

Would It Be Crazy For Pgi To Show Chassis Average Performance?
#1
Posted 25 October 2017 - 09:55 AM
Thoughts?
#2
Posted 25 October 2017 - 10:04 AM
You know how I know this? Because PGI has given us stats for mechs in the past and no one believed them. Reactions ranged from "PGI is lying" to "PGI is manipulating the statistics" to "PGI has made every single number up because Dev A was killed that one time by that one mech using that one weapon despite us saying that the devs never play this game."
Also nerfs and buffs are based on those stats and, again, no one believes PGI.
But sure, I'd love to see chassis average performance just to see how the community reacts.
#3
Posted 25 October 2017 - 10:10 AM
ForceUser, on 25 October 2017 - 10:04 AM, said:
You know how I know this? Because PGI has given us stats for mechs in the past and no one believed them. Reactions ranged from "PGI is lying" to "PGI is manipulating the statistics" to "PGI has made every single number up because Dev A was killed that one time by that one mech using that one weapon despite us saying that the devs never play this game."
Also nerfs and buffs are based on those stats and, again, no one believes PGI.
But sure, I'd love to see chassis average performance just to see how the community reacts.
Oh so true, insert any time a chassis is said to be over performing people flip out throwing claims at it not.
Or the any time Clan is Nerfed "Well Russ always has hated Clans and so does PGI"(btw this really is believed by many Clansmen) to the other side of "PGI hates IS and never gives us anything good! They only Buff Clans and Nerf IS!" (Once again many believe this as fact with illuminate like conspiracy tangents on it.
Fact is both sides have broken a$$ mechs and weapons and both sides have very under performing mechs and weapons. Both sides need some buffs and some nerfs....
Buy yes, PGI please let us see the numbers here, I am interested in see how things land.
Edited by CK16, 25 October 2017 - 10:11 AM.
#4
Posted 25 October 2017 - 10:12 AM
ForceUser, on 25 October 2017 - 10:04 AM, said:
You know how I know this? Because PGI has given us stats for mechs in the past and no one believed them. Reactions ranged from "PGI is lying" to "PGI is manipulating the statistics" to "PGI has made every single number up because Dev A was killed that one time by that one mech using that one weapon despite us saying that the devs never play this game."
Also nerfs and buffs are based on those stats and, again, no one believes PGI.
But sure, I'd love to see chassis average performance just to see how the community reacts.
Well the stats themselves are more just to compare to personal performance in the chassis.
The problem with the stats being used to make balance adjustments is they end up nerfing things that don't need to be nerfed because proper context isn't considered.
Its not that I don't believe the stats themselves exist. But like, Shadow Cats put up big numbers and low death rates because they sit and snipe with ECM. That doesn't make them OP, it just means they are being used in a standoffish role, which means low death rate and potentially solid damage numbers, but then they got nerfed.
This isn't about balancing the game, its just about seeing how people are doing in each mech.
CK16, on 25 October 2017 - 10:10 AM, said:
Yeah you tried to deny that the MCII is overperforming. LOL.
#5
Posted 25 October 2017 - 10:13 AM
#7
Posted 25 October 2017 - 10:14 AM
Gas Guzzler, on 25 October 2017 - 10:12 AM, said:
Well the stats themselves are more just to compare to personal performance in the chassis.
The problem with the stats being used to make balance adjustments is they end up nerfing things that don't need to be nerfed because proper context isn't considered.
Its not that I don't believe the stats themselves exist. But like, Shadow Cats put up big numbers and low death rates because they sit and snipe with ECM. That doesn't make them OP, it just means they are being used in a standoffish role, which means low death rate and potentially solid damage numbers, but then they got nerfed.
This isn't about balancing the game, its just about seeing how people are doing in each mech.
Yeah you tried to deny that the MCII is overperforming. LOL.
Only I defended the other Mk II's not the DS. Same as I did with the Kodiak back when that monster was young and healthy (all but the 3 were fairly balanced)
#8
Posted 25 October 2017 - 10:19 AM
#9
Posted 25 October 2017 - 10:21 AM
Their competitors like World of "fill in the blank" seems to have all that chassis performance information on 3rd party stat sites.
