80 replies to this topic

[Ziptop]

I'm sure MM (mostly) works as planned, but I really wish somebody from the dev side could explain how the Matchmaking handles specialised aircraft. How, (for example) is it possible for one team to have 5 specials and the other team to have no specials at all? - 5 is an extreme, but I have seen it. 3 specials on one side and none on the other is sadly quite common and it does feel like it skews the outcome, possible even more than being up-tiered.

What is the mechanism here? 

[apartclassic]

I don't think MM takes specialization into consideration at all. It was never stated directly it does, nor was it denied. However, looking at compositions, I believe MM doesn't even 'see' specialization. All that is taken into consideration is 1) tiers, 2) plane classes, 3) flights. That's it. MM matches those, nothing else matters.

[Ziptop]

apartclassic, on 01 April 2020 - 11:43 PM, said:

I don't think MM takes specialization into consideration at all.

 

Well, that is certainly the way that it feels. It would still be good to have some feedback from an official source

[0berhessen]

apartclassic, on 01 April 2020 - 10:43 PM, said:

nothing else matters.

Thanks for the catchy record

[blindfoId]

Ziptop, on 02 April 2020 - 01:40 AM, said:

I'm sure MM (mostly) works as planned, but I really wish somebody from the dev side could explain how the Matchmaking handles specialised aircraft. How, (for example) is it possible for one team to have 5 specials and the other team to have no specials at all? - 5 is an extreme, but I have seen it. 3 specials on one side and none on the other is sadly quite common and it does feel like it skews the outcome, possible even more than being up-tiered.

What is the mechanism here? 

 

The matchmaking system does not take aircraft configuration into account. More information about MM can be found here. 

[zgubny]

Lord knows you got to change , Baby ... Before i stop loving you ....

 

                                                        

[Ziptop]

blindfoId, on 02 April 2020 - 11:36 AM, said:

 

The matchmaking system does not take aircraft configuration into account. More information about MM can be found here. 

Ok thanks Blindfold. Nice to have some official feedback and confirmation.

 

Now for the next question. 

Will this ever be revisited by the Dev team? Specialised aircraft are a big part of the game now and facing off against 4 specialised aircraft when your own team has none leads to really quite predictable results. Possibly more so than being up-tiered. It would be nice to see some effort to balance this aspect.

 

Also, I'm pretty sure I see more people quit before game start or early in the game because they see so many special wings on the red team roster than I see quit because they simply got uptiered.  

 

Is this something that could ever be looked at and improved? 

[ScyThEoS]

Ziptop, on 03 April 2020 - 03:24 AM, said:

Ok thanks Blindfold. Nice to have some official feedback and confirmation.

 

Now for the next question. 

Will this ever be revisited by the Dev team? Specialised aircraft are a big part of the game now and facing off against 4 specialised aircraft when your own team has none leads to really quite predictable results. Possibly more so than being up-tiered. It would be nice to see some effort to balance this aspect.

 

Also, I'm pretty sure I see more people quit before game start or early in the game because they see so many special wings on the red team roster than I see quit because they simply got uptiered.  

 

Is this something that could ever be looked at and improved? 

 

Many players have a specialized plane after an endless number of battles, not after the quality of the game. Unbalanced battles in terms of no problem compared to outgoing players from the battle! This should be severely punished, taking a break from playing.

 

 

The machine is one, the pilot who controls it is something else.

Edited by ScyThEoS, 03 April 2020 - 08:59 PM.

[Frateras]

Specialist confiuration should become one criteria for the MM. We have too many new Players without any knowledge about the game. If they are not balanced and the specialists fly in the other team the result is predictable.

Edited by Frateras, 05 April 2020 - 05:27 AM.

[zen_monk_]

Frateras, on 05 April 2020 - 06:26 AM, said:

Specialist confiuration should become one criteria for the MM. We have too many new Players without any knowledge about the game. If they are not balanced and the specialists fly in the other team the result is predictable.

 

Quite frankly, if you have new players without any knowledge about the game, resulting meeting with experienced players would be fully predictable even without any specialization. And again frankly, that's how it is supposed to be,

 

Specialist configuration is important for the experienced vs experienced. But if the change in MM ever occurs, that's secondary. The main problem is flights vs non-flights team.

