Mobil 1 5W-20

[Matt Whiting]

You can do controlled, high-precision tests on few parts, or take the
empirical route with many samples.

Testing/experimentation IS the empirical route! Look up the meaning of
empirical...

If it were me, I would test it on a fleet of 200 identical cars. 100 with,
100 without synthetic oil. After 100K miles, tear them all down and measure
all ID's and OD's. Average them up, and there you have a valid test. Even
with that many samples, you may not get a statistically significant
variation between oil types.

I'd do something similar, but I'd run at least 200K miles and preferably
longer. Almost anything will last 100K these days and I'm not even
intested in engines that won't go at least 200K!

You'd also have to put extensive data recorders on each car to find out
the driving conditions each experienced so you could try to normalize
the data.

This would be a great experiment. When do you plan to start it?


Matt

 

[Bob Adkins]

But you then have no idea what driving conditions each car is seeing,
unless you heavily instrument each car. Again, this wouldn't be cheap.
And you'd have to ensure that none of the rental customers ever added
a quart of oil as that would contaminate your test.

The closest I've seen to this was a test that Consumer Reports ran with
a fleet of taxis many years ago.

With a sample size of 200, all conditions average out.

I don't mind it for the most part, but as I approach 50 it is getting a
little less fun each year. Then again, there is nothing like sitting in
front of a wood fire with a cup of hot chocolate or coffee in hand,
reading a good book, and watching the big snow flakes come down. It
doesn't get much better than that!

Ya, I miss that part.

Here in Louisiana, I have no excuse for not being out working.

 

[Bob Adkins]

You'd also have to put extensive data recorders on each car to find out
the driving conditions each experienced so you could try to normalize
the data.

This would be a great experiment. When do you plan to start it?

No, I'm afraid my quality assurance days are over. And I'm glad of it!

 

[Bob Adkins]

Super Tech oil is packaged by Warren Oil. They do not add or delete anything
from the oil that they repackage. I've been using Super Tech synthetic in my
vehicle for several years with no problems, and I called them to find out
what kind it was. They said it's made by specialty oil which is Pennzoil, or
Quaker State - they are the same, just different bottles.
http://www.wd-wpp.com/index.html

Just type super tech in the product name box http://msds.walmartstores.com/

Ah! So I WAS right!

 

[Matt Whiting]

With a sample size of 200, all conditions average out.

Not necessarily. And it would be hard to cover the full range of
driving conditions encountered in the USA with only 100 cars with each
type of oil. However, let me know when you plan to start the test and
I'll drive one of the cars for you ... no charge!


Matt

 

[Brian Nystrom]

The oil MAY be identical, but it may not be. Even worse is that it may
vary widely from lot to lot as often the oil is whatever is available at
the lowest price at a given time. All crude oils aren't created equal.

It comes from ONE source and it's blended to a standard specification.
The lab test I saw indicated that it was comparable to other synthetic
oils on the market. Your comments are just idle speculation with no
basis in fact. Fear mongering doesn't help anyone.

 

[Brian Nystrom]

Ah! So I WAS right!

If you search on "synthetic", you'll see that both Specialty and Warren
are listed for SuperTech synthetic oils. I'm not sure if one is the
manufacturer and the other is the distributor or what, but the dates on
the MSDS's overlap.

 

[Brian Nystrom]

I do as I read the study. And I gave you a direct reference as to where
to obtain a copy if you are really interested in further educating
yourself. However, you seem happy using cheap oils and if you are happy
then that is all that matters, right?


Again, I showed you where to get the full article with the context and
assumptions they made, who made the tests, etc.

I checked the MCN archives and the test appears to have been done in
2000, which is before the latests API specs came out and before
Supertech Synthetic was available. While it might be worthwhile as a
comparative study, it doesn't tell us much about current products.

How long in time doesn't matter much, but I meant how long as in how
many miles driven.

I know that and that's what I meant when I said it doesn't matter. ANY
oil will protect your car long term if it's changed at recommended
intervals and you use a decent filter. I don't care how much or how
little you spend on oil, as long as you use an API certified oil, it
will do the job. API specs are very exacting and effectively limit the
amount of variation that's possible in oils. That's the whole point of
the certification.

Yes, I was making a hypothetical argument to show how the number of
miles driven is directly related to whether different wear rates matter.

