fortunes only ever even out in the long run, which a single season of 30-40 matches is anything but. Analysis sites thus on occasion run articles how football decisions are often this short-sighted and more similar to somebody "investing" in the stock market, and then selling immediately by the first crash, whereas pro advice suggests the following.
In football this is naturally aided by public and media perception putting football's decision makers under additional pressure to finally do something. Plublic perception has it that by and large teams would end up in the table more or less exactly where they "deserved". That is, results would be an accurate reflection of performance and the table in particular by the end of each season would be a just ranking of teams (which bookies being delivered far better rankings by companies such as SmartOdds delightfully exploit). Even by simple and flawed purely xG data -- teams by the end of a season oft under and overperform that by a dozen points or more plus. I have a feeling that Brentford (run by Benham of SmartOdds) extended Frank's contract more recently despite the exceptionally poor run of results, as they realize all of this. (Brentford are currently the biggest xPoints underpeformer in the EPL, as Brighton were last season. Would be interesting what their own infamous and oft refered to "Table Of Justice" looks like. Given Frank's extended contract, it appears somewhat similar).
One of the questions for any dev of football simulations is whether they at all want to have this in the game, as obviously it's frustrating players as much as all those guys on X-Com raging about how they could miss 95% chances to hit their enemies in a row (which actually, has a huge chance of happening in succesion given the many attacks in an X-Com setting and the thousands playing the game). IIRC the X-Com guys at least at lower difficulties actually internally fiddle with those numbers so that this is less likely to happen, e.g upon a miss, the chance internally goes up, even though the 95% is still displayed.0 -
If (BIG IF) the AI actually consistently long-term does anything better, it's down to their overall tactics -- for which SI get numerous feedback each year, and sometimes even listen (even though they realize it will spring more complaints from players).
"Time wasting" in game to best effect has never been a single button solution (for better or worse, depending on whom you ask).
If you want to time waste by recycling the ball, for instance, where is that easier done? Way up in the final third or on deeper areas, where the ball can be recycled much easier, provided players don't run forward from their position and stay deep there? How is that achieved? What does the AI actually do? Can you copy it?1 -
FM12 too I played for quite a while, but I couldn't possibly go back to it (as there's little sense of achievement, in particular if you're aware of its flaws).
One of which being AI due to a selection bug frequently never shielding their backline and/or somebody in midfield protecting from counters/easy advances into their half. That's like playing a supposedly football management sim when your opponent isn't Klopp et. all, but somebody who hasn't ever watched a match of football and somehow still got the management job.
I reckon though that's actually a part of why it's still some popular.
Not that I'd go back to that anyway, currently happily playing the new Pathfinder game -- which unlike most mainstream computer RPGs these days delightfully made me wonder for hours in its character creation already what kind of character I should pick (there's like twenty classes, each with sub classes, the various races, traits, etc. to boot). These are fairly niche games nowadays on the RPG front, whereas the mainstream prefers open-world action/shooter games with barely characters systems such as the new Fallouts -- or The Witcher, a game full of barely interactive movie-like cutscenes witcher-sensing itself if you ever tackle a quest (hit a button and magically all the steps required to complete a quest appear in glowing red).
I mean, preferences, and that's all mighty fine in my book. But there might be a connection here.
0 -
1 hour ago, Ellis_D said:
I'm not talking about good runs or bad runs though. I've seen it a million times before where a team struggles to win, or can't stop winning, and then their form completely changes.
I was talking specifically about the fact I had started the season with 1 win, 7 draws and 0 defeats. Drawing 7 games out of 8 at the start of a season is just something I have never seen before.
Ah ok.
You're right, it's curious. But consider it like that: Given that the most common win margins in-game too are either by a point, er goal or two (in other words: they could oft easily be draws), I guess similar may be bound to happen to various people at some point.
PS: Given your assassment that all of these games could have been wins, your following run doesn't "Surprise" thus. If it indeed was "bad luck", that doesn't seem to repeat endlessly in-game as well (which is how it probably better should be).
PS: Bookies in real football too place much emphasis on a team's most recent WLD run -- if that would provide bettors any edge, they wouldn't do it. Unlike the general public they realize that any such run, in particular in such a low scoring sports of slight margins, is oft more random than it is perceived to be. In-game, they would have encouraged bettors to bet on that your team also draws the next game. Whereas they personally would look at more advanced data and probably realize that your team was just "unlucky" to not win more of those matches.