[apartclassic]

zen_monk_, on 05 April 2020 - 08:04 AM, said:

 

Quite frankly, if you have new players without any knowledge about the game, resulting meeting with experienced players would be fully predictable even without any specialization. And again frankly, that's how it is supposed to be,

 

Specialist configuration is important for the experienced vs experienced. But if the change in MM ever occurs, that's secondary. The main problem is flights vs non-flights team.

 

And flight anchoring, especially when one is a bomber. Total bs.

[Ziptop]

Look yes. I have seen plenty of pilots who must have found their specialized wings in a cereal packet. 

And yes, when it's a mild mismatch (and/or the cereal packet wings are in play) it might not be that big a deal. 

 

But, in higher tiers where the wings pretty much indicate the pilot does know what they are doing and/or the mismatch it ridiculous then there is very little you can do. 

 

I've been on both sides of this and I often see 3 specials Vs 1 which could be balanced really easily.

 

And yes..flights skew the game too. But I am willing to bet I'm not the only one that has been uptiered, against a team with a flight when Blue does not have a flight and all the Specialized aircraft are on red team also..

 

I don't quit battles.....but I do sometimes come close to it when the odds are so stupid that participation seems pointless.

(Note, I've been playing a fair bit today and have yet to have a match where I'm on a team that has more specials than red. - and I've been uptiered every time. Taking a break to rest my throat after all the shouting at the screen) 

[apartclassic]

What does the notion of 'skill-based MM' has to do with this topic, i.e. how does MM handle specialized planes?

[Ziptop]

apartclassic, on 05 April 2020 - 03:30 PM, said:

What does the notion of 'skill-based MM' has to do with this topic, i.e. how does MM handle specialized planes?

 

I'll be the first to admit that balancing games on player skill might be hard to achieve and also likely result in longer wait times.

I'm also willing to accept that not all pilots with wings are necessarily highly experienced. Pretty sure I saw a bunch of newly minted specialised spitfires during the Spitfire event that likely specialised their planes before they knew how to fly them properly.

However I was actually surprised to learn that it's not taken into account at all because when it's really mismatched it does feel like it skews the balance of the game

As for pilot experience. (pilot skill) hmmmm. Jeez. well how do you even quantify that when you are just playing? Sure, the server can see it, but as a player I can't. 

 

 

[Ziptop]

So. It seems there are other optimizations that people would like to see in MM.

That's fair enough to discuss too.

 

So, what would you like to see changed and in what order?  The list so far looks like 

 

If a team has a flight the other team should have one also.

Balance number of specialized planes across both teams.

Balance pilot skill across teams (pilot training) 

 

(We might be entering "be careful what you wish for" territory here) 

 

[apartclassic]

'Skill'? What skill? When you talk about skill points for a pilot, it has a semblance of viability. I'm thinking it doesn't have any utility at all though. Because a 13-point A6M1 pilot doesn't balance a 13-point He-111 pilot (who also has a reargunner, so you add another 13-point LF plane in the MM?). It also goes against the basic premise of the current MM, which is to shorten the time in queue as much as possible. I'd think even trying to include this sort of criteria would defeat the purpose.

 

'Player-skill'? Measured by what? Response time in miliseconds? WR? Average damage done to ground targets per sortie? Battles quit vs battles finished? Average number of battles per week? As soon as WG decides to pick a definition of 'skill' for the purposes of MM, we are going to have people crying out loud it's not THE definition, that it's wrong, that it doesn't take into account this and that, or that, etc etc. It could potentially solve some problems, while at the same time creating a whole slew of new ones. In my opinion it's a bad idea, nuff said.

 

From my perspective people tend to demonize some elements of this game. I agree MM has issues in regards to handling specialized planes, and I think that realistically speaking (low staff at Persha, low investment levels for this title, long production cycle, limitations of the original code) what we can ask for is something very simple. Like e.g. a hardcap for specced planes in a battle, or parity (1 for 1, on each side), or perhaps the same mechanism that tiering uses, i.e. +/- 1 in terms of specced per side. Now that specialized planes are rather common, and do usually mark a difference in experience between players (though definitely not always the case), MM should address the situation; looking for some abstract qualities, like 'skill', will end up in misconceptions and mishandlings more often than not. 

