Higby on Magriders - Page 7

MurderBunneh · 2013-02-19, 09:31 AM

I am pretty sure that if the k/d infantry statistics where shown for before/after GU2 it would be right in line with the other statistics we already have. In other words the Mag would have been in first place by a large margin pre-update and in second place post-update. The tank that wins fights gets to stay in the battle and do more damage IE camp more spawns therefore kill more infantry.

Those that say the amount of mbt spawns have something to do with it are also right.
CLEARLY the amount of spawns coincides with the power level of the mbt. This has been the case for ALL online games that I have ever been a part of. OP things are fun and therefore are spawned more and played more.

They could release all the statistics in the world and you would still switch your argument to some "fun factor" rant. Those that want a mobility boost for the MAG are you prepared to receive a firepower nerf to keep things in check?

MurderBunneh · 2013-02-19, 09:32 AM

Originally Posted by Rothnang

Of course its important, it's just not the "Haha gotcha, the magrider was super OP and now everything is great" piece of evidence people make it out to be.

Well it CERTAINLY doesn't show the Mag to be UP now does it??
I am sure that if the numbers were reversed you would see them in a new light.

Kerrec · 2013-02-19, 09:34 AM

Originally Posted by Rothnang

... it would show that certain vehicles are much harder to learn, but stronger when you do master them.

Yeah, nothing wrong with that. Isn't that the assymetrical design (factions not being copies of each other) goal? As long as everything is relatively equal in the big scheme of things, this is exactly what people want!!!

Originally Posted by Rothnang

The average simply doesn't tell the whole story, distribution does matter because it shows if everyone is clustered closely around the average, or if its a mass of people on one side that are balanced by some huge outliers on the other.

When sample sizes are large enough, every "combination" is accounted for. The frequency of those combinations tell the big picture... IE: allows someone to establish what the average is.

You can't force a bell curve. People that "suck" will be at one end, but as they practice, they'll move towards the middle. Maybe they'll even manage to get on the other side. Over time, the bell curve you forced, would fall apart.

Pella · 2013-02-19, 09:38 AM

Who cares if these stats correct or not. The fact is that since the Magrider nerf, Outdoor fights are allot more common. Less spawn camping. And less mass herds of Magriders.

To me its the best change they have made. And obviously tweaking the AA and lib have played apart. But the Magrider nerf was imperative!

psijaka · 2013-02-19, 09:41 AM

Originally Posted by Rothnang

Well, the data he has is the result of averaging out all the tank kills by Magriders, right?

So let's say we take all the individual datapoints in that average, and plot them out into a curve.

Let's say we have 100 Magrider samples.
39 of them scored 0 tank kills,
38 of them scored 1 tank kill,
14 of them scored 2 tank kills,
7 of them scored 3 tank kills,
2 of them scored 4 tank kills,

Then we have 100 Vanguard samples.
48 of them scored 0 tank kills,
20 of them scored 1 tank kill,
15 of them scored 2 tank kills,
9 of them scored 3 tank kills,
5 of them scored 4 tank kills,
3 of them scored 5 tank kills

In this hypothetical data set what you would see is that much fewer people do OK in a Vanguard than in a Magrider, but there are more people who do really well in a Vanguard than in a Magrider.

This kind of graph would tell us something about the skill distribution in the vehicle, and it would show that certain vehicles are much harder to learn, but stronger when you do master them.

The average simply doesn't tell the whole story, distribution does matter because it shows if everyone is clustered closely around the average, or if its a mass of data points on one side that are balanced by some huge outliers on the other... there are all sorts of things that only the distribution can tell you.

What if a tank with an average K/D of 1 had 99 out of 100 people getting 0 kills, and then one guy who got 100 kills. Don't you think that would be noteworthy, despite the fact that it doesn't show up on the average?

Oh, that would be noteworthy all right! But statistically meaningless.

The problem with producing a distribution curve as you suggest, is that SoE would have to track individual players rather than just collect raw data on each tank kill in isolation; I'm sure that this would be possible but the data capture is more complex and the survey would have to run for a lot longer to gather enough multiple kills for the distribution to be statistically significant. Would be good data to see though!

Assuming that we did get this data and it proved that the Magrider performs significantly better than the Prowler at the high end; the next question to ask is "is this really of consequence?". If the numbers involved are small (say only the top 5% of players), then in the scheme of things this is not significant; PS2 isn't an e-sport. But if say the top 30% perform significantly better, and the bottom 30% perform significanltly worse, then maybe there is an issue to be addressed. Then comes the difficult question - How?

Calisai · 2013-02-19, 09:58 AM

Originally Posted by Rothnang

Fewer people are using Magriders now than before, that's for sure. The number of prowlers and Vanguards doesn't seem to have gone up significantly as far as I can tell though.

Originally Posted by Pella

Who cares if these stats correct or not. The fact is that since the Magrider nerf, Outdoor fights are allot more common. Less spawn camping. And less mass herds of Magriders.

This is why, for me... the change isn't horrible. I've already adjusted to the new Mag. And getting all those zergling horrible 1/2 magriders off my roads gives me more targets and room to maneuver.

Personally, I am having a blast with all the footzerging that is going on now with the NC/TR.