0 -
Not as a start, but I remember an old save where for weeks going into the new year, we simply couldn't find a win despite matches being fairly even. As a football simulation, this should be actually fairly common place.
A more prominent example would be Juventus start of the 2015 season. I say more prominent because it's a) both a big club b) from a big league. And because c) the reast of the season actually was quite different.
Juventus started out the season with just two or three wins out of their first ~10 opening ten games. They sat mid-table until well into October. Media attributed this to the club to be "struggling", whereas metrics such as Expected Goals and Points hinted at that there might have been something else going on: They were dropping points every other week despite having the better set of chances (e.g. dropping points in matches they should have taken from something).
Luckily for the Juve manager, the club's board didn't do the same as the media. Because from that point on, they barely lost a match, the "spell" was lifted and they comfortably won the league being almost ten points clear by the end.
That's actually not that uncommon in isolated football seasons, where "good" and "back" luck will certainly never even out in any kind of way.Brighton last term have made an entire season out of it (how much that was down to a lack of "finishing" skill is naturallyl also up for debate -- in either case, their fortunes so far this term seems to have turned around rather curiously. e.g. they are currently sitting in a league positions their underlying xG data had always suggested last sterm already). Just how unlucky are Brighton? Expected goals analysis of Graham Potter's side (sportinglife.com)
0 -
11 hours ago, Piksi#10 said:
An exploit tactic if ever I saw one. Ridiculous.
IRL you'd get picked apart by Fulham and the scoreline might end up the other way around.
If this "makeshift back three" still positions the way it did -- all SI needed to code was an actual AI database which formation works actually "best" against which other one (whilst we're talking about possible AI improvements...). Because on previous releases, all the AI needed to do was fielding a three forward formation, as the wide forwards would just sit in all those giant half spaces opening up between a pairing of LB/ WB ---- CB ---- RB/WB. On that prior release, somebody reported like 7-1 cup win in the first leg, and a 8-0 AI comeback in the second (not made up, btw). This was connected to a very similar formation -- however, the AI manager in the 2nd leg chosing its "prefered attacking formation" in the db in a desperate attempt to get back in the tie (lol). By research, this happened to be a 3 central forward formation... Already in the first attack from kick off, it was like high danger.
(However, SI may not be interested in punishing their players too much). Which could still randomly frustrate them as there for sure are AIs prefereing 3 central forwards.
1 -
On 12/10/2021 at 23:18, janrzm said:
I feel like the biggest obstacle to improved AI transfers is that it may prevent a whole bunch of people being more successful than they currently are.
Personally, I want the game as hard as possible, but I feel those of us who think that way would be in a firm minority. Lots are just happy to not think too deeply about things and are equally happy winning the Champions League with Rotherham Utd after 4 seasons in charge.....If think about how easy it is the achieve some form of success in the game vs reality, its ridiculously out alignment. AI managers need to be much smarter, tactically and in their recruitment and squad management, but I doubt we'll ever see it.
The irony of course is that they then oft conclude being "so good at the game that it must artificially punish them to keep things interesting" -- precisely because of such achivements obviously outperforming any of the game's AI managers to such extents. If people were punished more frequently, they would question some of their decisions. However, as that's not a requirement, they never do, instead questioning the game. (Of course, improved AI would lead to added conspiracies all the same, as now the AI would keep more matches individually and seasons long-term closer affairs). It's kind of a damned if you do, damned if you don't.
Still. There's a new host of data analysis now introduced. Speaking of which, does it further only "spoon-feed" the human player or does the AI use it in some form all the same? I've been advocating for improved feedback for a very long-time, including xG and the like. However, that also must go in tandem with AI improvements. There was a time when those in "charge" of such decisions had actually acknowledged this, e.g. Paul Collyer himself. In very early iterations, SI didn't even do much match analysis whatosever, as it would unbalance the game world in favor of the human player. Which is/was against SI design ethos -- the human player is wholly coincidental to the game world, not the very special one centre piece of it all.
'On 13/10/2021 at 13:54, Freakiie said:People act as if SI implementing better AI would literally be altering a couple variables in the code and "BAM, AI BE GOOD!", but that's complete nonsense. There's a reason that difficulty in strategy games generally involves a combination of nerfing the player and giving the AI unfair advantages, instead off just "making a better AI".
