Quant & Hedge Fund Techniques for 15-Min BTC Prediction Markets

Research compilation of quantitative trading techniques applicable to Kalshi 15-minute BTC up/down markets. Focused on techniques implementable by retail traders with public data.

Executive Summary

After analyzing hedge fund and quant trading techniques, the highest-value additions to our current TAP/MGA/MRP/VRC system are:

Order Flow Imbalance (Easy, High Impact) — Already have some of this, but can refine
Regime Detection via HMM (Medium, High Impact) — Would improve VRC significantly
Funding Rate Integration (Easy, Medium Impact) — Free sentiment proxy
Fractional Kelly Sizing (Easy, High Impact) — Likely biggest immediate edge
Cross-Asset Correlation Windows (Medium, High Impact) — ETH/SPY as leading indicators

1. Market Microstructure Techniques

1.1 Order Flow Imbalance (OFI)

What it is: Measures the difference between aggressive buying and selling by analyzing how limit orders are consumed. The formula:

OFI = Σ(Buy Market Orders at Best Ask) - Σ(Sell Market Orders at Best Bid)

A positive OFI indicates more aggressive buying; negative indicates selling pressure.

Application to 15-min Kalshi trading:

Track BTC order flow on major exchanges (Binance, Coinbase, Kraken) in real-time
Aggregate OFI over rolling 1-5 minute windows
Strong OFI in one direction often precedes price moves by 30-120 seconds
Use as a confirmation signal for TAP threshold crossings

Data/Tools Needed:

WebSocket connections to exchange order books (free tier available)
Binance: wss://stream.binance.com:9443/ws/btcusdt@depth@100ms
CCXT library for multi-exchange aggregation
~100ms update frequency sufficient (not HFT)

Expected Edge Improvement: +3-5% accuracy on directional calls

Implementation Difficulty: Medium

Need to handle WebSocket connections reliably
Requires aggregation logic across exchanges
Storage for historical analysis

Key Insight: OFI is most predictive when it diverges from recent price action. Price going up but OFI turning negative = potential reversal signal.

1.2 Bid-Ask Spread Dynamics

What it is: Spread widening typically indicates:

Increased uncertainty (market makers stepping back)
Impending volatility
Low liquidity (worse for trading)

Spread compression indicates:

High conviction / low volatility expectations
Good execution environment

Application to 15-min Kalshi trading:

Wide spreads before your entry = avoid the trade (execution risk)
Sudden spread widening mid-window = volatility incoming (helps with prediction)
Track "spread velocity" — rate of change matters more than absolute level

Data/Tools Needed:

Same order book feeds as OFI
Simple calculation: spread = best_ask - best_bid
Normalize to basis points: spread_bps = (spread / mid_price) * 10000
Track rolling z-score of spread vs 1-hour mean

Expected Edge Improvement: +1-2% (mostly risk management)

Implementation Difficulty: Easy

1.3 Volume Profile Analysis (VPOC, VAH, VAL)

What it is:

VPOC (Volume Point of Control): Price level with most volume traded
VAH (Value Area High): Upper bound of 70% volume zone
VAL (Value Area Low): Lower bound of 70% volume zone

Price tends to gravitate toward VPOC and reverse at VA boundaries.

Application to 15-min Kalshi trading:

Calculate rolling 4-hour volume profile
If current price is near VAH and momentum fading → favor "Down" bet
If current price is at VAL with absorption (high volume, no breakdown) → favor "Up" bet
VPOC acts as a magnet — mean reversion target

Data/Tools Needed:

Historical candle data with volume (free from most exchanges)
Aggregate trades into price buckets (typically $50-100 buckets for BTC)
Recalculate every 5 minutes

Expected Edge Improvement: +2-3%

Implementation Difficulty: Medium

1.4 VWAP/TWAP Deviation

What it is:

VWAP: Volume-Weighted Average Price — where the "average" buyer bought
Price above VWAP = buyers in profit, may hold or add
Price below VWAP = buyers underwater, may panic sell

Application to 15-min Kalshi trading:

Calculate intraday VWAP (reset at midnight UTC or 00:00 exchange local)
Distance from VWAP indicates stretched conditions
Mean reversion more likely at >1% deviation from VWAP
Trend continuation more likely when riding along VWAP

Data/Tools Needed:

