Book 2 - Page 72

Book 2 - Page 72
Comic - Book 2  Page 72
Recent posts... (See full thread)
Sorry to double post, but I think I need to deal with these separately

Kreistor wrote:
sleepymancer wrote:
2. My girlfriend is a research chemist who used to work in industry developing new formulations. A few years ago we decided to re-construct some medieval-style inks from eleventh-century recipes. My interests were in how did the recipes work, or indeed, did they work, what was the ink like to write with, and so forth. She read one of the recipes, frowned and said 'this won't work. We will have to add [x] to make it functional'. Her approach revolved around improving the formulation to make a better ink in the present, rather than investigating the historical one.


Is that what you think I meant by "re-engineering the past" with a view from modern engineering? No, it was not.

What I meant was, because historical records are often not contemporary, the described events can be blown out of proportion. The Archimedes Death Ray is one of those. The text describing it was written more than 100 years after the defense of that port, and nothing contemporary wiht the battle describes such a weapon used any other time. In trying to reconstruct the weapon, we can begin with modern materials and methods, because if we fail with far superior materials, we would certainly fail with period materials. You can work backwards from modern methods to historic methods, in order to make the effort both cheaper and easier. The ultimate goal is a period accurate device, but we don't have to waste our time with period accurate problems (ie. bad sealing methods, inadequate oils, weak metals) if it doesn't work with better materials.


I actually meant that piece regarding my better half as an illustration regarding disciplines and people on a slightly different thread in the conversation, rather than an illustration of what I thought your methodology was. Having read it in that light, however, I can see why you would be offended, and I think it comes across as worst-possible practice from your methodology or mine, (which is why I let that project slip quietly away, lol )!!!

As a person who works with texts, my interest is far more often on, what was trying to be done with the text itself rather than the events it purports to describe. I can see the benefits of what you are doing with your methodology, I guess its what you'd call a teleological approach beginning with the end-product and re-building it that way. It sounds both sensible and fun. However, I can also imagine the response you get from the academic community (and imagine I came across entirely as a proponent of that way...)
Kriestor" wrote:
It's Occam's Razor, of course. We've spoken at length about it, and you described the process wihtout stating the common name for it. And I mena the original definition, "The answer that introduces the minimum number of unknown entities is the most likely solution." rather than the more modern misinterpretation that "The simplest answer is the most likely solution."


The modern statement, presented properly in a Bayesian context, would be that more complex solutions have a lower prior probability, and so must have correspondingly greater explanatory power to be as likely as a simpler theory. Let's see how this applies in this case:

Kreistor wrote:
drachefly wrote:
Additional information, as I said! I mean this in a technical sense. How many bits of information are required to specify the addition to the theory? Each bit comes with a prior probability cost.


That's the problem... for most history, there is no new information forthcoming. The library of Alexandria burned and took enormous amounts of history with it, leaving us somewhat blind. We aren't going to get too many Dead Sea Scroll finds. So, we're stuck with what we've got.


Your response is based on a misreading: I clearly stated that this additional information is the information used to specify the addition to the theory, NOT data of evidence. To clarify, as it seems you really don't understand this: You have a theory here. How many bits of information are required to state the theory, with the evidence in hand?

Let's get a bit more concrete and abstract at the same time. You have this set of 12 data points on a graph, and you're trying to perform a blind curvefit. There are several candidate models - a straight line fit through the origin (1 parameter), a power law (2 parameters), and a power law plus an offset in X (3 parameters).
The chi-squared is the sum of the squared number of standard deviations off of the model that each data point is. You expect the chi-squared figure for a good fit to be around the number of data points minus the number of fitting parameters (for data sets much larger than the number of parameters).
The chi-squared figures for these fits are:
Ax (1 parameter): 13.6
Ax^P (2 parameters): 10.2
A(x-X)^P (3 parameters): 9.6

So, which fits most tightly? The 3 parameter fit - it had the most knobs to turn, and it was in a strict hierarchy with the others - it couldn't do worse!

But which of these is best? Each parameter is only said to have explanatory power to the degree by which it decreases the chi-squared. If it doesn't decrease the chi-squared by more than 1, it's not paying for itself in terms of explanatory power, and if it only decreases it by around 1 the model is barely improving on the state of knowledge before. The smaller the gain, the more trivial the effect explained.

Adding the power law let the chi-squared drop by 3.4, a big win. Adding in an X offset only dropped it by 0.6 - the X-offset did not provide as much explanatory power. And, worse, the number of points is getting to be a noticeable fraction of the total data set size - overfitting alert! If the effect is real but the small increase is due to the small number of points, then if we can take more data, we'd expect that improvement in chi-squared to increase linearly with the number of points until it was larger than 1.

Another example would be you have a long historic curve, then after a few parameters, the best additions to the model will be bumps 'explaining' individual events. They can still pay for themselves, especially if there's a rational basis for their inclusion. They can lead to better understanding.