I'm currently playing Wrath Of The Righteous, an RPG based on the Pathfinder tabletop. Whilst the game deals in difficulty options that change the composition of enemy mobs -- and make them use their abilities differently: The game indeed also has options that indeed just nerf/boost enemy stats. In parts, this is actually a necessity. The Pathfinder RPG system (based on Dungeons & Dragons 3rdish edition) in itself is anything but balanced. It allows character builds that wouldn't be able to hit a cow's arse on the one end of the spectrum. Whereas on the other end there are builds possible that are completely OP (which long-term players of the tabletop are naturally familiar with). Thus, to simplify, the hardest difficulty settings make it incredibly harder to hit enemies, etc. Whilst there is a pop-up warning if you would pick such a difficulty, many players chose these options regardless. They then complain that enemies would be "impossible" to hit -- perhaps because "hard/er" difficulties in modern games else tend to be what "don't hurt me" mode was during the Doom1 era, but who knows.
Anyway, FM itself too is aynthing but balanced. In fact, FM is hilariously unbalanced. Whilst it's claims to simulate a sports/job all about shifting the odds some: There are reports every single year how players would be sacked early on almost every other save even with half-decent sides of their level. Whereas others consistently get crap teams competing for European spots on every new save immediately. And all of this largely based on tactical picks! The reasons for this naturally are manyfold:
- A tactical UI that without much further ado (or warning) allows tactics that are absolutely crap and/or toothless (including the classic of sticking 10 burly men behind the ball and letting them boot the ball upfield to a lone forward -- wish him luck he may need it even in the game engine). Perhaps worse yet: Same as this tactical UI easily allowing contradictions / bad picks that may not even be immediately apparent, as it's such a puzzle box of design
- An AI that is limited as it cannot specifically react to any tactic pick in particular
- An engine that is not quite 100% real football yet, thus always has either flat out flaws to "exploit" or at least a bias towards certain kind of tactics
SI won't ever deal in difficulties as such, however they may look like. They also don't seem to be overly fussed, unless somebody gets a crap team winning the league immediately. However, even improving just the engine as well as tacitcal UI may one day lead to a more "level" playing field all itself -- no less as the AI may "suffer" from that tactical UI all the same.0 -
Going purely by the Poisson maths, stuff like this then should actually be all over the place (those are chances that are far removed from winning a lottery, but people still win that someplace every week)...
(25:45 mins in)
Reboot Develop 2017 - Tim Cain, Obsidian Entertainment / Building a Better RPG: 7 Mistakes to Avoid - YouTube[If those people would all gather say in an online forum and share their genuine experience of similar kinds, you'd get the idea it was even more common than it actually would be Currently again very into that kinda stuff, having just bought Pathfinder Wrath Of The Righteous, which RNG according to google research appears as "broken"and/or rigged as Baldur's Gate's 3s and Solasta's, which interestingly would make that 3 D&Dish games with a "broken" RNG in a row... BG3 aiming for mainstream appeal subsequently even introduced an option that would fudge the game's dice in the player's favor and make horrible streaks impossible to less likely to happen -- but some people still perceive that as broken.
]
0 -
Going purely by the Poisson maths then, stuff like this then should actually be all over the place (those are chances that are far removed from the desperate attempt of winning a lottery, but people given the sheer volumes and players still win that someplace every week)...
(25:00 mins in).
Reboot Develop 2017 - Tim Cain, Obsidian Entertainment / Building a Better RPG: 7 Mistakes to Avoid - YouTube[If those people would all gather say in an online forum and share their genuine experience of similar kinds, you'd get the idea it was even more common than it actually would be. Currently again very into that kinda stuff, having just bought Pathfinder Wrath Of The Righteous, which RNG according to google research appears as "broken"and/or rigged as Baldur's Gate's 3s and Solasta's, which interestingly would make that 3 D&Dish games with a "broken" RNG in a row... BG3 aiming for mainstream appeal subsequently even introduced an option that would fudge the game's dice in the player's favor and make horrible streaks impossible to less likely to happen -- but some people still perceive that as broken
]
0 -
On 21/08/2021 at 21:38, Svenc said:
Juve in 2015/2016 by match day #10 had just won 3 of their 10 opening games, drawing 4 and equally losing 4.
In doing so, they had underperformed their Expected Points by ~10 points in just those ten matches, losing and drawing matches they could have won based on xG in sequences (they actually managed to win that season, as xG had always suggested their performance had been much better than their results all along. There's a saying that the league table never lies. xG is one of those stats that's always challenged that (which is why bookies or bettors have always been supportive). LFC analyst staff also had something to say about Klopp's last season at Dortmund for sure... Still waiting for the Statsbomb season analysis of Brighton tho.
