Jump to content

Svenc

Members+
  • Posts

    5,137
  • Joined

  • Last visited

  • Days Won

    1

Everything posted by Svenc

  1. 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. ]
  2. 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 ]
  3. 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...
  4. 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).
  5. 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.
  6. 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.
  7. 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, but needed reloads point drops left and right from the very start. Oops, I did it again.
  8. 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). https://imgur.com/ED734b8
  9. 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.
  10. 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.
  11. One thing I'd actually be interested in is the Expected Point Table by the end of the seasons compared to the actual point table. (Constantly happening and such). The Expected Point table lets one compare actual points taken to what should have been statistically taken going by the xG in matches. In a sense, it's meant to be a measurement how many points were won/lost by luck/bad luck. (xG is supposed to be a measure of chance quality btw --- the CCCs make everything equal, which they have always been far from, which is one reason they've never really worked.) The supposedly biased AI on the occasion underperforms this by ~15 points+ -- I've seen it getting one of the Istanbul clubs relegated that way first season. [That aside, I'd have to look at the positioning (not sure about player PPMS and individual player instructions) but in tendency the above tactic similar to the kind of stuff this an attacking PSG AI played against this Ajaccio player successfully parking the bus against it on FM 2017 (the only thing missing was a winner goal) . Let's Play Football Manager 2017 #18 - Paris Saint-Germain zu Gast! [AC Ajaccio / Deutsch] - YouTube That's not the most recent edition, but in this match, the AI actually created very few quality, as it makes basically all the attacking players push towards the same space with two inside forwardish players cutting inside to occupy a simliar space as the AMC -- and both central midfielders pushing upfield also into the samish kinda space. (That said, 2.05 xG from 23 shots looks about alright, mind, as for the individual match, but then it's impossible to read what was going on from stats -- there is an AI manager on the pitch too.). 4-2-3-1 is a top heavy formation as is, I'd personally actually frequently employed two holding/staying deeper roles in both centre mids positions (not a must). However, this oldie video was meant to illustrate what then happens if the ball actually goes back to them -- every time the two Barca centre mids staying real deep get the ball (Busquets and Mascherano respectively), the opposition is actually forced to push up, opening space between its lines. Barcelona 4-0 Cordoba - Match Highlights - YouTube This is the kind of stuff if you're able to "read" it, where you can gain the biggest edge over the AI, as obviously SI won't be able to code an AI that can "read" matches on a second to second level, as else they wouldn't be making video games but be in another field altogether. PS: I've been barely following the tactical community much these days, as I found it provides too much an edge over AI. To me football (management) is about manakers roughly seeing eye to eye as they have the same/similar education. However, links to more recent threads analyzing matches and match decisions were made by players may also prove useful.
  12. That's been one of Garry's problems. If he's using one of knaps, whilst knaps likewise throws "this is my tactic 24/7, end of" at the engine and tests it, I think knap at least provides a couple of variants, eg. SUS tactics (Shut Up Shop) and similar. However, the AI managers likewise, are managing matches dynamically. It can't read a match, but it's been given match context to react to in more or less logcial ways (changing things up at HT if results may not go to plan, going for broke if there's but a few minutes left and a goal down, etc.) But in general, this is about the match dynamicism, and back then also being able to spot when things don't work as well as they usually do. Even knap's may have an AI nemesis somewhere, as they did on FM17 where most his tactics too likewise exploited the glaring ME weakness that central spaces were bust wide open (wide midfielders barely coming inside to defend centrally, huge half spaces in between wide defenders and central ones). The key is that the AI only defended this some by chance (AI managers are edited to have different prefered formations and traits, some of which HAPPEN to defend central spaces a bit better). If it would be intelligent, it would have tried such every game. (Second half of this post of mine): Of course, this exploiting too is never seen as glaring "cheating" or reported as such, this always goes one way. There's only ever been one bias, which is in favor of the players, in particular with feedback becoming ever more "spoon-feeding".
