87 replies to this topic

[M R T]

Problem with pure WLR is player behavior. We have already observed that players consider tiers a reward, and act selfishly as a result. This is nothing new, in QP you have always had this behavior. However, if win is all that matters, this behavior will turn into players throwing lost matches and rushing for objective to secure a win. As long as matchscore matters, players will at least have an incentive to keep on fighting even if the match is lost.

Edited by Morte Nilsum, 10 August 2020 - 01:26 PM.

[V O L T R O N]

Nightbird, on 10 August 2020 - 06:15 AM, said:

Thanks for your input, [redacted].

Experiments are still designed using mathematics. People that test the first ideas that pop into their head are doomed to accomplish nothing in a lifetime of experiments.

And Mathematics are formed from experience and REALITY.

The problem is PGI wont make other changes other than PSR. Like drop zones/lances, organizing them by speed instead of groups or tonnage. Fixing the huge structure in maps that push nascar, and I can name about 100 other things that would IMMEDIATELY improve the game that has nothing to do with match maker.

I would also make it WAY harder to reach tier 1. If thats even possible.

and thank you for your response [redacted].

Edited by Ekson Valdez, 13 August 2020 - 09:24 AM.

[Nightbird]

V O L T R O N, on 10 August 2020 - 01:41 PM, said:

And Mathematics are formed from experience and REALITY.

The problem is PGI wont make other changes other than PSR. Like drop zones/lances, organizing them by speed instead of groups or tonnage. Fixing the huge structure in maps that push nascar, and I can name about 100 other things that would IMMEDIATELY improve the game that has nothing to do with match maker.

I would also make it WAY harder to reach tier 1. If thats even possible.

and thank you for your response, [redacted].

Great, you have one opinion. When you can put together the opinion of every player in the player base, analyze it, model it, and make informed business decisions out of it, you'll have demonstrated thinking.

Edited by Ekson Valdez, 13 August 2020 - 09:24 AM.

[Cluster Fox]

BackShot, on 10 August 2020 - 06:46 AM, said:

...
At 500 AMS you can even en up with a negative w/l and still be in the tier 1 bracket. ( 0.95 )
...
That being said, group queuing in the soup padding W/L so much and not padding AMS, a heavily weighted W/L would be flawed too cause of that.
...
So think it is less an issue than match score padding.

Backshot, you hit the nail on the head. W-L padding is much harder than MS padding.

Where's the balance? That's where I scratched my head for a while. The trendline is natural across the player base so that's what I used.

I'm not saying my system will work for all the players, nor would NightBird's; but if it works for at least 80-90% of the players out there, it benefits the full player base in the end. Closer matches, less stomps.

I'd like to point out that the 400+ AvgMS and 0.95 WLR player you are talking about doesn't exist on Jarl's list. Theory crafting only gets so far and yes this model has it's limits - can't deny that.

About the access to Tier 1....
How hard should it be to get in Tier 1? Well that's a question for PGI to answer, not me. If they mention "we'd like to see X% of the players in Tier1", well I can tweak the A-factor and make it happen.
I tuned the amplitude so that not too many players "top off" either at 0 PSR or at 5000 PSR.

Edited by Cluster Fox, 10 August 2020 - 04:37 PM.

[Nightbird]

Cluster Fox, on 10 August 2020 - 04:29 PM, said:

I'm not saying my system will work for all the players, nor would NightBird's; but if it works for at least 80-90% of the players out there, it benefits the full player base in the end. Closer matches, less stomps.

Speak for your self... WLR will make it as hard as possible to stat pad by grouping, since after a point, your group of 4 will always be teamed with the 8 worst players out of the remaining 20. If you can still rack up wins with that, great, you deserve it. Most 4 mans can't and therefore will reach some equilibrium. We'd see very few people with over 3WLR under my system. Under Jay Z's, 30WLR, 50WLR, the sky's the limit. The same will be true under your system because you cap PSR and so under-estimate the impact of grouping and high WLR players.

Edited by Nightbird, 10 August 2020 - 09:18 PM.

[BackShot]

I know a player with 500 AMS and less than 1 w/l doesnt exist, and thats obviously a good thing. I am just using that extrapolated exemple to show up what i find being a flaw in this system.

I just dont see why a player with 50 more AMS than another would requires like 0.20 less or more w/l to be in the same bracket.

AMS gives way too much weight to raw damage, spread, uneffective damage.

I dont see why someone using LBx, Lrms, Mrms, ect.... would require less w/l to be in the same bracket as someone using gauss, ppc, lasers, Acs...

If you win as much as another guy, you sould be in the same bracket as him, whatever your raw damage output. This is how i see it.

[Cluster Fox]

Nightbird, on 10 August 2020 - 08:56 PM, said:

Speak for your self... WLR will make it as hard as possible to stat pad by grouping... The same will be true under your system because you cap PSR and so under-estimate the impact of grouping and high WLR players.

You are totally right about Win Loss padding being hard and the cap.

