Knowledge Base
Understand the Edge
Every signal explained. Every decision unpacked. The more you know, the better you'll trade.
01 — Foundation
What Is This Bot?
The bot runs 4 independent strategies simultaneously. Each one looks at different aspects of the market:
Regime Trend
Timeframe: 1h + 6h
Identifies the direction of the trend using WaveTrend and MACD. Only trades in the direction of the dominant regime.
Monte Carlo Zones
Timeframe: Daily
Simulates 1,000 possible future price paths to find statistical support/resistance levels. Trades near high-probability reversal zones.
Confidence Scorer
Timeframe: 1h
Scores each setup based on 5+ confluence factors: ADX trend strength, MACD, Bollinger/Keltner squeeze, RSI divergence. Higher score = stronger setup.
Multi-Tier Quality
Timeframe: 1h + 6h
Uses EMA crossovers, VWAP anchoring, and multi-timeframe regime alignment. Only trades when the 1h signal and 6h regime agree.
Strategy Performance by Market Condition
| Strategy | Trending ↑ | Trending ↓ | Ranging | High Vol | Low Vol |
|---|---|---|---|---|---|
| Regime Trend | Excellent | Excellent | Poor | Poor | Good |
| Monte Carlo | Good | Good | Excellent | Poor | Excellent |
| Conf. Scorer | Good | Good | Good | Good | Good |
| Multi-Tier Quality | Excellent | Excellent | Good | Poor | Good |
"Excellent" = strategy fires reliably and accurately in this condition. "Poor" = avoid using this strategy here.
Strategy → Outcome Flow
Flow width is proportional to each strategy's contribution to wins and losses. Wider green paths = stronger edge.
What Would the AI Do?
Adjust the market conditions to see how the 7 agents would respond. All logic is based on the bot's actual decision rules.
RSI
620 Oversold100 Overbought
ATR %
1.8%0% Low5% High
Signal Score
720 Weak100 Strong
Regime
Price Zone
SMA Alignment
Setup Quality89
Regime
Haiku
CONFIRM TREND93%
Regime: trend · RSI: 62
↓
Trade
Sonnet
SKIP — INSUFFICIENT CONFLUENCE47%
Zone: accum · SMA: bullish · Score: 72
↓
Risk
Haiku
Leverage 3×
Normal ATR — standard sizing
↓
Critic
Sonnet
SKEPTICAL — PROCEED WITH CAUTION
Moderate confidence — requires tighter risk
↓
— SKIP
Insufficient confluence — agents recommend waiting
02 — Signals
Understanding Signals
03 — AI Brain
The LLM Multi-Agent System
When multi-agent mode is enabled, every trade goes through 5 specialist Claude AI agents in sequence:
Regime
Haiku
Market regime classification
→
Trade
Sonnet
Go / Skip / Flip decision
→
Risk
Haiku
Position sizing + risk flags
→
Critic
Sonnet
Stress-tests thesis, veto power
→
Learning
Haiku
Post-trade lessons (offline)
Regime
Claude Haiku
Classifies market regime + directional outlook. Sets the context for all downstream agents.
Trade
Claude Sonnet
Forms a directional thesis. Decides Go / Skip / Flip based on all available signals.
Risk
Claude Haiku
Sizes the position. Flags portfolio risk overlaps. Sets leverage tier.
Critic
Claude Sonnet
Stress-tests the Trade agent's thesis. Must provide a counter-thesis if it wants to veto.
Learning
Claude Haiku
Post-close only. Extracts lessons and tracks thesis accuracy per setup type.
Memory Architecture — How the Bot Remembers
The AI brain reads from both memory stores each cycle and writes new observations after each decision. Long-term patterns graduate from short-term notes via the Learning Agent.
04 — Safety
Risk Management
Circuit breakers are automatic kill switches that prevent catastrophic losses:
Daily Loss Limit
If the bot loses more than X% of current equity in a single day, it stops trading until the next day.
Consecutive Losses
After N losses in a row, the bot pauses and waits for conditions to improve before re-entering.
Position Limits
Maximum number of open positions at once. Prevents over-exposure to correlated assets.
Drawdown Guard
If equity drops below a rolling 30-day peak by more than X%, the bot enters a cautious mode with reduced sizing.
