Strategy Composition Guide¶

This guide shows how to combine Katbot's three configuration layers — Research, Trading Strategy (Persona), and Quick Prompts — to build powerful, cohesive trading agents.

The Three Layers¶

Every recommendation your agent produces is shaped by three distinct layers, each with a different scope and purpose:

Layer	Where You Configure It	What It Controls	How Often You Change It
Research	Agent Detail → Research Configs	Which tokens the agent considers	Weekly or when market regime shifts
Trading Strategy	Agent Detail → Recommendation Configs → Persona	Risk rules, leverage limits, entry checklists, position management	Rarely — defines the agent's core personality
Quick Prompt	Portfolio → Recommendations panel	Tactical focus for a single recommendation request	Every request

%%{init: {'theme': 'dark'}}%%
graph TD
    %% Input Layer
    subgraph Input_Layer [SYSTEM INPUTS]
        direction LR
        A[<b>QUICK PROMPT</b><br/><i>The Intent</i><br/>'Find a bullish pullback in<br/>DeFi tokens with RSI < 40'] 
        B[<b>TRADING STRATEGY</b><br/><i>The Persona</i><br/>Capital Guardian: 1% risk,<br/>5x max leverage, 3:1 R:R]
        C[<b>RESEARCH BUCKET</b><br/><i>The Filter</i><br/>MOMENTUM: Tokens with<br/>7d+ uptrend + volume]
    end

    A -->|Natural Language| D{<b>KATBOT ENGINE</b>}
    B -->|Risk Constraints| D
    C -->|Asset Selection| D

    D -->|Combined at Request Time| E

    subgraph Output_Layer [ACTIONABLE RECOMMENDATION]
        E["<b>BUY ARB/USD</b><br/>Entry: $1.12<br/>Stop Loss: $1.08 | Take Profit: $1.24"]
    end

    style D fill:#ea580c,stroke:#fb923c,stroke-width:2px
    style E fill:#064e3b,stroke:#34d399,stroke-width:2px,color:#ecfdf5
    style Input_Layer fill:#1e293b,stroke:#334155,stroke-dasharray: 5 5

How they combine at runtime:

Research narrows the token universe (from ~80 tokens down to 3–8 high-signal candidates)
The persona sets hard guardrails (max leverage, required R:R, position limits)
The quick prompt tells the agent what to look for right now

The agent cannot violate persona rules regardless of what the quick prompt asks. Think of it as: Research picks the menu, the persona sets the budget, and the quick prompt places the order.

Persona Overview¶

Katbot provides five risk personas. Load one in the Trading Strategy editor, then customize it if needed:

Persona	Risk/Trade	Max Leverage	Min R:R	Best For
Capital Guardian	1%	5x	3.0:1	Conservative accounts, drawdown-sensitive capital
Base Trader	2%	10x	2.5:1	General-purpose trading (default)
Tactical Alpha	2.5%	15x	2.0:1	Active traders seeking capital efficiency
Scalp Specialist	0.75%	20x	1.5:1	High-frequency, minutes-to-hours holds
Swing Operator	1.5%	3x	4.0:1	Multi-day to multi-week positions

Strategy Examples¶

Below are five complete agent configurations showing how to combine all three layers for different trading styles.

Example 1: Conservative Momentum Rider¶

Goal: Catch established trends with minimal risk. Hold for hours to days. Never chase.

Research Config¶

Bucket: MOMENTUM

Methodology Prompt:

Focus on tokens with the strongest directional price momentum across 24h, 7d, 
and 30d windows. Use rate of change as the primary ranking signal. Require all 
three timeframes to agree on direction. Penalize tokens where momentum is 
decelerating (24h ROC much weaker than 7d ROC). Require minimum $2M daily 
volume. Only include tokens with score >= 0.7.

Trading Strategy¶

Persona: Capital Guardian

Customization (optional): Add this below the existing position management section:

## ADDITIONAL RULES
- Only enter tokens where 7d momentum and 30d momentum agree on direction
- If BTC is in a downtrend regime, decline all long entries regardless of altcoin signals
- Prefer tokens with funding rate < 0.02%/8h (cheap carry)

Quick Prompts to Use¶

Situation	Quick Prompt
Bull market	"Find the strongest trend continuation setup. Prefer tokens pulling back to their 20 EMA with rising OI."
Uncertain market	"Analyze all tokens but only recommend if there is an A+ setup. Otherwise HOLD."
After a dip	"Look for tokens that held above their 50 EMA during the pullback and are now resuming the trend."