Totally missed this season Preview of Brighton of theirs. (SI gerally can take a couple cues off such pieces and how to improve the data analyst module in the future).
Brighton & Hove Albion: Season Preview 2021/22 | StatsBomb
For the record, Brighton according to xG/xPoints should have finished fifth rather than the relegation zone, showing some of the league's best xG difference all throughout and undperforming their Expected Points tally by a whopping ~20 points EPL xG Table and Scorers for the 2020/2021 season | Understat.com. In other words, time and time again they were denied points despite creating the better chances. How much this was down to simply "bad luck" and how much down to a lack of skill (SEVERAL players underperforming their xG, including their forward Maupay), will probably be seen this term, for which they have started a tad better in terms of points (but 3 games really ain't much to analyze about). As said, a single season is anything but long-term, for which such data is most useful. They seem on the lookout to add yet another forward to their roster, so...
0 -
40 minutes ago, FulchesterFred said:
Interesting. So your formation is your shape when the opponent has the ball. But it’s pretty similar when my team has the ball!
It stays that way if you'd give every player a hold position instruction/defend duty (and the player has no PPMs that "make him run forward at every opportunity", etc. Otherwise, eventually, they're going to leave their position in the formation during the attack. An attacking wing back bombing forward, a box to box central midfielder not staying in the CM position but eventually venturing forward towards the edge of the box, etc.
Otherwise, the side would just keep its (defensive) formation and advance into the opposition final third that way, with no movement in between the lines. E.g.
A 4-2-3-1 with every player told to hold position/given a defend duty. Top heavy formations in general advance the most advanced players deep into opposition territory by total default. However, both CMS as well as the wide backs stay back in position as instructed.
A 4-1-4-1. Aside of the lone forward, nobody is prone to getting anywhere near to the box.This has always been pretty basic FM101. Therefore, if you'd be intent on keeping things "tight" at some point of a match or from kick off with lesser shots, going with less top heavy formations and advancing less players has always been a good place to start. By doing the 4-1-4-1 kinda thing on a prior release from kick-off in every match, and additionally just playing keep-ball when in possession to dawdle time, I was getting this string of results -- against oppostion the AI considered underdogs, so likewise didn't pick exactly attacking tactics all the same. As so very few players would advance, most of the shots were entirelly set piece based affairs too (shots which you can influence likewise, like -- by actually not advancing much players into the opposition area under set piece tactics).
1 -
On 20/08/2021 at 12:49, themadsheep2001 said:
I still don't think users get enough quality in game feedback, but this is what Xg was introduced for, working out your actual shot quality.
xG as well as any other analysis won't matter for as long as people don't have an understanding proper how to use it, like many in here, including more recent @Bowachas. Whilst I personally had applauded various TV programs to include it -- it seems not being used proper, likely just showing the stat for individual matches and no much talk or analysis about it to boot, whilst the biggest gain is in the long-term and quality analysis is to be found elsewhere.
Juve in 2015/2016 by match day #10 had just won 3 of their 10 opening games, drawing 4 and equally losing 4.
In doing so, they had underperformed their Expected Points by ~10 points in just those ten matches, losing and drawing matches they could have won based on xG in sequences (they actually managed to win that season, as xG had always suggested their performance had been much better than their results all along. There's a saying that the league table never lies. xG is one of those stats that's always challenged that (which is why bookies or bettors have always been supportive). LFC analyst staff also had something to say about Klopp's last season at Dortmund for sure... Still waiting for the Statsbomb season analysis of Brighton tho.
And that's with respecting managers never solely managing by stats, but actual play they witness. If the game were able to replicate at least some of this, it would be a simulation of football.
Additionally, I stand by what I said about AI on the former page, despite the crappy and laughable AI conspiracies, this has to happen in accordance to Ai improvements too, otherwise it's just a one horse race to (spoon-)feed the human player's ass.1 -
People still don't get xG or stats, in particular the "long-term" part of it (which a single season is anything but). Whilst they probably did improve quite a bit, in particular defensively, it's probably also no coincidence that City were back in contention after last year's (where they had COMFORTABLY the best xG difference but still finished like 20 points behind LFC).
