// Android App · Rules-Based Crypto Investing
Dollar-cost averaging works because it removes the decision. Smart DCA keeps that discipline and adds the part a calendar reminder cannot: how much to buy today. Statistical forecasting fused with a Transformer network reads the market each morning and sizes the buy against your own risk profile and capital plan.
// Watch the Demo
// Features
One recommendation each morning, with an amount attached. No dashboard to interpret and no indicators to weigh yourself — the analysis runs server-side overnight, so you open the app to a decision rather than to a pile of charts.
Holt-Winters exponential smoothing captures trend and seasonality, a Transformer neural network learns the structure it cannot, and a valuation layer arbitrates between them. Three independent readings of the same market, reconciled into one number.
The MVRV Z-score measures market value against realized value — how far price has run from what holders actually paid. It is the on-chain signal that has historically framed cycle tops and bottoms, and it is what tilts the app toward buying harder into weakness.
Fixed DCA buys the same amount whether the market is euphoric or capitulating. Smart DCA scales the buy against valuation and your stated risk tolerance, so a cheap market draws more of your capital and an expensive one draws less — without ever exceeding the plan you set.
Each recommendation carries a suggested limit price rather than a market order — a small edge, taken consistently, on a purchase you were going to make anyway.
Bitcoin and Ethereum do not respond to the same indicators in the same way. Each coin carries its own calibrated thresholds rather than inheriting Bitcoin's and hoping for the best.
A written daily briefing explains what moved and why the recommendation landed where it did — so the number is not a black box you are asked to trust on faith.
// Under the Hood
The strategy was backtested across five years of market data before it shipped, and the full forecasting stack reruns every day on Google Cloud Run — PyTorch for the neural layer, Firebase for authentication, live market and on-chain feeds for the inputs. Nothing is computed on your phone, and nothing is left stale.