BETI: Bitcoin Exponential Trend Index and technical analysis

[segeln]

The last data point is yesterday, just not shown on the x-axis

With this graph, we may even say we are still in a big bull-phase starting from Mar 2013. The price is still >10x than 1 year before, anyway.

thanks again and continue with Graphs,please

[1714470909]

1714470909

[1714470909]

1714470909

[1714470909]

1714470909

[jl2012]

We are about to hit the long-term trend line. First time since last October

[podyx]

long term trendline is valued at about $680 currently right?

[jl2012]

y-axis is ln(price)
Blue line is the daily VWAP
Red line is the expected price of the day. For each day, a regression is fitted with all data of and before that day, so it is not a straight line.
Green line is the current regression line

Yesterday we broke the trend line but the VWAP was still a bit over it. We were only 4 days ahead of the trend

Date: 25 Feb 2014
VWAP = 480.66
x = 1319
a = 0.006044
b = -1.82447
Rsq = 0.874307
Today's expected price = 467.8862
Predicted date for today's price = 1 Mar 2014
Days ahead = 4
Daily price rank = 100
Predicted date for ATH ($1126) = 25 Jul 2014

(See OP for explanation)

[oda.krell]

First thing I want to say first: thanks for the diligent work. Seriously.

Second thing: Care to make the (R, I guess) script available so others can play with it? Wouldn't be unusual to tie it to a btc donation though, btw.


Anyway... what I really wanted to say: I've said several times in different threads that also use log-linear model for price analysis that I'm generally critical about using this method, at least if the goal is active trading. The problem is of course how actual price relates to the price predicted by the regression -- in the sense that any actual trading signals are very imprecise, and can be off by an order of magnitude if interpreted naively.

(sorry if I'm not getting my point across concisely, big flaw of mine) What I mean is this: most of the time when I've seen people posting regression analysis, they take price going *above* the regression prediction as a sell signal ("We're above the target"), and price going below as a buy signal. If they would have traded like that, they would have ended up at a huge loss compared to buy&hold (e.g. they would have had to re-buy much higher in 2013, after selling way too early). Note please that I'm not saying you're giving that advice, but my point is: regression analysis has to be interpreted somehow, and it is not obvious how to interpret it, in my opinion.

Here is one preliminary observation, or attempt at a slightly less "naive" trading model based on regression: Except for 2011, we tend to stay below the loglinear trendline much shorter than above. So selling if we go *above* the line is probably less warranted than buying when we go *below*. Very crude first attempt, I know, but I'd like to get the discussion started.

EDIT: hahaha, just noticed, you're a member since 2012. Probably don't feel the need to ask for mBTC donations :D

[jl2012]

First thing I want to say first: thanks for the diligent work. Seriously.

Second thing: Care to make the (R, I guess) script available so others can play with it? Wouldn't be unusual to tie it to a btc donation though, btw.


Anyway... what I really wanted to say: I've said several times in different threads that also use log-linear model for price analysis that I'm generally critical about using this method, at least if the goal is active trading. The problem is of course how actual price relates to the price predicted by the regression -- in the sense that any actual trading signals are very imprecise, and can be off by an order of magnitude if interpreted naively.

(sorry if I'm not getting my point across concisely, big flaw of mine) What I mean is this: most of the time when I've seen people posting regression analysis, they take price going *above* the regression prediction as a sell signal ("We're above the target"), and price going below as a buy signal. If they would have traded like that, they would have ended up at a huge loss compared to buy&hold (e.g. they would have had to re-buy much higher in 2013, after selling way too early). Note please that I'm not saying you're giving that advice, but my point is: regression analysis has to be interpreted somehow, and it is not obvious how to interpret it, in my opinion.

Here is one preliminary observation, or attempt at a slightly less "naive" trading model based on regression: Except for 2011, we tend to stay below the loglinear trendline much shorter than above. So selling if we go *above* the line is probably less warranted than buying when we go *below*. Very crude first attempt, I know, but I'd like to get the discussion started.

EDIT: hahaha, just noticed, you're a member since 2012. Probably don't feel the need to ask for mBTC donations

That's simply Excel.

I tried different models: square root-linear, log-quadratic, square root-quadratic. The log-linear is the best fit model and with Rsq = 0.874307 is really amazing

I do not try to give trading recommendation with the graph. However, go above some SD above the trend line could be a sell signal

p.s. Someone here did give me 1000 mBTC donation in 2012. But smaller amounts are always appreciated.

[oda.krell]

First thing I want to say first: thanks for the diligent work. Seriously.

Second thing: Care to make the (R, I guess) script available so others can play with it? Wouldn't be unusual to tie it to a btc donation though, btw.


Anyway... what I really wanted to say: I've said several times in different threads that also use log-linear model for price analysis that I'm generally critical about using this method, at least if the goal is active trading. The problem is of course how actual price relates to the price predicted by the regression -- in the sense that any actual trading signals are very imprecise, and can be off by an order of magnitude if interpreted naively.

(sorry if I'm not getting my point across concisely, big flaw of mine) What I mean is this: most of the time when I've seen people posting regression analysis, they take price going *above* the regression prediction as a sell signal ("We're above the target"), and price going below as a buy signal. If they would have traded like that, they would have ended up at a huge loss compared to buy&hold (e.g. they would have had to re-buy much higher in 2013, after selling way too early). Note please that I'm not saying you're giving that advice, but my point is: regression analysis has to be interpreted somehow, and it is not obvious how to interpret it, in my opinion.