Trade-level data or candle data with volume
Simple calculation: VWAP = Σ(Price × Volume) / Σ(Volume)
Track deviation: vwap_deviation = (price - vwap) / vwap

Expected Edge Improvement: +1-2%

Implementation Difficulty: Easy

2. Statistical Arbitrage Techniques

2.1 Ornstein-Uhlenbeck Mean Reversion Model

What it is: Mathematical model for mean-reverting processes. Key parameters:

θ (theta): Speed of mean reversion (higher = faster reversion)
μ (mu): Long-term mean price will revert to
σ (sigma): Volatility of the process

dS = θ(μ - S)dt + σdW

Application to 15-min Kalshi trading:

Fit O-U process to recent price data (rolling 2-4 hours)
Calculate "half-life" of mean reversion: t_half = ln(2) / θ
If half-life < 15 minutes → mean reversion likely within our window
If half-life > 60 minutes → trend following more appropriate
Use z-score from μ as entry signal

Data/Tools Needed:

Python statsmodels for estimation
Rolling window of 1-min candles (240 data points for 4 hours)
Refit every 15 minutes

Estimation method (simplified):

# Using Augmented Dickey-Fuller test or OLS regression
# S(t) - S(t-1) = θ(μ - S(t-1)) + ε
# Regress: ΔS on S(t-1) to estimate θ

Expected Edge Improvement: +3-5% (especially for MRP signals)

Implementation Difficulty: Medium

Key Insight: This is exactly what your MRP signal is trying to capture. O-U formalization gives you a rigorous way to estimate mean reversion speed.

2.2 Hidden Markov Model (HMM) Regime Detection

What it is: Statistical model that assumes market operates in hidden "states" (regimes) that you infer from observable price/volume data. Common regimes:

Regime 1: Low volatility, mean-reverting
Regime 2: High volatility, trending
Regime 3: Choppy, unpredictable (avoid trading)

Application to 15-min Kalshi trading:

Train 2-4 state HMM on recent data (rolling 24-48 hours)
Use returns and volatility as emission variables
Infer current regime probability
Adjust strategy: MRP in regime 1, MGA in regime 2, sit out in regime 3
This supersedes your VRC with a more principled approach

Data/Tools Needed:

Python hmmlearn library
Features: 5-min returns, 5-min volatility, volume
Retrain every 4-6 hours

Sample Implementation:

from hmmlearn import GaussianHMM
model = GaussianHMM(n_components=3, covariance_type="diag")
model.fit(features)  # features = [returns, volatility, volume_change]
current_regime = model.predict(latest_features)
regime_probs = model.predict_proba(latest_features)

Expected Edge Improvement: +4-6%

Implementation Difficulty: Medium-Hard

Key Insight: The biggest edge is knowing when NOT to trade. HMM gives you principled "regime 3" detection.

2.3 Cross-Asset Cointegration (ETH, S&P Futures, DXY)

What it is: While BTC and ETH aren't perfectly correlated, they're often cointegrated — they move together over time with predictable deviations. When ETH moves and BTC doesn't (or vice versa), BTC often catches up.

Application to 15-min Kalshi trading:

Track BTC/ETH ratio in real-time
When ratio deviates >1 standard deviation from recent mean:
- Ratio up (BTC outperforming) → slight favor to "Down" on BTC
- Ratio down (ETH outperforming) → slight favor to "Up" on BTC
Also track SPY futures (ES) for macro risk-on/risk-off signals
DXY (dollar strength) inversely correlated with BTC

Data/Tools Needed:

ETH price feed (same as BTC)
ES futures or SPY (from Tradier, Yahoo Finance, or futures data)
DXY index or calculate from major currency pairs

Expected Edge Improvement: +2-4%

Implementation Difficulty: Easy-Medium

Key Insight: ETH often leads BTC by 30-120 seconds in correlated moves. This is free alpha.

2.4 Intraday Seasonality Patterns

What it is: BTC exhibits predictable intraday patterns:

Asian open (00:00 UTC): Often volatile as new liquidity enters
London open (07:00-08:00 UTC): Volume spike, trend establishment
US open (13:30-14:00 UTC): Highest volatility period
US equity close (20:00 UTC): Often reversal time
Sunday night / Monday morning: Weekend gap fills

Application to 15-min Kalshi trading:

Weight signals differently based on time of day
Avoid mean reversion bets during US open (trends are stronger)
Favor mean reversion during Asian session (lower volume, more ranging)
Track which hours have historically favored your signals

Data/Tools Needed:

Timestamp awareness in trading logic
Historical performance by hour-of-day

Expected Edge Improvement: +1-3%

Implementation Difficulty: Easy

3. Machine Learning Approaches

3.1 Feature Engineering for Short-Term Prediction

What it is: Transforming raw price/volume data into predictive features. Quality of features matters more than model complexity.

