Works really well in practice, but requires a WT who wins the headers vs opposing defensive wingers. Thus, I'd be surprised if this works well in this specific tactic testing scenario.
( TYPE 1 ) Fast, short and weak striker ( 137 CA ) as Target Forward
( TYPE 2 ) Strong, tall and slow striker ( 140 CA ) as Target Forward
RESULT:
As you can see despite having a higher "star" rating as TF and a higher CA with TYPE 2 attributes, it still produces a worse result.
Btw, there's nothing surprising in that result because as I said, we've tested that many times already, just read through this thread or some others.
If you're getting any different results in your own tests then it's because you test for not enough matches or your testing mythology has issues.
Cheers. Expand
<3 Thanks for the test! Have your stored the resulting save games by chance? I would love to compare some of the stats from both test cases to see where it differs from the one I've ran!
TommyToxic said: I remember Zippo (or a different admin) tested a 442 I submitted with floated crosses and a TF. One tested with a proper TF (high jumping reach etc), and one with standard testing league strikers. The one with a proper TF did much worse in tests. This was in FM23. Expand
Do you remember if the test setup used two TFs as forwards and both were "proper" without a speedy one?
Yep, looks bad enough, I think if the AI uses some target forward archetype as I've proposed in that TF spot and does not manage to beat RNG in this test, it's not worth it even thinking about adding that archetype @Zippo
Two Target Forwards does not make sense and does not fit the custom coding SI did for TFs. All the settings about "To Target Forward" only works properly with one from what I've seen, and in addition to that you need the AI to select one fast and one tall striker. Please try it with this tactic, I think without a tall one it will be terrible enough to clearly notice the difference despite rng.
Zippo said: Unforutanlty, I can take into consideration the result of your test because 500 matches is really nothing and your testing methodology/environment is unknown for me and even if it were know then I still can't be sure that your test were done properly.
As I said it's been tested many times before but I can test it specially for you one more time.
I'm assuming the loadout of the strikers in your tests is the same as in Dzeks database here:
Is that correct? If yes, here's a loadout for a target forward archetype with the same positions and recommended ca:
Basically just decreasing some stats like acceleration, pace, technique and dribbling a bit, and increasing jumping, strength, heading, aggression and bravery until I got the same recommended CA.
Note that I've only changed the players that only have "Striker" position instead of those that can also play wings (if I understood it correctly, thats also how it's set up in the fm-arena test), so the AI has both the fast option for the advanced striker role and the slow and tall one for the target forward. Since both have the same recommended CA, the AI selects the correct one for each position.
If something in the fm-arena setup is different feel free to adjust the exact stats
Delicious said: If you don't have API of SI, good luck to teach to the "AI" to how select players for builds scenario. Expand
I've just tried it, it wasn't very difficult. I've adjusted the values of the TF archetype until the Recommended CA of both players matched. As soon as that was the case, the AI correctly selects the player with the highest star rating for that role - which is the target forward archetype.
Delicious said: The only thing i never understood is why players like Beto/Adebayo/Chermiti/Lewin/Doumbia and many others are just not included in the scenario.
Try Adebayo in a lone striker build and compare it to another gnome monster like Karim Konatè.
I mean do you prefer a gnome that can't jump or a Space-Cow that can even Jump?
That's why Haaland is a cheat-code. But you can achieve crazy results even with Pio Esposito Terminator. Expand
Completely agree, Davie Selke is (or was especially in FM23) just another one of those strikers with insane physicals with aggression and bravery, in our online save he had multiple seasons with >40 goals. The very best strikers in the highly competetive online game are those that can also jump, not just run.
Delicious said: You can play a Target Forward as AF, roles are roles, there is no sky-rocket science being it, sadly. Expand
I mean roles are pretty much just some presets on the individual instructions with very minor custom coding for some roles (like inverted defenders, segundo volante, etc), but even these minor changes can be fun to use. For the target forwards, you have some unique mechanics, like setting him as priority target for crosses or for the keeper. This, in combination with some PPMs, can make tactics work that just won't work in these simplified test-scenarios. So I do think there are some incentives to at least try to capture some of this in testing, even if not everything can be modelled correctly if we don't get some amazing "sandbox" editor (maybe the switch to unity engine will allow a few more things like that).
Delicious said: Don't get biased by "average rate", it's totaly cosmetic on some roles, why?
Because if a specific player lose alot of "headers" aka Aerial Battle, they will simple get negative rate. Expand
I agree, it's the very same thing in real life ratings like sofascore. But still, the top scores in my test case posted above where the TF archetypes, despite them having only 133 suggested CA, while the speedy ones hat 141.