Why aren't Manchester City better? | StatsBomb
Still, stats alone will never tell the *fully* story.0 -
1 hour ago, MBarbaric said:
after all these years, i don't know who is to blame here. people who ask these questions or those who come to answer
Touché.
You can also already predict that there will be no uploaded save, as (mostly) usual.
The core argument actually seems similar to some of GarryHammerWHUFC10000, who under his new alias sadly seems back to being banned, at least muted, for about the 10000th time. Namely, there being ME mechanisms that ensure point drops, to keep things "interesting". That must be why Rashidi went like several hundred matches unbeaten on a prior release, basically even beating "bad luck", say in the form of 3 DFKs conceded in a single match against all odds.
It's likely the the guy is having a similar issue, namely for some reason being unable to read the match engine. The added clue is also in the random tactical switches, which will always be random unless you can read the ME. Not only for the added challenge, I'd actually wish SI would code the AI managers to be brutal, so that there would be no such supicious, butneeded reloadspoint drops left and right from the very start.Oops, I did it again.
1 -
39 minutes ago, Grifty said:
I never see the point in looking to the game this in depth. So what if you are correct and there is something deep in the code that means you HAVE to lose a particular game.
Spoiler: There isn't. 30 time reload, but you can't win - Football Manager General Discussion - Sports Interactive Community (sigames.com)
Would the game in question be uploaded, somebody would win it immediately.
(As the Match Engine always has issues, sometimes it may appear though a game is "unwinnable". For example, if you didn't realize back then that three central fowards didn't track back, you always had a troubled time defending them when the AI switched to such a formation (can be mid-match, can be from the start). They hung at the half way line and overran your defense on each interception.
Still remember AI de Boer managing Chelsea on a save way back uploaded due to an "unwinnable" Chelsea match. Due to his editor traits, he oft switched to 3 Central Forwards at HT if results didn'T go his way at that point. The 3 CFs were Hazard (who would average tons of dribbles per match anyway), Costa et all respectively. Curiously, it was easier to win that match if you didn't lead at HT but later on, as that switch oft then wouldn't come... ) Back then, the Barcelona assistant employed similar tactics from kick-off if you holidayed and let him take over. He broke all kinds of La Liga records in terms of goals scored (the AI managers mostly couldn't defend the 3 CFs very well either).
0 -
9 hours ago, MasterFolke said:
I wish sigames had folks making the knap-tactics unplayable.
Whilst I'd much rather have competent AI tactical assistants helping players: I don't mind about anything like that personally as such. Anybody can play the game whatever way they like -- plus, if they're having fun, all fine. (That is apart of the fact that there's typically always an ME/AI issue behind the wonder results, which of course is never reported as a possible bug -- or perceived of the players in question "cheating" the AI, but that#s another matter
.)
What I do mind is that:- players downloading inherently won't have a clue about any tactic's strenghts and weakness (and thus when to slightly change a few things)
- never learn to manage matches (which the AI does all the time, even though it obviously cannot "read" a match as such)
- perceive the game as inherently unfair in favor of the AI in spite of all the stats/results/evidence going their way generally
Take a lookie at this. One of these replies was to be found in the biggest download thread (considering the barely defense there appears to be at least "some" realism occasioinally, although from experience this is vastly overstated how often "it" happens:Or this, from a user playing the same way and posting how unfair the game was in favor of the AI on like the last couple pages (I think he's back to being banned).
I mean, dude's right. Several downloads certainly can win you the league with West Ham (even if you never do anything but belting continue to progress the time -- technically, that may still count as "playing a game", I guess). This is never about the overall success. It wasn't back then also, when under a thousand different alias he replied with the same thing ("I'm generally successful, so how can it be tactical?). But being unable to see when in individual matches -- or spells during individual matches -- things don't go as well as usual.
However, what would happen if the AI one day were coded to prove an added challenge, say by managing the visibly space on the FM pitches, e.g. by accessing the positional heat maps to check for whether there is space available against any one tactics? It would be finally delight for anybody yearning for an added challenge sure, but it would also prompt even more posts of the game being inherently unfair and guys getting "FM'd constantly". SI naturally know this, so may approach their AI with cautious for as long as this perception exists.wwfan fan (former mod) was right on this eons ago already too.