  13. Additionally: [quote]My proof is from the horses mouth and that is that SI's claim that there is no AI bias is just not true and that is because the AI have all the information in the form of code to know how to react[/quote] This is kinda true. It's called coding AI logics and/or receiving feedback on it an improving on it (that is to a degree -- I don't think SI will ever code the AI to exclusively challenge their tactical hardcore as that may mean a drop in sales and/or a large increase in frustration). Case in point the popular FM12/13. FM12 had an unresolved AI tactics bug that meant the AI frequently did not employ somebody protecting their backlines. IIRC, this actually had continued to early FM13 betas, despite being fairly basic stuff. When fixed, the main coder/s already anticipated that some would find the game harder. Coupled with that up to FM12 it was a feasible (and popular) schtick of human managers to keep 6-7 outfield players always behind the ball (defend duty or "runs from deep" set to never) and launch long balls to isolated forwards, you had a lot of rage quits on early FM13. This was feasible because up to that point forwards would visibly off the ball run through their markers to pick up long/through balls as if they didn't exist. (The concerned and cynical guy in me is wondering whether that isn't a big part of why FM12 is still that popular -- you'd have to do pretty lousy things to actually underperform FM12's AI accross the board. Ditto actually FM17, where players simply needed to flood central spaces due to wide midfielders always sticking out wide, GG). Any release typically has issues. There's a difference between that and perceiving the game as inherently unfair / biased towards any AI though. SI may never aim to challenge their tactical "hardcore". But it's of importance to challenge such perceptions. Think of it like this: If you yourself can generally outperform the AI this much despite all (and still perceive the game of being biased towards the AI) -- where does that leave anybody else, e.g. me? My Stoke example was hardly any tactical masterclass, but team sports basics 101. (I don't consider myself a particularly tactical guru, btw.) If teams, in any team sports, aim to stretch defenses, they typically make the pitches the defending team has to cover bigger, and vice versa. The AI back then suscribed to such logics as that a) is fairly basic stuff and b) as such it was fed that logics to do such into its tactical decision making. If you can't think in such terms (nor read what would be happening in any particular match against any particular AI formation/approach), then you will never get to grips with the game's ME. It's not about trial&erroring through a variety of wide, narrow, blue, purple tactics on every release. It is about watching what's going on and reacting accordingly to it.
  14. It's not. I'm tied to NDA which I won't break, but I was briefly a part of the internal Beta, which is where it was communicated how the AI operated back then pretty openly (and it was pretty apparenty to see anyways, in particular for vintage editions IIRC the public tactic forums even had threads on learning how to read what the AI was employing in terms of roles, duty, even mentality and instructions at any one period of a match). Additionally, all the feedback given back then if there had been any bias would have been for nought, there was also quality analysis in terms of AI vs AI matches, also in terms of engine bugs, the likes of which you rarely see in public bug posts. (I later on criticized the AI tactics in public posts such as this and a couple others, which evidently cost it points). As far as I remember, you had an invitation back then also, but refused to take it. Had you taken it, you would have additionally noticed how wrong you were on how the game worked (I'm not disputing it's quality, just how it works in general). Additionally, it was you who got the poor results from wwfan's then tactics set, which you also said. As always, you speak for everybody ("FM'ers"), but ignore counter evidence that doesn't suit your argument (Perception bias - Catalog of Bias), such as back then many being able to win matches despite having far fewer shots on your perennial posted final match screens. That is because you can't read the ME for some reason and never could, and as such neither time nor pick your decisions accordingly. I actually think that's the best bet you have: Outline and criticize why you can't read a thing, and rely on stats alone this much (at least there's starting to be a few half decent ones in xG and Expected Points now now even though they may need tuning), and one day you may be getting somewhere. FOr the record, I don't think you're alone in this (does anybody remember helping a guy named looping)? Not sure what your aim is, but by now it appears it's a mental protection of kind of yours. Whereas in formative years you at least wondered what you may be contributing to your experience (vastly outperforming the AI in general, btw, as recorded in any of your season screenshots), it now seems you're not prepared to be challenged anymore. Doing so would be equal to painfully admitting to yourself that despite all these many hours poured in, you didn't fully get it all along. (Which you don't, 100% btw).