Here is one preliminary observation, or attempt at a slightly less "naive" trading model based on regression: Except for 2011, we tend to stay below the loglinear trendline much shorter than above. So selling if we go *above* the line is probably less warranted than buying when we go *below*. Very crude first attempt, I know, but I'd like to get the discussion started.

EDIT: hahaha, just noticed, you're a member since 2012. Probably don't feel the need to ask for mBTC donations

That's simply Excel.

I tried different models: square root-linear, log-quadratic, square root-quadratic. The log-linear is the best fit model and with Rsq = 0.874307 is really amazing

I do not try to give trading recommendation with the graph. However, go above some SD above the trend line could be a sell signal

p.s. Someone here did give me 1000 mBTC donation in 2012. But smaller amounts are always appreciated.

Sure, I know you're not giving trading advice. I was more thinking out loud what my own thoughts are on the topic. I've played with loglinear models myself, can see how they seem to fit impressively, but at the same time, it's not at all obvious what to *do* with them. I'm still working on it

I'm going to make a donation then. Maybe not 1 btc, I'm not that loaded. But I really appreciate your work. That said, could you perhaps run me through how to get the results? Not the single loglinear trendline, I can do that myself (in R). But I have no idea how you got the "running" calculation, i.e. the red trendline. Some 'for' loop, that works on an increasingly bigger subset of the price data? Maybe wait with the explanation until my transaction gets through

[jl2012]

This is the residual plot, i.e., the difference between the Series 1 and Series 2 of my last chart. It seems a bubble is likely to pop when the price is e^1.5 = 4.48 times over the trend line.

For the first bubble, one would have sold at $17.16 on 04/06/2011, 5 days before the bubble pop but only 53% of the bubble top

For the second bubble, one would have sold at $211.97 on 09/04/2013, just the day before the bubble pop with 79.7% of the bubble top

For the latest bubble, one would have sold at $976 on 28/11/2013, 7 days before the bubble pop with 84.6% of the bubble top

So the advice could be to start selling when it is over by 4.48 times. Sell more when it goes higher.

For buying, it seems any pirce below or near the trend line would be okay, except for the bear market after first bubble

[jl2012]

Please PM me if you want a copy of the Excel spreadsheet. Also many thanks to the one who donated 0.05XBT to me

[seleme]

Is there any place where I could check trend line and where it is updated automatically?

[jl2012]

Is there any place where I could check trend line and where it is updated automatically?

I will update once a few days but not automatic. As this is a long-term trend, you don't really need a daily update. The trend line is growing by about 0.6062%, or $3 per day at current price.

[seleme]

Is there any place where I could check trend line and where it is updated automatically?

I will update once a few days but not automatic. As this is a long-term trend, you don't really need a daily update. The trend line is growing by about 0.6062%, or $3 per day at current price.

ok, thanks

[hlynur]

following & thanks for your efforts (also thanks for your SecondMarket Observer thread)
donation to follow during the next days.

[seleme]

mccoyspace made an autoupdating link for rpietilla's trendline. We're on 742$ today on his one. Could you use wolframalhpa to make a link like this for your trend?

Post - https://bitcointalk.org/index.php?topic=322058.msg5391481#msg5391481

Link for trendline price -

https://www.wolframalpha.com/input/?i=10+%5E+%28+%28+3073%2F1000000+%29+*+%28+number+of+days+since+jan+3%2C+2009%2F+days+%29+-+2.909514+%29

[jl2012]

mccoyspace made an autoupdating link for rpietilla's trendline. We're on 742$ today on his one. Could you use wolframalhpa to make a link like this for your trend?

Post - https://bitcointalk.org/index.php?topic=322058.msg5391481#msg5391481

Link for trendline price -

https://www.wolframalpha.com/input/?i=10+%5E+%28+%28+3073%2F1000000+%29+*+%28+number+of+days+since+jan+3%2C+2009%2F+days+%29+-+2.909514+%29

It includes price before MtGox opened, which I don't think the data is reliable nor practically meaningful

[seleme]

mccoyspace made an autoupdating link for rpietilla's trendline. We're on 742$ today on his one. Could you use wolframalhpa to make a link like this for your trend?

Post - https://bitcointalk.org/index.php?topic=322058.msg5391481#msg5391481

Link for trendline price -

https://www.wolframalpha.com/input/?i=10+%5E+%28+%28+3073%2F1000000+%29+*+%28+number+of+days+since+jan+3%2C+2009%2F+days+%29+-+2.909514+%29

It includes price before MtGox opened, which I don't think the data is reliable nor practically meaningful

Could you use your formula on Wolfram so we could have an autoupdated link for yours?

[theonewhowaskazu]

Shit its growing by $3 per day and is $150 ahead of us already?

Not sure whether to be happy or sad about that

[seleme]

Shit its growing by $3 per day and is $150 ahead of us already?

Not sure whether to be happy or sad about that

At rpietilla's is 150 ahead, at OP's is some 100$ below.

[theonewhowaskazu]

Shit its growing by $3 per day and is $150 ahead of us already?

Not sure whether to be happy or sad about that

At rpietilla's is 150 ahead, at OP's is some 100$ below.

Ok then we're right smack in the middle. Maybe that indicates the current trading range

[jl2012]

Shit its growing by $3 per day and is $150 ahead of us already?

Not sure whether to be happy or sad about that

At rpietilla's is 150 ahead, at OP's is some 100$ below.

Wolfam link: https://www.wolframalpha.com/input/?i=e+%5E+%28+0.006044++%28+number+of+days+since+jul+17%2C+2010+%2Fdays+%29+-1.82447+%29