However TBH since my fixes are stable, a cap is not required. I should test and add that too, thanks for the indirect suggestion.

As for my "it's valid for 80-90% of the players" claim, I meant it as "there's no perfect system".

BackShot :
I'm currently running a 100% W-L based simulation to see if it improves over my own tweaks. (Different X,Y,C) If it does improve I'll try to go halfway. Results expected later this week.

[Nightbird]

Cluster Fox, on 11 August 2020 - 10:33 AM, said:

You are totally right about Win Loss padding being hard and the cap.

However TBH since my fixes are stable, a cap is not required. I should test and add that too, thanks for the indirect suggestion.

As for my "it's valid for 80-90% of the players" claim, I meant it as "there's no perfect system".

BackShot :
I'm currently running a 100% W-L based simulation to see if it improves over my own tweaks. (Different X,Y,C) If it does improve I'll try to go halfway. Results expected later this week.

Just make sure the end result is better than WLR... what's the point of a more difficult and higher effort system that delivers less performance than a stupid WLR system? I were to make any more complex suggestions, like adaptive design systems or bayesian systems, they would outperform WLR.

I see all of this as making Jay Z's formula as close to WLR as possible, and while this is paying some dividends, the parts that are not WLR based are actively working against you and making even the best of your attempts 30% worse.

[Cluster Fox]

Well, my goal was to test out my tweaks on JayZ's model and simulate them to compare. Because no one else stepped up to try and I don't like suggesting stuff if I can't back it up. Then of course I fine tuned my suggestions with the results - and still am.

Getting a better result than WLR would be great, but providing an alternative that would work using most of whatever PGI already chose, is my aim here.

Neither of us has the final say. Sadly PGI might very well do nothing about it either way and decide that JayZ's system is completely fine for their purpose...

Edited by Cluster Fox, 11 August 2020 - 02:22 PM.

[Nightbird]

The deafening sound of feet wiggling in the air... yep.

[thievingmagpi]

Cluster Fox, on 11 August 2020 - 02:18 PM, said:

Well, my goal was to test out my tweaks on JayZ's model and simulate them to compare. Because no one else stepped up to try and I don't like suggesting stuff if I can't back it up. Then of course I fine tuned my suggestions with the results - and still am.

Getting a better result than WLR would be great, but providing an alternative that would work using most of whatever PGI already chose, is my aim here.

Neither of us has the final say. Sadly PGI might very well do nothing about it either way and decide that JayZ's system is completely fine for their purpose...

They haven't even followed up on their claim to re-examine how match score is calculated.

This was a done deal as soon as they could demonstrate they put forward the tiniest bit of effort

[Nightbird]

You can't fix match score because you cannot program when an action taken is beneficial to the team or harmful to the team. You can only tell that the action was taken. In a match, now matter how you weight actions, you are encouraged to survive as long as possible as this give you more chances to take actions that earn MS = playing selfishly.

As far as evaluating the skill of a pilot, MS is as useful as dendritic water crystals.

[thievingmagpi]

Nightbird, on 11 August 2020 - 03:23 PM, said:

You can't fix match score because you cannot program when an action taken is beneficial to the team or harmful to the team. You can only tell that the action was taken. In a match, now matter how you weight actions, you are encouraged to survive as long as possible as this give you more chances to take actions that earn MS = playing selfishly.

As far as evaluating the skill of a pilot, MS is as useful as dendritic water crystals.

That's lovely.

However, PGI hasn't followed up on their claim to re-evaluate how match score is calculated.

[Tarl Cabot]

Nightbird, on 10 August 2020 - 08:56 PM, said:

Speak for your self... WLR will make it as hard as possible to stat pad by grouping, since after a point, your group of 4 will always be teamed with the 8 worst players out of the remaining 20.

If you can still rack up wins with that, great, you deserve it. Most 4 mans can't and therefore will reach some equilibrium. We'd see very few people with over 3WLR under my system. Under Jay Z's, 30WLR, 50WLR, the sky's the limit. The same will be true under your system because you cap PSR and so under-estimate the impact of grouping and high WLR players.

Simply pointing out, as per Nighbird's post, of how a MM setup would work towards keeping low W/L ratio, for both sides of a drop. So in this drop, one side would have a killer group with "drones" or meatshields/cannon fodder, players with the lowest W/L ratio of that drop, and potentially the lowest Matchscore, Damage, preferably accurate dmg > followed by kills/kmdd, etc wins most games, even more so while working as a team.

But how would matching up a killer team with the worse players of a match actually be sorta out? What other criteria? Essentially how would the MM determine which team a player goes to? Presentation is everything but this is the first time, in your own words, how the MM would work based on a W/L ratio...Then what of the other factors, such as team tonnage/weight Class, Group size and its tonnage (or weight class), etc?

No disrespect but you keep tooting your horn while not even providing examples, from avg to extremes, just like PGI fails to present, only laying out a generalize outline. As a reminder, PGI was asking for MS thresholds/triggers while anything else would likely have needed a MM rewrite itself.

