Frequently Asked Questions

BitPredict is an institutional-grade algorithmic trading platform. It combines proprietary market data (15+ bar types, AI regime detection, on-chain analytics), hundreds of pre-built transparent strategies, a visual strategy builder, backtesting and optimization infrastructure, and multi-exchange execution — all in one platform. It's built for quant researchers, algo traders, developers, and institutional teams.

No. The Strategy Builder uses a visual drag-and-drop DAG editor — you connect blocks to compose signal pipelines without writing a single line of code. Data exploration, backtesting, and execution are all point-and-click. If you do code, the REST API and official Python/JavaScript SDKs give you programmatic access to everything.

Start with the Free plan. It gives you full access to browse all data, strategies, and analytics. You only need to upgrade when you want to run backtests, activate strategies, or train ML models — all of which consume credits. Upgrade to Pro when you're ready to use compute features regularly.

Credits are BitPredict's compute currency. 1 credit = $0.10. They're consumed only by compute-intensive actions: running a backtest (1 credit), activating a strategy (10 credits), saving a strategy draft (1 credit), running an optimization job (1 credit per 3 minutes of compute), training an ML model (1 credit per 2 minutes of compute), and exporting CSV with regime metadata (1 credit). Your subscription plan unlocks access; credits pay for usage.

Yes. The Free plan lets you explore all market data, browse and analyse every strategy, view backtesting history, and read the full API documentation — all without spending credits. You'll need credits (and potentially a paid plan) to run your own backtests, build and activate strategies, and access execution features.

BitPredict provides 15+ proprietary bar types: Time, Dollar, Volume, Tick, Range, Renko, Volatility, Imbalance, Run, and several Hybrid variants. Each type aggregates price data differently — Time bars close on fixed intervals, Dollar bars close when a fixed dollar value trades, Volume bars close on fixed volume, and so on. The Data Quality page scores each bar type's ML suitability so you know which are safe for model training.

BitPredict uses a proprietary AI model to classify market conditions into four states: Bull, Bear, Range, and Transition. The model runs at bar level — every bar in the system carries a regime label, a confidence score, and a full probability breakdown (bull score, bear score, range score, transition score) plus secondary metadata: trend direction, volatility percentile, momentum, trend strength Z-score, and transition pressure. The model is continuously updated.

Data is updated continuously in production. The Market Data Explorer shows the latest available date for each symbol and bar type. On-chain metrics are refreshed on a schedule aligned with the underlying data providers.

Yes. From the Market Data Explorer, you can export clean OHLCV data as CSV. A second export option includes full regime metadata — 11 additional columns per bar including regime label, confidence, trend direction, volatility percentile, momentum, trend strength, transition pressure, and the four regime probability scores. The annotated export costs 1 credit.

The ML Suitability score (0–100) on the Data Quality page measures how appropriate a bar type is for training machine learning models. It checks 7 criteria: low return autocorrelation, variance ratio near 1.0, adequate sample size (≥1,000 bars), low kurtosis tail risk, bar size uniformity, low volatility clustering, and high data integrity (≥95% valid bars). A score above 70 is generally safe for ML training.

Pre-built strategies are live, actively running strategies created and maintained by the BitPredict research team. They continuously generate trading signals on real market data. Every pre-built strategy is fully transparent — you can inspect its entire signal pipeline (the DAG), its complete trade ledger, regime performance breakdown, and execution trace. You can assign any pre-built strategy directly to a virtual account, demo account, or live execution without modification.

Yes — full transparency is a core principle. Every strategy's Pipeline tab shows the complete signal DAG: data sources, indicators, filters, logic operators, and entry/exit rules. The Ledger tab shows every historical trade. The Regimes tab shows performance broken down by Bull, Bear, Range, and Transition market states. The Trace tab replays signal generation step by step so you understand exactly why each trade fired.

Check the current regime on the Data → Market Regimes page, then filter strategies by "Best Regime" on the Strategies page. The Regime Champions leaderboard surfaces strategies that perform best in the current regime. You can also use AI natural-language search — e.g. "trend-following BTC strategy that performs well in bear markets" — and the system will surface the closest matches.

