[from stockcharts.com]

It's a way to gauge the money flow over some time span (20 periods in the formula, I think 21 in my chart, period is 2h in my case), which means the formula takes into account the (relative) *change of the price* and the corresponding *volume*.

I'm not an expert in reading these, but the way it looks to me is that, on mtgox at least, money flow has been overall positive for the past week(s), and is now recovering from a dip a few days ago (that didn't go into the negative, but was a sharp drop).

Bitstamp on the other hand looks less good, money flow is negative or near zero, for about a week now, and no real sign of reversal yet. However, historically mtgox has led and bitstamp has followed, so it's possible that this will be the case again now.

Forget trend lines. They're bullshit anyway :3

This is what make me cautiously optimistic, for the coming days at least:



Then again, there's this:

Hey, it was your choice to enter academia, now you gotta deal with peer review.

Valid point. I forgot that trading started below 1 USD.

Anyway, to make it a bit more interesting for the general population of this thread, summary time!

(1) sarc's exact modelling techniques aside, the graph he posted is not entirely different from similar results that have been posted before. sarc's regression analysis sees the price below 30 USD right now (correct me I'm wrong, please), while the others had us in the realm of 50 USD (does anybody have a link to one of those?).

(2) regression analysis is a well-founded statistical technique, but by no means guaranteed to accurately predict the future price.

(2) Math is hard. Let's go shopping.

I know close to nothing about EW analysis. In any case, I don't think we'll fall back to 50, and certainly not sub-30. I can see us going back to 60-80, though, I admit :/

Do you really mean:

log(x+1)

as in

log_(n)(x+1), where x = USD/btc, and n is the base you chose? If so, maybe that's a technique used in statistical modeling, but it certainly is not correct per se.

As I wrote in my 'EDIT' already, there is no disagreement between us re: correct use of log, only a confusion of notation (more due to my part, I admit, my way of writing it was highly misleading)

log_(n)(x) and log_(n)(x+1) are of course not the same. If sarc stated that, he was wrong. Funny enough, for his results it shouldn't really matter much, the difference will be minimal I think.

Correction to my own post above:

30 USD is (probably, I didn't calculate it myself) the result of linear regression and, as I understood sarc, removing some outlier data (i.e. the bubbles).

The other charts I've seen that use the entire data have as at around 50 USD, IIRC.

I never said you should add a constant to the value. I was talking about the equivalence of mapping to a log n and log n+1 chart. And that part is certainly true.

EDIT: I see the problem now. Damn sloppy notation :) you're talking about 'x', and I'm talking about base 'n' in log_(n)(x).

My point was, since the entire discussion between the two was sort of messy and unstructured, I'm not entirely sure what the claims of each party are, and I though at some point the two were arguing whether the above matters.

Anyway, unless we start doing the mathematical part of this discussion in a more systematic fashion, I'm afraid we will keep misunderstanding each others' points.

Let one thing be said, though: whether sarc's graph or method was entirely accurate or not, the prediction "in the range of 30 USD" seems to be the one you get in fact, if you do a simple regression on the entire historical data.

This is not due to some massaging the data by sarc or a mistake in his method, but simply the result of that particular analysis. Whether that type of analysis is actually predictive of the future price is a different matter.

@prof7bit

Calm down, please. You're going as far as accusing sarc of posting false information on purpose, when all that goes on in reality is a disagreement about the math behind what was posted here.

My take on it (already mentioned above, shortened version to follow):

I don't know exactly what sarc did in his model, but if the dispute between you and him boils down to plotting the data on a chart with a log n y-axis vs. plotting the data on a chart with a log n+1 y-axis, then there is no dispute. The two are, for our purposes, equivalent.

If that's not the point of your dispute, I'm sorry that I missed it.

Have to agree with sarc on this one. Whether you use log n or log (n+1) for graphing, as long as you use it consistently it won't matter for the purpose this graph is used for: drawing a straight line to represent an exponential growth function. Depending on the log value you chose, that line (of best fit) will rise more or less sharply, but the actual *function* it depicts stays the same, since the line (of best fit) is based on the data plotted on the same log value.

EDIT: spelling

Yes, sorry, completely forgot about that :/

Couldn't you simply place the legend outside the box? Or make the font smaller, so that it takes up less space? On my screen, the legend covers almost half-way the horizontal dimension of the box. For comparison, in a single, bigger box like on clarkmoody it covers, maybe 5%, 10% of the box, on my screen.


Heh, I was wondering how popular your service was so far. I'll mention it a few more times on the speculation forum.

Why would I? I'm a mildly formal philosophy person, basically :P

EDIT: sorry for the off topic conversation. it's just that nothing happens right now.

It actually does. It's just that his historical data spans from the beginning of btc trading to now, while the log charts you see more often begin in January 2013.

Biology? Bioinformatics?

Last reply I'm gonna make on this topic. Promised.

Had you bothered to read my post history, you could have noticed that I *consistently* make the point that I don't believe in *the* trendline/growth function/etc governing: price = market behavior = human behavior.

Which is why I don't like it if you accuse me of selectively appllying that rigor only when it is convenient for me. I make this point all the time, when talking about bullish trendlines and bearish ones. Assuming that there is *the* magical growth function that holds once and forever is delusional, therefore any analysis that plots such a line needs to be taken with huge grains of salt. (Note: I'm not saying "should be completely dismissed")

eh, sorry. my field (well, of sort) is foundations of mathematics, so I can get worked up over not knowing the (formal) premises of a method one uses.

anyway, I'm not even completely denying your point. I'm extremely cautious right now, and far from ruling out price will drop further. But I still consider it a distinct possibility that the fundamentals and public perception have changed enough for growth to continue from (roughly) where we are now. Time will tell.

You're funny when you try too hard: I didn't suggest any method. I simply formulated the assumption made in sarc's analysis.

And "the result wouldn't differ much"? You're kidding, I hope. Simply (yet unrealistically) assuming that, say, growth was determined by one function up to January 2013, and another one following January 2013, would probably put us into the 500? 5000? (can't be arsed to calculate this now) range right now.

Which is obviously not where we are. Hence: unrealistic assumption as well, as noted above.

Dude, you either don't know the math behind what you do, or you're willfully obstinate to make a point: if you run a simple regression analysis like you did, you de facto work from the assumption that there is one and only function underlying those data points. Otherwise, you "identified" jack shit.