I wouldn't call him a fraud either if he just made predictions that go wrong. The big difference is that he never admits to have been wrong where he was wrong. He simply lies about his past predictions. In that Gold example that you mention, he later actually claimed that he and his computer predicted the bounce:

Gold Report 2016 Part 1:

Another anomaly was that at the end of 2015, we elected the Quarterly Bearish
Reversal at 1112 closing at 1060.20. At the same time, that low generated a
minor Quarterly Bullish Reversal at 993. So we generated a long-term sell signal,
but a short-term buy signal. The next two Quarterly Bullish generated were at
1308 and 1347. That meant we should first rally typically for 2 to 3 months in a
reaction and if they failed to be elected by the close of March, then the longterm
trend should resume to retest support.

That IS fraud. That minor Quarterly Bullish Reversal came out months after he was still publishing bearish messages in his blog as you remember. This behavior goes so far that it is embedded in the source code of Socrates, where, if Socrates goes wrong, in the following report it claims it was right, by injecting after-the fact signals that it had never reported earlier.


Read this blog starting at page 273 to find out more about computerized fraud