How much is the electricity cost for the newest equipment, btw? 10%?
Depends, but USD 0.06/kWh (approx. Washington State?) and 0.9 J/GH (Spoondoolies/AM) suggest an electricity cost of USD 5.53 per XBT. |
|
|
|
How would you do variable size?
"Variable" in the sense that each forecaster would be permitted to choose the interval size s/he prefers. Example: (GT/E=greater than or equal to, LT=less than) Forecaster 1 | GT/E | LT | Prob | | 100 | 200 | 35% | | 200 | 400 | 50% | | 400 | 500 | 15% |
Forecaster 2 | GT/E | LT | Prob | | 100 | 700 | 50% | | 700 | 1500 | 50% |
When the outcome is ready, the forecasts can be ranked as that other smart guy suggested. Say the outcome is 450. The probability predicted for this by F1 is 0.15% (distributing 15% uniformly over an interval of 500-400=100), and the probability predicted by F2 is 0.08333% (distributing 50% uniformly over an interval of 700-100=600). F1 would win. The rest of the comments (base 10 or e, log or not, the scoring system, etc.) is really only a matter of taste. To each his own! Edit: Let me clarify: I suggested base e because of the formula for continuously compounding interest. It's a very useful choice in very, very many circumstances, although it may not be as easy on the eyes as base 10. |
|
|
|
Partnership? It was only one project and ASICminer was a customer. Now AC can sell their technology to anybody. No share from profit for ASICminer, its completely different company.
What's your point? It's obviously a strategic partnership that they both stand to profit from. Promises. And what after that?
Well, you asked if AM is "officially building any mining farms", and the answer is "Yes, they are officially doing so." I imagine that after they are officially finished building it, it will officially be put to some use. Because I see no results and no communication about whats happening? No news were always bad news here.
Engineering samples sent out to customers doesn't count as "results"? In fact, there have been plenty of "results" and FC actually keeps us relatively updated on the progress and is pretty transparent about the operations. Three weeks may be longer than you (and I, for that matter) would like to wait between updates, but considering the pace at which these things move it's not too bad. |
|
|
|
What has ASICminer in common with immersion cooling from Allied Control other than one small room filled with "proof-of-concept" which is anyway in every day loss right now?
A partnership. Is ASICminer officially building any mining farms ? No. Unofficially behind our backs? Maybe.
They have (repeatedly) said that the HK datacenter will be used. What is crucial is timing. And I see only delays and empty promises over last 6-8 months here ! Timing is everything and Friedcat is very slow. He knew like 6 months about tape-out timeline, he should have everything in place ... and now, when chips are finally out he is unable to do anything for 3 weeks? And no communication with shareholders? What the... ? So much for "overpromise and overdeliver" ... yeah.  If he hasn't been communicating, how can you conclude he hasn't been doing anything? |
|
|
|
I read the proposals with enthusiasm and thought, which one would be easier for the forecasters - producing a continuous probability distribution, or producing a discrete distribution like the one below. The ones who can handle continuous distributions, can easily process them to give piecewise results, but not necessarily the other way round.
Sensible. I agree. I would propose that the ones who want to participate, need to produce a table that gives the probabilities (adding to 100.00%) for each log slot. The slots would be with 0.05 intervals like below. Eg. 3.00-3.05 == $1,000-$1,122. Once the correct result is known, we take a geometric mean of the probabilities given to that slot in the predictions. Then you get + or - points depending if your prob was higher or lower than the average. Eg. if your prob was 8% and average was 4%, you get 8/4-1 = 1 point. If your prob was 1% and average was 3%, you get -3/1+1 = -2 points.
Mostly sensible. Comments/Questions: 1) I don't see why the intervals would need to be of a fixed size 2) I don't understand why you need to logarithm (I understand why it's used for modelling, that's fine, but for reporting and interpreting results I don't see the reason) 3) Furthermore, I find you choice of base 10 unnatural (base e is the natural choice!;) 4) What's the exact logic of the point system? Why is there a division and a subtraction/addition? What do the numerators and denominators represent? Why do the points seem asymmetric, i.e. lose more points on a bad guess than you gain with a good guess? 5) The intervals need to be in absolute terms, not relative to the price at posting time - otherwise you would also need to keep track of the price at posting time for each prediction (not difficult, just an avoidable pain in the ass) |
|
|
|
Ok in that case people will have to guess distributions rather than intervals, then just compare the probability density at the price. If people want to do intervals, then they can use uniform distributions.
This sounds like a solid, sensible and fair solution. Loved the illustrations too! Edit: May be a bit cumbersome for each forecaster, but allows each forecaster to put as much work into it as he wants and still remains fair. |
|
|
|
Predicting a range is clearly useful, but having the value fall into that range or hit the mean value of the range does not always mean the prediction is better.