05 — Process
How a Trade Flows
A signal must pass through 6 sequential gates before becoming a trade. If it fails any gate, it's rejected and logged.
📡 Signal Generated
↓
1
Validity
SL width ≥ 0.3%, proper direction
✓ PASS
↓
2
Circuit Breaker
No daily loss limit breach
✓ PASS
↓
3
Position Limits
Max open positions not exceeded
✓ PASS
↓
4
Leverage
Calculated leverage within safe range
✓ PASS
↓
5
Liquidation
Liquidation price buffer adequate
✓ PASS
↓
6
Sizing
Position size within risk limits
✓ PASS
↓
✅ Trade Executed
After the Trade Opens
IDLE
No position
→
OPEN
Position active, stop loss set
→
TP1_HIT
First profit target reached, size reduced
→
TRAILING
Stop moved to BE+, trailing up
→
CLOSED
Trade complete, results logged
06 — Tools
Trading Calculators
Enter your account size, risk tolerance, and stop distance. The calculator shows exactly how many units to buy and what leverage that implies.
Account Size ($)
Risk Per Trade (%)
Entry Price ($)
Stop Loss Price ($)
Stop distance: 0.84% = $800 per unit
Max Dollar Risk
1.5% of account
$150
Position Size
BTC / SOL / HYPE etc.
0.1875 units
Notional Value
Total position exposure
$17,813
Effective Leverage
Notional ÷ account size
1.8×
Pro tip: The bot uses 1.5% risk per trade by default. At $10K account, that's $150 max risk per trade. Effective leverage of 3-7× is healthy for most setups.
07 — Copy Trading
How to Copy-Trade This Bot
1
Watch the Signals page
Go to Signals → watch for a signal card to appear with a score ≥ 75. The higher the score, the stronger the bot's conviction.
2
Check the regime
Look at the current regime badge in the top nav. "TREND" or "RANGE" regimes are typically safer. Avoid copying trades in "PANIC" or "HIGH_VOLATILITY" regimes unless you understand the risk.
3
Calculate your position size
Use the Position Size Calculator above. Risk 1-2% of your capital max. Enter the signal's entry, SL, and TP levels into the calculator.
4
Set your order on Hyperliquid
Open Hyperliquid → select the symbol → set a limit order at the entry price, or market order if price is currently at the entry zone. Set your stop loss immediately.
5
Set your take profits
Place a TP1 order at the first target (typically 50% of your position), and TP2 for the remainder. Consider setting a trailing stop after TP1 is hit.
6
Follow the bot's exit signal
Check the Activity Feed for exit signals. If the bot shows a regime change or veto on the same symbol, consider closing early.
08 — Reference
Glossary
Total Terms
54
definitions
Topics Covered
7+
RSI, ATR, Regime…
Last Updated
Mar 2026
kept current
strategyriskAI / Agentstechnical
Key Concepts
RegimeATRRSIVetoConfidence ScoreConfluenceDrawdownEnsembleBacktestCircuit BreakerR:RThesisMACDBollinger BandsTrailing StopAdvisory ModeCounter-ThesisSMAEMAKeltner ChannelVWAPWaveTrendADXOpen InterestWalk-Forward TestingMonte CarloMax DrawdownLiquidation PriceMarginProfit FactorSharpe RatioCalmar RatioSlippageFunding Rate
ATR (Average True Range)
Measures how much an asset typically moves in a single candle. High ATR = high volatility. The bot uses ATR to size stop losses so they are proportional to actual market movement.
SMA (Simple Moving Average)
Average of the last N closing prices. SMA20 (fast) crosses above SMA50 (slow) = uptrend signal. SMA20 crosses below SMA50 = downtrend signal.
RSI (Relative Strength Index)
Momentum oscillator on a 0-100 scale. Above 70 = potentially overbought (caution on longs). Below 30 = potentially oversold (caution on shorts). The bot uses RSI as one confluence factor among many.
Confidence Score
The bot's internal 0-100 rating for how strong a setup is. Based on how many strategies agree and how strongly. Score ≥ 70 is considered a strong signal. The LLM adds its own confidence layer on top.
Regime
Market state classification (trend/range/panic/high_volatility/low_liquidity). Different regimes call for different strategies. The Regime Agent classifies this each evaluation cycle.
Ensemble
The process of combining votes from multiple strategies. The bot uses "weighted_veto" mode: strategies vote, but any strategy can veto a trade if conditions are too risky.
