.

[YoYa]

Good to see you back chodpaba, It's good to see a thread where I'll be learning something new.

[||bit]

edit above.

Do you know where to find the data on how many wallet users were there were when it peaked about $32 in 2011?

[||bit]

edit above.

Do you know where to find the data on how many wallet users were there were when it peaked about $32 in 2011?

Maybe this? http://blockchain.info/charts/n-unique-addresses

That only goes back to April 2012.

[||bit]

edit above.

Do you know where to find the data on how many wallet users were there were when it peaked about $32 in 2011?

Maybe this? http://blockchain.info/charts/n-unique-addresses

That only goes back to April 2012.

Look at the bottom of the page. There is a link for "All Time"

Thanks. I was interested in that for reasons that might suggest a cause to any coincidence with your plotted ratio and price converging to a point.

On another thought. Not sure if it means anything. I noticed the wallets immediately spiked by a factor of about two to two and half times around the 2011 peak price. And when the bitcoin price was basing at about $10, prior to this recent uptrend, there were around 40,000 wallets. We are now about two time that value now. Maybe there's a possible clue to a near peak in that too?

For what it's worth. Based on simple technical reasons, I was already anticipating an at least short term top near around $181 to $189. This is one reason your thread caught my attention. My reasoning, however, has nothing to do with money flow. Just prior movements up to this point.

I've also guessed that if we will see at least a scaled up spike of 2011. The 2011 run was from $1 base to $32 peak. We will need to see it run or spike to $320 from our recent basing at ~$10 (mathematically convenient numbers )

Our recent baseline at $10 also occurred with an average about 40 times more wallets than the 2011 base. Does that mean 40 times more interested bitcoin buyers as 2011 base time? Interesting coincidence that 40 x the $32 prior peak = $1280 is very near your next peak at $1274.

[||bit]

edit above.

Do you know where to find the data on how many wallet users were there were when it peaked about $32 in 2011?

Maybe this? http://blockchain.info/charts/n-unique-addresses

That only goes back to April 2012.

Look at the bottom of the page. There is a link for "All Time"

Thanks. I was interested in that for reasons that might suggest a cause to any coincidence with your plotted ratio and price converging to a point.

On another thought. Not sure if it means anything. I noticed the wallets immediately spiked by a factor of about two to two and half times around the 2011 peak price. And when we bitcoin price was basing at about $10 prior to this bull run, we had close to 40,000 wallets. We are now about two time that value now. Maybe there's a possible clue in that too?

For what it's worth. Based on simple technical reasons, I was already anticipating an at least short term top near around $181 to $189. This is one reason your thread caught my attention. My reasoning, however, has nothing to do with money flow. Just prior movements up to this point.

I've also guessed that if we will see at least a scaled up spike of 2011. The 2011 run was from $1 base to $32 peak. We will need to see it run or spike to $320 from our recent basing at ~$10 (mathematically convenient numbers )

Our recent baseline at $10 also occurred with an average about 40 times more wallets than the 2011 base. Does that mean 40 times more interested bitcoin buyers as 2011 base time? Interesting coincidence that 40 x the $32 prior peak = $1280. Very near your next peak at $1274.

Well, it's not wallets really, but new Bitcoin addresses. As people decide to move their coins to/from the exchanges they create new addresses.

Oh, your right! I wonder if that could still be a roughly proportionate to unique wallet users. And unfortunately, the data for the number of my wallet users doesn't seem to start accumulating until near the end of 2011.

[||bit]

As of the last complete candle, for the same calibration, the limit is calculated at a about $176.

I have not tried to recalibrate. Just waiting to see what happens.

I've been thinking about this, and still can't think of any reason the number would or should converge. Well, at least your testing it. My hope is almost that it fails the test, simply because if it works, it will drive me crazy trying to figure it out despite the obvious potential benefit.  

Can you describe better how you calc/draw the ratio line?

Is this it:

initial raw data set:
{hour bar #1: O,H,L,C, Volume;
  hour bar #2: O,H,L,C, volume;
  hour bar #3: O,H,L,C, volume;
  hour bar #4: O,H,L,C, volume;
   ...etc...}

process the initial data to make this data set:
{hour bar #1: average(H+L+C)*volume;
  hour bar #2: average(H+L+C)*volume;
  hour bar #3: average(H+L+C)*volume;
  hour bar #4: average(H+L+C)*volume;
 ...etc...}

using that processed data set, is this then the final data set?:
{|(#1 - #2) divided by (#1 + #2)|;
  |(#2 - #3) divided by (#2 + #3)|;
  |(#3 - #4) divided by (#3 + #4)|;
  |(#4 - #5) divided by (#4 + #5)|;
 ...etc...}
where |term| properly => absolute value (term)

[||bit]

As of the last complete candle, for the same calibration, the limit is calculated at a about $176.