Example:
Prediction 1: 90-100; Prediction 2: 50-150. Actual: 101
The first is clearly the better prediction, although the actual is outside the range and further from the mean than the second.
What you really want to do is 1) reward a predicted mean value (i.e. max+min/2) that is close to the actual; 2) reward a smaller range preferentially to a larger range.
Actually, it is not clear that the first prediction is better. It depends. Also, to the point 1), consider that the probability density function can be asymmetric within the range, i.e. the mean can be different from 0.5*(max+min). No. D is the deviation from the interval edge and is added to the weighted interval width. I think you misread it.
Sorry, I misread your AND as an OR in the conditional and I didn't look too closely at the inequalities. In that case, this proposal would give the same score to two forecasts with the same I, regardless of where within the interval the actual outcome falls. For example: a prediction of 90-110 would be ranked equal to a prediction of 99-119 with an outcome of 100. I don't find that intuitive either. |
|
|
|
What you want to do is reward for the prediction being within the interval and penalize for distance from the interval as well as for the width of the interval. For example: I= upper_limit -lower_limit
W=Weight of Interval width penalty
P= Actual
if( P>lower_limit AND P<upper_limit){
D=0
}else{
D=min [ abs ( P - upper_limit); abs ( P - lower_limit) ]
}
Score= D+W*I
W could even be 1, IDK. Lowest score wins. Sorry for coming off as very negative today, I really don't mean to smack down everything, but this wouldn't work either. My counterexample: I would choose upper_limit = lower_limit = 0 and win every time (assuming P is not zero)! I would get D=0 and I=0, which would give me Score=0, regardless of W. Furthermore 0 is the lowest possible score since D>=0 and I>=0, so I would win every time!  |
|
|
|
Suppose rpietila's formula was amended to yield zero in the case that the prediction is within the range. Then your offered counterexample fails. Did you have others that would prove the amended scoring formula invalid?
Note, to anyone interested, that the reduction of the price to log10 form allows the predictions to be compared on widely differing timescales, in which price values might be 10x larger or smaller. rpietila has been talking about deviation from the log10 trendline in terms of these log10 deltas.
Fair enough, let's do another one: Imagine the actual outcome is 95. A prediction of 50-100 would score 5. A prediction of 89-100 would also score 5, although it is obviously much better than the first one. So the amended formula does not correctly rank the predictions either (at least not intuitively). Let me see if I can think of another one... Log scale or not doesn't really matter for comparing predictions to each other (by which I mean ranking/ordering them). Edit: Came to think of another counterexample for the amended formula: if the prediction is 50-100 and the outcome is 101, the amended formula gives score 0. If the outcome is 150, the amended formula also gives score 0, although the first case is obviously better than the second. (You might argue that this is "fair", though, but that is a subjective argument: the amended formula still clearly fails to correctly rank/order the two predictions). |
|
|
|
|
RM is apparently expecting to sell 10PH this year. Meaning profit of somewhere between XBT 0.014 and 0.039 per share from just one customer.
How many more customers are there?
|
|
|
|
Eureka! It is this simple: - Every predictor gives two prices in log scale eg. "In 2014-5-16 the price is between 2.7 and 2.85 (roughly 500 and 700)" - When the actual price is known, you take min [ abs ( actual - upper_limit); abs ( actual - lower_limit) ]- Whoever has the lowest average error after a reasonable number of predictions (predictions can be renewed as often as you wish regardless of their maturity) is the best!  - Proof omitted I would be very grateful if you could explain this to a simpleton like myself. Whoever was the closest to the actual price with the narrowest range was the best. I think he's being a little facetious here, because this is, of course, obvious. Except that it doesn't actually work. Proof by counterexample: Imagine a forecast range of 50-100. If the outcome if 95, i.e. within the range, the formula produces a score of 5. However, if the outcome is 105, i.e. outside the range, the formula produces a score of 5. But clearly, the first situation should score better, but with this formula it does not! QED? Edit: I can think of more examples where it doesn't work too, can I leave those as an exercise to the reader? Edit 2: For those wondering how to do it properly, I suggest searching the meteorology literature - it's much more comprehensive on this issue than the financial/economic/econometric literature. |
|
|
|
Agree. The unoptimized test results were good news. I don't know what these people who bailed were expecting? Nice opportunity to acquire. Hopefully more reasonably good news will drive it down further in the short term. A power consumption 100% above expected is "good news"? You can argue that it's not entirely unexpected, you can argue that it still beats many competitors, and you can argue that it may improve. But calling it "actually good news" sounds somewhat disingenuous, don't you think? |
|
|
|
|
It's been a bit quiet in here lately.
Is everyone fed up, or is everyone sitting on the edge of their seat waiting for performance numbers?
|
|
|
|
I suppose that when you say the historical distribution of log returns, you mean without regard to phase of the bubble cycle. Right?