  15. Yep, and in retrospect, he's been given 100%, make that 10000% correct replies like 10 years ago already in like all the big threads he created (in particular, but not limited to, by wwfan, who back then actually also had an influence over AI tactical decision making by providing the basis of the then tactical creator). He would have it much easier nowadays to explain as the game allows ME highlight uploads, has much more analysis tools and stats (e.g. xG, Expected Point Tables et all). He, naturally too, bettered the AI on all fronts (which was much simpler then anyways). Not a Rant, But.... - Football Manager General Discussion - Sports Interactive Community (sigames.com) Lets Assume It's "MY" Fault (time to ask for help?) - Football Manager General Discussion - Sports Interactive Community (sigames.com) (Funnily, wwfan even mentioned Garry's tendency to go all narrow in a bid to better possession, which he also still did in the FM 2015 game against a Stoke AI packing it and sitting deep I posted earlier. Same as his reliance on set pieces seen in the vid, that is back then when the threads were created actually buggy set piece exploit goals to score the crucial opener against an AI manager on the defense -- which he never perceived as "cheating" of course, unlike anything the AI did to him. That was FM15, but I don't think it's a coincidence. Garry's been always a stats man (which is a big part why he well never "get" the game nor its ME), and whilst he doesn't think as highly about possession anymore (good for him Football is a numbers game but only Gary Neville seems to realise it | Football | The Guardian), the quote under "Biography" in my profile suits him perfectly, even if it wasn't from him). If the AI is actually better at x, it's because it's better at doing that, and the reason has always been on him and always always will be. Simple as that. The AI has blatantly always had the same tools -- and there's been oft plenty enough to criticize about how it uses it as it naturally would affect its long-term performance (choice of roles, duties, the combination of both in particular formations, the timing when it goes for a more attacking tactic when trailing an opposition lead, substitutions etc. etc.) Back on those old releases you actually had AI for entire matches playing narrow formations but never advancing a single wide player of the formation and similar stuff thus never stretching any defense much -- the "AI bias" theories have only ever distracted from such actually game issues of there's typically plenty enough. tldr; Garry has always been the perfect example of that there's a big difference between winning (or losing) and understanding why you do so.
  16. Recently I've been thinking again that FM, for better and worse, is a game where you can actually have success without understanding why. Which is where a lot of frustrations come from. This is not meant to directly adress anything on the last page/s, but how the game can be perceived in general. Addressing that will be vital -- I personally think the introduction of xG (and xG plots!) for instance is a decent step, in particular if further tweaked by SI and used and understood proper by players (it's most useful as a long-term tool, as, to quote this article "In soccer, the rate at which shots are scored is massively variable. A club might win 1-0 on a perfect long-distance strike one week and then lose 2-1 the next while firing a dozen shots from good positions right into the keeper's chest. ") Borussia Dortmund's crisis isn't a crisis at all, and stats prove they will rebound (espn.com) I still think needs to communicate that an xG difference between teams in isolated matches of but 1-1.5 is hardly a guarantee for a win; and that introducing a stat such as xG per shot, probably even a league table of xG per shot, would also be benefitial. E.g. Germany at the World Cup 2018 with their gazillion supposedly "chances", most shots of any team in the group stages by far, but none of them actually of the highest calibre. Articles such as these also hint at where analysis in general could improve. The Anatomy of Germany’s World Cup Disaster - The Ringer However, stats alone will never be enough, which needs to be communicated too. Nobody is using exclusively stats in football, naturally. Case in point FM 2017 and a large portion of the FM tactical download community, as both are also very popular. I'm picking FM17 as that is a more obvious example likely to be understood by most, not because I like to pick on it again. As wide midfielders positioned the way they did (as in the screenshot: mostly sticking out wide and isolating their central peers), eventually by either trial&error or else a large portion of the tactics on offer honed in on the fact that the central spaces where pretty wide open. Most tactics thus swarmed the middle of the pitch (keeping it narrow and/or having a lot of players in the CM/DM/AMC positions, where they could have "training matches" against the oft undermanned central midfield of the opposition, such as this 4-4-2 with its 2 CMs covering it all alone.) Naturally, as human player wins and goals are human player wins and goals, this is never to rarely reported as a bug or ME defensive weakness (or unfairness against the AI, for that matter :D), but that's the human mind for you. Success tends to be attributed to own skill -- failure to something else. Generally as usual, the success of the tactics was not in doubt. However as with any tactic there's a caveat. For as long as SI don't code every AI to play exactly the same way with the same formations and players (and making the same in-match decisions), ditto the data research not entering the same values into the editor for any AI manager; eventually, tactics may perform differently against different resulting AI. This type of tactic back then came unstuck in particular if an AI manager sat deep, packed their backline with a back five and fielded a couple DMs/CMs to shield that backline. (If the AI could actually read the match play, it would have done this every match). This naturally plugged all the "holes" exploited so thoroughly. For every centre back pulled out of the backline so easily against most AI approaches, opening the central direct route to goal bust wide open, there was another spare man to cover that hole. The end result, in particular if the crucial opening goal didn't come and the AI was forced to go more attacking, oft was not very pretty (loads of long shots and/or attempts inside the box almost exclusively from the set piece). Few, if any providers offered a plan B for such scenarios. What I said earlier about the difference between winning and understanding why (or being somewhat able to recognize if things aren't working as well as they usually do). Another more obvious example from recent edition was the "Three Fishermen" from back when if a team (either humans, who typically systemically exploited this and thought they were clever rather than unfair), or AI, fielded three central forwards. And upon defending wouldn't track back, so would immediately face a 3-2 advantage against most formations upon every interception of the many there are in every match of FM. I still remember a couple of threads where players conceded AI comebacks sometimes in the space of but a few minutes (3 central forward formations are prefered attacking formations edited into the db for quite a few AI managers). Again, I'm not concentrating on these editions to have another criticism at them (all of this was criticized enough -- even without anyone getting banned ). I'm concentrating on them because they offer the more obvious examples of that stats alone aren't enough, same as there may be a difference between winning and understanding why -- and in general, at least optionally better analysis may be vital, maybe by an optionally tactical analyst/assistant proper. Actually somewhat understanding results is the key imo in both getting some enjoyment out of the game either way, or else things would feel quite "random" (which actual football, naturally quite often is).