I have not tried to recalibrate. Just waiting to see what happens.

I've been thinking about this, and still can't think of any reason the number would or should converge. Well, at least your testing it. My hope is almost that it fails the test, simply because if it works, it will drive me crazy trying to figure it out despite the obvious practical benefit. haha

Can you describe better how you calc/draw the ratio line?

Is this it:

initial raw data set:
{hour bar #1: O,H,L,C, Volume;
  hour bar #2: O,H,L,C, volume;
  hour bar #3: O,H,L,C, volume;
  hour bar #4: O,H,L,C, volume;
   ...etc...}

process the initial data to make this data set:
{hour bar #1: average(H+L+C)*volume;
  hour bar #2: average(H+L+C)*volume;
  hour bar #3: average(H+L+C)*volume;
  hour bar #4: average(H+L+C)*volume;
 ...etc...}

is this the final data set?:
{|(#1 - #2) divided by (#1 + #2)|;
  |(#2 - #3) divided by (#2 + #3)|;
  |(#3 - #4) divided by (#3 + #4)|;
  |(#4 - #5) divided by (#4 + #5)|;
 ...etc...}
where |term| properly => absolute value (term)

The first thing you have to do is to redimension the time series. I am starting with hourly candles just because it is convenient for computing budget, but you could use anything on up to tick data if you wanted to. Group that data into candles of a constant quantile of BTC volume, the exact quantile you use to calibrate the model will depend on the resolution of the data you start with.

Then, keep a running total of $volume from each of your redimensioned candles and divide that sum by the $volume of each individual candle. Then adjust the BTC volume quantile until low spots on the curve are approximately equal to the price peaks. About as close as I could get to this calibration with hourly data was with a 98.24% quantile. Raw data of higher resolution will enable you to get a more accurate calibration.

I see - moreso.

But since you intentionally modified the the volume quantile sizes to match the ratio minimum with price peak, why do you think the next major price peak should happen when it converges again with the ratio plot? Price could just blow past it. Or are you just being curious that it might.

[||bit]

I am always trying out different priors for forecasting. And when I hit upon this it just seemed so odd that the peaks would tend to line up like that. Sure, I calibrated it, but the odd thing is, regardless of the calibration that the current peak and the $32 peak would approach the money flow ratio curve at about the same ratio. Perhaps it will seem less odd if there is no close confirmation, it just seemed so crazy I had to check it out.

Could you describe what you think the ratio might say in descriptive layman terms.

[phelix]

What about this simple interpretation: People trade less BTC when the price is high?

Do you have the same chart multiplied by price? That would be interesting to compare.

[oakpacific]

Welp, we have reached $190/BTC which is outside the range of error for my working hypothesis. Which tells me that past price bubbles do not really have much of a bearing on the current activity. The tail for the probability for price growth is becoming longer, moving into rotation territory.

Feeling like doing a update? I want to see what it looks like right now.

[Crypt_Current]

Somebody out there—replicate this.

[phelix]

this.

[Crazy]

Nice, thread looks like A+. Gotta come back and read it today.

[Crypt_Current]

definitely a visible divergence downward in the more brightly colored lines

[bb113]

Are you posting the updates for fun?  I only half understood everything, but I got the impression that you thought the model was broken as the price had trampled through your error bounds.

It only means that, instead of two calibration points, I may only have one. It remains to be seen what the error is. Even with the error, I can apply a posterior analysis, I just have not gotten that far.

Regardless, there is still an obvious covariance that I think can tell us something about price development.

What exactly are you calculating now? I am not sure but I think you may be detecting that during "bubbles" there is a divergence between the rate of change in USD volume and the USD volume volatility.

[TiagoTiago]

Does that "DDoS" on Gox yesterday, and the associated dive in the price, fit on your model in any way?

[TiagoTiago]

Is it this one? https://bitcointalk.org/index.php?topic=173313.0

[YoYa]

Hey chodpaba, you gotta admit this is an interesting one. Again, convergence correlated with decline but in this case with no fundamental reason other than market panic. To me this would suggest that your model at the least would suggest unsustainablity in a parabolic.

[rothschild_666]

Have you thought of normalizing the values?  Say, per $1000?

[Adrian-x]

Have you thought of normalizing the values?  Say, per $1000?

I have. But normalizing to BTC volume seems to work better. It is simply the case that there is less variance in BTC volume than $volume, for obvious reasons.

Just from a lay persons perspective, a growth in Bitcoin's value  is equal to deflation in the BitCoin Economy, and that fact alone will cause people to save (less BTC activity), however when there is a large increase in Dollar value that will cause a greater imbalance (more BTC Activity) To me it seems that it is both that are important to model.