Chinese Slumber Method prediction for Feb/20
Prediction valid for: Thursday 2014-02-20, 19:00--19:59 UTC (not before, not after)
Huobi's predicted price: 3800 CNY
Bitstamp's predicted price: 621 USD
For yesterday's prediction I had used a least-squares-fitted line over the last 4 True Slumber Points (mean prices at 19:00--19:59 where Huobi had nearly zero trade). It turns out that the forecast would have been more accurate if I had put more trust in the method, and used all yellow dots since Feb/10 in the fit.
Thus, tomorrows forecast for Huobi, above, was obtained by fitting a straight line to the last seven True Slumber Points (yellow dots) at February 10, 12, 14, 15, 17, and 18, and 19, and extrapolating that trend line for one more day. The prices are ~4080, ~4006, ~4004, ~3961, ~3905, ~3817, and ~3832, respectively; which extrapolate to ~3800 CNY.
The prediction for Bitstamp, as usual, is that for Huobi divided by the normal currency factor R = 6.12.
NOTE 1: Those predictions are conditional to Huobi's clients and their robots being all in bed by Feb/20 19:00 UTC. Otherwise the Method offers no prediction.
EDIT PS. the "confidence intervals" in previous forecasts were not based on any logic; I should have not posted them, sorry. I could compute a "nominal confidence interval" based on the deviation of the yellow dots from the fitted line, but that would not serve any purpose other than providing an excuse for errors. Therefore, from now on I will post only the central value of the prediciton, without any "confidence inerval". The prediction error will be mesured post-facto from that central value, as it has been so far.