If the argument was completely bogus, what's the point? Exaggeration
like that is deliberately misleading. I can make up all kinds of "what
if" scenarios too. For example, what if the actual difference in wear
rate was 0.1%, which is probably closer to the truth?

You had suggested that the amount of miles driven didn't matter, I was
showing that it matters greatly if the wear rates are different.

Only in your world of exaggerated wear rates.

I have no data to show if the wear rates are different.

No kidding.

And often engines
don't fail from wear per se, they fail from the rings getting stuck due
to varnish and carbon build-up, oil passages getting blocked with crud,
etc. I have seen enough engines torn down to know that synthetic oil
keeps an engine a LOT cleaner than dino oil.

That I can agree with.

Not identical, as the trainer engines are smaller, typically 200 cubic
inches whereas most other singles are 360 cubes or larger. However, the
engine designs are virtually identical within a family (Lycoming or
Continental).

Then it's an apples to oranges comparison and it's largely pointless.

Test data is an empirical result.

Sorry, I meant anecdotal.

You may wish to refresh your memory
on the meaning of empirical. I've never heard of impirical and don't
believe that to even be a word.

OK, it was a typo. So shoot me.

Fine, but we all know how unreliable anecdotal data can be.

 

[Brian Nystrom]

Actually, that isn't the basis of the scientific method, at least not
for sophisticated scientists. In many "real world" situations, this
simply isn't possible, yet much science is still accomplished. Look up
Taguchi for more information.

Fair enough, but there are limits to how far you can stretch this before
the results are meaningless. Comparing different engines under different
loads, then trying to draw correlations between continuous running and
frequent stops/starts seems pretty far-fetched. Results from such a test
could might indicate that a more definitive test may be worthwhile, but
in and of themselves they'd be largely meaningless.

 

[A. Sinan Unur]

Unfortunately, without knowing the scale of the bar grapha, a 2X
difference is meaningless. To make an analogy, an amplifier with .002%
total harmonic distortion has 2X as much as one with .001%, but
neither is audible. In practical terms, it makes no difference.
Perhaps the oil study is different, but we have no way of knowing
that.

The height of bars can also be deceiving due to the choice of origin: If
the vertical axis of the graph starts at 10 and goes to 15, the bar for 11
will be half the size of the bar for 12.

Not that I know anything about oil, but I do teach how to lie with
statistics.

Sinan

 

[A. Sinan Unur]

Testing two engines doesn't mean squat statistically. I don't know
what sample size you would need to ensure statistical significance,
but I know it is a lot more than one for each condition being tested.

If the engines are otherwise identical, you would not need a lot more
than, say, 15 tested with each kind of oil to detect a difference that
is practically as well as statistically significant.

On the other hand, if we are setting up a test of engine lifetimes on
different oils, the experiment may have to be run for a long time.

Sinan

 

[Brian Nystrom]

The height of bars can also be deceiving due to the choice of origin: If
the vertical axis of the graph starts at 10 and goes to 15, the bar for 11
will be half the size of the bar for 12.

Not that I know anything about oil, but I do teach how to lie with
statistics.

Sinan

Thanks for the clear example. That's what I was trying to get across.

 

[Matt Whiting]

It comes from ONE source and it's blended to a standard specification.
The lab test I saw indicated that it was comparable to other synthetic
oils on the market. Your comments are just idle speculation with no
basis in fact. Fear mongering doesn't help anyone.

Again, I ask to see the mysterious data you keep referring to, but can't
seem to produce a reference to. I provided a clear reference to the
source of the data that I saw that is in direct conflict with your
claims that all oils are created equal.

Matt

 

[Matt Whiting]

If you search on "synthetic", you'll see that both Specialty and Warren
are listed for SuperTech synthetic oils. I'm not sure if one is the
manufacturer and the other is the distributor or what, but the dates on
the MSDS's overlap.

And you accuse me of fear mongering for not wanting to use an oil where
the source can't even be determined?


Matt

 

[Matt Whiting]

I checked the MCN archives and the test appears to have been done in
2000, which is before the latests API specs came out and before
Supertech Synthetic was available. While it might be worthwhile as a
comparative study, it doesn't tell us much about current products.

2000 sounds about right. I knew it was a few years ago. I'd be happy
to have more current data, but I've been unable to find any.