Thanks for your reply @Zippo <3 I'll try to keep my points as concise as possible, but if something is unclear, I can elaborate further.
* Attributes that are used for 1v1 comparisons have the highest impact (e.g. physical attributtes + dribbling) * Impact of strength in actual games is severely undererstimated in fm-arena because set pieces cannot be set (Kurt Zouma with 20 stength can easily do >35 goals a season while others with the same jumping reach cannot) * Using Dzek's version of the test league, I've tested TF tactics with default vs TF archetype (e.g. increased jumping, strength, aggression and bravery, but reduces pace, acceleration, dribbling and technique) and the results were way better (High crosses + wingers are useless if strikers can't get to the ball). Here are some results without using set pieces, only a 4-4-2 tactic with reliance on wing play, one TF and one AF in the front:
Vanilla Setup:
TF Archetype for the TF striker:
As you can see, the results were significantly better, more crosses connected and more goals were the result. The used archetype looked like this:
When looking at the average ratings and goals, most player will be this archetype. This might also be due to some slight balancing issues (maybe some other stats have to be reduced more, since jumping reach got a +7), but it the point should still be valid.
Summary: I agree that strength won't do much in this specific test scenario, and I'm 100% sure introducing the archetype wouldn't suddenly change the meta tactics much - but I'm also quite sure we'd see more diverse tactics uploaded that use roles like wingers, target forwards, targeted crossing and high crosses with mid-level success. The only difficulty I see is getting the archtetyp just right so that the AI will assign them correctly.
- - -
Some other points about biases in the attribute testing:
* Using a highly specialiced tactic for attribute testing might introduce some major biases into the resulting data. The selected tactic using "dribble more" on every role, uses the low-possesion, faced-paces and high-press meta. This will likely lead to overestimating the impact any related attributes, while underestimating attributes required for slow build-up play. * A plain increase of some attributes does not account for the different impact that attribute has on CA (for the specfic position) * Increasing/Decreasing some attributes affects all players, while others only affect a few (Unsurprising that physical attributes are relevant for everyone, while some mental attributes barely have any effect for most)
Thanks for your work @Zippo! From your experience, is there any chance if you would add some more archetype players to the teams (like a target forward where the jumping/strenth stats matches those of the defenders, at the cost of speed/dribbling), would the automated team selection be smart enough to use them if such a role is selected in a tactic?
Hi, thanks a lot for sharing this dzek! There are a few quirks of the fm-arena test setup that I wanted to analyze, so this save file is extremely valuable to me.
Some stuff I want to test: * Low crosses only meta because all strikers are smaller than defenders? * Can roles like target forwards, false 9, etc. work if you design players that fit that position better? * If so, do roles like wingers profit from this? * Attribute Testing has been done using some meta tactic with lots of vertical passing and dribbling - does the impact of the "creative" passing attributes increase with some tactic that relies on slow build up play? * etc. etc.
Cheers for those tests, very interesting results! I've seen quite a few hefty discussions about the implications of your tests, but I haven't been able to find some relevant info that would be required to actually understand what they mean.
In your test above, you simply "hard set" and froze the attributes, so the game does not distribute a few more attributes to get back to the CA, is that correct? Can you confirm that you did the same thing in the "Player Attribute Testing", e.g. setting +5 Pace does not decrease other attributes so the CA stays the same?
Because depending on that, the meaning of results changes quite a bit.
If +5 for an attribute means its plain +5 and everything else stays the same, this wouldn't account for the fact that some attributes affect CA more than others. On the other hand, this would also mean that it's very odd to see barely any changes for most attributes compared to the "default" test, as the actual CA would be increased.
If +5 for an attribute means its increased by +5 and the game scales the other attributes so that the CA stays the same, this would mean that the increased "meta" attributes in that specific scenario provide the most benefit. This would also mean, if there's no change to the overall CA, it would be expected that we don't see any changes in the end result for most attributes, as it should roughly balance out. I was hoping that this was the case, as this would mean that most attributes would be farely balanced.
------
Irregardless of this, I think your statement:
Zippo said: Also, there's one important thing to understand when in our tactic testing league we set Acceleration attribute of a player to "20" then his Accelerating attribute will be 5-6 points higher than the Acceleration attribute any other player in the league. Expand
Cannot be stressed enough. The current meta tactics all use very high tempo and lots of dribbling, so setting some attribute like passing or overview +5 is maybe relevant to the DMs and AMs, but mostly irrelevant for most other players. Changing the physical attibutes, however, affects each and every position by quite a lot, as their direct counterpart will ALWAYS be inferior in a direct comparsion (ceteris paribus). The only "odd one out" is the dribbling attribute, but all meta tactics use "more dribbling" in general and for each individual role, so it would be interesting to see the results for some possession based tactic.
