Most bots fire on rules. Vektor argues. Three independent AI agents debate every setup and vote before a single order moves. 92 markets, 24/7, so you don’t have to watch.
Every setup gets interrogated before it fires. Most get rejected. That's the point.
10 parallel layers — VWAP, vol delta, MACD, TTM Squeeze, market structure, liquidity sweeps, OBV+CVD, Supertrend+Pivots, pre-market range (stocks), ATR+funding — produce 35 dimensions per symbol before any AI call runs.
Every symbol is classified A/B/C by live win rate. Proven performers scan every cycle. Underperformers rotate or get excluded entirely — re-evaluated every 50 trades.
DeepSeek pre-screens with all 35 dimensions + live news context. ~22 → 4–8 high-probability candidates. Saves ~.50/scan. Bad news upstream kills bad setups before expensive AI runs.
DeepSeek (50%), Technical (40%), GPT-4 Vision (10%) vote independently — all with full indicator wiring. News/regime confidence penalties applied. Platt calibration corrects overconfidence from model scores at 50+ trades.
Claude writes the strongest bear case against each trade before deciding. Default is scrutinise, not approve. APPROVE / REJECT / REDUCE_SIZE — with bear case logged for every signal. Portfolio-aware: concentration, correlation, drawdown context.
R:R gate (TP/SL ≥1.5:1), regime alignment, unknown regime block, drawdown confidence boost. Position sized by 7 multipliers: volatility, regime, asset class, OI liquidity, indicator alignment, drawdown scalar, MTF confidence.
Three independent AI agents score each candidate from different angles. Their weighted scores combine, then news and regime penalties are applied, then Platt calibration corrects model overconfidence. Only setups clearing the tightened threshold proceed — then Claude writes a mandatory bear case before approving, rejecting, or sizing down each trade.
Every retail platform claims “AI” and “news sentiment.” Here’s what that actually means — and what Vektor does differently.
| Manual Trading |
Generic Bots |
● Vektor | |
|---|---|---|---|
| Trades 24/7 without fatigue Always-on executionVektor runs continuously on a dedicated VPS. No sleep, no distractions, no missed setups at 3am. Every market session covered automatically. | ✕ | ✓ | ✓ |
| Multi-agent AI consensus 3-agent vote systemThree independent AI agents — DeepSeek Strategy, GPT-4 Vision, and a final reasoning model — vote with weighted scores. Combined score must clear 65% to proceed. Exception: a fresh liquidity sweep or squeeze firing triggers a strong-pass, bypassing the threshold. | ✕ | ✕ | ✓ |
| Regime-aware strategy switching Market context detectionThe bot classifies current market conditions (strong bull, bear, choppy, neutral) and adjusts trade bias accordingly — only longs in bulls, only shorts in bears. | ✕ | ✕ | ✓ |
| Circuit breakers & drawdown limits Auto-stop protectionDaily loss limit: 10%. Max drawdown: 20%. When either trips, the bot halts new entries automatically and sends a Telegram alert. | Manual | Basic SL | ✓ |
| Self-improving meta-learner Learns from every tradeGradientBoosting + SGD online learning run in parallel — the batch model learns from history, the online model adapts after every closed trade. SHAP explains each decision. The system gets smarter over time. | ✕ | ✕ | ✓ |
| GPT-4 Vision chart recognition AI sees the chartBefore every entry, a live candlestick chart is rendered and sent to GPT-4 Vision. It identifies patterns, support/resistance, and assigns a risk level as a final visual gate. | ✕ | ✕ | ✓ |
| Adversarial final gate (APPROVE / REJECT / REDUCE) Last AI checkpoint — adversarial by defaultBefore approving any trade, the final AI agent must construct the strongest argument against it. Only if that argument can be refuted does the trade proceed. Portfolio-aware: concentration, drawdown %, regime, correlation. Bear case logged per signal in the dashboard. Introduced in v2.0 after data showed the original gate was too permissive. | ✕ | ✕ | ✓ |
| Directional news intelligence MOAT No bot does this natively18 signed directional keyword buckets (~300 keywords) cover oil/gold/crypto/macro as separate bearish and bullish categories. G7/IEA/SPR reserve releases score oil bearish. DeepSeek applies 30+ event→bias rules. Context injected upstream — filters setups before costly AI analysis runs. | ✕ | ✕ | ✓ |
| Professional-grade risk & execution engine 10 systems working together
| Partial | Partial | ✓ |
From symbol tier filtering to contrarian Claude gate — every v2 layer documented in detail. Including 7-factor Kelly sizing, dynamic TP widening, and Platt calibration.
Vektor v2.0 launched March 15, 2026 — a full quant overhaul with corrected indicator wiring, R:R gating, symbol tier filtering, 7-factor Kelly sizing, and contrarian Claude validation. Stats below reflect v2 trades only.
Pre-v2 history (32 trades, −$78.56) is preserved as training data and excluded from displayed metrics.
Test any strategy configuration against historical market data before going live. The example below uses illustrative data — real backtesting is in active development.
Start free for 14 days. No credit card. Cancel anytime.
Everything you need to know before committing.
14-day free trial. No credit card. Full dashboard from day one.
⚠ Risk Disclosure — Trading perpetual contracts, leveraged instruments, and other derivatives involves a substantial risk of loss and is not suitable for all investors. Leverage magnifies both potential gains and losses, and you may lose more than your initial deposit. Past performance of any strategy, signal, or algorithm — including any results displayed on this platform — is not indicative of future results and provides no guarantee of profitability.
Vektor is an automated research and execution tool. It does not constitute financial advice, investment advice, or a recommendation to buy or sell any instrument. All trading decisions and their consequences are the sole responsibility of the user. You should only trade with capital you can afford to lose entirely, and you should seek independent financial advice if you are unsure whether trading is appropriate for you. By creating an account and using this platform, you confirm that you have read, understood, and accepted this disclosure in full.