Assist · 2013-02-19, 10:00 AM

Originally Posted by Kerrec

When sample sizes are large enough, every "combination" is accounted for. The frequency of those combinations tell the big picture... IE: allows someone to establish what the average is.

You're right and wrong. You're assuming the sample size is large enough that it will account for the spikes. Using this assumption in Planetside 2 is wrong. Boiling down what you said, you're saying the pre-patch when one server had 100 Magriders another server at some point had 100 Vanguards. Your assumption is that because one group has an outlier than another group will have an outlier that counters it, and that's just not true even in large sample sizes. Even in large statistical analysis outliers are thrown out because they're considered invalid data for being so far from the mean/average.
Unless you assume that the same number Magriders, Prowlers, and Vanguards were spawned pre-patch than not every possibility is accounted for. I'd be surprised if there wasn't near double the number of Magriders spawned compared to Prowlers and Vanguards pre-patch. Assuming that information, statistics wouldn't even be used to compare the two. It'd be like attributing global warming to petroleum prices, there's just too many variables to take those kinds of statistics seriously.

Pella · 2013-02-19, 10:07 AM

Originally Posted by Calisai

This is why, for me... the change isn't horrible. I've already adjusted to the new Mag. And getting all those zergling horrible 1/2 magriders off my roads gives me more targets and room to maneuver.

Personally, I am having a blast with all the footzerging that is going on now with the NC/TR.

Good man. Now only if half the people who whine and dont actually play the game had the same mentality the game would be even better.

Kerrec · 2013-02-19, 10:34 AM

Originally Posted by Assist

You're right and wrong. You're assuming the sample size is large enough that it will account for the spikes. Using this assumption in Planetside 2 is wrong. Boiling down what you said, you're saying the pre-patch when one server had 100 Magriders another server at some point had 100 Vanguards. Your assumption is that because one group has an outlier than another group will have an outlier that counters it, and that's just not true even in large sample sizes. Even in large statistical analysis outliers are thrown out because they're considered invalid data for being so far from the mean/average.
Unless you assume that the same number Magriders, Prowlers, and Vanguards were spawned pre-patch than not every possibility is accounted for. I'd be surprised if there wasn't near double the number of Magriders spawned compared to Prowlers and Vanguards pre-patch. Assuming that information, statistics wouldn't even be used to compare the two. It'd be like attributing global warming to petroleum prices, there's just too many variables to take those kinds of statistics seriously.

Absolutely not. IF Magriders were being spawned at double the rate of either of the other two tanks, across all servers at all times, THAT in itself is an indication that something is wrong. It is captured in the statistics.

Data mining a data set to define WHAT is wrong is an art. Almost anyone can manipulate statistics to show what they want to show.

But you look at trends before a change, and trends after a change.... with a large sample size, it proves that the thing that was changed WAS or WAS NOT affecting the results in a significant way.

You keep going on about how Magriders had way more numbers and that was "the problem". That's pretty shallow. Did you ever ask yourself WHY did Magriders have more population with respect to other tanks? Obviously not, because that would imply there is something about them that is superior.

You can't just throw out data you don't like because you don't approve of what it implies.

Bear · 2013-02-19, 11:40 AM

The data that Higby used for his assessment is taken from every tank kill over every server for an entire month. There simply isn't a larger set of data available. Someone with even the most basic understanding of statistics understands that a large sample size mitigates variables. What many VS seem to be arguing for is some form of perverse heteroscedasticity is happening that paints them as more powerful than they really are. It's an absurd argument.

Arguing that your tank was fine in the face of overwhelming statistical evidence is called denial. The Mag is fine where it is right now, the Prowler needs to be toned down a bit and the Vanguard buffed. End of story.

MrBloodworth · 2013-02-19, 11:52 AM

If you don't like the numbers, Claim they are cooked. Vote Republican!

MrBloodworth · 2013-02-19, 12:03 PM

Originally Posted by Rothnang

This video shows EXACTLY what people are talking about when they say the Magriders dominance was mostly up to people playing stupidly.

Every single tank he kills, no attempts to angle their armor toward him, just sitting their with their turret turned taking it in the side and back.

That video doesn't in any way showcase an overpowered magrider, it showcases idiotic opponents.

You are out of your god dam mind if that's your take away from that video.

No tank can match the mag in this regard. No tanks turret can turn as fast as a mag can while always facing forward. That would not have helped. No tank can turn its body as fast as a mag while always facing forward. That would not have helped. NO OTHER TANK CAN FLOAT OVER ANOTHER TANK like the mag can AND SPIN AROUND FACING FORWARD AFTERWORDS( Pre-patch ).

You are in full denial and spin zone.

Quovatis · 2013-02-19, 12:31 PM

Higby screwed up the graphs. The Mag vs Prowler and Mag vs Vanguard are identical. That's statistically impossible. One of them is wrong.

psijaka · 2013-02-19, 12:46 PM

Originally Posted by Quovatis

Higby screwed up the graphs. The Mag vs Prowler and Mag vs Vanguard are identical. That's statistically impossible. One of them is wrong.

Yeah, my guess is that they cocked things up and plugged the same data set into excel or whatever they used to produce the graphs.

Doubt that this affects the main graph at the top though, but it does devalue the data somewhat if they can't get this right.

DirtyBird · 2013-02-19, 12:50 PM

new one