Have you tried these exact same tests with a "Clean Sheet" Tactic?
Thanks a lot for your work and Merry Christmas Svonn
This tactic - https://fm-arena.com/thread/8910-wingplay-target-forward-test-tactic/
( TYPE 1 ) Fast, short and weak striker ( 137 CA ) as Target Forward
( TYPE 2 ) Strong, tall and slow striker ( 140 CA ) as Target Forward
RESULT:
As you can see despite having a higher "star" rating as TF and a higher CA with TYPE 2 attributes, it still produces a worse result.
Btw, there's nothing surprising in that result because as I said, we've tested that many times already, just read through this thread or some others.
If you're getting any different results in your own tests then it's because you test for not enough matches or your testing mythology has issues.
Cheers.
<3 Thanks for the test! Have your stored the resulting save games by chance? I would love to compare some of the stats from both test cases to see where it differs from the one I've ran!
Do you remember if the test setup used two TFs as forwards and both were "proper" without a speedy one?
As I said it's been tested many times before but I can test it specially for you one more time.
Please, share any tactic that you think would work better with a slow/tall/strong striker instead of a fast/short striker - https://fm-arena.com/board/12-football-manager-2024-tactics-sharing-section/
The tactic will be tested through out our regular testing then I'll test it with the strikers you suggested
and you will be able to see what difference it'll make.
Good idea, cheers for running those tests! I've uploaded the tactic I've used for my testing here:
https://fm-arena.com/thread/8910-wingplay-target-forward-test-tactic/#thread-name-anchor
I'm assuming the loadout of the strikers in your tests is the same as in Dzeks database here:
Is that correct? If yes, here's a loadout for a target forward archetype with the same positions and recommended ca:
Basically just decreasing some stats like acceleration, pace, technique and dribbling a bit, and increasing jumping, strength, heading, aggression and bravery until I got the same recommended CA.
Note that I've only changed the players that only have "Striker" position instead of those that can also play wings (if I understood it correctly, thats also how it's set up in the fm-arena test), so the AI has both the fast option for the advanced striker role and the slow and tall one for the target forward. Since both have the same recommended CA, the AI selects the correct one for each position.
If something in the fm-arena setup is different feel free to adjust the exact stats
Cheers!
https://fm-arena.com/thread/5351-should-you-follow-the-highlighted-attributes-of-the-roles/page-5/
This will be used to check if there's any benefit to this tactic using a taller but slower striker.
I've just tried it, it wasn't very difficult. I've adjusted the values of the TF archetype until the Recommended CA of both players matched. As soon as that was the case, the AI correctly selects the player with the highest star rating for that role - which is the target forward archetype.
Delicious said: The only thing i never understood is why players like Beto/Adebayo/Chermiti/Lewin/Doumbia and many others are just not included in the scenario.
Try Adebayo in a lone striker build and compare it to another gnome monster like Karim Konatè.
I mean do you prefer a gnome that can't jump or a Space-Cow that can even Jump?
That's why Haaland is a cheat-code. But you can achieve crazy results even with Pio Esposito Terminator.
Completely agree, Davie Selke is (or was especially in FM23) just another one of those strikers with insane physicals with aggression and bravery, in our online save he had multiple seasons with >40 goals. The very best strikers in the highly competetive online game are those that can also jump, not just run.
Delicious said: You can play a Target Forward as AF, roles are roles, there is no sky-rocket science being it, sadly.
I mean roles are pretty much just some presets on the individual instructions with very minor custom coding for some roles (like inverted defenders, segundo volante, etc), but even these minor changes can be fun to use. For the target forwards, you have some unique mechanics, like setting him as priority target for crosses or for the keeper. This, in combination with some PPMs, can make tactics work that just won't work in these simplified test-scenarios. So I do think there are some incentives to at least try to capture some of this in testing, even if not everything can be modelled correctly if we don't get some amazing "sandbox" editor (maybe the switch to unity engine will allow a few more things like that).
Delicious said: Don't get biased by "average rate", it's totaly cosmetic on some roles, why?
Because if a specific player lose alot of "headers" aka Aerial Battle, they will simple get negative rate.