  17. Curious whether there will be "proof" -- I mean, it could point to actually game issues too, so why not post it? It's not like the game never had any issues... I'd personally have never much played the likes of FM17 despite its popularity (wide midfielders barely helping to defend central areas, and as such AI vs AI matches where 3rd division opposition dominated the park against 1st division opposition and pegged them back most of the match simply by a choice in formation). And was criticizing early FM15 (advanced players given an attack duty not tracking back to defend, and as such some teams, including AI, barely being able to defend with hockey scorelines every week) and a couple others as well until they were patched. The AI is completely unaware of such dynamics, leading to such freak matches, as unlike a player, it cannot actually "read" the engine play. AI bias is big fat lie -- if there's a bias, it's all for the human player. Including all the analysing tools SI by now provide which in formative years they never did, as it would give players another edge over the AI if used proper. Garry, if this is you: The last time I personally saw you uploading anything, that is outside of the perennial final match stats of hand picked matches (shots, possession, and the ill-fated CCCs...) which you always mass post to "prove" something, it was actually a couple of pkms. Don't think you ever do that anymore as "smart asses" or similar would rip it apart. That was an old release, thus. I'll be putting this in a spoiler tag as it goes a bit beyond "FM 21" feedback. Hopefully it still works. This is a typical quote of yours: [quote]The amount of "lucky strikes" and "fluke goals" the AI score(from a lot less chances than i'm having) is the first suggestion of Bias. [/quote] [spoiler] I saw the set of pkms rather late back then, but whilst there were some decent guys replying -- they somehow missed a couple things. Firstly, you claimed that you never scored off "lucky" goals. Now, the definition of a lucky goal may be subjective, but to me your West Ham goal in the following vid certainly applies. Secondly, for all the shots recorded inside the box, there was absolutely no space created whatsoever. I went into the analysis and basically recorded every single shot in the video. Almost all of them were exclusively from the set piece (little space by definition). West Ham 2-1 Stoke - Match Highlights - YouTube Now, there may not a solution to everything (game issues and all). However, this was mainly tactical. The combinations of roles/duties, instructions as well as attacking width barely stretched Stoke. If you pause at 1:06, which is where the easy interception starts leading to the Stoke counter and subsequent goal -- if you'd draw a circle around all your players having advanced, that's a really small area a very deep sitting Stoke AI has to cover. Keeping the pitch small allows you to keep possession easier (a stat you've put much emphasis on at least in formative years). However, as a consequence, Stoke easily got a foot into almost every open play move -- no matter who receives the ball next, he's immediately pushed as the Stoke defenders have no distance to cover. IIRC correctly, you perceived this and similar of "AI cheating" just the same, as the AI would have it much easier to trawl through your defensive without a tackle or any defensive action in sight. It's true that the AI had it easier to go unchallenged. However, this was all of your own making. The AI back then (generally) when using an attacking tactic and pushing for a goal used the width and depth of the pitches, and as such made defenders cover a much larger distance to close down the next ball carrier. Granted, I'm not sure you actually WANT help. If your intent is trying to prove an AI bias which doesn't exist (neither back then nor now, players have ALL the edges over it, and that's not defending the game), so be it. [/SPOILER]
  18. Art imitating life. Top 10 weirdest football excuses (besoccer.com) It's also much easier to blame a game mechanic than yourself (and oft times, it can actually be a combination of both).. Ironically, many suspect the AI of "cheating" as they generally overperform with their clubs this much. This isn't actually seen for what it is (that is, outperforming the same AI they accuse of cheating, which would never perform the same way with the same clubs). But a mechanism that supposedly kicks in so that things stay interesting rather than boring. The logic is: I am too good for this game, so the game must punish me (and everybody else doing so well) on the occasion to keep me interested. Unfortunately, that completely ignores players who don't have the same issues this regularly. Online, it's like an echo chamber of perception bias. Somebody suspects this mechanism to be "true", and everybody looking for confirmation of their beliefs chimes in, so things repeat at naseum. There's naturally also