Methodology

State-adaptive signals — not rigid indicators

Algo Bot is designed as a state-adaptive trading system: a Hidden Markov Model infers the market regime; a Markov Decision Process decides what to do. Patterns and technicals feed the model; rated signals are what you see.

01

Market observations

Candles and derived features — log returns, volatility, RSI, ATR, volume, and pattern tags — become the observable inputs for the model.

02

HMM: infer the regime

A Hidden Markov Model estimates the hidden market state (e.g. bullish, bearish, sideways) as a probabilistic belief — not a single fragile indicator.

03

MDP: choose the action

That belief feeds a Markov Decision Process that picks buy, sell, or hold by optimizing a risk-adjusted reward (e.g. Sharpe-style), instead of fixed RSI/MA rules.

04

Publish rated signals

Only stronger decisions become product signals — side, levels, and a success rating — then delivery to Telegram, Discord, and the public ledger.

What is what

HMM vs MDP in plain language

These are the two research building blocks behind the product. One reads the market’s hidden state; the other chooses an action under that state.

Hidden Markov Model

HMM

Markets have states we can’t directly see (regime). An HMM treats price/volatility features as “emissions” from those hidden states, then estimates which regime we’re likely in right now.

Regime probability

Belief state

Instead of “the market is bullish / done,” we keep a probability mix across regimes. That belief updates as new candles arrive.

Markov Decision Process

MDP

Once we have a state, we need an action. An MDP maps state (+ position context) to buy / sell / hold by maximizing expected risk-adjusted reward — the control policy.

Product confidence

Success rating

Product-facing score for how publishable a setup looks: regime clarity, policy confidence, and expected quality vs noisy firehose alerts.

Why pair them?

An HMM alone describes regimes — it doesn’t say what to trade. An MDP alone needs a useful state — otherwise it optimizes noise. Together they form a state-adaptive loop: adapt when crypto regimes shift, then act with a policy aimed at risk-adjusted quality, not every pattern that appears on a chart.

Research → product

How the thesis becomes a signal

Same pipeline, two audiences: thesis language for the model; product language for Telegram, Discord, and the ledger.

Research layerWhat you get in the product
  • Emissions (returns, vol, RSI…)

    Pattern & technical inputs you recognize on charts

  • HMM regimes + belief

    Market regime context behind each signal

  • MDP action (buy / sell / hold)

    Signal direction — or no publish if hold wins

  • Risk-adjusted reward

    Why we prefer quality over raw win-rate

  • Walk-forward evaluation

    Public ledger outcomes you can audit

Primary research pair is BTC/USDT, with walk-forward evaluation against buy-and-hold and simple rule-based benchmarks. We do not custody your funds — v1 is signals you can execute yourself.

Free vs Paid

What we show publicly

Transparency is the product. Actionable timing is the upgrade.

Free receives direction, regime/pattern label, success rating, and the performance ledger — enough to judge whether the system is honest before paying.

Paid ($14.99/mo) receives the full signal at publication: entry, stop, take-profits, and instant delivery on Telegram and Discord.

We do not delete losing signals from the public history. Expired and stopped trades remain visible alongside winners.

See the ledger yourself

Review sample performance, then request an invitation for live channel access.

Performance