Where @Whufc10000 actually has a point is that the AI managers indeed have "Knowledge" insofar as they are coded to make logical decisions that can work. That's because of SI knowing their engine. But also because of them receiving criticism for AI manager decisions by players who can read the ME. Players have to figure this out all themselves. Some can, some cannot. Thus I personally still think a tactical assistant /match analyst module proper could help close the gap. Both between AI match management and human player match management in general. And in between players who "get" it and those who don't. (Sure, the tactical UI also remains a bit of a puzzle box, but that's another topic).
Which is also crucial if there's ever going to be any much further headroom in terms of long-term AI development.4 -
- Popular Post
The Blatant Human Player Bias of FM --
A concerned plea from your fellow AI Managers to SI
Dear SI,
we've been very concerned lately. You've received much feedback on how to improve us throughout the years. During the dark ages of FM, we used the then sliders, just as anybody else. Granted, to our severe disadvantage, we didn't know anything about Mentality theorems used by advanced players. However, you coded us to make a couple half-decent tactical switches, such as going 4-2-4 when being a goal down and ten minutes to go.Later on, by introducing the Tactics Creator in FM10 (which had translated the rather complicated sliders into football concepts), you gave us a few additionally legs to stand upon. As you did by introducing the style pre-sets into ca. FM20. However, not enough is not enough. We need more.
We've recently witnessed a fellow Human Manager, and it became blatant to us how much you code the game to be in the player's favor. He took West Ham, an about decent squad, googled the internet for one of the latest super tactics (which are also published on your site in spades), pushed continue, unlike us never made a single switch mid-match, holidayed in between matches, let the assistant do all the talking, never even subbed a player (!), and this was the outcome.
Top of the table 12 matches in, with a whopping 2.5 goals scored.
Not one, but two attackers with finishing 12 respectively outperforming their xG by a whopping +4 to +5 (how are we supposed to compete with that?)Despite ANY kind of match management: Throughout the 12 matches, his side needing an average of 7.69 shots to score a goal (3.41 on target), whilst us needed 8.08 (us actually 4.08).
And that despite his tactics throwing everybody and the kitchen sink forward, 90 minutes, 24/7, every single spell of possession.
Now, we understand the ME as well as us AI will always have issues. Additionally, you don't want to alienate fellow FM'ers who casually watch match Of The Day, who don't read a single thing about football tactics -- or whose idea of match management is having more shots/chances on a final match report. I'm sure you're also concerned about that one or two matches in there in which this human manager dropped the points despite a statistical domination -- after all, this always tends to be perceived as us cheating on him. In essence, you want this to be a mass market game appealing to anybody who even has the faintest interest in football. That's ok.
However, give us a bit more tools to fight back. Perhaps us even being able to recognize the huge cans of space left behind on every other attack -- and dropping players there accordingly. Maybe even improve the ME by introducing actual "physically" collision detection between players, so that crowding the areas comes unstuck, and these kinda tactics are punished -- after all, in actual football managers are being punished for far less, with better players than this alright West Ham side as well. By now, you are even spoiling the actual roles we are using on any given position to that human bloke!!!
Give us a rest.
And do something for us too.Forever yours,
your AI managers.
22
(And I've personally reported stuff that clearly hurt AI chance conversion in the past, mainly tacitcal).
You're right, it's curious. But consider it like that: Given that the most common win margins in-game too are either by a point, er goal or two (in other words: they could oft easily be draws), I guess similar may be bound to happen to various people at some point. 
*Official* Football Manager 2023 Feedback Thread
in Football Manager General Discussion
Posted · Edited by Svenc
Yes, they did.
Real Madrid 2 - 2 Valencia (August 27 2017) | La liga | 2017/2018 | xG | Understat.com
Real Madrid 1 - 1 Levante (September 09 2017) | La liga | 2017/2018 | xG | Understat.com
Real Madrid 0 - 1 Real Betis (September 20 2017) | La liga | 2017/2018 | xG | Understat.com
Which would continue until well into January, with CR7 alone not hitting a cow's arse with a banjo, scoring but 4 goals from over 100 shots (two of which penalties). At which point Barcelona were already pretty much confirmed champs.
Despite managed by someone doing a bit more than looking at final match stats to gauge how well his team played... Which the AI on this game luckily hasn't ever done either. Speaking of AI managers:
If the AI were more capable at match tactics, it would pull of these kind of matches even more. General rule of the thumb: If the AI can do something you cannot, the AI has an edge over you, end of story. Imagine it was able to read the ME play the way the actually superior players of this game can (flaws included)... dynamically adjusting match tactics to space available during matches. SI aren't going to do that, naturally, as that would lead to enraged players.