a more practical explanation. In a (typically) successful FM save, over the course of it, AI opponents oft tend to be the only managers who actually would play defensive, shut-up-shop football. The reputation of the successful human team rises, and with it, how AI managers perceive it. More and more see it as the match favorite, and tactically adapt "accordingly". As the AI for many is the only manager ever employing park the bus tactics for prolonged periods of a match (content with a draw or just not getting trashed), it's the only manager who ever (and if by luck or even a bug) either scores from few shots, the very first of such few shots, or even wins matches with few shots . That is, if they score at all. Looking through knap's threads, despite his rather attacking tactics (and oft mediocre teams picked by the downloaders), that doesn't appear to happen that wildly often. This is then seen as "cheating", (the term getting "FM'd), as many aren't able to replicate the same -- however, If you'd try yourself to park the bus against such an AI parking the bus, well, I had a couple interesting sequences on a prior release. (If Whufc1000 is indeed Garry, and I suspected as much due to many many many reasons, he's got history on these boards and many FM boards, btw. Just google for hammer1000, garrywhufc or Gazhammer+Football Manager. During his heydays he was this infamous, he even had his own FM myth attached to his namesake (admittedly invented by me). At one point it seemed he was on a much better track as far as I remember. Rather than accusing the ME/AI of a bias/cheat, which simply has never been there and actually decent players have proven that every release, he's questioned the game's feedback (as well as, I think, sometimes even actual ME flaws and bugs). Indeed, to me, improved game feedback is the key of getting rid of such theories -- or at least lessen them. However, this has to be a careful balancing act, as some may be alienated if the game just spoon fed everything and gave it all away. (I've been arguing for optionally tactical assistants proper for a very long time which would go through the same "thought" process as any AI manager would both pre-match as well as in-match and actually explain it, and many more).
  19. You can already tell he's exaggerating a few things, as well as second-guess the conversions of both AI as well as his team are nowhere near his perception from looking at all the screenshots he's posted in knap's thread in like the last couple of months. Shots showing actual seasonal results rather than hand-picked matches (speaking of which, Guardiola likewise always concedes or drops the points in tendency when his opposition has much fewer shots and, naturally, lower xG, given that his sides sometimes concedes like 6 shots seasonal average -- and he's probably adapting a tad more). Speaking of which, if knap's wouldN't outconvert any AI left right front and centre even with just alright players on their level (which naturally is never worth reporting for a game issue as wins are wins), they wouldn't be as popular. His set piece settings for instance may naturally very well trigger marking bugs the AIs' don't, which is all very well possible and worth a report. For a good chunk, and this is long-term experience, this will be another case of wonderful perception bias (own goals are never checked for whether they may be somewhat buggy or "lucky" -- and some of them inevitably will be given the sheer volumes), as well as possible AI tactical issues I was hinting at before. E.g. if the AI were playing too defensively for a bit too long periods of a match (and that argument was brought up for FM21 by at least some, linking it to a perceived lack of difficulty as the AI would try too late to get back into matches etc.), it means they won't regularly construct moves from open play, and as such get more of their (few/er) shots throughout the season from set pieces, flukes, errors, and the like than "realistically" they should have. Some of those will still inevitably during a season lead to goals. Some of them will then cost points and matches. Back then with the AI at their most rigidly defensive, I was starting fun saves with a top team considered match favorite every week from the start anyway (so the AI expected to be beaten or at best draw the match and chosed tactics "accordingly"), trying how many match days I could go without conceding a single goal. However, by around match day 10, the AI was awarded another corner (which was the only time they put volumes in the box), or whatever, from which eventually a goal was scored. That was ca. FM 2015ish, btw. Even AI managed top sides, e.g. Bayern Munich in the Bundesliga, could go with conceding less than 10 goals throughout the season with a bit of luck, as too, they likewise faced the same rigidly defensive AI opponents every week (admittedly counter attacks and the like were also pretty tame on that release).
