Mitchs best woodie poll

My sister and her family lived in Paris briefly last year when my brother-in-law ended up having to work there for about 9 months. I got an email from her after their visit to Parc Asterix and her comments about TdZ were the same: horribly rough. Same about Goudrix, of course. They loved the Mack bobled!

There are four issues with Tonnerre de Zeus:

1- the ride never got retracked. Only thing that got done to it was support reinforcement.

2- the park refuse to grease the ride, arguing it "makes it go too fast!". The result is that the ride sounds like 3 subways are pulling into a station whenever the train turns.

3- CCI messed up a section of track . Basically, halfway through the ride, you got this horrible right jolt. It feels like CCI realised that the train was misaligned, so they just built it correctly after, not bothering to fix the previous section.

4- the maintenance staff cannot maintain the PTC trains. Must be a record that one of the original train needed to be replaced 7 years after opening!

All of that gives a very aggressive and violent ride...

This is what I don't get. Most coasters ranking on Mitch's poll are similar to thier ranking on the golden tickets. At least to some degree. You can see sentimental favourites do a little better on golden tickets than on mitch's, but only relativly. Why is it that the Beast is always 7 or 8 on GT's, and in the 50's or 60's on mitch's. Is it because people who don't like the ride put it at the bottom of thier list to really bring it down? Just curious.

You could try reading the explanation on Mitch's page.

A short summary: The voting mechanism for the golden ticket is biased in favor of good coasters ridden by a lot of people, whereas the voting mechanism for Mitch's poll is biased towards the collective preferences of those voters who have ridden lots of coasters.

For an additional comparison, you could look at the wooden rankings at CoasterFanatics. In this case, the Beast is 21st in the overall poll (as of today), but doesn't crack the top 25 amongst those that rated more than 100 coasters.

Unfortunately, that bias can cause a small sample of experienced riders with eclectic tastes to throw off the results. A small sample is statistically irrelevant.

You're becoming 'that guy' ;)

GTs, by their very methodology, are always going to favor coaster that have a higher number of riders. Mitch's poll attempts to account for that by its comparison method, making every rider/ride combination a separate data record and thereby drastically increasing the power of the statistical analysis...but it does have its own inherent flaw by (potentially over-)correcting and giving a ride with VERY few riders an inside leg on a top spot...

I think that's a fatal flaw. I think there should be some cut off that disqualifies any ride with too few riders. Instead of a popularity contest, it becomes an anti-popularity contest. It's like those people who think music is cool because they're the only ones who listen to it.

If there's only 20 people in the world who've ridden coaster A and coaster B, then the opinion of 10 or 15 is a hell of a sample size. Because, remember, each vote is a comparison of two coasters - not just a vote for one. It's not about one coaster having very few riders, it's about the number of people that have ridden any two coasters. If I've ridden coaster A, but not coaster B, then I'm not qualified to make a judgement on which is better. Mitch's poll eliminates that aspect (which to me is the fatal flaw in the 'popularity' method) from the equation.

I still say that for some of the comparisons the sample size is adequate because the overall number is incredibly small, especially for the goal at hand - ranking enthusiats preferences of roller coasters. Let's keep some perspective here. It's a really inventive method for trying to quantify something that's silly and irrelevant.

Well, there are the '#' rides which are those that have less than 5 riders (I think now, but it used to be 3), but that's still a pretty small sample. If all 5 went on the same day (fairly likely for some rare coaster getting hit by a TPR or ACE trip) then the results will be skewed.

There's also the natural skew that a coaster that is harder to get will skew higher on the polls (for a couple of different psychological reasons).

In response to the "Fall from grace" of ST, TZ, and Megafobia, how many of the spots that they've slipped were taken by coaster built after them? I'm sure the ride experience has suffered, but the relative positions are probably as much to do with better coasters being built.

That's what has always impressed me about Phoenix's place in the polls (and BD's to some extent). It's been around and seen hundreds of new coasters built since its installation and yet still manages to consistently hang around the top 10 and even crack the top 5 some years.

http://groups.google.com/group/rec.roller-coaster/browse_thread/thread/b57d2fde4823217c

Not to mention there is a whole lot of people out there who have ridden a decent amount of coasters but think that filling out that ballot takes way too long and is not worth it (raising hand.)

In the end I only care about polls not named "Touchdown's Coaster Poll" when evaluating what new parks I might want to work into my next trip.

What Andy said. +1. We can go around with this, as we do every year, but statistical significance is not a matter of opinion, it's math.

Yes...and the math says that if you sample 80% of the potential population(such as 16 of 20 people who've ridden both Coaster A and Coaster B), that it's pretty damn close - plenty close for the purposes at hand.

It's when you start comparing it to other samples in an attempt to rank them that it becomes even remotely hinky...and that's just a matter of interpretation. Mitch chooses to interpret the data one way...it's certainly not the only way.

And still it comes back to the fact that it's not that important and for as important as it is, this is plenty valid...probably more than is necessary.

Don't overthink it. :)

Twenty people is not statistically significant by definition, whether you survey one or all of them, particularly in a field of tens of thousands of people who are into roller coasters. It has nothing to do with over-thinking it, that's just math, dude. Would you be comfortable taking a cure for cancer that was tested on 20 people? Not me.

Seriously, this is high school statistics.

We aren't curing cancer though, we are rating roller coasters.

I knew someone would get hung up on that, and that's not the point. Read up my friends...

http://en.wikipedia.org/wiki/Statistical_significance

Exactly. The relative importance of whatever you're trying to sample is irrelevant. The power of an experiment or survey depends on how large the sample. The smaller the sample, the weaker the test and the harder it is to prove x. Bigger samples = better. What Mitch's does is pseudo-statistics.

(Not that there's anything wrong with that.)

Jeff said:
Twenty people is not statistically significant by definition, whether you survey one or all of them, particularly in a field of tens of thousands of people who are into roller coasters.

But what if the population is only 20 people? That's my point and the thing that you keep overlooking. It's not a poll of tens of thousands of people into coasters. It's a lot of individual little polls of people who've ridden any two given coasters...and in some cases that number is going to be extremely small.

If there's only 20 people who meet the parameters (people who've ridden Coaster A and Coaster B) and you sample 14 or 16 of them, then you have a pretty valid sample of what people who've ridden both Coaster A and Coaster B think.

I know it's high school statistics and I don't understand how sampling 80% of a population is statistically insignificant.

Mitch is asking which is better, Coaster A or Coaster B. The only people qualified to answer that are people who've ridden both. If you ask 15 of 20 people what they think, you have a pretty good idea of what those 20 people think as a whole. Anyone outside of those twenty is outside the population we're interested in sampling.

Would you be comfortable taking a cure for cancer that was tested on 20 people?

Absolutely, if those 20 people it was tested on were the only 20 to ever have this very specific type of cancer that I'm interested in a cure for.

The population isn't 20 people. It's thousands. You're missing the point... 20 people who went on some trip together to ride some obscure coaster don't make a significant enough sample to reach any conclusion. Seriously, read the article on statistical significance. I'm begging you. That they compare one ride to another doesn't matter, it's still 20 opinions, period. Sample size matters. If more people don't ride the coaster, the pool of opinions is too shallow to draw any conclusions.

Like the cancer analogy, the 20 people cured are too small of a sample to account for a thousand different variables that better describe the population at large.