Jump to content

Lexis

Members+
  • Posts

    266
  • Joined

  • Last visited

  • Days Won

    3

Posts posted by Lexis

  1. 30 minutes ago, _Ben_ said:

    Not had much luck so far!

    classes/custom/attribute descriptors/good

    <class class="acceleration_descriptor" parent="position_role_duty_star_range_based_picture" image_alignment="left" >
        <list id="value_list">
    
            <record value="0" file="" />
            <record value="1" file="" />
            <record value="2" file="" />
            <record value="3" file="" />
            <record value="4" file="" />
            <record value="5" file="" />
            <record value="6" file="" />
            <record value="7" file="" />
            <record value="8" file="" />
            <record value="9" file="" />
            <record value="10" file="" />
            <record value="11" file="" />
            <record value="12" file="" />
            <record value="13" file="" />
            <record value="14" file="TEST/quick" red_replacement="white"  />
            <record value="15" file="TEST/quick" red_replacement="white"  />
            <record value="16" file="TEST/quick" red_replacement="white"  />
            <record value="17" file="TEST/quick" red_replacement="white"  />
            <record value="18" file="TEST/quick" red_replacement="white"  />
            <record value="19" file="" />
            <record value="20" file="" />
        </list>
    </class>

    You'd need to go through each attribute in the elite, good and poor area and change this so that it shows for different attribute values.

    See above!

    Hey,

     

    I think I misspoke. I meant the circle ratings, where you can see your attributes in 'half a circle' increments when inspecting your own players.

  2. There are browser and mobile games with graphics 10x that of FM. For me, in today's world, with unreal engines and what not, to stick to what they re using to support specs from 15 years ago is just mind boggling.

    Look at the above video with FM 11. Can you really say today s version showcases a 11 year evolution of the graphics engine? Can you really say you see 11 years worth of difference? I honestly honestly can't .

  3. On 13/10/2021 at 14:54, Freakiie said:

    The biggest obstacle to better AI is that... it's really damn hard to do!

    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".

    On top of that, better AI would most likely mean more processes it goes through before making a decision, which would mean a ton of extra processing because there's countless decisions made by the AI over the course of the game. Heck, I've played a couple strategy games with modded improved AI and the performance loss was absurd. In a game with way less decisions from the AI this might not be that bad when an AI's move goes from split seconds to a couple seconds, but imagine if AI decisions in FM were slowed down, it would massively increase processing time.

    There's definitely more than a few quirks in the AI that can be looked (and SI is looking at them), that could probably be adjusted without complicating things too much, but an improvement in the AI to the degree where it is even somewhat challenging to long time FM players is just completely unrealistic.

    If only there were more computing power available for the game to use......

    They are literally keeping this running on bricks, more computing is literally the last thing they are short of

  4. 9 minutes ago, XaW said:

    Both of these things are (sort of) implemented and talked about in Part 1. Nic talks about a new animation engine specifically and improved player AI.

    That is not a new graphics engine. It's a far cry from that. We keep patching the engine we got with FM10/11 or whatever. And truth be told, it's not better in all areas. 

    Better AI both inside the match and in the long term career. Don't we hear that year after year? Is that not worth a revamp already and significant improvements? I was referring more to the coaching AI inside a game.

  5. Have not seen one thing to get me fired up besides the WCB and hopefully other roles or some sort of tactical engine revamp. The data hub looks interesting so I'll give it a plus. Everything else is just a reskin. You did the same last year, however the ME was exceptional and it flied by. I can only hope ME is exceptional this year as well.

  6. 55 minutes ago, forameuss said:

    They can't physically sell a previous version when the new one comes out as far as I'm aware.  In a world where you have to actually pay to get the correct ordering of fixtures, I'd turn it around and ask what the licensing issue wouldn't be?

    Ultimately this is something that is likely far harder than even initially meets the eye (and I would see significant complexities on the surface).  Plus they already give you the means to recreate an older database.  Fire up the editor and spend the next few years doing it, and there you go.  

    There are a lot of reported limitations with the editor, including significant performance issues past a number of edits. 

    I'm not even saying give us official out of the box retro databases.

    I'm merely saying create a framework for the community to ensure the process is smoother.

