Ultimatum X, on 08 August 2014 - 10:29 AM, said:
We've already told you, and others repeatedly why the tests were flawed.
Flawed testing is bad, it leads to bad conclusions and bad decisions.
If something is over performing you need to know exactly what is the root cause, you need clean data for this not polluted data.
What we have is completely polluted data, drawing conclusions from it are pointless.
Here's what there is to show why the testing is flawed (this doesn't mean "clans are fine" it means "testing was flawed".
1) Public Queue, 0 controls
2) No accounting for ECM in MM. More ECM on teams often leads to wins. Clans only have 2 choices of lights, one of them sucks really hard, the other one has ECM. Most clan teams probably had multiple ECM units.
3) No accounting for tonnage differences. Clan mechs are all at, or near their weight bracket tonnage limit. This is not the case IS side. Higher tonnage often skews towards winning matches.
4) Along with number 3, Clans only have 2 assaults. One is really popular, weighs 100 tons and brings 50 tons of raw firepower, the other is not as popular. This is both potentially a weight, and raw firepower advantage.
5) Clan mechs have fixed engines, most travel at the same speed (except for Super Fatty). If you ever watch competitive teams, you'll see this is actually something they build into their drop decks. It's a very subtle advantage, as the whole team moves together.
6) IS side is the only side with trial mechs.
What all of this says, and I'm going to spell this out for you, is not "clans are fine - nothing to see here".
It spells out:
The testing was flawed.
1. Assumption. You don't know what any of the testing parameters were. You don't know a single thing regarding their testing procedures. All you know is it was about Clan v. IS. Any comment about flawed testing is absurd.
2. Assumption. You don't know what Mechs were in each match recorded.
3. Valid. They did 2 testing periods and are planning on doing more. However, the results of the 2 tests was shocking enough for them to post a hard number and decide that balancing changes are required.
4. Irrelevant. How can you consider having 1 Assault, the DWF, as a detrimental variable? Lack of choice is a detriment? I can understand that, but would you really pick a WHK over a DWF? Doubtful.
5. Obfuscation. Your "sh*tty" XL engine offers you survivability and maneuverability that the IS can't match. I haven't seen a single build where, if possible, it would be desirable to swap an XL for a STD. I've asked people to provide examples of when a STD engine would be desirable over an XL.
6. Cherry-pick assumption. If you're trying to balance a game you have to view it from the perspective of the majority. If the Group Queue players are really only 16% of the gaming population then having your "controlled" environments is a case of sampling bias. In order to get valid numbers you have to include everyone in the testing process. That includes Trial Mechs and Cadet Bonus players. If you think balancing a game around 16%(?) of the population in an environment that is the polar opposite of how the majority of matches are fought then your definition of "balanced" is different than mine. We've all seen what happens when PGI attempts to "balance" aspects of the 16%.
Again, this point of "balance" and "skewed results" is based off the assumption that you know what PGI is doing. Unless you work for them or unless they outline their testing process, an assumption is just an assumption and not fact.
I understand why you're arguing this. If Clan Mechs currently win 90% of the matches then it will be a field day for you all when CW happens. I wouldn't want to give up 90% win-rate. What isn't being mentioned is that the IS side will always have Trial Mechs and "bads" in Solo Queue so completely disregarding them in your "controlled" environment is pretty shameful and not relevant to the game that is actually being played.
Edited by Be Rough With Me Plz, 08 August 2014 - 10:55 AM.