  20. The AI has never operated under different rules. It's always had the same options available as the human player, visibly, except that the human player can use those options far more creatively. That's an excuse for own failings (or at least a testament to a lack of proper official game documentation / pitfall tactical UI). And that's not defending the game but stating facts in the hopes of ever improving at least optional AI that may also provide a challange for players who delve a little deeper than the next download. It may appear that the AI had an "edge" sometimes, sometimes even due to bugs, but that's different. Whilst the AI back then was much more immediately apparent to outperform (seasonal ratio of Shots on Target per goal as recorded clearly being much inferior to a decent player), as it used the Tactics Creator in really simple ways: This thread is still a classic on the perception at play, btw. Why is the AI so much better at finishing? - Football Manager General Discussion - Sports Interactive Community (sigames.com) (Heavy spoiler: Despite some severe tactical issues that were adressed by users, the OP's conversion was actually much superior to the AI right from the start). I don't know how many weird tactical picks or nonsensical changes at HT (or during a match) I have witnessed or reported in the public bug forums throughout the years (also in AI vs AI matches), clearly costing the AI points and/or leading to freaky results. It is only when they were adressed and acknowledged by SI staff that the AI became more robust. Additionally, until a couple iterations ago, the AI became this defensive this readily, IF IT SCORED AT ALL, it was always either from a fluke, set piece, or even a bug, as AI manager tactics rarely pushed enough men forward for long enough to construct decent moves from play during a match (technically lower mentalities plus like half a team plus including both wide backs on "defend" hold position duties for like 90 minutes even when the AI was managing Portugal in an Euro semis). User types similar to Whufc1000 perceived this as a "cheating AI" all the same as the AI always exclusively scored from flukes, if at all, never any decent play. The more common "complaint" is naturally that it would be always the AI that would score off few/er shots. If you can't do that likewise, the burden is on you. And always will be. (Luckily the game has introduced a bit better (statistical) feedback by now, even though it may need furhter tuningn throughout the years). An introduction and a question about (contextual) feedback and stats in the game - Tactics, Training & Strategies Discussion - Sports Interactive Community (sigames.com) I reckon anybody involved with reporting AI glitches could attest to similar stories. And players like Rashidi et all, if you would face them online, would do things to you far more nasty than any AI manager, as actually good players don't merely make somewhat "logical" decisions based on match situations and scorelines and make switches in-match accordingly (as an AI manager does). Not merely for reporting bugs, they can actually read the second by second match play and draw conclusions, which the AI will never be able to.
  21. The perceived quality was not what I was addressing. I don't care whether somebody thinks this would be the worst game; or the best thing ever. (If you asked me, it was neither, and I skipped quite a few iterations, but that's besides the point). What "triggers" me was the claim the AI would have the ability to "score whenever it wants"* and that the game had a coded bias towards AI. It doesn't and it never had. Why do I care? Simple: If the AI one day finally were actually coded to challenge us players who have sussed it all out in spite of the many myths and cheap player excuses going around, the accusations the game wouldn't be playing fair would only increase in numbers, as the AI would become ever more efficient and players increasingly alienated. And key SI staff naturally know this from years of experience and developing their AI, so this isn't quite a conspiracy. From my end, it's worth battling this AI myth for this reason alone. I'm glad this stuff is usually deleted at least here, thus (I consider this site to be one of the few with reliable game advice, for which SI's lack of documentation is also to blame, mind). I reckon it's because most mods likewise know how the game works (and quite a few of them outperform AI likewise). Some may also be part of beta testing, and know that any AI bias would render feedback, such as on AI tactical picks, useless. The insistance on such popular game myths also distracts from adressing actual game issues all the same, of which there may be well enough already. *And that from somebody simply downloading a tactic and then performing in a way no AI manager with West Ham would ever even in a hundred new saves (and by sheer lucky rolls of the internal "match dice" in sequence), which by any sane logic would be ample evidence of the contrary alone.