Correct. If that is what you did, then perhaps by analyzing each of the previous bubbles you could obtain 10 deciles by partitioning each bubble. Then you could run ten times your existing simulation on each of the 10 decile's distribution.
My own projection of future bitcoin price possibilities heavily depends upon guessing where we are in the bubble cycle. Because I think we are near a bottom, I expect much higher prices this summer as a result.
I can think of a complication or two fitting the deciles, so if that is something that you have no interest in doing, then I have the data to perform a stochastic monte carlo simulation myself.
Not entirely sure what you mean, but in my current (mis-)interpretation of it, I'm not sure it would be a very useful exercise, since I don't consider bubbles to be so predictable and/or uniform. I would be more inclined to consider that a Monte Carlo simulation based on historical data takes on an appropriate amount of bubblyness, which, when averaged out over 50k simulated price paths, are generally smoothed out. If you look at a single generated price path, you clearly see bubbles etc. with characteristics similar to what we have seen historically -- but if you generate 50k paths and summarize them, then the bubbles disappear and the general trend remains. That was kind of the point, and I sidestep any personal opinions on where we are in "the bubble". Nice. Another refinement might be to take a vol trend line and scale log returns accordingly.
Scale how? And where can I find reliable and representative numbers for volume? In this market, I'm very skeptical to volume figures: I believe the exchanges reflect the general price quite well, but I don't believe the volume on the exchanges is representative for volume traded in total. |
|
|
|
Sorry for interrupting the off-topic discussion. No offence, but the TA in this thread ain't what it used to be. Here's some price forecasts for the next 90/180/365 days you might find interesting:    The charts were made by simulating 50k paths on the log returns based on the historical distribution of log returns. The shaded areas represent percentiles (dark blue for 25-75 and light blue for 10-90), the lines represent the mean and the median. Notable features: - Only 10% chance of hitting $1200/previous ATH before June 2014
- 50% chance of hitting $1200 by October 2014
- The probability distribution is massively skewed to the upside: e.g. 10% chance of being lower than current price (~$440) in December 2014 and 10% chance of being higher than $8000
- The uncertainty is huge, i.e. the shaded areas are large
How is that for a forecast, eh? |
|
|
|
Specs are nice - 12GH at 4+ Watts
Seems right on target, 4 W/12 GH/s is just about 0.35 W/GH/s... |
|
|
|
ASICMINER should rethink their approach here, and sell the chips to the highest bidders as well as mine for themselves. While I understand the concern, I totally disagree with selling the chips by auction: the buyers of the chips need to make investments too (parts, shop space, etc.), and selling the chips by auction would create too much uncertainty around their operation (would they win the auction, would they manage to get a reasonable price, etc.). In the end, I think the uncertainty would scare potential buyers/partners away. Auction would be a good form for end-user devices, or perhaps one-time buyers, but to build stable working relationships based on auctions is not possible. In either case, we have no reason to believe that RM is treated any differently than other buyers, I think some of you guys have too much imagination. Are there any indications whatsoever that RM is getting preferential treatment? (But then again, with so many scams around, I don't blame you for being cautious.) Questions for the list I would like answered: 1) What are future plans? 2) Size of batch 1? 3) What challenges does the regulatory uncertainty in China pose and how is it being dealt with? |
|
|
|
|
Desperately looking forward to seeing final chip specs and testing results.
Hurry up, FC & team!
|
|
|
|
The last 5 times china "banned" bitcoin seem to have had no effect on AM why will it this time?
I don't believe the gradual restrictions on BTC in China has had "no effect" on AM - if nothing else, at least through the exchange rate and the size of the potential market for chips. It's no secret that it's getting gradually harder for bitcoin businesses in China, and it would be unwise to ignore potential regulatory threats. Their wavers are not made in China as far as I know, so they don't need the money in China. Also, Hong Kong is not China. It' the freest economy in the world. it's an off-shore tax haven, it's where Chinese and businesses around the world keep their money save from their own government (seemingly, lol). Hong Kong retains top spot of world's freest economy http://www.bbc.co.uk/news/business-16523643Hong Kong Bitcoin Exchange: https://anxbtc.com/Bitfountain is, as far as I know, incorporated in Shenzhen and not in Hong Kong. So unless they have moved, they are still subject to the normal Chinese regulations. I'm not spreading FUD here, I would just like to know if it is creating any complications for AM's operation (and, if so, how it will be addressed). |
|
|
|
|
In the next round of shareholder questions, I would like to ask how the crackdown on Chinese exchanges will affect the operation. I imagine the lack of Chinese exchanges would be an impediment to moving in/out of BTC/fiat that might complicate the operation.
And also, does the crackdown mean increased business risk for AM/Bitfountain? What is the risk, for instance, that an overzealous bank closes or freezes AM bank accounts?
|
|
|
|
|
Show Posts