Edited by Tarl Cabot, 13 August 2020 - 06:02 AM.

[Nightbird]

Tarl Cabot, on 13 August 2020 - 05:58 AM, said:

Simply pointing out, as per Nighbird's post, of how a MM setup would work towards keeping low W/L ratio, for both sides of a drop. So in this drop, one side would have a killer group with "drones" or meatshields/cannon fodder, players with the lowest W/L ratio of that drop, and potentially the lowest Matchscore, Damage, preferably accurate dmg > followed by kills/kmdd, etc wins most games, even more so while working as a team.

But how would matching up a killer team with the worse players of a match actually be sorta out? What other criteria? Essentially how would the MM determine which team a player goes to? Presentation is everything but this is the first time, in your own words, how the MM would work based on a W/L ratio...Then what of the other factors, such as team tonnage/weight Class, Group size and its tonnage (or weight class), etc?

No disrespect but you keep tooting your horn while not even providing examples, from avg to extremes, just like PGI fails to present, only laying out a generalize outline. As a reminder, PGI was asking for MS thresholds/triggers while anything else would likely have needed a MM rewrite itself.

Point to one such example for Jay Z's system.

No offense but you guys hold yourself to no standards at all and keep setting them for mine. The results from two simulations already show WLR (which I call a stupid system) outperforming Jay Z's system by 300%, so what do you call something that much worse than stupid?

I've put more than 10x the effort into proving the viability of my system than Jay Z's "presentation", not gonna put a second more into it. QP aren't worth the effort, I'm a FP guy anyways.

Edited by Nightbird, 13 August 2020 - 06:30 AM.

[C H E E K I E Z]

I have nothing important to add, i just came to say hi to nightbird and jay <3


Continue with your discussions

[VonBruinwald]

thievingmagpi, on 11 August 2020 - 05:05 PM, said:

That's lovely.

However, PGI hasn't followed up on their claim to re-evaluate how match score is calculated.

I'm not sure you understood what he wrote. Matchscore is inherently based upon the actions of the individual, which is a bad determinator of "skill" in a team based game as it encourages the individual to prioritise their own performance over that of the team.

Adjusting matchscore values will do nothing to fix its inherently broken nature.

Nightbird, on 13 August 2020 - 06:24 AM, said:

No offense but you guys hold yourself to no standards at all and keep setting them for mine. The results from two simulations already show WLR (which I call a stupid system) outperforming Jay Z's system by 300%, so what do you call something that much worse than stupid?

It's called double standards, these guys pushed their idea through back channels to get it implemented. Now they're desperately trying to save face. Unfortunately your models (and reality) aren't on their side.

[thievingmagpi]

VonBruinwald, on 13 August 2020 - 12:32 PM, said:

I'm not sure you understood what he wrote. Matchscore is inherently based upon the actions of the individual, which is a bad determinator of "skill" in a team based game as it encourages the individual to prioritise their own performance over that of the team.

Adjusting matchscore values will do nothing to fix its inherently broken nature.

Nope, the point stands and it's entirely relevant.

PGI hasn't even followed up on their claim to re-evaluate how match score is calculated.

It's pretty straightforward. Try using your head.

VonBruinwald, on 13 August 2020 - 12:32 PM, said:

It's called double standards,

No double standard. One system exists, the other doesn't.

VonBruinwald, on 13 August 2020 - 12:32 PM, said:

these guys pushed their idea through back channels to get it implemented. Now they're desperately trying to save face.

Which back channels were these exactly?

Go ahead.

I'll wait.

VonBruinwald, on 13 August 2020 - 12:32 PM, said:

Unfortunately your models (and reality) aren't on their side.

Actually, being able to present a system in an appropriate manner is the most important element at hand. That's what was done.

You can continue pretending otherwise all you want.

Edited by thievingmagpi, 13 August 2020 - 12:57 PM.

[VonBruinwald]

thievingmagpi, on 13 August 2020 - 12:52 PM, said:

Nope, the point stands and it's entirely relevant.

PGI hasn't even followed up on their claim to re-evaluate how match score is calculated.

It's pretty straightforward. Try using your head.

System is inherently broken.

You believe PGI's claims......

I think it is you who is not using ones noggin. Do you seriously believe everything they say and claim considering their track record?

thievingmagpi, on 13 August 2020 - 12:52 PM, said:

No double standard. One system exists, the other doesn't.

I'm not sure you understand how double-standards work.

[thievingmagpi]

VonBruinwald, on 13 August 2020 - 01:23 PM, said:

System is inherently broken.

Of course, it's PGI.

VonBruinwald, on 13 August 2020 - 01:23 PM, said:

You believe PGI's claims......


I think it is you who is not using ones noggin. Do you seriously believe everything they say and claim considering their track record?

Which claims of PGI's do I believe exactly?

Go ahead, I'll wait.

VonBruinwald, on 13 August 2020 - 01:23 PM, said:

I'm not sure you understand how double-standards work.

Double means two.

One and none is one, not two.