#10
Posted 25 October 2017 - 11:00 AM
Gas Guzzler, on 25 October 2017 - 09:55 AM, said:
Thoughts?
With the small player base, the stats would be biased. If we had the population of DoTa or LoL, then it would be a pretty good representation of the good verse poor mechs.
But then we would need a true "ranked" mode.
Edited by mogs01gt, 25 October 2017 - 11:00 AM.
#13
Posted 25 October 2017 - 11:17 AM
ForceUser, on 25 October 2017 - 10:04 AM, said:
You know how I know this? Because PGI has given us stats for mechs in the past and no one believed them. Reactions ranged from "PGI is lying" to "PGI is manipulating the statistics" to "PGI has made every single number up because Dev A was killed that one time by that one mech using that one weapon despite us saying that the devs never play this game."
Also nerfs and buffs are based on those stats and, again, no one believes PGI.
But sure, I'd love to see chassis average performance just to see how the community reacts.
Which stats have they shown for mechs before?
They said they've nerfed the SadCat because it was "overperforming", but they never gave us information
We just naturally assumed the most logical part, as Gas has said
It's not dead because MGs no longer suck, thankfully.
#14
Posted 25 October 2017 - 11:22 AM
#15
Posted 25 October 2017 - 11:25 AM
ForceUser, on 25 October 2017 - 10:04 AM, said:
You know how I know this? Because PGI has given us stats for mechs in the past and no one believed them. Reactions ranged from "PGI is lying" to "PGI is manipulating the statistics" to "PGI has made every single number up because Dev A was killed that one time by that one mech using that one weapon despite us saying that the devs never play this game."
Also nerfs and buffs are based on those stats and, again, no one believes PGI.
But sure, I'd love to see chassis average performance just to see how the community reacts.
So IF PGI releases the info and it points out that they are "flawed" in their decision making, it makes them look bad.
If PGI releases data that justifies their decision making, no one will believe them and they look bad.
Looks like a no win scenario for them.
So PGI needs some other metric they can use to justify their data and position. Something like their record of pushing out timely patches, fixing the "little details" that Russ is all about, and how well their balancing efforts have worked out.
You see what I'm getting at here? Huh, huh, huh?

In order to be credible, you have to have credibility to begin with.
Edited by TLBFestus, 25 October 2017 - 11:26 AM.
#16
Posted 25 October 2017 - 11:28 AM
Quote
If PGI releases data that justifies their decision making, no one will believe them and they look bad.
Looks like a no win scenario for them.
If PGI doesnt release the info they dont win either
#17
Posted 25 October 2017 - 11:32 AM
CK16, on 25 October 2017 - 10:14 AM, said:
Eh, the MCII-1 is only a half step behind the DS, IF THAT. I like that one.
#18
Posted 25 October 2017 - 11:49 AM
-sarcasm font on-
First off that is just so unnecessary. Over two years ago Paul stated that the whole goal of balance was to make every mech down to the the variant level viable and of equivelant value. I am sure that after two years they have gotten so close to this design goal that a chassis to chassis “average” comparison is redundant; since once you know one all the others must be pretty much the same.
Wait a sec...are you suggesting that they have failed in this goal? That different chassis might actually be performing better or worse than others!!? This is a shocking idea. I mean if this is even remotely true, I would have expected that PGI would be iteratively buffing any and all chassis that are of lower average performance. Since they don’t do such a thing, I can confidently conclude that all chassis must be pretty close in average performance.
Now excuse me, I am going to go prove this average equivelance by taking on every Deathstrike and Mistlynx G I can find with my St. Ive’s Blues because I believe in PGI!
-font off-
#20
Posted 25 October 2017 - 12:09 PM
Gas Guzzler, on 25 October 2017 - 09:55 AM, said:
Thoughts?
You'd need a lot of data to make it worthwhile. The guy with 2x LB10x, 2 ERLL and 2 ersmls on his Deathstrike I saw the other day is going to deadweight the chassis performance.
If you do it you need chassis by loadout by performance. So Deathstrike with X weapons is winning this amount and averaging this score. Y loadout is this, etc.
I would love to see that. Straight mech stats without context would be of limited use though.
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users