[Winter Balance Update] UKF Feedback
- This thread is locked
Posts: 143
That said, Comps on allied side will behave different compared to Comps on Axis side. You will always see a pattern.
Also bots from expert to easy have a really stupid behaviour when it comes to using buildings as cover or the cover mechanic at all. - sometimes they switch going in and out rapidly - sometimes they stand still when being flamed.
Comps dont use the tech tree as designed nowadays. That means there are units a player/ the community would think are mandatory and a Comp starts to use them not even the slightest way they were designed for.
Bots dont play the game as people do. They are designed to "know" enemy field presence through the fog of war. They are designed (experts) to flood the player with numerous amounts of units. Hell, they sometimes even have a fixed tactic for a certain map thats not going to change. Their commander choice is not as random as you would think.
I only use exp battles for "training purposes". The more efficient i use my units and ressources against AI the more confident i grow in PvP where i tend to be more careful overall. Its a good "training" to get comfortable with multitasking and timings. Though, Comps are predictable: you always know when they start spawning their first vehicle whereas in pvp a player might fool you with that.
To sum it up. With what i said above, it is possible for a UKF bot to behave in a such inefficient way that even being on exp level they act like little children. (Man, wasn´t it the same in CoH1?)
Posts: 680
If there are any maps you think I should avoid then let me know.
For instance I just ran a USF hard versus OKW standard AI and USF did indeed win, after about half an hour. Russia on the same settings dispatched the OKW in a 20 min rofl stomp.
My feeling is that balance patch has made the UKF considerably weaker, will run UKF hard versus OKW standard on the live version to check.
Did run an AI battle with 300 popcap and unlimited resources once... The axis built loads of Leigs, Stukas and Pak 43s in addition to the normal PIVs and Pumas.
Posts: 17914 | Subs: 8
Doesn't matter how mong the AI is... It is equal mong. Unless someone comes up with a mathematical model which perfectly predicts the power and cost efficiency of units the only possible way to compare equal players is via the AI.
And the only possible way to compare good drivers is to put them in a car and roll from a mountain, certainly not against each other with their respective cars.
There is no less reliable method to compare anything anyhow then vs AI, even flat map void tests have more merit.
For instance I just ran a USF hard versus OKW standard AI and USF did indeed win, after about half an hour. Russia on the same settings dispatched the OKW in a 20 min rofl stomp.
Ai doesn't play like players.
Doesn't know current meta.
Doesn't micro like players.
Has access to ALL doctrines at ALL times.
Doesn't know how to do cap orders.
You literally could take a random person from the street, sit it behind your PC and tell it to play, you would receive equally valid results.
Posts: 680
And the only possible way to compare good drivers is to put them in a car and roll from a mountain, certainly not against each other with their respective cars.
The way to compare cars would be with the same driver...
In live the UKF Hard AI comfortably beats the OKW standard AI. In the 'balance' patch OKW beats the better driver.
Unless you can find two players of completely identical skill then the AI results do show something.
Posts: 8154 | Subs: 2
The way to compare cars would be with the same driver...
In live the UKF Hard AI comfortably beats the OKW standard AI. In the 'balance' patch OKW beats the better driver.
Unless you can find two players of completely identical skill then the AI results do show something.
No, it just shows which AI was programmed better. Different factions have different AI behaviours. It's no different that pairing different players with different skills.
Posts: 13496 | Subs: 1
Doesn't matter how mong the AI is... It is equal mong. Unless someone comes up with a mathematical model which perfectly predicts the power and cost efficiency of units the only possible way to compare equal players is via the AI.
Relic did come up with formula that they used, but I doubt it was perfect. Not sure I that there many "perfect" things.
Posts: 112
All emplacements are highly situational and hard-countered by off-map and arty.
Furthermore, with the new stuka, it will become completely obsolete against OKW (they have leig too) and unusable in team games.
Thoughts:
- retracted original suggestion for mortar pit
- adjust brace to a toggle ability. When braced, as well as the damage reduction, the crew mans the garrison positions (from back when we could do that) and fires out with rifles at nearby enemies. If no units are within firing range and brace is activated, they will repair damage to their emplacement.
