Chriswin4 said: That's true, I don't think anyone thinks Wemby should be rated higher, but in terms of what's important to the game, he is better to have. That's why we want a GS rating to reflect that, so we can future identify good Match Engine players. Expand
You say Wembangomo shouldn’t be rated higher, but then you say you he *is* better and you want a rating system to rate him higher. So I don’t think you can mean both. In my opinion he will not perform better in the match engine, in fact he will be substantially worse. You shouldn’t interpret single attribute tests to mean that only pace and acceleration matter. If there was a test of increasing all attributes except pace and acceleration I’m confident that would outperform the increase only in pace or acceleration. And this is what is reflected in the the various ratings that are available. There are strikers faster than Haaland in the game but they don’t perform as well because they don’t have stats close to him in all the other areas. If they had 20 pace and acceleration and a current ability in the 190s and great hidden attributes then maybe they would be better, but no player with a 140-150 CA can be close to him.
NandaldiaN said: Using your rating system in my current save, there is one in particular that catches my attention and I would like to ask the people on the forum which of these 2 they would choose to play on their team.
Currently your rating puts Odriozola as a 68% rating, compared to Wembangomo with a 61% for the DR/WBR position.
It is obvious that in everything other than physical statistics, Odriozola is better, but if we stick to the tests here, wouldn't Wembangomo be much superior in the match engine? Expand
I think any reasonable ranking system would have Odriozola better (and this is the case with all 4 rankings I have set up in GS right now). He’s a legit Premier League player. The other guy is a League One player that’s a bit faster.
CBP87 said: Well that is exactly what is happening with some of the tactics, well the majority of them, it is rare to see a score go up in a 2400 match testing, the majority have a reduction in score Expand
That is what happens with top tactics because the reason they are top after 1200 is partly due to luck. The score decreases with more testing because of regression to the mean. If you looked at the worst rated tactics you would find the exact opposite happening. That’s just how the statistics of large samples work, and also why the more tests done the more accurate the result … even if you will only play one season in a real save .
You could reasonably conclude that the three or four point difference between a tactic after 1200 vs 4000 matches is not likely to be meaningful in a save that lasts only a few hundred matches. But that isn’t a reason to choose the highest rated tactic based on 1200 matches; instead you should pick any tactic near the top that fits the best players you have available.
Does anyone remember which tactic was top after 1200?
@Zippo@Droid do you have this information available please? is it possible to add the result of 1200/2400/4000 (if applicable) to the results table?
The reason I ask is because (and I don't mean any disrespect with this) but I only really take note of results from 1200 matches, the reason for this is because by the time I've played 1200 matches in game, I've probably already got an OP squad.
Im pretty sure that over 1200 matches there was 1 tactic that achieved 65/66 points but for the life of me can't remember which one it was
Thanks Expand
I think you’ve missed the point of doing more than 1200 matches, it has nothing to do with how long a save is played for. When a tactic is tested there is a lot of randomness and just by luck a tactic can rank higher or lower than it should. The more matches the closer the points ranking is to the “true” value. What matters to you is the true value because that is what most accurately represents what you experience in your save.
With the new set piece system you don’t necessarily have to have the big CBs; as long as you have a couple of players with high jumping reach and headers at some position they will get put into the right place. Regarding the original question, what I would do on console is look at the average pace and acceleration for the league youre in and then set a filter one or two points higher than that. That can be a starting point at any outfield position. Then by position take the next 2 or 3 most important attributes and add those (again 1-2 points higher than the league average). Now adjust the filters up one tick at a time until you get a list short enough to evaluate in detail (of course put on a price filter depending on budget, or end of contract etc).
One other thing I do is look for that one slow guy that is exceptional in other areas. In particular if you don’t have a set piece taker I think it’s worth picking up a player with great technicals even if he doesn’t fit the overall meta.
Yarema said: I mean, opening the correct view in FM, copying and pasting, opening the right file and all the other minor things you have to do also take like a minute or two in total. Plus you actually need to be doing stuff in that minute instead of 1 click to load database and do other things while it finishes. Expand
I was only commenting on GS, not making a comparison. The point is that something that can be coded to run almost instantly (and is in other tools) takes minutes in GS. The rating part of it is (to me) easier in GS than changing a python script, and can get the same result.
Gracolas said: Ok, makes sense.. I dont mess up with the ykykykyk ratings, so i just leave them unchanged. In this case i find GS really smooth and quick. Expand
It’s loading the db from the game that’s slow; it takes a couple of minutes for something that should take 1-2 seconds. Otherwise you can easily make your own ratings for GS to match any particular python script.