Example 2: Aggressive Scalper¶

Goal: High-frequency trades lasting minutes to hours. Small wins, strict time discipline. Never hold through funding.

Research Config¶

Bucket: VOLUME_BREAKOUT

Methodology Prompt:

Find tokens with volume spikes exceeding 2.5x their 7-day rolling average in 
the past 4 hours. Require minimum $5M daily volume and tight spreads. Include 
both long and short signals. Penalize tokens that have already moved >5% from 
their breakout level (don't chase). Score >= 0.6 minimum.

Trading Strategy¶

Persona: Scalp Specialist

Customization (optional):

## ADDITIONAL RULES
- Hard time stop: close any position open > 90 minutes without TP1
- Never enter if next funding snapshot is within 45 minutes
- If 3 consecutive losses occur within 4 hours, pause for 2 hours
- Prefer tokens with bid-ask spread < 0.05%

Quick Prompts to Use¶

Situation	Quick Prompt
Volume spike detected	"The volume breakout tokens are active. Find a momentum entry on the 5m chart with clear invalidation within 0.5 ATR."
Ranging market	"Look for range fade setups at the extremes of 1h S/R levels. Mean reversion only."
After news event	"Identify the strongest directional move in progress and find a pullback entry on the 5m timeframe."

Example 3: Swing Trader with Funding Awareness¶

Goal: Capture multi-day structural moves while keeping funding costs low. Patient entries on daily structure.

Research Config¶

Bucket: FUNDING_RATE_ANOMALY

Methodology Prompt:

Identify tokens where funding rate is significantly negative (< -0.03%/8h) 
suggesting overcrowded shorts, or significantly positive (> 0.05%/8h) 
suggesting overcrowded longs. These extremes often precede reversals. Include 
tokens where funding has been extreme for > 24 hours (crowded trade about to 
unwind). Minimum $3M daily volume. Score based on duration and magnitude of 
funding anomaly.

Trading Strategy¶

Persona: Swing Operator

Customization (optional):

## ADDITIONAL RULES
- Mandatory funding check: projected funding cost over expected hold must be < 10% of TP1
- If going LONG, prefer tokens with negative funding (you get paid to hold)
- If going SHORT, prefer tokens with positive funding (you get paid to hold)
- Re-evaluate thesis on every daily close — if funding normalizes, the thesis may be done

Quick Prompts to Use¶

Situation	Quick Prompt
Funding extreme detected	"Funding is extremely negative on several tokens. Find the best mean-reversion long where shorts are overcrowded and price is at 1D support."
Trend continuation	"Identify tokens in a strong 1W uptrend where funding just turned negative (late shorts entering). This is a trend continuation signal."
Risk-off	"Funding is normalizing across the board. Check if any existing positions should be closed as the unwind thesis is complete."

Example 4: Sector Rotation Tactician¶

Goal: Identify which crypto sectors are heating up and position early. Moderate aggression with diversification.

Research Config¶

Bucket: SECTOR_ROTATION

Methodology Prompt:

Analyze sector performance across L1s, DeFi, AI/ML, Gaming, and Memes over 
7d and 30d windows. Identify the sector showing the strongest relative 
outperformance vs BTC. Within that sector, rank tokens by: (1) 7d relative 
strength vs sector average, (2) volume trend, (3) market cap > $100M for 
liquidity. Include 3-5 tokens from the leading sector and 1-2 from the 
second-strongest sector for diversification.

Trading Strategy¶

Persona: Tactical Alpha

Customization (optional):

## ADDITIONAL RULES  
- Maximum 2 positions in the same sector (diversification requirement)
- If a sector has outperformed BTC by > 30% over 7 days, it may be extended — require extra confirmation
- Consider sector beta: high-beta sectors (Memes, Gaming) get 50% size reduction
- Track sector rotation weekly — if the leading sector changes, existing positions in the old sector get tighter stops

Quick Prompts to Use¶

Situation	Quick Prompt
New sector emerging	"AI tokens are showing sector rotation strength. Find the best entry in the sector with pullback to support and rising volume."
Taking profits	"The current leading sector has been running for 2 weeks. Find a token in the next sector showing early rotation signals for a fresh entry."
Hedging	"I'm long 3 DeFi tokens. Find a short opportunity in a weakening sector to reduce portfolio correlation."

Example 5: Correlation Divergence Arbitrageur¶

Goal: Trade tokens that are decoupling from BTC/ETH. When correlation breaks, mean-revert or follow the divergence.