Veto
When the Critic agent (or a strategy) overrides a trade signal. Usually triggered by conflicting signals, high risk conditions, or poor risk/reward ratio.
Circuit Breaker
An automatic stop-trading mechanism that activates when losses exceed a daily threshold or a certain number of consecutive losses occur. Prevents catastrophic drawdowns.
Drawdown
The peak-to-trough decline in account equity. Max drawdown is the worst historical decline. Lower is better. The bot aims to keep max drawdown below 15%.
Profit Factor
Gross profit divided by gross loss. A ratio above 1.0 means the strategy makes more than it loses. 1.5× is good; 2.0× is excellent.
R:R (Risk/Reward Ratio)
How much profit you aim for vs how much you risk. The bot requires R:R ≥ 1.0 before entering. A 2:1 setup means potential gain is twice the potential loss.
Accumulation Zone
Price level below current price where the bot considers the asset "cheap" relative to its volatility bands. Good levels to consider longs.
Distribution Zone
Price level above current price where the bot considers the asset "expensive". Good levels to consider profit-taking or shorts.
Funding Rate
The cost of holding a perpetual futures position. Positive funding = longs pay shorts. Extreme funding rates signal crowded trades the bot may fade.
Liquidation Price
The price at which your leveraged position gets automatically closed by the exchange. The bot always checks that the stop loss is hit long before the liquidation price.
Trailing Stop
A stop loss that moves up with a winning trade (for longs). Locks in profit as price rises, closes only if price falls back by a set amount. The bot uses progressive trailing after TP1 is hit.
Sharpe Ratio
Return per unit of risk taken. Higher = better. Adjusts for volatility so a steady 10% return is rated better than an erratic 15% return.
Open Interest (OI)
Total number of outstanding futures contracts. Rising OI + rising price = strong trend. Rising OI + falling price = distribution. The bot monitors OI divergence as a signal.
Advisory Mode
The LLM operates in advisory mode by default: it analyses and logs what it would trade, but does NOT execute trades. Useful for monitoring AI judgement before giving it execution power.
Slippage
The difference between the price you expected to trade at and the price you actually got. Higher volatility and lower liquidity = more slippage. The bot accounts for 0.05-0.1% slippage in its cost model.
VWAP (Volume-Weighted Average Price)
The average price weighted by volume, resetting each day. Institutional traders use VWAP as a benchmark. Price above VWAP = bullish bias; below = bearish. The Multi-Tier Quality strategy anchors its entries to VWAP.
MACD (Moving Average Convergence Divergence)
A momentum indicator that shows the relationship between two EMAs. When the MACD line crosses above the signal line = bullish momentum. Used by the bot's Regime Trend strategy for trend confirmation.
Bollinger Bands
Volatility bands placed 2 standard deviations above and below a moving average. Price near upper band = overbought. Near lower band = oversold. When bands squeeze (narrow), a big move is typically coming.
Keltner Channel
Volatility envelope based on ATR. When Bollinger Bands are inside Keltner Channels, the market is in a "squeeze" — coiling energy before a breakout. The Confidence Scorer uses squeeze detection.
EMA (Exponential Moving Average)
Like SMA but gives more weight to recent prices, making it more responsive to current conditions. The bot uses EMA crossovers (short EMA crossing long EMA) as trend signals.
ADX (Average Directional Index)
Measures the strength of a trend, regardless of direction. ADX above 25 = strong trend; below 20 = range/choppy market. The bot uses ADX to avoid trading range-bound markets with trend strategies.
WaveTrend
A custom momentum oscillator that combines channel-based smoothing with momentum. The Regime Trend strategy uses WaveTrend to detect oversold/overbought conditions within the context of the current regime.
Paper Trading
Simulated trading with real market data but no real money. The bot runs in paper mode by default. All signals, decisions, and "trades" are executed in a virtual account to test performance before going live.
Live Trading
Trading with real capital on the exchange. Requires the bot to be configured with exchange API keys and LLM_MODE ≥ 1. Only activate after extended paper trading validation.
Entry Price
The price at which a trade position is opened. The bot calculates entry based on current market price at the time the signal is confirmed. For limit orders, the entry may differ slightly from the trigger price.