I think the question may have been more along the lines of what test conditions FM-Arena uses, such as:
-on holiday, first choice XI for all games with no transfers - player condition and morale maximized for all games, no injuries - staff controls set pieces and training
There are definitely people that are rude to devs and react like babies when they don’t get what they want, when they want it. But I think it is normal to feel some irritation at things that don’t work properly year after year. Is there really a good reason why when I customize something and size the columns to fit perfectly the game can’t remember that when I advance the day? It cant even remember when you used the game's own auto size feature.
I don’t think you can go wrong with either the machine learning weights, Mark’s, or the more recent weights from squirrel_says. The AI in the game is really not very smart and seems to value players based on CA and reputation. So if you play a half decent tactic with players selected based on what matters in the ME, and managing fitness and morale properly, you can win everything with an average CA substantially less than the top AI teams. My point was more a theoretical one that a “better” rating system should consider how the attributes are correlated since its easy to over- or under-weight attributes based on tests done one variable at a time.
A bit sad for those that remember when SI had a policy to only release when the game was ready. Last year we find out that they removed features so they could launch on time. This year they launch with the most bugs I can remember. It’s the price of becoming a larger company, more overhead, more employees to get disgruntled if they don’t get their bonus $$$.
I appreciate the effort put into this, so I feel bad about giving negative feedback. The problem of trying to work from the attribute testing is that these are single variable tests; any positional differences are averaged out. Also when you try to combine the attribute tests you are not accounting for the correlations between the different attributes that exist in the player database; a player with 20 pace is going to be very likely to have high acceleration and very unlikely to have extremely low acceleration. So if you treat the effects of pace and acceleration as additive you will be overestimating their combined effect. The value of the machine learning approach is that it tests all variables simultaneously and accounting for position.
dzek said: The truth is I haven't tried it in this year's version. Last year (though I believe the same is true this year) I came to the conclusion that for some reason the left side is somehow more efficient than the right. @alex
I also saw other evidence in my own tests of this in addition to the results above.
But we can, if anyone wants to of course, try it again this year maybe something has changed. Expand
dzek said: The truth is I haven't tried it in this year's version. Last year (though I believe the same is true this year) I came to the conclusion that for some reason the left side is somehow more efficient than the right. @alex
I also saw other evidence in my own tests of this in addition to the results above.
But we can, if anyone wants to of course, try it again this year maybe something has changed. Expand
A two point difference in 1200 matches with exactly the same GF and GA. So surely just noise not a real difference between left and right. In practice there could be a difference based on a particular opponents having a weak left or right sided defender, otherwise I’d just choose based on which side fits your own players strengths.
But as I said before you can’t test a tactic without set pieces, so what is going to be your default? Do you want to have a default that deliberately aims headers away from CBs? If you play with 3 CBs are you not allowed to have them all in the box because that’s “unfair”? Some people are just jumping to conclusions about rankings without even trying to present evidence that there’s a problem. (Like showing that some completely different tactic would move from a mediocre ranking to be ranked best if you just aimed corners at the other CB). Yes, in FM23 you could create a corner routine that scored a lot of goals … IF the player you aimed at had extreme jumping reach and headers, and if your corner taker was accurate. If you prioritized fast CBs and they weren’t good at jumping and heading you could have that routine and not even know there was anything unusual about it. The issue was just as much a problem of how the match engine dealt with those attributes as it was the positioning. So should tactics also not exploit pace and acceleration because that is a much bigger advantage than any corner routine?
I don’t see how the changes in FM24 affect anything. To test a tactic you have to have set pieces, and set pieces are always dependent at least on formation and the propensity of the tactic to lead to set pieces. It’s not like the rankings are skewed because one group is submitting tactics with deliberately bad set pieces to not take advantage of known exploits, so I don’t think the ranking of tactics would change at all. If you don’t want to score a lot from corners you can just pick a top tactic and change the set piece routine, or just not sign CBs with 18 jumping reach and 16 headers. It’s just as much a problem of attributes as it is of the set pieces themselves.
You say Wembangomo shouldn’t be rated higher, but then you say you he *is* better and you want a rating system to rate him higher. So I don’t think you can mean both. In my opinion he will not perform better in the match engine, in fact he will be substantially worse. You shouldn’t interpret single attribute tests to mean that only pace and acceleration matter. If there was a test of increasing all attributes except pace and acceleration I’m confident that would outperform the increase only in pace or acceleration. And this is what is reflected in the the various ratings that are available. There are strikers faster than Haaland in the game but they don’t perform as well because they don’t have stats close to him in all the other areas. If they had 20 pace and acceleration and a current ability in the 190s and great hidden attributes then maybe they would be better, but no player with a 140-150 CA can be close to him.