[zen_monk_]

The most imba is one side with flights vs other without. Especially when you go up the tiers. Would be easy to bring into MM, only the waiting time for another similar pair would increase it to insane levels.

 

The other major fail of current MM is one side with, say, 3-4 HF, other with 1. Or 2 human bombers on one side and 2 bot bombers on the other, etc..  Could be solved easily, without increased waiting time.

 

And the mythical skill-based MM?

OK, let's make MM on player's skilz - let's say it's an universally coveted WinRate. You take all the players above ~70% and force them to fight exclusively each other. After a hundred battles all of them would be 48-52 %. So you drop them into 45-55% category. They are better and make 56%, jump into to 55-65% group from where they again are drown down to 45-55%... where everybody is!  And the same would go with every other skill criteria. And the same goes for people with 35-45% winrate.  After a while we would all end up in 48-52, permanently, every rise in success would only serve to be placed in a higher group from where you must drop down again.

 

No.

 

WG, fix first two lines from here... and while at it, split the damn Token Missions per tiers, and make HoF only for tiers IV to X, solo. The "after battle loot according to your performance, not per win" couldn't hurt, and almost literally all the troubles would be solved, including leaving the battles, deserted tier 10 an' all...

 

 

Edited by zen_monk_, 05 April 2020 - 06:03 PM.

[jakub_czyli_ja]

zen_monk_, on 05 April 2020 - 03:17 PM, said:

And the mythical skill-based MM?

OK, let's make MM on player's skilz - let's say it's an universally coveted WinRate. You take all the players above ~70% and force them to fight exclusively each other. After a hundred battles all of them would be 48-52 %. So you drop them into 45-55% category. They are better and make 56%, jump into to 55-65% group from where they again are drown down to 45-55%... where everybody is!  And the same would go with every other skill criteria. And the same goes for people with 35-45% winrate.  After a while we would all end up in 48-52, permanently, every rise in success would only serve to be placed in a higher group from where you must drop down again.

 

No.

No, it isn't win rate, and surprise, but there are skill criteria that can be used.

 

Skill level indicator has to be developed basing on criteria that aren't finite during the battle - number of frags, amuont of damage, personal points and things like that dealt simply won't work exactly as WR won't.

 

There are other criteria, like plane handling (evasion, aiming, target picking...), reaction time, fire efficiency (distance, accuracy)  - these can be measured, quantified and players can be ordered by these criteria in a good enough manner, to separate noobs with tunnel vision starting to fire from 1km.

For development time, analytics should find correlations between these criteria and those finite criteria. It can be done and be reliable, because now skilled players compete with unskilled for these finite resources, but because of skill, they grab more.

And after finding criteria correlating best, future classification should be done on these found factors.

 

So players with good situational awareness, good in dodging, firing, would fight with similar ones. None of these factors would significantly decrease like WR in irrecoverable way. If player falls down due to criteria decrease (after all meets targets harder to hit and shooting better, so it's harder to dodge) would fall into lower ladder, would quite fast go back up, because has better skill than those on ladder he felt - they are worse shooters, worse dodgers and have more tunnel vision.

Simple as that.

[apartclassic]

jakub_czyli_ja, on 05 April 2020 - 07:53 PM, said:

No, it isn't win rate, and surprise, but there are skill criteria that can be used.

 

Skill level indicator has to be developed basing on criteria that aren't finite during the battle - number of frags, amuont of damage, personal points and things like that dealt simply won't work exactly as WR won't.

 

There are other criteria, like plane handling (evasion, aiming, target picking...), reaction time, fire efficiency (distance, accuracy)  - these can be measured, quantified and players can be ordered by these criteria in a good enough manner, to separate noobs with tunnel vision starting to fire from 1km.

For development time, analytics should find correlations between these criteria and those finite criteria. It can be done and be reliable, because now skilled players compete with unskilled for these finite resources, but because of skill, they grab more.