I know that and that's what I meant when I said it doesn't matter. ANY
oil will protect your car long term if it's changed at recommended
intervals and you use a decent filter. I don't care how much or how
little you spend on oil, as long as you use an API certified oil, it
will do the job. API specs are very exacting and effectively limit the
amount of variation that's possible in oils. That's the whole point of
the certification.

No, the point of most certifications, and I believe API falls into this
category, is to provide MINIMUM standards. It doesn't prevent a
manufacturer from going ABOVE the standards and many manufacturers do
this. Sure, many will skirt just above the minimums, but the MCN test
showed that many of the reputable names, Mobil being one, have products
that are well above the minimum requirements. So, certification doesn't
limit variability, it just places a lower limit on the variability
range. The upper end is generally not limited by specification.

If the argument was completely bogus, what's the point? Exaggeration
like that is deliberately misleading. I can make up all kinds of "what
if" scenarios too. For example, what if the actual difference in wear
rate was 0.1%, which is probably closer to the truth?



Only in your world of exaggerated wear rates.

No, in any difference of wear rate. If the rate of wear is different AT
ALL, then the amount of wear between two oils will be completely
dependent on the mileage driven.

No kidding.

Yes, my data is just as good as yours claiming that Supertech is a good
oil.

That I can agree with.

So now you agree that better oils are better for your engine? I thought
you were saying that all oils were essentially equal and thus buying a
better oil was a waste of money.


Matt

 

[Matt Whiting]

Fair enough, but there are limits to how far you can stretch this before
the results are meaningless. Comparing different engines under different
loads, then trying to draw correlations between continuous running and
frequent stops/starts seems pretty far-fetched. Results from such a test
could might indicate that a more definitive test may be worthwhile, but
in and of themselves they'd be largely meaningless.

Yes, I agree that it is very difficult and that is the reason that I
believe it has never been done. A test worth doing would cost literally
multiple millions of dollars and just isn't worth it to anyone.


Matt

 

[Matt Whiting]

The height of bars can also be deceiving due to the choice of origin: If
the vertical axis of the graph starts at 10 and goes to 15, the bar for 11
will be half the size of the bar for 12.

Not that I know anything about oil, but I do teach how to lie with
statistics.

I saw the data. I own the "How to Lie with Statistics" book that was
required reading in my statistics class at Penn State a couple of
decades ago. The MCN data was extremely well done.

It is funny how some of you like to throw stones at data that you are
too cheap to spend a few bucks to get a copy of and actually see for
yourself. You must be a statistician...


Matt

 

[Matt Whiting]

Thanks for the clear example. That's what I was trying to get across.

He's not saying at all what you were saying. I'm surprised you can't
tell the difference. You are talking about a difference in the data and
whether that difference is of significance. He's talking simply about
the presentation of that data.

I now understand why you have such a hard time following my arguments.
If you can't tell this difference, then the concepts I'm explaining
won't be understandable either, so I'll stop wasting my time now.


Matt

 

[A. Sinan Unur]

A. Sinan Unur wrote: ....


I saw the data. I own the "How to Lie with Statistics" book that was
required reading in my statistics class at Penn State a couple of
decades ago. The MCN data was extremely well done.

It is funny how some of you like to throw stones at data that you are
too cheap to spend a few bucks to get a copy of and actually see for
yourself.

On the other hand, my comment was not specifically about the data, but
only about the fact that without knowing the scale of the vertical axis,
the heights of the bars in bar graph do not convey meaningful
information.

You must be a statistician...

Frankly, I do not know anything about oil, and I don't much care. As
such, I am unwilling to invest any time or money in researching the
article.

That should make it obvious that I am an economist who occasionally
teaches statistics.

Sinan

 

[Brian Nystrom]

He's not saying at all what you were saying. I'm surprised you can't
tell the difference. You are talking about a difference in the data and
whether that difference is of significance. He's talking simply about
the presentation of that data.

I now understand why you have such a hard time following my arguments.
If you can't tell this difference, then the concepts I'm explaining
won't be understandable either, so I'll stop wasting my time now.

Nice try, but Sinan points out exactly what I was saying before. A
difference of "2X" is meaningless without context. Depending on the
context it can be a big difference, or completely insignificant. Do you
know what the context of the data in the study you referred to is? For
that matter, do you even know what parameters were measured? Was it
frictional resistance? Viscocity vs. temperature? Levels of chemicals as
in the above analysis? Something else. Your claim of "twice as good" is
meaningless unless we also know "compared to what?"