High-Value Features for 15-min BTC:

Feature	Description	Predictive Power
Return momentum	Σ(returns) over 5, 15, 60 min	Medium
Volatility ratio	Vol(5min) / Vol(1hr)	High
Volume ratio	Vol(5min) / Vol(1hr)	Medium
RSI (14-period)	Overbought/oversold	Medium
Order flow imbalance	OFI over 5 min	High
Spread z-score	Current spread vs 1hr mean	Medium
VWAP deviation	Distance from VWAP	Medium
Funding rate	Perpetual funding	High
ETH-BTC divergence	Residual from regression	Medium
Hour of day (encoded)	Cyclical encoding	Low-Medium

Application to 15-min Kalshi trading:

Use these features as inputs to your signal generation
Track feature importance over time (it shifts!)
Create interaction features: momentum * volatility_ratio

Data/Tools Needed:

All readily available from exchange APIs
Python pandas for calculation
Storage for feature history

Expected Edge Improvement: Foundation for all ML approaches

Implementation Difficulty: Medium

3.2 Gradient Boosting (XGBoost/LightGBM) for Binary Classification

What it is: Ensemble of decision trees optimized for prediction. Excellent for tabular data with mixed feature types. Outputs probability of "Up" vs "Down."

Application to 15-min Kalshi trading:

Target variable: Did price go up or down in next 15 min?
Features: All from 3.1 above
Train on rolling 7-30 days of data
Output probability directly usable with Kelly criterion

Training Approach:

import lightgbm as lgb

params = {
    'objective': 'binary',
    'metric': 'auc',
    'learning_rate': 0.05,
    'num_leaves': 31,
    'max_depth': 5,  # Keep shallow to avoid overfitting
    'min_data_in_leaf': 50,
    'feature_fraction': 0.8,
    'bagging_fraction': 0.8,
}

# Use time-series cross-validation (walk-forward)
# Never use future data in training!

Expected Edge Improvement: +3-7% (if done carefully)

Implementation Difficulty: Medium-Hard

Key Warnings:

Overfitting is the #1 risk
Use walk-forward validation ONLY
Retrain frequently (daily or every 1000 samples)
Start with fewer features, add gradually
Track out-of-sample performance religiously

3.3 LSTM/Transformer for Sequence Prediction

What it is: Neural networks designed for sequential data. Can capture complex temporal patterns.

Application to 15-min Kalshi trading:

Input: Sequence of last 60 1-minute candles (OHLCV + features)
Output: Probability of up/down
Transformers (attention mechanism) can identify which past periods matter

Honest Assessment: For 15-minute BTC prediction with retail resources:

Not recommended as primary approach
Requires significant compute for training
Prone to overfitting on small datasets
XGBoost typically outperforms for this use case
Consider only after simpler methods are exhausted

Expected Edge Improvement: +1-3% over XGBoost (marginal)

Implementation Difficulty: Hard

3.4 Online Learning / Adaptive Models

What it is: Models that update incrementally as new data arrives, adapting to regime changes.

Application to 15-min Kalshi trading:

Use online gradient descent variants
Update model after each trade with new outcome
Implement "learning rate decay" to balance stability and adaptation

Simple Implementation:

# Exponentially weighted moving update
# After each trade:
new_weight = old_weight + learning_rate * (actual - predicted) * feature
# where learning_rate decays over time

Expected Edge Improvement: +1-2% (mostly prevents decay)

Implementation Difficulty: Medium

4. Risk & Position Sizing

4.1 Kelly Criterion & Fractional Kelly

What it is: Optimal bet sizing that maximizes long-term wealth growth:

Kelly % = (p * b - q) / b
        = (p * (1 - loss%) - (1-p) * loss%) / (1 - loss%)

Where:

p = probability of winning
q = 1 - p = probability of losing
b = payout ratio (for Kalshi, typically ~0.85-0.95 after fees)

Application to 15-min Kalshi trading:

