Blade test of nine different steels, M4 wins .

[tvenuto]

Jack

This is just my thoughts on it, but I think you're getting a little lost on the specifics of the topic (steel) while Cliff's point is more general. It's less about understanding the nuances of metallurgy, than it is about understanding testing and statistics. I'll throw out a completely made up example, which may or may not be a great analogy.

Let's say you and your friend were arguing about baseball. You claimed that it was harder to hit home runs out of Yankee stadium than it was out of Camden yards. To settle it, you look up the last game at each stadium. Lo and behold, there were 3 home runs in Camden yards, and only 1 at Yankee stadium. So you win, right? No, this one game doesn't tell you anything about how hard it is to hit a home run in each stadium. This is because the other factors such as game situation, skill of the batters, skill of the pitchers etc, are far more important for one game than the stadium shape and size (which are relatively close to each other). That is, the effect of the stadium is being hidden by the stronger effects mentioned. The only way to tease out the effects of the stadium is to build up a large body of games so that every team has played there in every situation and you get enough data so that those other factors average out.

To bring this home, you can look at steel composition like the stadium shape and size. It has a relatively minor affect on edge retention, while other factors that are hard to control are having larger affects. Thus, you need to do your best to control them, and you need a lot of data so that they average out.

This stuff is difficult to comprehend, and most people make the above error constantly in their daily lives. A great book on the topic is called The Drunkard's Walk: How Randomness Rules Our Lives. I actually envy Cliff for working with something so simple as steel. I'm concerned with human performance, and I'm sure you can imagine the random factors that give us fits when trying to say anything about the data we collect on a particular athlete (or small group of athletes).

[DougC-3]

I expect a truly skilled swordsman could easily slash harder or slack up a bit to insure that his consistency remained intact :)

At least the bodies they used may have been armed at some point, unlike those in the Blade magazine test :)

After a squirrely start this has turned out to be a pretty good thread :D

[MCM]

"I thought all knife afficianados subscribed to Blade magazine"
Thanks for the chuckle.

[Evil D]

I used to be a diehard magazine guy but I think the internet and forums have made the format obsolete. I also trust the opinions of some people I know on here more than any magazine type people, besides folks like Mike J. who are both

[RadioactiveSpyder]

Yeah, what are these "magazines" you speak of?

I'm going to go make a wash cloth-based squirrel now for testing my Spydies...

[DougC-3]

A group of samurai photographed between 1860 and 1880, apparently not in a very good humor that day...

[jackknifeh]

A group of samurai photographed between 1860 and 1880, apparently not in a very good humor that day...

Actually, I think this is their "HAPPY FACE". This is my "HAPPY FACE". :D


Jack

[Evil D]

Yeah those guys ARE smiling. You don't wanna see their angry face.

[DougC-3]

haha, if I met them on a mountain trail, my happiest face would look something like this ---> :eek:

[GCG199]

Yes, for the most part, forums and the various social networks have replaced knife magazines for me too.

I still subscribe to Blade because it is fairly good for an overall view of the knife industry.

It only costs $20 a year for the digital subscription too.

I wouldn't want to see it go out of business like what has recently happened to Tactical Knives.

[Cliff Stamp]

I actually envy Cliff for working with something so simple as steel.

That's not work, that is just a hobby, my actual background is in nonlinear, highly correlated systems with dozens to hundreds of highly correlated parameters. However, they were still physical systems, if you are collecting data on people, either performance or especially behavior based, you have to use completely different methods to take into account non gaussian scatter that people put into the equations. If you are doing that, respect to you.

Nice example.

I agree. I used to wish you would give a bullet statement to a question.

Jack, many years ago a number of people asked for a short summary at the end of posts/commentaries so I did them. Then a few people started quoting them without context and it generated misinformation, I still see those references.

For example recently I did a comparison on 1095 blades from 58 to 66 HRC :

- http://www.spyderco.com/forums/showthre ... e-project)

Now if you just take the very last point, that the edge retention increase by about 8% per HRC point, without understanding what was done and how that can influence the results then it is very likely that what you would walk away with would be incorrect.

Aside from context, there is also the issue of terminology. I can explain things using just lay speech, the problem is that the explanation would take a tremendous amount of time if instead of saying aggregate, carbide, grain size, etc. I had to physically had to explain / write out all of those terms.

If you want to understand a topic you do have to learn a certain amount of the lingo, if you want to understand physical systems you have to learn a certain amount of empirical methodology. How much you need to learn just depends on what you want to know.

For example if you wanted to determine if 440C had higher corrosion resistance than 1095 then the methods required are extremely basic because the difference in performance is so very large. It doesn't take a lot of statistics to figure that out out.

However if someone wants to understand why SuperBlue is different than 1095 or wants to do an experiment to measure the differences then you have to use a decent amount of statistics because the performances are so close. If you don't take proper controls you will get misleading results.

If you do that experiment and you want to talk about it then you have two choices :

-just say what happened

-describe what you did and why

The first one is much simpler but it doesn't get the audience any ability to review your work they have no rational reason for accepting it. Now how much detail you give is provided based on feedback. If you watch my posts you will see I will escalate detail based on questions/interactions.

The lack of even basic controls is why there is so much contradiction in the knife industry because the vast majority of conclusions are essentially just formed at random.

Now don't walk away from this thinking you need a graduate degree in statistics, the more math you put into it the better conclusions you will get, however if you just do two simple things :

-repeat any experiment at least three times
-have some measure of blinding

then your conclusions will be far more likely to be correct.

For example if the deer guy simple did this :

-skin three deer with a knife
-have the sharpness checked on a simple scale of 1-4 by someone who wasn't part of the skinning

repeat this until all deer were skinned for that year.

Now he would have a very small table with a list of steels and basic descriptors which could have been similar to :

Benchmade S30V : sharp, dull, sharp

Spyderco ZDP-189 : very sharp, sharp, very sharp

Gerber 420J2 : very dull, dull, dull

(1, very dull; 2, dull; 3, sharp; 4, very sharp)

Now you would have some interesting results and you could draw some tentative conclusions.

If you wanted to step up the experiment and make it a bit more robust :

-measure the sharpness and get a number

and moving even further you can look at issues of edge angles, handle ergonomics, etc. .

However often times these articles are not meant for actual information but entertainment, and the people who write them are often very skilled at that.