To make tommies more balanced:
- Remove medic kit from tommies, reduce pop cost for medic squad. Medic tommies are just unjustified.
- Perhaps add a scoped rifle (which takes bren slot) to the pyro upgrade (re-name it to spotter training or whatever). Just something so its not an automatic choice.
- Move gammon bomb to royal engineers
Posts: 680
No, it just shows which AI was programmed better. Different factions have different AI behaviours. It's no different that pairing different players with different skills.
Regardless it is Relic's game and Relic's AI indicates that this 'balance' patch has significantly nerfed the UKF.
Live version both feels stronger than the 'balance' patch and the AI matches confirm this. The same AI on live performs significantly better without the new patch.
You really can't get a fairer situation or comparison than the computer playing itself both pre and post patch. Any limitations of the AI are the same for both teams.
Posts: 112
Activated ability and told it to attack ground. The effect and (possibly) the rounds remained in effect after the ability expired.
Posts: 8154 | Subs: 2
Regardless it is Relic's game and Relic's AI indicates that this 'balance' patch has significantly nerfed the UKF.
Live version both feels stronger than the 'balance' patch and the AI matches confirm this. The same AI on live performs significantly better without the new patch.
You really can't get a fairer situation or comparison than the computer playing itself both pre and post patch. Any limitations of the AI are the same for both teams.
The factions is clearing getting nerfed, that's no doubt.
But using the AI as any parameter of balance is frankly stupid. Because the AI doesn't play by the same rules as normal players nor does it follow any normal build order.
Posts: 680
Quite the opposite though. It is moronic and toxic to argue that this is anything other than a valid test as the only parameters which change are those created by the balance patch.
Maybe the computer is biassed? Or UKF only lost to a lower ranked oppononet because Pluto is in Uranus?
Do any of the balance team have any actual commercial experience of testing software or balancing games? What procedures are used? What is the methodology?
Posts: 143
Ok, give me another try to explain AI behaviour.
In the patchmod UKF will feel weakened because the earlier engineers will build bofors and mortar pits sooner. That means less map presence overall. If they didnt have the engineers put into T0 they would build more maniline infantry until the tech is done. This is an example and i can asure you they will do it more likely on other maps then the one i forgot the name of lol
You simply cannot compare AI to player power levels. They dont use units the intended way SU76 going into your infantry, flak halftracks hiding behind enemy lines in fog of war and trying to stay hidden, they even have a different approach in using ressources for tech/nades. They don´t build mines - never will, they´re infantry won´t vault over obstacles, ukf for example won´t even build medium tanks on some maps and waste the fuel for bofors until they start spamming Churchills.
This is more or less scripted behavior - not artifical intelligence that "thinks" clever in a way to overcome your playstyle, not even adapt to your playstyle. They just build, spam, avoid grenades, and run around you AT-gun with HE-Shermans no matter how many snares you throw at them.
It wont work to balance the game around that behaviour. Players are way stronger in playstyle, tactics, positioning etc. Players re-evaluate (or atleast should) situations, npcs can be brought to a point where you almost know what they are going to do next because of their "script"
Posts: 13496 | Subs: 1
Build order seemed to be 3 Tommies, engineer then AEC so nothing untowards about that..
Quite the opposite though. It is moronic and toxic to argue that this is anything other than a valid test as the only parameters which change are those created by the balance patch.
Maybe the computer is biassed? Or UKF only lost to a lower ranked oppononet because Pluto is in Uranus?
Do any of the balance team have any actual commercial experience of testing software or balancing games? What procedures are used? What is the methodology?
I could try using AI vs AI fight for checking some things but overbalance is not really one of them.
For instance one could test AI vs AI same faction and level to check if a map give a faction an advantage.
Or may one could try to run AI same faction from different patches to see it something really stand out but one would have to create a mod for that...
Any of these test would need a large number of repetitions to come a solid conclusion and more like indication than anything else.