  22. Not really. Which is why this makes me despair a bit. Player who downloads a knap-a-ttactic (which like any tactic will always have a flaw unless he finds a way to field like 30 outfield players, second half of this post), scores 3 goals average per match with an alright squad and still claims the AI were rigging the game. Not that this would be the sole example. Reddit, et all, are full of posts like that. I don't have any issue with any playing style and I don't consider any playing style to be superior over any other, mind. Still, the question stands, where would that leave anyone who's engaged a bit deeper into that part of the game and still wants a bit of a challenge? The AI actually evidently dropped points because of crappy tactical picks on a couple recent releases, with a particular highlight being Guardiola's Bayern ca. FM 2016ish, regularly barely scoring more than 60 goals per season with one of the most domestically dominating squads at that time (and losing up to 8, 9 league games). Barely any competition from AI management like that. The fix to such major bugbears wasn't to code the AI to be more "unfair" or make it "scoring a goal whenetever it wants" (because that's not how the game was ever coded). It was adressing AI tactical flaws. And it would be still like that to this day. Mind you, as an actual "simulation", the game shouldn't allow there to be actual crappy tactical picks. FM's UI easily allows this, including major contradictions, for both AI as well as human player as the entire tactical UI is basically a puzzle box full of pieces (which don't necessarily all fit well together), which likely makes this a nightmare on the coding front, considering that every AI manager too has edited preferences in formations, favorite roles used, etc. as well as every AI manager making dynamic switches throughout every match based on current scoreline, time left on the clock, etc. Also, in your post supposedly showing how AI Liverpool would win matches despite barely creating chances -- there are like two matches in there they won despite having the lower xG. (SI should finally get rid of the broken CCCs and half chances now that this is in-game). Speaking of which, there are xG tables as well as Expected Point tables in-game with which you can compare actual goals scored vs what should have been scored, as well as actual points taken vs. what should have been taken (statistically). The supposedly goal rigging AI on the occasion underderpforms Expected Points by as much as ~15 by the end of a season. Still, should Liverpool (and their forwards) actually outperform their xG over the seasons, that should actually be the case, as the few elite forwards in football long-term actually do score a couple more goals from the chances they have, compared to, well..... bloody West Ham's. That's why clubs are spending those gazillions on those few elite forwards. Opta Expected Goals - YouTube
  23. Top of the league with a GD of +36 after 17 matches with bloody WHUFC in 2021 -- but clearly it's the AI rigging the game when in need of a goal. (I'm sure none of those 3 goals average scored for WHUFC ever have any issue, they're all brilliant legit football goals scored by definition no doubt). This is why this game will never get a more competitive AI btw. Every time in a hundred a perceived "****" hits the fan, it's the game being unfair. Gamasutra - Analysis: Game AI & Our Cheatin’ Hearts That's not a dig, btw. It's just that the human mind in general is a thing of many wonders. It's the same psychology at work that regards the RNG of Baldur's Gate 3 (Early Access release) as borked, even when collecting numbers it shows it isn't. The devs are actually optionally coding in a bias in the form of loaded dice in favor of the player, similar to the Civ developers in the article linked to above (and even that is perceived as suspicious as the human mind is what it is). Baldur's Gate 3 rolls out a hotfix for players who have 'angered the RNG Gods' | PC Gamer
  24. xG is best used for long-term evalution (football is a super low scoring sports, with the most common win margins being by a point or two likewise. Which means actual performances and results aren't a match sometimes for weeks -- even months. The table may not "lie", but as a definite ranking of teams it's misleading. Bookies, club analysts as well as bettors have realized this long ago already). Borussia Dortmund's crisis isn't a crisis at all, and stats prove they will rebound (espn.com) As predicted, Dortmund actually managed to bounce back that very season, and still qualified for Europe being dead last in the point tables by as late as February -- however, there was no "guarantee" for such an actual comeback. It's just that xG had identified a team that was exceptionally unlucky for like 6 months running. Coincidentally or not, the LFC data and analyst staff had anaylzed much the same, with Klopp even arguing they are the reason he's at all a LIverpool manager, when most other clubs perceived him as declining. The game likewise collects such long-term data in the xG tables as well as the Expected Point tables, making it possible to compare actual goals scored vs. what should have been scored on average, as well as actual points taken vs. what should have statistically been taken.
  25. There's an "Expected Point" table where you can get a bit of a clue of how much luck may have been involved. E.g. Expected point table vs. actual point table.
×
×
  • Create New...