That is a very interesting site. I had NOT previously seen it.
The data seems to be inconsistent between clicking on the beta predictor and the other 24 hour, 5 day and 20 day charts.... I mean the beta live version seems to be a bit more bullish than the non-live version. In this regard, the live version predicts mid-$700s in the coming 24 hours; while the non-live versions predicts lower $700s in the coming 24 hours (though granted, the NON-live version seems to update on the hour, so seems to have NOT yet accounted for the price spike that occurred within the last hour).
The developer, Kevin asserts that short -term predictions are likely more accurate than long-term predictions, so long as there are NO major news events affecting the prediction outcomes. And, the algorithm seems to be based on historical performance, which we likely know is NOT necessarily a good predictor of future performance.
Further, his 20-day prediction seems to place BTC prices in the $300 to $400 territory, which seems very bearish and failing to take into account that BTC has been in a down trend for more than 6 months, which will more likely cause BTC prices to go up from here (our current price-point) rather than going down from here. I could NOT really gather from the website how he came to what seems to be a hair-brained idea that BTC prices are going to plummet anywhere below $600 in the coming month. Seems highly unlikely given the pent-up at-the-moment BTC situation.
NONETHELESS, I do like the concept of the site, if he is able to take into account more variables, then maybe the predictive aspect (especially longer term) may improve both long-term and short-term prediction(s).
historical performance has been a great predictor