Pyros on Steroids

[Ahab]

I am an engineer. I know statistics.

The Pyros are jacked up, probably the witches too.

There is no doubt in my mind. It's due to either a changed baseline probability or a bad random number scheme (pardon ... Pseudorandom), but one things for certain: it's jacked.

Don't make me perform a hypothesis test on it!

Just load up on pyros...

[Jeffery]

Wanna explain how they are jacked????
The do exactly what they are designed to do, no more no less.

[Ahab]

Put it this way: I predicted before the game started that this guys pyros (Dragon20011) would block more than usual.

He blocked the first seven of eight times.

Predictability is a strong indicator of a change in the true mean.

Do it yourself: note the number of blocks versus the number of hits in your next game. Report back to me!

I would not be surprised if it isn't better than even.

I would be surprised if its around 1/3, very surprised if its less than that.

EDIT

(By the way: 3^7 is 9*9*9*3=2178 I think. Its one thing to have that happen; its quite another to be able to predict it)

[drakonfire]

actually the pyros have been acting normal for me... its my assassins that r the problem, 2 times they have been shot from the front in the last 2 days... both were hits

[...]

oooooo, ouch
an assassin being hit from the front?
that is totally unheard of:rolleyes:

[xyxaxyz2]

Which units do you use to attack Pyros? Scouts seem to miss mages a lot.

[...]

yea, ive really noticed that too, i think assassins hit em better, well fer me anyways

[Merforga]

Blocking is what makes it so fun though ^^. You cant equate chance and statistics. Of course, theoretically it should block only 1/3 times but again it is CHANCE! It is RANDOM! =P

[Ahab]

Yup. My four stoned pyro/dsms are killin now. ^^

[Mystic Wanderer]

My test data:

Knight hit Pyro
Front: 25/12
side: 30/7
back: 999999999999999999999999/0.00000
Scout hit Pyro
side: 28/7
Front: 7/2

Got tired of testing, so skip DMW.

Basically, the blocking rating matches test data.

People usually remember the times Pyros block and when Pyros get hit, they took for granted and don't remember. Our memory can sometimes trick us. So no bug in Pyros blocking, unless somebody provide a more detailed annalysis.

[Twelve]

There are times when pyros block too much, it seems, and then there are times when pyros don't block enough(when they're on my side...lol).

It all averages out properly in the end.


12

[Ri'Orius]

Originally posted by Merforga
You cant equate chance and statistics. Of course, theoretically it should block only 1/3 times but again it is CHANCE! It is RANDOM! =P

Actually, yeah, you can use statistics on random variables. It's what a lot of statistics is about.

You collect data, find a sample mean, then calculate the probability that this sample mean would be reached if the true mean is X. That's what he's talking about, hypthesis testing.

Statistics works a lot with random variables. After all, are statistics gathered by asking absolutely everyone about something? No: they use random samplings to estimate the actual value of what they're looking for.

But yeah, I'll try to keep track next time I play and get the results up here, but for now, until data is shown against it, I agree with Mystic Wanderer's opinion that people just notice it more when a Pyro blocks.

[Ahab]

Well, all I can say is that the variance is high, the number stream is highly stochastic (look that up in your Funk and Wagnall), and it 'trends'; that is, it seems to run hot and cold.

If dice did this in Vegas you would be either very rich very quickly, or calling the cops.

Yes, the pyros did settle down in later games. However, as my Elite buds will tell you, for this game, I predicted this and then observed it, calling out "3 for 3", "4 for 4", etc, as it happened.

It would fail the hypothesis test at a pretty high level of confidence; that's why I'd recommend you keep tabs on it. We need more samples!

[Northwind]

This might be cheating, but I just pasted a previous response regarding blocks into this discussion.

I know people are going to disagree with me here, but . . .

Random often does not look random. They have actually done studies in which people were presented with lists of numbers that were random and lists of numbers that were simply made to look random. When asked to choose which list was truly "random", people invariably chose the "fake random" lists - forgetting that in a truly randomized string, you are going to see sequences that do not "appear" random.

I think it's likely that you have the same thing is happening here. People selectively attend to those times that they are blocked, rather than when they actually hit. When is the last time someone posted saying "this game must not be random because my guys kept blocking my opponent's a ridiculous number of times!".

My guess is that if you kept track of your hits/blocks over a long (key word here) period of time, you would see that things were very close to the stated stats. However, within that there would be a few games that would appear ridicuously non-random.

Of course, there might be a bug in the program. What the hell do I know?

And, I might add, even hypothesis testing is wrong a pre-determined number of times (depending on what alpha is set at). It is possible that your example would simply have been one of those times. I am in total agreement that more data is needed to determine this, but think that the likeliest explanation here is that people post only when they are really annoyed by a particular game. You just happened to predict yours.

[Ahab]

Yah, but it's the prediction (remember this is supposed to be a 1-in-2100 occurrence) that bothers me. I predicted it based on the previous game, which was also quite blocktastic.

However, when I switched to a four pyro/dsm setup, they stopped blocking so well. Go figure... I even got beat by doorsofperception, a grey. But a good grey!

[DUI Warrior]

What we have here is a nice mix of Murphy's Law and Law of Large Numbers, so it's hard to tell what's going on.

Also, I agree with the fact that we only notice the strange occurances and take regular ones for granted. We may take our unblocked attacks for granted and only notice the times when opponent's pyros block, so it seems that we're constantly getting screwed. I'll just experiment with 2 fake accounts (on the same computer :O) and see what my results will be, but I think it should be 1/3.

[Jeffery]

I have HAD IT.
There HAS to be something wrong with the blocking. I know the "chance means lucky and unlucky streaks" and blah blah blah. And I have always supported percentages.
BUT THIS IS IT!
When I am playing and GET NOT ONE HIT TO THE SIDE and my opponent hits repeated front shots, it is way too much.
This has made me want to quit altogether more than ANYTHING else.