    - Better editor performance

    - Ability to import previous DBs and give you a starting point at least, etc.

    - Ability to import parts of previous DBs (eg only import players, only import clubs, etc). Assist with autogenerating dependencies where required.

    Give the community power. In the end, it is serving your own product. Look at what Skyrim achieved. Look at WoW with the Classic and now BC Runs. You are doing a service to your own game.

  7. 54 minutes ago, FrazT said:

    SI will undoubtedly tell you that the resources they have are limited and those that they have are working on FM 22 to make that game better.

    I am not asking for some sort of continuous support that would increase their overhead, rather some sort of framework that is low maintenance and a one-off effort.

    Not sure what the licensing issue would be?

  8. I'm just really curious, why isn't SI providing more support in terms of retro databases?

    - Most of the recent databases could probably be ported relatively easily between versions. With some amount of work to potentially fill gaps, etc, I wouldn't imagine this would be too hard. I imagine the older DBs could potentially be ported as well with some massive JSON export or similar and then the newer editor filling in blanks.

    - Add the feature of "programming" regens coming through up to whatever point we desire, with the attributes we want, etc. This exists in a very limited way but would be simple to add it as a more fleshed out feature.

    Why are we not taking a page out of Blizzard's book? They've been adamant against classic for so long, up until the point they did it and it's been a huge success.

    If SI provides the "framework" for the community to do this, in ways that are more efficient than what is currently possible (think editor performance, etc), then that burdain can be left on the community with minimal to possibly no official support from SI.

    You can only gain new players / added retention from such a move. 

  9. 11 minutes ago, Rashidi said:

    So here's a suggestion:

    The human is always having an advantage over the AI because the human is generally more creative. Let's identify some ways.

    1. Does the AI use match preparation the way the human does? I know I use it like some super bonus cookie and will swallow the maximum number I can. So why is the AI not doing the same? Shouldn't it? Whenever the AI manages my training it does appear that it fails to do that, forcing me to add them myself. So it follows that if the BEST tactical manager i can find as assistant manager doesn't do it, then perhaps the AI is also failing to take advantage. It's a possibility.

    2. This could be more a research issue than anything else. Some AI formations are particularly imaginative, but not too many. The AI manager seems to have a few preset tactics that it seems to go over, and these are dependant on the manager attributes. Perhaps managers should have a bit more versatility in their options.  This would mean someone taking a deeper look at how the formations are chosen.  I know some of the AI teams I have played against have stunned me with their versatility. To be fair we can't really expect every team to play like that, but choosing to go attacking with 10 minutes to go after being 2 nil down for 80 mins doesn't sound right,  Perhaps the AI should have better options, like:
     

    - Human manager draws yellow cards down one flank, AI focus play to that flank, just one example. There are other options too. These tactical options could be embedded into a playlist of sorts.

    This brings me back to how we used to play FMLive. I now have a system of play that evolves how I combine various team instructions for different periods in a match, much like how we used to play FMLive.

    An example:

    Start of the game, the AI could start Balanced, short passing, narrow width, normal/quick tempo, This is done to keep possession of the ball while trying to suss things out. Then if it needs to open up, it goes to a different set of instructions, but these are done, not based on time or score, but based on things like whether its controlling midfield or losing the battles at certain areas of the pitch. 

    A lot of this depends on internal research from SI and how accurate  they want different systems to be. I know I liked Bielsa's Leeds this year, and I also know who researched that team, it was versatile and good. 

    Frankly speaking, as humans we are super creative. The very best players have one tactic with a few plans ready to go. They can do different kinds of things with their one tactic, but the AI manager plays off a template of tactics. Perhaps it's time to revisit how the AI manager adapts during the game.  At the moment the game feels too much "painting by numbers", get the right players with the right attributes and chances are you are going to win most of the time. I remember talking about this a few years ago. The AI manager needs to be more adaptable, it has to be able to spot weaknesses in your system and exploit them intentionally.

    Doing this though has its downsides, the game is meant to be a simulation of human behaviour. And this is the hardest to incorporate into the game. Take for example, Salah gets a new bumper contract...chances are he is going to have a dip in performance for his next few games before it goes back up again. This is common for footballers, but how do you code this kind of behaviour - the random human element. 