Currently your rating puts Odriozola as a 68% rating, compared to Wembangomo with a 61% for the DR/WBR position.
It is obvious that in everything other than physical statistics, Odriozola is better, but if we stick to the tests here, wouldn't Wembangomo be much superior in the match engine?
I think any reasonable ranking system would have Odriozola better (and this is the case with all 4 rankings I have set up in GS right now). He’s a legit Premier League player. The other guy is a League One player that’s a bit faster.
That is what happens with top tactics because the reason they are top after 1200 is partly due to luck. The score decreases with more testing because of regression to the mean. If you looked at the worst rated tactics you would find the exact opposite happening. That’s just how the statistics of large samples work, and also why the more tests done the more accurate the result … even if you will only play one season in a real save .
You could reasonably conclude that the three or four point difference between a tactic after 1200 vs 4000 matches is not likely to be meaningful in a save that lasts only a few hundred matches. But that isn’t a reason to choose the highest rated tactic based on 1200 matches; instead you should pick any tactic near the top that fits the best players you have available.
Does anyone remember which tactic was top after 1200?
@Zippo @Droid do you have this information available please? is it possible to add the result of 1200/2400/4000 (if applicable) to the results table?
The reason I ask is because (and I don't mean any disrespect with this) but I only really take note of results from 1200 matches, the reason for this is because by the time I've played 1200 matches in game, I've probably already got an OP squad.
Im pretty sure that over 1200 matches there was 1 tactic that achieved 65/66 points but for the life of me can't remember which one it was
Thanks
I think you’ve missed the point of doing more than 1200 matches, it has nothing to do with how long a save is played for. When a tactic is tested there is a lot of randomness and just by luck a tactic can rank higher or lower than it should. The more matches the closer the points ranking is to the “true” value. What matters to you is the true value because that is what most accurately represents what you experience in your save.
One other thing I do is look for that one slow guy that is exceptional in other areas. In particular if you don’t have a set piece taker I think it’s worth picking up a player with great technicals even if he doesn’t fit the overall meta.
I was only commenting on GS, not making a comparison. The point is that something that can be coded to run almost instantly (and is in other tools) takes minutes in GS. The rating part of it is (to me) easier in GS than changing a python script, and can get the same result.
It’s loading the db from the game that’s slow; it takes a couple of minutes for something that should take 1-2 seconds. Otherwise you can easily make your own ratings for GS to match any particular python script.
-on holiday, first choice XI for all games with no transfers
- player condition and morale maximized for all games, no injuries
- staff controls set pieces and training
I don’t think you can go wrong with either the machine learning weights, Mark’s, or the more recent weights from squirrel_says. The AI in the game is really not very smart and seems to value players based on CA and reputation. So if you play a half decent tactic with players selected based on what matters in the ME, and managing fitness and morale properly, you can win everything with an average CA substantially less than the top AI teams. My point was more a theoretical one that a “better” rating system should consider how the attributes are correlated since its easy to over- or under-weight attributes based on tests done one variable at a time.
Here are my tests from last year:
Left - https://fm-arena.com/thread/6199-fm23-the-last-dance-v1-7-left-by-dzek/
Right - https://fm-arena.com/thread/6200-fm23-the-last-dance-v1-7-right-by-dzek/
I also saw other evidence in my own tests of this in addition to the results above.
But we can, if anyone wants to of course, try it again this year maybe something has changed.
dzek said: The truth is I haven't tried it in this year's version. Last year (though I believe the same is true this year) I came to the conclusion that for some reason the left side is somehow more efficient than the right. @alex
Here are my tests from last year:
Left - https://fm-arena.com/thread/6199-fm23-the-last-dance-v1-7-left-by-dzek/
Right - https://fm-arena.com/thread/6200-fm23-the-last-dance-v1-7-right-by-dzek/
I also saw other evidence in my own tests of this in addition to the results above.
But we can, if anyone wants to of course, try it again this year maybe something has changed.
A two point difference in 1200 matches with exactly the same GF and GA. So surely just noise not a real difference between left and right. In practice there could be a difference based on a particular opponents having a weak left or right sided defender, otherwise I’d just choose based on which side fits your own players strengths.