Research Config¶

Bucket: CORRELATION_DIVERGENCE

Methodology Prompt:

Find tokens whose 7-day rolling correlation with BTC has dropped below 0.3 
(normally > 0.7). Distinguish between: (1) tokens diverging UP while BTC is 
flat/down — potential alpha, and (2) tokens diverging DOWN while BTC is 
flat/up — potential distress. Score based on magnitude of divergence × 
duration × volume support. Minimum 30-day average correlation > 0.5 (so the 
divergence is meaningful). Include both long and short signals.

Trading Strategy¶

Persona: Base Trader (default)

Customization (optional):

## ADDITIONAL RULES
- For tokens diverging UP from BTC: treat as momentum signal, enter long with trend
- For tokens diverging DOWN from BTC: wait for confirmation of capitulation (volume spike + RSI < 25) before mean-reversion long, OR short if structure is breaking down
- If BTC correlation re-couples (returns above 0.6), close the position — the divergence thesis is over
- Maximum 1 divergence trade per direction at a time

Quick Prompts to Use¶

Situation	Quick Prompt
Bullish divergence	"SOL is diverging up from BTC with increasing volume. Find an entry on the 4h pullback with stop below the divergence start point."
Bearish divergence	"TOKEN is breaking correlation to the downside. Determine if this is distress (short it) or a buying opportunity (wait for capitulation)."
Re-coupling	"Correlation is returning to normal. Recommend closing any divergence positions and assess if a new opportunity exists."

Tips for Combining Layers Effectively¶

Match your research bucket to your persona's timeframe¶

Persona	Best Research Buckets
Capital Guardian	MOMENTUM, SECTOR_ROTATION
Base Trader	Any bucket
Tactical Alpha	VOLUME_BREAKOUT, SECTOR_ROTATION, WHALE_ACTIVITY
Scalp Specialist	VOLUME_BREAKOUT, LIQUIDATION_CLUSTERS
Swing Operator	FUNDING_RATE_ANOMALY, CORRELATION_DIVERGENCE, MOMENTUM

Keep your methodology prompt time-insensitive¶

Your research methodology is saved and reused across runs. Write it for general conditions:

Good: "Find tokens with 7d momentum > 5% and rising volume"
Bad: "Find tokens pumping after today's Fed meeting"

For time-sensitive focus, use the runtime directive field when running research, or the quick prompt when requesting recommendations.

Let the persona handle risk — let the quick prompt handle tactics¶

Don't repeat risk rules in your quick prompt. The persona already enforces them. Instead, use quick prompts for:

Which direction to focus on (long/short/both)
What pattern to look for (pullback, breakout, mean reversion)
Urgency or patience ("only if setup is A+" vs "find best available")
Specific technical conditions ("RSI < 40", "near support")

Iterate on one layer at a time¶

When refining your strategy, change only one layer per research/recommendation cycle:

Start with the default Base Trader persona and MOMENTUM bucket
Run research, get tokens, request recommendation
If results are too aggressive → switch to Capital Guardian persona
If results are wrong tokens → adjust research methodology
If results miss opportunities → refine quick prompt specificity

Quick Reference: Configuration Checklist¶

Before your first recommendation, ensure all three layers are configured:

[ ] Agent created and assigned as primary to a portfolio
[ ] Research config created with bucket + methodology prompt
[ ] Research run completed and tokens applied to portfolio
[ ] Trading strategy (persona) loaded and optionally customized
[ ] Model selected appropriate to your subscription tier

Then request recommendations using quick prompts or custom instructions.

Strategy Composition Guide¶

The Three Layers¶

Persona Overview¶

Strategy Examples¶

Example 1: Conservative Momentum Rider¶

Research Config¶

Trading Strategy¶

Quick Prompts to Use¶

Example 2: Aggressive Scalper¶

Research Config¶

Trading Strategy¶

Quick Prompts to Use¶

Example 3: Swing Trader with Funding Awareness¶

Research Config¶

Trading Strategy¶

Quick Prompts to Use¶

Example 4: Sector Rotation Tactician¶

Research Config¶

Trading Strategy¶

Quick Prompts to Use¶

Example 5: Correlation Divergence Arbitrageur¶

Research Config¶

Trading Strategy¶

Quick Prompts to Use¶

Tips for Combining Layers Effectively¶

Match your research bucket to your persona's timeframe¶

Keep your methodology prompt time-insensitive¶

Let the persona handle risk — let the quick prompt handle tactics¶

Iterate on one layer at a time¶

Quick Reference: Configuration Checklist¶

Further Reading¶