Take Profit (TP)
Pre-set price target where a portion of a position is closed to realize profit. The bot uses two targets: TP1 closes ~50% of the position and moves the stop to breakeven; TP2 is the final target for the remainder.
Breakeven Stop
Moving the stop loss to the entry price after TP1 is hit. This guarantees the trade cannot lose money on the remaining position. The bot does this automatically after TP1.
Confluence
When multiple independent signals all point in the same direction. High confluence = higher confidence trade. The bot requires at least 2 strategies to agree, plus regime alignment, before considering a trade.
Perpetual Futures (Perps)
Futures contracts with no expiry date. Unlike regular futures, they track the spot price via a funding rate mechanism. Hyperliquid trades are all perpetual futures, allowing both long and short positions with leverage.
Long / Short
Long = betting price goes up. Short = betting price goes down. Perpetual futures allow you to profit in both directions. The bot can take either direction depending on signal and regime.
Mark Price
The "fair" price of a futures contract, calculated using a weighted average of spot prices across exchanges. Used by Hyperliquid to calculate unrealized PnL and trigger liquidations — not the last traded price.
Margin
The collateral deposited to open and maintain a leveraged position. Cross margin shares your full account balance across positions. Isolated margin limits risk to a fixed amount per position. The bot uses isolated margin.
Unrealized PnL
The profit or loss on an open position if it were closed at the current price. Shown in the dashboard as "open position PnL." Not banked until the position closes.
Realized PnL
Profit or loss that has been locked in by closing a position. This is what counts for your account equity.
Basis Points (bps)
1 basis point = 0.01%. Used to express small percentage changes. A 50bps stop loss = 0.5% below entry. Fees are often quoted in basis points (e.g. 3.5bps maker fee = 0.035%).
Thesis
The reason for taking a trade. The Trade agent constructs a thesis: "BTC is in an accumulation zone, regime is trend, RSI is 58, 3/4 strategies agree. Thesis: continuation long." The Critic then stress-tests it.
Counter-Thesis
The opposite argument to the trade thesis. The Critic agent is required to provide a counter-thesis before it can veto a trade — it cannot just say "bad trade" without explaining why.
Backtest
Running the bot's strategies on historical price data to see how they would have performed. Results include total return, win rate, drawdown, and per-trade breakdown. Used to validate strategy changes before going live.
Walk-Forward Testing
A more rigorous form of backtesting where the strategy is optimized on one period and tested on the next, then the window moves forward. Helps detect overfitting — strategies that work on past data but fail on new data.
Overfitting
When a strategy is tuned so precisely to historical data that it fails on future data. The bot avoids overfitting by using robust, logic-based indicators rather than curve-fitted parameters.
Monte Carlo Simulation
Running thousands of random simulations of possible future price paths based on historical volatility. The bot uses this to find statistical support/resistance zones that hold across many possible scenarios.
Hyperliquid
The decentralized perpetuals exchange the bot trades on. Key advantages: fully onchain, sub-second execution, deep liquidity, low fees (0.035% maker / 0.05% taker), and a wide range of tradeable assets.
API Key
A credential pair (API key + secret) that allows software to interact with an exchange on your behalf. Required for live trading. Never share your API key — store it encrypted in your .env file only.
Hot Wallet
A wallet connected to the internet, used for exchange trading. Contrasted with cold storage (hardware wallets). Only fund your trading wallet with what you are willing to risk.
Equity Curve
A chart showing account equity over time. A smooth upward-sloping curve with small drawdowns is the goal. The bot's backtest results page shows the equity curve for all tested periods.
Max Drawdown
The largest peak-to-trough decline in equity during a trading period. A strategy with a 30% max drawdown would have caused an account to fall 30% from its peak before recovering. Lower is better.
Calmar Ratio
Annual return divided by maximum drawdown. A ratio above 1.0 means the strategy earns more annually than its worst drawdown. Higher is better.
Win Rate
The percentage of trades that are profitable. A strategy can be profitable with a win rate below 50% if the average win is much larger than the average loss. See the live track record at /forensics for the bot's actual win rate.
Expectancy
The average profit per trade, accounting for both win rate and average win/loss size. Formula: (WinRate × AvgWin) − (LossRate × AvgLoss). Positive expectancy means the system is profitable over many trades.
Ready to see it in action?
View live signals, copy trades, or explore backtest results.