How our predictions work
No black box. This is what actually happens between clicking “Predict” and seeing a forecast.
Most retail prediction tools fall into one of two traps: they are either a single indicator dressed up with marketing (“our proprietary signal”), or a raw AI text model asked to guess a price — confident, articulate, and unaccountable. Predixova is built differently. A forecast here is an ensemble of four independent evidence channels, each with a bounded vote, combined under weights that are continuously re-checked against real outcomes.
The channels are deliberately different in kind. Pattern matching asks “what happened the last thousand times a chart looked like this?” Technical indicators ask “what is the measurable state of momentum, trend and volatility right now?” Sentiment asks “is the information environment pushing with or against the price?” And calibration asks the question most tools never do: “when we said 65% before, were we right 65% of the time?”
Every forecast ships with its evidence attached — the matched historical analogue, the indicator readings, the sentiment tier, and the confidence interval — because a probability you can't interrogate is just an opinion with decimals.
Historical pattern matching →
Every chart you submit is converted into a numerical fingerprint and compared — in milliseconds — against a large library of historical market windows whose outcomes are already known.
Technical indicator analysis →
Momentum, trend, volatility and mean-reversion indicators computed on the exact bars you're looking at, with a regime filter deciding which of them deserve a vote.
News, social & macro sentiment →
Language models read the news flow, social chatter and macro backdrop around an asset and reduce them to bounded scores that can never override the price evidence.
Calibration & honesty →
Every resolved prediction is scored against what actually happened. Brier scores, hit rates and per-channel information coefficients feed back into the ensemble weights.
Parameterized by what you trade
A memecoin and a blue-chip stock do not move for the same reasons, so they are not analysed the same way. The engine adjusts channel weights, volatility expectations and confidence bounds per asset class — crypto majors, memecoins, equities and currency pairs each get their own profile — and per timeframe, from one-minute scalping charts to daily swing charts. The same discipline applies to screenshot mode: when you upload a chart image, we first ask you to confirm what instrument and bar size it shows, because the right analysis depends on it.
Predictions from a screenshot
You can also submit a screenshot of any chart. A multimodal vision model reads the candlesticks directly off the image — open, high, low and close for each bar — sanity-checks them (wicks must contain bodies, prices must be positive), and feeds them into exactly the same pipeline described above. Extraction precision is limited by image quality, and the interface says so rather than pretending otherwise.
What we will not claim
No system predicts markets reliably, and anyone who says otherwise is selling something. What a well-built ensemble can do is estimate probabilities with honest error bars, be right about its own confidence, and show its work. That is the entire product. Predixova is a research tool, not financial advice — position sizing, risk management and the decision to trade remain yours.