    We can make the AI play like deep blue, but we also need to be able to mimic human predictability and sometimes human stupidity. I do hope that over time this happens. It's a fine line to walk, and I for one disagree on difficulty levels. It's never happening and SI have been very clear on that. 

    Totally in agreement. That's also what I wanted to point out. That while we give the AI access to more tools, they should be prone to misusing them just as much as we are.

  10. Also, because it was brought earlier. Let's imagine a world where a supercomputer and super AI is developed for the NPC managers in FM. And they would be able to use every single feature in the game with a very high degree of efficiency. Let's also assume that the supercomputer is a master tactician and would know the insides of most tactics and how to counter them.

    Even in this scenario, that computer will likely be beaten in 10-20% simply due to the fact that, like us, the manager simply will try to maximize the chances of winning his team has, but the playing itself and the carrying out of those tactics will be done by the players.

  11. Couple of things on my mind:

    1. I don't think we'll ever see difficulty modes and they would not make sense to begin with. I think SI stated this themselves so why don't we just cross this off the list?

    2. What this thread asks for (and I imagine a lot of us) is give the AI the capability to use the exact same tools and depth of usage that a player has. This does NOT necessarly always have to increase difficulty. Given they are tools, they can also be used in a wrong way. An AI manager with bad Motivating would still suck at motivating his team through pep talks and interaction, but he has the tool at his disposal. Yes, fantastic managers will probably use a lot of these tools in an efficient and positive way more often than not which might make some of the top managers harder to beat. But you are not increasing the difficulty artificially, you are only leveling the playing field between AI and Human. Hence why I started an analysis to see what exactly can an human manager do and where the AI is lacking.

    3. Yes, there is more than enough computing power to achieve this, whoever claims otherwise has no clue what he is talking about.

    4. Handicapping yourself (by not doing X and Y) is not the answer. The AI needs to be better. If you want to handicap yourself, sure go ahead, but that is a separate thing and should be treated as such. I am not referring here to not exploiting serious game bugs, obviously that should be avoided.

  12. I've just finished my first season with Barcelona. My feedback for the beta is as follows:

    - The main features announced, as suspected, were simply a rehash. The new body gestures are simply a renaming of the old voice tones, you can even see that in their description.

    - The match engine looks solid in the Beta. Probably best version I've seen in years. I would say there's still some issues with inside forward movement and tendency to not go inside that much but still a very good ME.

    - First time I've actually seen the AI do solid transfers. After the first season, most top teams have done some very solid transfers which is encouraging for long term play.

    - I was unable to win the first season with barca anything (got beaten by one point in the league by Real and Lazio trashed everybody in the UCL) which is absolutely fantastic. I wanted the challenge to be there. I also didn't go around poking gegenpressing, I opted for a tiki taka style.

    - Unsure if gegenpressing is still OP.

    - Unsure if match condition still has no impact in pressing / etc.

    - General feeling outside the match is good usually.

    Provided the beta ME only IMPROVES from now on, we should be having ourselves a fantastic FM version. I can only imagine if they actually pushed out some features this year as opposed to rehashing. 

  13. 7 hours ago, zyfon5 said:

    1. I can ensure you squad building is not the biggest problem with the AI. I can ignore all of these things and still outperform the AI. The AI can have the best scouting and coaching team in the league and yet still underperform. A good backroom team does not win your titles, a good player does. If you cannot utilize your backroom team to find you good players, your backroom team is useless. The important thing is to teach the AI how to use their staff. The AI are decent at bringing enough cover for their players sometimes even a bit too much. In fact, the first thing I do every time I manage a new team is to clear the deadwood. The AI more often than not buys too many players.

    2. Rotation is a big issue with the AI so I can agree on that. I am not sure teaching the AI how to counter tactics is going to help them if their tactic is trash in the first place.

    3. AI can use opposition instructions from what I observe. If your assistant can do opposition instructions then I am sure the AI can do it as well. AI after FM20 tinkers quite a lot in match sometimes even better than human. Just glance through the number of posts that claims simming through matches is easier to win than playing in full detail. I am not sure if mentality is that big of a deal for big teams like Man City, Liverpool et al. They are already playing at higher mentality most of the time and yet still underperform occasionally. My title winning team is the 4th highest scorer in the league and not the most attacking team from stats analysis and despite only predicted to finish 7th managed to win the title. In fact I am playing more defensive than the AI only conceded 17 goals in the whole season.