Calculate Kelly % for each signal based on your estimated probability
Use Fractional Kelly (25-50%) to account for:
- Probability estimation error
- Variance reduction
- Psychological comfort
- Model uncertainty

Example:

Your signal says 60% probability of "Up"
Kalshi payout is 90 cents on the dollar (b = 0.9)
Full Kelly = (0.6 * 0.9 - 0.4) / 0.9 = 0.156 = 15.6% of bankroll
Half Kelly = 7.8% of bankroll
Quarter Kelly = 3.9% of bankroll

Data/Tools Needed:

Accurate probability estimates (the hard part)
Payout ratios from Kalshi (varies by contract)
Bankroll tracking

Expected Edge Improvement: +5-10% on risk-adjusted returns (Sharpe ratio)

Implementation Difficulty: Easy (once you have probabilities)

Key Insight: This is probably your biggest immediate edge. Most retail traders bet flat amounts regardless of edge strength.

4.2 Drawdown-Constrained Sizing

What it is: Adjust position size based on recent performance to limit maximum drawdown.

Rules:

After consecutive losses, reduce size
After hitting daily/weekly loss limit, stop trading
After recovery, gradually increase size back to normal

Implementation:

# Daily loss limit: 5% of bankroll
# Weekly loss limit: 15% of bankroll
# After 3 consecutive losses: reduce to 50% size
# After 5 consecutive losses: stop for the day

def get_size_multiplier(recent_trades, daily_pnl, weekly_pnl):
    if weekly_pnl < -0.15:
        return 0  # Stop trading
    if daily_pnl < -0.05:
        return 0  # Stop for day
    
    consecutive_losses = count_consecutive_losses(recent_trades)
    if consecutive_losses >= 5:
        return 0  # Cool off
    if consecutive_losses >= 3:
        return 0.5
    
    return 1.0

Expected Edge Improvement: Prevents blow-ups (survival > optimization)

Implementation Difficulty: Easy

4.3 Edge Decay Detection

What it is: Monitoring whether your signals are losing predictive power over time.

Implementation:

Track rolling hit rate (last 50-100 trades)
Track rolling Brier score (calibration of probabilities)
Compare to baseline expected performance
If significantly underperforming → reduce size or pause

Warning Signs:

Hit rate dropped >10% from historical average
Brier score increased (worse calibration)
Consecutive losing days
Market regime appears to have changed

Expected Edge Improvement: Prevents giving back profits when edge deteriorates

Implementation Difficulty: Easy

4.4 Correlation-Adjusted Sizing

What it is: If you're making multiple bets in the same 15-minute window (e.g., on different platforms or overlapping contracts), they're correlated.

Rule: Reduce individual bet size when making correlated bets.

# If making N correlated bets:
adjusted_size = base_kelly / sqrt(N)

Expected Edge Improvement: Reduces variance on correlated outcomes

Implementation Difficulty: Easy

5. Behavioral/Sentiment Edge

5.1 Funding Rate as Sentiment Proxy

What it is: Perpetual futures funding rate indicates whether longs or shorts are paying:

Positive funding = longs pay shorts (market is leveraged long)
Negative funding = shorts pay longs (market is leveraged short)

Extreme funding often precedes reversals.

Application to 15-min Kalshi trading:

Track funding rates on major perps (Binance, Bybit, dYdX)
Extreme positive (>0.05%) → slight favor to "Down" (crowded long)
Extreme negative (<-0.03%) → slight favor to "Up" (crowded short)
Funding resets typically 00:00, 08:00, 16:00 UTC — volatility around these times

Data/Tools Needed:

Free APIs from major exchanges
Binance: /fapi/v1/fundingRate
Update every 15 minutes sufficient

Expected Edge Improvement: +2-3%

Implementation Difficulty: Easy

Key Insight: This is free, easy to implement, and surprisingly predictive. Should be in your system.

5.2 Open Interest Changes

What it is:

Rising open interest + rising price = new longs entering (trend confirmation)
Rising open interest + falling price = new shorts entering (trend confirmation)
Falling open interest + price move = position closing (potential reversal)

Application to 15-min Kalshi trading:

Track OI changes on major perp exchanges
OI spike + strong move in one direction → trend likely continues
OI collapse + move → likely short-term exhaustion

Data/Tools Needed:

Binance /fapi/v1/openInterest
Track rate of change, not absolute level

Expected Edge Improvement: +1-2%

Implementation Difficulty: Easy

5.3 Liquidation Data

What it is: Track large liquidations on leveraged platforms. Liquidation cascades cause waterfall price movements.