I agree, it's the very same thing in real life ratings like sofascore. But still, the top scores in my test case posted above where the TF archetypes, despite them having only 133 suggested CA, while the speedy ones hat 141.
* Attributes that are used for 1v1 comparisons have the highest impact (e.g. physical attributtes + dribbling)
* Impact of strength in actual games is severely undererstimated in fm-arena because set pieces cannot be set (Kurt Zouma with 20 stength can easily do >35 goals a season while others with the same jumping reach cannot)
* Using Dzek's version of the test league, I've tested TF tactics with default vs TF archetype (e.g. increased jumping, strength, aggression and bravery, but reduces pace, acceleration, dribbling and technique) and the results were way better (High crosses + wingers are useless if strikers can't get to the ball). Here are some results without using set pieces, only a 4-4-2 tactic with reliance on wing play, one TF and one AF in the front:
Vanilla Setup:
TF Archetype for the TF striker:
As you can see, the results were significantly better, more crosses connected and more goals were the result. The used archetype looked like this:
When looking at the average ratings and goals, most player will be this archetype. This might also be due to some slight balancing issues (maybe some other stats have to be reduced more, since jumping reach got a +7), but it the point should still be valid.
Summary: I agree that strength won't do much in this specific test scenario, and I'm 100% sure introducing the archetype wouldn't suddenly change the meta tactics much - but I'm also quite sure we'd see more diverse tactics uploaded that use roles like wingers, target forwards, targeted crossing and high crosses with mid-level success. The only difficulty I see is getting the archtetyp just right so that the AI will assign them correctly.
- - -
Some other points about biases in the attribute testing:
* Using a highly specialiced tactic for attribute testing might introduce some major biases into the resulting data. The selected tactic using "dribble more" on every role, uses the low-possesion, faced-paces and high-press meta. This will likely lead to overestimating the impact any related attributes, while underestimating attributes required for slow build-up play.
* A plain increase of some attributes does not account for the different impact that attribute has on CA (for the specfic position)
* Increasing/Decreasing some attributes affects all players, while others only affect a few (Unsurprising that physical attributes are relevant for everyone, while some mental attributes barely have any effect for most)
Cheers! :-)
Some stuff I want to test:
* Low crosses only meta because all strikers are smaller than defenders?
* Can roles like target forwards, false 9, etc. work if you design players that fit that position better?
* If so, do roles like wingers profit from this?
* Attribute Testing has been done using some meta tactic with lots of vertical passing and dribbling - does the impact of the "creative" passing attributes increase with some tactic that relies on slow build up play?
* etc. etc.
I hope to find some time for this, cheers!
Cheers for those tests, very interesting results! I've seen quite a few hefty discussions about the implications of your tests, but I haven't been able to find some relevant info that would be required to actually understand what they mean.
In your test above, you simply "hard set" and froze the attributes, so the game does not distribute a few more attributes to get back to the CA, is that correct? Can you confirm that you did the same thing in the "Player Attribute Testing", e.g. setting +5 Pace does not decrease other attributes so the CA stays the same?
Because depending on that, the meaning of results changes quite a bit.
If +5 for an attribute means its plain +5 and everything else stays the same, this wouldn't account for the fact that some attributes affect CA more than others. On the other hand, this would also mean that it's very odd to see barely any changes for most attributes compared to the "default" test, as the actual CA would be increased.
If +5 for an attribute means its increased by +5 and the game scales the other attributes so that the CA stays the same, this would mean that the increased "meta" attributes in that specific scenario provide the most benefit. This would also mean, if there's no change to the overall CA, it would be expected that we don't see any changes in the end result for most attributes, as it should roughly balance out. I was hoping that this was the case, as this would mean that most attributes would be farely balanced.
------
Irregardless of this, I think your statement:
Zippo said: Also, there's one important thing to understand when in our tactic testing league we set Acceleration attribute of a player to "20" then his Accelerating attribute will be 5-6 points higher than the Acceleration attribute any other player in the league.
Cannot be stressed enough. The current meta tactics all use very high tempo and lots of dribbling, so setting some attribute like passing or overview +5 is maybe relevant to the DMs and AMs, but mostly irrelevant for most other players. Changing the physical attibutes, however, affects each and every position by quite a lot, as their direct counterpart will ALWAYS be inferior in a direct comparsion (ceteris paribus). The only "odd one out" is the dribbling attribute, but all meta tactics use "more dribbling" in general and for each individual role, so it would be interesting to see the results for some possession based tactic.
Have you tried these exact same tests with a "Clean Sheet" Tactic?
Thanks a lot for your work and Merry Christmas
Svonn