    4. Transfer is a tricky question in my opinion. I managed a Leicester side that finished 10th last season and turned them into league Champions by playing pretty much the same squad the AI had (both seasons expected finish is 7th) and playing the same 4231 formation they used last season. The only player I brought in is a new starting striker and I sold their star winger and replaced him with a younger and worse winger. That is the only changes I made and the end result could not have been much different. The real problem is not that the AI is bringing in bad players but underutilizing players they have and I am not sure the AI has developed far enough to address this problem. 

    5. Pep talks has minimal effect. I have delegated pep talks to assistant and there is barely any noticeable difference.

    Currently there is really only one core issue with the AI and that is utilizing the players they have to maximum effect. Improving this will require improvements in a few aspects of AI management in the game. 

    1. Better tactics especially managers that plays a 'more defensive' tactic. The default tactics for catenaccio and park the bus is poorly designed. The best default tactic in the game is vertical tiki taka. AI managers that has this style tends to do well from my observation.

    2. There needs to be a transfer logic tied to the tactics used by the AI manager. Too often I see some AI manager that has route one as their playing style and do not have a target man in their team. Or the AI buys a good winger for 60M and just let him rot at the bench. If I have the same player he will be a star player in my team.

    3. Better scouting intelligence by the AI to look beyond star ability and rely more on stats, attributes, personality and playing style. Currently in my save two of the best defenders in the league is not even four stars and plays for mid table teams in the Premier league. The AI will not be able to realize this if they do not look at their stats. Meanwhile I am able to grab them for cheap myself and improve my team immensely.

    Good input. I don't know what to say about pep talks, I always go for the full greens, I never tested what happens if I go full red on the players.

    Default tactics is an interesting subject. Does the AI have a set of default tactics they adhere to and can't go beyond that tactical constraint? Eg let's say there s a coach with 4-2-3-1 preferred formation and gegenpressing style. He gets a default tactic with some player roles and he cannot go beyond those player roles / instructions?

    Given the amount of possible permutations it certainly would seem like a big limiting factor.

    I was really hoping SI would engage with us in this discussion as well to help clarify stuff.

  14. The suggestion is rather simple.

    For corners, free kicks and throw ins, let us add different routines for each tactic mentality.

    The reasoning is rather simple:

    On a 2-0 lead if I'm playing cautiously, I'd want to keep a few people at the back on set pieces to not get caught on the counter.

    If I'm chasing a 2-0 lead and playing on attacking / very attacking, I would want the opposite, to have more people load the box.

    Doing this at the moment means manually going and changing routines while in match.

    Let us pick specific routines for specific mentalitites.

    Simple and effective.

    Thanks.

  15. 2 minutes ago, mp_87 said:

    At a very quick glance my issues would be that you have three other playmakers + a ball playing CB wanting to do their own thing and be the focus for the play. Then there is aggressive wide roles and specific instructions (overlap) to go with that.

    Also if you want to funnel the play through Messi, maybe a change of position? Instead of starting high and dropping deep towards the play with his back to goal, consider he starts deeper and pushes forward with the play ahead of him....

    I had a hunch with too many playmaking roles, will certainly give a try with less.

    You are suggesting a shadow striker role, correct? 

    Thanks

  16. Hello all,

    I'm trying to recreate a Barcelona-style tactic based on possession. I'm getting success with possession play, however I can't seem to be able to get Messi into play, neither in a F9 playmaking role, neither in a more offensive-ish F9 Pep-style. Whatever I try to play with him (F9 or trequartista) centrally, he just seems to not be involved anywhere near enough and when he is, he is always facing away from goal and passing back.

    I've tried some various roles on the wide men on the tactic below, I am unsure at this point what the issue is. Possibly should change the CMs to non playmaking roles in the idea that they won't be considered playmakers and thus more balls would go to Messi?

    Thanks

    image.thumb.png.83efd91394baf5cf03ec0f8bd984ec9a.png

×
×
  • Create New...