Application to 15-min Kalshi trading:

Large long liquidation → price likely to continue down (cascade)
Large short liquidation → price likely to continue up (cascade)
Monitor via exchange WebSocket or aggregators like Coinglass

Data/Tools Needed:

Coinglass API or similar aggregator
Exchange-specific liquidation streams

Expected Edge Improvement: +2-3% (mostly for avoiding bad entries)

Implementation Difficulty: Medium

5.4 Whale Wallet Monitoring (Lower Priority)

What it is: Track large BTC wallet movements for potential market impact.

Application to 15-min Kalshi trading:

Large exchange inflows often precede selling
Large exchange outflows often indicate accumulation

Honest Assessment:

15-minute timeframe too short for most wallet signals
By the time on-chain data confirms, move may be over
Lower priority than other signals

Expected Edge Improvement: +0.5-1%

Implementation Difficulty: Medium

6. Execution Alpha

6.1 Optimal Entry Timing Within Window

What it is: For a 15-minute window, when you enter matters for expected value.

Key Insights:

Don't enter at very beginning — first 1-2 minutes often see mean reversion within window
Don't enter at very end — spread often widens, bad pricing
Sweet spot: 2-5 minutes into window — early trends visible, still time for trade to work
If signal is strong and market moving fast — enter earlier to get in before move

Implementation:

def entry_timing_adjustment(minutes_into_window, signal_strength, volatility):
    # Base: prefer 2-5 minute window
    if minutes_into_window < 2:
        return "wait"
    if minutes_into_window > 12:
        return "skip" if signal_strength < 0.7 else "enter"
    
    # High volatility + strong signal → enter earlier
    if volatility > 1.5 and signal_strength > 0.65:
        return "enter"
    
    return "enter" if 2 <= minutes_into_window <= 5 else "wait"

Expected Edge Improvement: +1-2%

Implementation Difficulty: Easy

6.2 Limit Orders vs Market Orders on Kalshi

What it is:

Market orders: Immediate execution, pay the spread
Limit orders: Potentially better price, risk non-execution

For Kalshi specifically:

Spreads can be wide (5-15 cents on dollar contracts)
Limit orders at mid-price often fill
Use limit orders when you have time, market when signal is urgent

Implementation:

def order_type_decision(signal_strength, time_remaining, current_spread):
    # Strong signal + limited time → market order
    if signal_strength > 0.7 and time_remaining < 60:
        return "market"
    
    # Wide spread + time → limit order at better price
    if current_spread > 0.05 and time_remaining > 120:
        return "limit"
    
    return "limit"  # Default to limit

Expected Edge Improvement: +1-3% on execution costs

Implementation Difficulty: Easy

7. Retail Trader Advantages

7.1 Capacity Constraints Work For You

What it is: Hedge funds managing billions can't trade small markets like Kalshi 15-min BTC contracts. The market is too small for them.

Your Advantage:

You're fishing in a pond too small for whales
Less competition from sophisticated players
Inefficiencies persist longer

Implication:

Kalshi 15-min markets likely have more alpha than large, liquid markets
Your edge may persist longer than in traditional markets
Focus on this niche rather than trying to trade where big funds operate

7.2 Speed of Deployment

What it is: Hedge funds have investment committees, risk approvals, compliance reviews. You can update your strategy immediately.

Your Advantage:

Saw a new pattern? Implement it today
Strategy not working? Adjust in real-time
New data source? Add it immediately

Implication:

Iterate fast, fail fast, learn fast
Don't over-engineer — simple improvements compound quickly

7.3 Flexibility in When to Trade

What it is: Hedge funds must deploy capital continuously. You can sit out.

Your Advantage:

No edge? Don't trade
Uncertain market? Wait
Model underperforming? Pause

Implication:

Your best trades are the ones you don't make when there's no edge
Sitting out is a valid strategy that big funds can't easily do

7.4 No Career Risk

What it is: Hedge fund managers worry about quarterly performance, investor redemptions, and job security. You don't.