Activating a strategy (in the Strategy Builder) promotes your draft from saved-only status to live status. A live strategy runs in the BitPredict system, generating signals on real market data in real time. Activation costs 10 credits. You can deactivate at any time at no cost — the strategy stops generating signals but its history is preserved.

BitPredict's backtesting engine models realistic execution costs: configurable slippage, commission/fee model, leverage, and position sizing. However, no backtest perfectly replicates live trading — slippage models are approximations, and past performance is not a guarantee of future results. For the most realistic pre-live evaluation, follow up backtesting with paper trading in a Virtual Account.

Backtesting is compute-intensive — the engine processes every bar over the selected date range, applies the full signal pipeline, and computes all performance metrics. At $0.10 per run, credits are a lightweight cost that prevents abuse and funds the infrastructure. A thorough testing workflow of 10–20 runs costs $1–2.

Overfitting is when a strategy is tuned so precisely to historical data that it performs well on paper but fails in live trading. It happens most often during optimization — if you pick the trial with the highest return, you're almost certainly overfitting. Instead, sort by Sharpe ratio, use out-of-sample testing (reserve the last 20% of your date range and never include it in optimization), and prefer parameter sets that are robust across a range of values rather than pinpoint-optimal ones.

A backtest runs a fixed strategy configuration over a historical period and reports performance. An optimization runs many backtests automatically, systematically varying parameter values (stop loss levels, indicator periods, etc.) within ranges you define. Optimization finds which configuration performs best — but use its results carefully to avoid overfitting.

A Virtual Account is a pure simulation — no exchange is involved. It uses BitPredict's internal signal data and simulates P&L mathematically. Demo Trading connects to Bybit's testnet and executes real orders on a real exchange using fake capital. Demo Trading is more realistic because it tests actual order placement, latency, and exchange-side execution — but it requires a Bybit Demo API key.

Live trading supports Binance, Bybit, OKX, Kraken, and Hyperliquid. Live trading is currently invite-only. You can request access via the in-app contact form.

Yes. Bybit Demo API keys are encrypted at rest using Fernet symmetric encryption before being stored. Keys are never logged or transmitted in plaintext. The BitPredict backend uses internal secret authentication to prevent cross-service key leakage. You can revoke a key from Settings at any time.

Live trading is currently invite-only while we scale the execution infrastructure. Submit a request via the Contact page, selecting "Live Trading Access" as the reason. The BitPredict team will review your request and reach out.

Telegram, Discord, Slack, WhatsApp, and Email. Each channel can be independently enabled or disabled per strategy subscription. You can receive alerts on multiple channels simultaneously.

Yes. You can subscribe to as many strategies as you want. Each subscription can send to any combination of your connected channels. Alert history is filterable by strategy, channel, and date range so you can review past signals easily.

Each alert contains: signal type (ENTRY or EXIT), direction (LONG or SHORT), symbol (e.g. BTC/USDT), bar type, strategy name, and timestamp. For entries, the approximate entry price is included. The format is consistent across all channels.

Credits are purchased in bundles via Stripe checkout from your Settings → Credits page.

Save Strategy as Draft (1 credit), Activate Strategy (10 credits), Run Backtest (1 credit), Run Optimization Job (1 credit per 3 minutes of compute), Train ML Model (1 credit per 2 minutes of compute), Run ML Optimization (1 credit per 2 minutes of compute), Export CSV with Regime Metadata (1 credit).

Compute actions will be blocked until you purchase more. Your existing strategies continue running and generating signals — only new compute jobs (backtests, optimizations, ML training) are paused. Purchase more credits instantly from Settings → Credits.

Yes. Credits do not expire and roll over indefinitely. They are tied to your account, not your billing period.

The full API reference is available in-app at /api-docs. It includes the full endpoint reference with request/response schemas, a quickstart guide, authentication instructions, rate limits documentation, SDK installation guides, and a changelog.

Yes. Rate limits vary by plan tier and endpoint type. The API documentation at /api-docs/rate-limits has the full breakdown. When you hit a rate limit you'll receive a 429 response — implement exponential backoff before retrying.

Yes. Official Python and JavaScript/TypeScript client libraries are available. Installation instructions and usage examples are in the API documentation under /api-docs/sdks.