And after finding criteria correlating best, future classification should be done on these found factors.

 

So players with good situational awareness, good in dodging, firing, would fight with similar ones. None of these factors would significantly decrease like WR in irrecoverable way. If player falls down due to criteria decrease (after all meets targets harder to hit and shooting better, so it's harder to dodge) would fall into lower ladder, would quite fast go back up, because has better skill than those on ladder he felt - they are worse shooters, worse dodgers and have more tunnel vision.

Simple as that.

 

And while the 'analytics' are at it, why don't they calculate albedo of bald people in NY, average muscle mass of 3rd graders in Wales, ratio of deer with longer left antlers to those with longer right antlers, and a couple more things. They are all quantifiable and countable, after all, so it 'can be done and be reliable'. Oh, while they are at it, how about they also find a reliable and verifiable way of measuring and either incorporating or discounting variations in network performance over given span of time, in a given area, as a constant and non-variable factor in the algorithm. Plus, as I'm sure they will have the tools available and a lot of extra time on their hand, how about they present the findings to some external, independent standardizing agency, so that it can be verified, reproduced, and rated reliably - something akin to PEGI. Also, because these might be relevant, how about they calculate some sort of a coefficient accounting for differences between old and newer generation machines players use to play WoWP, degree of hardware enthropy, flukes and spikes in ISP performance - you get the idea. PLUS - in cooperation with psychologists, let's have them quantify daily/weekly/monthly variatons of neuronal activity, perceptional acuity, nutritional levels of the brain, and 'natural' cycles of human body (as, I'm sure you know, there are times when we feel and perform better than our average, and there are times when we just can't accomplish jack) - as they are all quantifiable, finite criteria, undoubtedly influencing those petty numerical values of evasion, aiming, target picking, reaction time, fire efficiency and others.

 

Sounds like a fun job to do in the afternoon, inbetween watching some vids on YT. And it shouldn't be expensive, all the data is there, it's just a matter of data-mining and putting it all together.

 

You nailed it, mate. Brilliant idea.

[levlos]

hey guys,

 

good news, I have a friend who knows a friend who is the boyfriend of someone's aunt who has access to the IBM Power System AC922, IBM POWER9 22C 3.07GHz, NVIDIA Volta GV100, Dual-rail Mellanox EDR InfiniBand super computer at the Department of Energy, in the Oak Ridge National Laboratory. I forwarded the last suggestion of ApartClassic, and I secured some computing time on the POWER9: an entire day ! Mind you, the machine has only 2.41 million cores and is only capable of 148.6 petaflops, so the computing power could not be sufficient.... but it is possible we might get an answer to the question "what is skill in WoWP" ! Aren't we a lucky bunch ?

 

Now if you will allow me, I want to share with you guys a science fiction short story of Frederic Brown:

 

Answer

by Fredric Brown, 1954

 Dwar Ev ceremoniously soldered the final connection with gold. The eyes of a dozen television cameras watched him and the sub-ether bore through the universe a dozen pictures of what he was doing.

He straightened and nodded to Dwar Reyn, then moved to a position beside the switch that would complete the contact when he threw it. The switch that would connect, all at once, all of the monster computing machines of all the populated planets in the universe – ninety-six billion planets – into the super-circuit that would connect them all into the one super-calculator, one cybernetics machine that would combine all the knowledge of all the galaxies.

Dwar Reyn spoke briefly to the watching and listening trillions. Then, after a moment’s silence, he said, “Now, Dwar Ev.”

Dwar Ev threw the switch. There was a mighty hum, the surge of power from ninety-six billion planets. Lights flashed and quieted along the miles-long panel.

Dwar Ev stepped back and drew a deep breath. “The honor of asking the first question is yours, Dwar Reyn.”

“Thank you,” said Dwar Reyn. “It shall be a question that no single cybernetics machine has been able to answer.”

He turned to face the machine. “Is there a God?”

The mighty voice answered without hesitation, without the clicking of single relay.

“Yes, now there is a God.”

Sudden fear flashed on the face of Dwar Ev. He leaped to grab the switch.

A bolt of lightning from the cloudless sky struck him down and fused the switch shut.