Dev branch for iguana now has most of the initial LP node code active. It currently generates realtime raw features in the SVM folder and there is an API command to generate an SVM. The SVM will then make predictions as to the future price direction.
There are currently 8 SVM’s for different timeframes, from 10 minutes to 12 hours in the future. You can configure things to track a large number of trading pairs and it even detects arbitrages that are possible if two or more exchanges are active. See tests/arbinit script.
The realtime predictors are not yet active, but SVM generation is averaging 85%+ accuracy for today’s BTC price action. A very very small sample and as we get more and more different type of days this accuracy will reduce, however the more data there is the more accurate the overall realtime forward predictions will be. It calculates prices for all BTC pairs into USD and CNY. It also monitors the NXT blockchain for the various asset prices and maps them to BTC prices, so things are all setup for automated LP node quotes for all the assetchains, plus any supported BTC trading pair.
it really is quite a lot of new code and some pretty fancy AI level predictor generation as SVM have much better performance than neural networks for large scale datasets.
Update were written to #tradebots on SuperNET Slack.