Now, back in history. We have three models here: a model that doesn't mention footwear in any way, and one each that affirms the presence or absence of footwear. Even if each fits the data to the same degree, the one that doesn't mention footwear has less information in it, and thus a smaller probability cost.

~~~~
* That's not the same thing as an answer revealing many new questions, like the discovery that stars are so distant that they must be of sizes comparable to the sun. That just revealed our prior ignorance; it didn't manufacture new ignorance.
Ummmm . . . I think you're arguing apples and crescent wrenches here . . . my original statement that the archers at Agincourt fought unshod was based upon a contemporaneous statement made by a (disgruntled) survivor of the battle on the losing side who felt that the English had behaved in a base and less-than-honorable manner which somehow constituted cheating. This has been also picked up upon by other historians, as I recall Keegan made reference to this in his book on the subject, as well as others who had noted that the English were more agile than their French opponents under the unique conditions of this particular battle.

Under normal circumstances, no, it would be foolish to remove your footgear or your greaves, and your officers would quite properly ream you for doing so. But when you're ankle / midcalf / knee deep in mud, then it makes a lot of sense. Especially if you're not all wrapped up in the notion of 'proper' behavior or 'winning', and are more concerned with 'getting the hell out of there alive'.

As for speculations, I've been fighting in armored combat for close to 30 years, though now that I'm getting old & feeble and the bones aren't knitting like they used to, I've switched over to light weapons and study late medieval / renaissance technique. Meyer, Talhoffer, Cappo Ferro, etc, and dabble a bit in Spanish School. Got pics if anyone is interested. So I know whereof I speak on this topic.

The group I'm with (the SCA) does extensive re-creations (with varying degrees of scholarship, of course) on a very wide range of subjects, cultures, and time periods (6th to 16th centuries). Though our re-creation of period heavy combat is limited by modern constraints (lawyers), what we do is extremely close to period Hastaludes and Tourney combat. If anyone would like to see it up close or (heh) try their hand at it, we have events and practices all over the country (world, actually) on a very regular basis. We usually have loaner gear and love tenderizing fresh meat . . . Or go to YouTube and search for 'SCA fighting', 'SCA combat', 'Pennsic battles', or for fencing 'SCA rapier combat' (for real fun, though, do a search for 'SCA rapier kill shots' . . . ).
That's awesome Sleggy! I was in the SCA during my teens and really enjoyed it. I've always wanted to go back, but am overwhelmed with other pursuits. I need a second self, at least.
drachefly wrote:
Kriestor" wrote:
It's Occam's Razor, of course. We've spoken at length about it, and you described the process wihtout stating the common name for it. And I mena the original definition, "The answer that introduces the minimum number of unknown entities is the most likely solution." rather than the more modern misinterpretation that "The simplest answer is the most likely solution."


The modern statement, presented properly in a Bayesian context, would be that more complex solutions have a lower prior probability, and so must have correspondingly greater explanatory power to be as likely as a simpler theory. Let's see how this applies in this case:

(snipping lots of math)

Now, back in history. We have three models here: a model that doesn't mention footwear in any way, and one each that affirms the presence or absence of footwear. Even if each fits the data to the same degree, the one that doesn't mention footwear has less information in it, and thus a smaller probability cost.

~~~~
* That's not the same thing as an answer revealing many new questions, like the discovery that stars are so distant that they must be of sizes comparable to the sun. That just revealed our prior ignorance; it didn't manufacture new ignorance.


Trust the quantum physicist to frame it all in the terms of what MIGHT be true. You guys are never really sure about anything these days huh...

(Nothing but love, Drache...)
Housellama wrote:
Trust the quantum physicist to frame it all in the terms of what MIGHT be true. You guys are never really sure about anything these days huh...


:lol: On a somewhat related note, I always imagine neutrinos as wearing Arlecchino (aka Harlequin) suits, for some reason. Those jokers!

My bet'd be on an error in the electronics btw, but really that's all I can say. That whole thing is way out of what I'm vaguely knowledgeable about.
HeineeBreadia wrote:
Please Ban Me! Please Ban Me!

Please Ban Me

Please Ban Me Please Ban Me Please Ban Me Please Ban Me


I assume, then, that you are taking the role of King of France for the battle, and rather than raising the arriere ban yourself, are relying on your nobles (the moderators? us?) to do it for you instead???

:lol:
Sieggy wrote:
Ummmm . . . I think you're arguing apples and crescent wrenches here . . . my original statement that the archers at Agincourt fought unshod was based upon a contemporaneous statement made by a (disgruntled) survivor of the battle on the losing side who felt that the English had behaved in a base and less-than-honorable manner which somehow constituted cheating. This has been also picked up upon by other historians, as I recall Keegan made reference to this in his book on the subject, as well as others who had noted that the English were more agile than their French opponents under the unique conditions of this particular battle.


This I can respond to.

The enemy seeing the Archer unshod do not know the reason they were unshod. They can believe it was intentional, but that may not be true.

Henry marched his army 360 miles in the days leading up to Agincourt, and the timing of this was at the end of the campaign season, when equipment had already begun breaking down.

They were unshod because they wore out their boots. Shoes with gaping holes in them fill with mud and become hindrances, so if your shoes are worn out, you take them off,

As I said, experiments in the mud show good leather boots do not sink kneedeep in Agincourt mud, so they didn't remove them for that. But there may be other perfectly good reasons why some were without shoes.
Sieggy wrote:
Ummmm . . . I think you're arguing apples and crescent wrenches here . . . my original statement that the archers at Agincourt fought unshod was based upon a contemporaneous statement…


That's why I emphasized 'even' towards the end. That way it layers on top of your argument. Of course adding evidence can tip the scales one way or another.

HouseLlama wrote:
Trust the quantum physicist to frame it all in the terms of what MIGHT be true. You guys are never really sure about anything these days huh...


This isn't exactly new - the math has been around for around 250 years. I guess I've been primed on it since I'm a quantitative scientist, but it's nothing to do with quantum mechanics. It just has to do with not being omniscient, something everyone has experience with.
drachefly wrote:
This isn't exactly new - the math has been around for around 250 years.


Indeed. Here we are though, still reading Erfworld in the 22nd century ;)