Your Advantage:

Can take Kelly-optimal positions (most funds bet sub-optimally)
Can hold through temporary drawdowns
Can pursue strategies that look bad short-term but have long-term edge

8. Prioritized Implementation Roadmap

Immediate (This Week)

Technique	Why	Effort
Fractional Kelly Sizing	Biggest impact, easy to add	1-2 hours
Funding Rate Integration	Free data, proven signal	2-3 hours
Time-of-Day Adjustment	Simple, effective	1 hour
Drawdown Limits	Risk management essential	1 hour

Short-Term (1-2 Weeks)

Technique	Why	Effort
Order Flow Imbalance	Refine existing microstructure signals	1-2 days
ETH-BTC Divergence	Easy cross-asset signal	4-6 hours
Entry Timing Optimization	Execution improvement	2-3 hours

Medium-Term (1 Month)

Technique	Why	Effort
Hidden Markov Model Regimes	Replace/upgrade VRC	1 week
O-U Mean Reversion Estimation	Formalize MRP	3-4 days
XGBoost Probability Model	Unified signal framework	1-2 weeks
Volume Profile (VPOC/VA)	Support/resistance refinement	2-3 days

Long-Term (When Basics Are Solid)

Technique	Why	Effort
Open Interest Integration	Additional confirmation	1 day
Liquidation Monitoring	Cascade detection	2-3 days
Online Learning Updates	Prevent edge decay	1 week
VWAP Integration	Additional mean reversion signal	1 day

9. Specific Recommendations for Your System

Based on your current TAP/MGA/MRP/VRC framework:

A. Upgrade VRC to HMM-Based Regimes

Your VRC (Volatility Regime Context) can be formalized with Hidden Markov Models:

Current: Likely using volatility thresholds
Upgrade: 3-state HMM (low-vol trending, high-vol trending, choppy)
Benefit: Probabilistic regime assessment, better "sit out" detection

B. Formalize MRP with O-U Model

Your MRP (Mean Reversion Pressure) can use Ornstein-Uhlenbeck:

Current: Likely using z-scores or RSI-type indicators
Upgrade: O-U half-life estimation
Benefit: Know WHEN mean reversion should occur, not just that price is stretched

C. Add Funding Rate to MGA

Your MGA (Momentum Gap Analysis) can incorporate funding:

Current: Likely price momentum based
Upgrade: Combine with funding rate as crowding indicator
Benefit: Momentum + crowded positions = higher reversal risk

D. Convert TAP to True Probability for Kelly

Your TAP (Time-Adjusted Probability) needs calibration:

Current: Probability estimate
Upgrade: Verify calibration with historical data
Benefit: If TAP says 60%, should win ~60% of time. If not, recalibrate.

E. Add ETH Lead Signal

ETH often leads BTC in correlated moves:

New component: "ETH Lead Indicator"
Signal: ETH moved but BTC hasn't yet → expect BTC to follow
Integration: Confirmation for TAP signals

10. Data Sources Summary

Data	Source	Cost	Update Frequency
BTC/ETH OHLCV	Binance, Coinbase	Free	1 min
Order book (depth)	Binance WS	Free	100ms
Funding rates	Binance, Bybit	Free	8 hours (use predicted)
Open interest	Binance, Coinglass	Free	5 min
Liquidations	Coinglass, Binance WS	Free	Real-time
SPY/ES futures	Yahoo Finance, Tradier	Free/Cheap	1 min
DXY	Yahoo Finance	Free	1 min
Kalshi pricing	Kalshi API	Free	Real-time

11. Metrics to Track

Performance Metrics

Hit rate: % of trades that win
Brier score: Calibration of probability estimates
Sharpe ratio: Risk-adjusted returns
Max drawdown: Worst peak-to-trough

Signal Metrics

Signal accuracy by regime: Does MRP work better in low-vol?
Signal decay: Are older signals still predictive?
Feature importance: Which inputs matter most?

Execution Metrics

Slippage: Expected vs actual fill price
Entry timing: Did we enter at optimal time?
Spread at entry: What did we pay?

Conclusion

The biggest edges available to retail Kalshi traders are:

Proper position sizing (fractional Kelly) — most underutilized
Knowing when NOT to trade (regime detection) — discipline is edge
Free sentiment data (funding rates, OI) — easy alpha
Cross-asset signals (ETH leading BTC) — simple, effective
Niche market focus — you're in a pond too small for whales

Start with the immediate priorities. Each incremental improvement compounds. Don't over-engineer — a simple, well-executed system beats a complex, poorly-implemented one.

Good luck! 🎯