Posts: 680
Whataboutery snipped"
So I've played a few games, found numerous obvious bugs, felt that an already weak faction felt significantly weaker and gotten the impression that this is untested garbage based upon a wishlist of 'how do I win easier if playing a different faction'.
Testing it with the AI merely confirmed my suspicions.
So, what is the methodology for testing this patch? What data was used? All those famous tournament wins where someone picked the UKF in the finals? It feels completely untested.
Your assertions that this is merely down to quirks of the AI are moronic. It is the SAME AI being used pre and post patch and it didn't look like bofors were used earlier to me. I'm not even sure the AI built engineers in tier 1, or that players would.
Still that this patch puts the balance out to such a degree that UKF needs a +2 AI level advantage to win ( haven't actually tested that... But I'll gracefully assume that it can't be THAT bad) is pretty shocking.
I'm just trying to help out here, would be a good game if it was even vaguely balanced. Thought I'd check to see whether things were going in the right direction and whether the game was likely to be fun post patch. But......
Posts: 167
In short: They shouldn't use the AI for balancing.
Posts: 680
Using AI for balance is really a bad idea, the difference between easy AI and Expert AI is that Expert AI gets more resources and have vision and maybe it has damage bonces and so on, but it doesn't have better thinking and strategy, what AI do is basically spawning Random units like when u give a controller to a child and the child keep clicking the buttons randomly.
In short: They shouldn't use the AI for balancing.
It's an indication...
What metrics or methodology do they use?
Posts: 167
It's an indication...
What metrics or methodology do they use?
I don't really know but, what ever they are, they made the game way more balanced than before.
But I think the metrics that they use are experience, testing as much they could and of course the community feedbacks.
Posts: 8154 | Subs: 2
It's an indication...
What metrics or methodology do they use?
Discord > Custom game > replays
Playing vs AI at least gives you a rough impression of how units FEELS, not how necessarily they balance out against a real player.
Not sure what you are trying to prove when you set up a match between AI vs AI as it has been stated before that they don't play like human players and follow different rules of gameplay.
A player dropping and been replaced by an AI =/= game vs AI.
Posts: 1594
Both AI were on expert... Have run it again on hard, same result in 19 minutes. Can't recall which I played against but it wouldn't have been expert.
I've also run hard AI UKF versus standard AI OKW and.... OKW won after an hour and a half.
I'll run Hard UKF versus standard Ostheer to see if that is as bad.
I really think the balance team should make greater use of AI to balance the game. Regardless of some self important poohbah or other 'thinking' that this or that faction or unit is OP the only possible way to pit two equally skilled opponents against each other is via the AI.
Edit: Ostheer on Standard also beats UKF on hard. After about 1 hour 20. Won quicker than OKW did but looked like a slightly more even game.
The AI is incapable of effectively using many strategies that a human player can. They're really not that useful to gauge balance.
Posts: 680
Empirical, evidence based and substantive data driven figures? With automated testing over a large number of iterations? Mathematical modelling of units to ascertain compounding advantages? Or for that matter the asymetric advantages of Lanchester?
Or....
Is it just all pulled out of someone's ass?
IF, as appears to be claimed, AI fights have absolutely no relevance or bearing... Because, erm, it's playing a completely different game like... Or it smells or something...
Then what is the oh so superior model or tool being used?
Livestreams
122 | |||||
22 | |||||
2 | |||||
1 | |||||
62 | |||||
57 | |||||
12 | |||||
10 | |||||
1 | |||||
1 |
Ladders Top 10
-
#Steam AliasWL%Streak
- 1.829222.789+35
- 2.34857.859+13
- 3.587233.716+3
- 4.1095612.641+19
- 5.882398.689+4
- 6.280162.633+8
- 7.996645.607+4
- 8.379114.769+1
- 9.300113.726-1
- 10.717439.620+1
Replay highlight
- cblanco ★
- 보드카 중대
- VonManteuffel
- Heartless Jäger
Board Info
4 posts in the last week
23 posts in the last month
Welcome our newest member, asherllc
Most online: 2043 users on 29 Oct 2023, 01:04 AM