Live Trading Monitor

Paper trading simulation only. Not financial advice.
Data delays: Crypto ~2-5s (Kraken/FreeCryptoAPI) • Forex ~15s (TwelveData), cross-pairs ~60s (CurrencyLayer) • Stocks real-time quotes (Finnhub, NYSE hours only) • Candles from Kraken (w/ volume)/CoinGecko/Finnhub. Kraken = Ontario-valid exchange with real bid/ask spreads.

System Health

Prices: checking...

Signals: checking...

Trades: checking...

Portfolio: checking...

Win Rate: checking...

Auto-Execute: checking...

Market Regime: checking...

Loading system diagnostics...

Portfolio Value

$10,000

+$0 (0%)

Open Positions

No active trades

Win Rate (24h)

--%

No data

Today's P&L

No trades today

Drawdown

From peak

Circuit Breakers

ALL OK

No limits hit

Crypto

₿

Loading crypto prices...

Forex

💱

Loading forex prices...

Stocks MARKET CLOSED

📈

Stocks fetch during NYSE market hours (9:30 AM - 4:00 PM ET).
Click Refresh Prices to force-fetch.

Active Signals

	Time	Symbol	Algorithm	Type	Strength	Entry	TP / SL	Hold	Expires	Action
Loading signals...

Manual Trade ▼

Asset Class

Symbol

Direction

Position Size: 5%

Take Profit %

Stop Loss %

Max Hold Hours

Open Positions

Symbol	Dir	Entry Price	Current	P&L %	P&L $	Hold Time	SL / TP	Action
Loading positions...

Trade History

Exit Time	Symbol	Dir	Entry	Exit	P&L %	Reason	Algorithm	Hold
Loading history...

Top Crypto Movers (Dynamic Discovery)

Scans Binance for crypto pairs with >3% 24h change and >$500K volume that aren't in our static watchlist. Runs 8 algorithms on each mover to find actionable signals.

Symbol	Price	24h Change	Volume	Direction	Signals	Details
Click Scan Movers to discover trending crypto pairs

Trading Summary

Total Trades

Wins

Losses

Win Rate

Avg Return

Avg Win

Avg Loss

Profit Factor

Total P&L

Max Drawdown

Best Trade

Worst Trade

Algorithm Leaderboard

Algorithm	Trades	Win Rate	Avg Return	Total P&L	Profit Factor
Loading algorithm leaderboard rankings...

Equity Curve

Trades by Asset Class

Trades by Algorithm

Trades by Time of Day (UTC)

System Analysis DIAGNOSIS

Current Status

• Real-time multi-asset paper trading
• 14 crypto pairs, 10 forex pairs, 12 stocks
• $10K starting capital, 5% position sizing ($500/trade)
• 20 algorithms active with regime gating
• 12 closed trades, 58.33% WR, $45.58 PnL
• Profit factor: 2.61

Issues Identified

• Tiny sample: Only 12 closed trades — no statistical significance
• Zero overlap: Top backtested algos not deployed forward
• Walk-forward: DEPLOYED — walk_forward_validator.py + cross_asset_ml_intelligence.py
• Ensemble weighting: DEPLOYED — cross_asset_ml_intelligence.py computes optimal weights
• EGARCH sizing: DEPLOYED — egarch_position_sizer.py (daily 4:15 PM EST)
• Some algos at 0% WR: Consensus regime-gated (Feb 12). RSI Reversal still learning

What's Working

• Multi-asset diversification (crypto + forex + stocks)
• 20 algorithms with regime gating on 19/20
• ML grid search parameter learning is ACTIVE
• Transparent P&L with real-time monitoring
• Automated exit rules (TP/SL/max-hold)
• Sector concentration cap (max 3 per sector)

Algorithm Performance Breakdown (Forward)

Algorithm	Win Rate	PnL	Status
Ichimoku Cloud	80%	+$44.97	TOP
StochRSI Crossover	100%	+$16.39	TOP
MACD Divergence	--	--	LEARNING
Consensus	0%	-$3.28	WORST
RSI Reversal	0%	-$0.17	WORST

ML Parameter Learning Status

Grid Search Learning: ACTIVE

Ichimoku learned: TP 8.5% (was 5%), SL 5% (was 2.5%). StochRSI learned: TP 7.75% (was 5%). Total signals: 2,491 learned vs 500 original. The system is adapting parameters based on forward trade outcomes, but only grid search is used — no full ML pipeline (gradient boosting, neural nets) yet.

System Architecture

Data Sources LIVE

FreeCryptoAPI — 14 crypto pairs (BTC, ETH, SOL, etc.)
TwelveData — 10 forex pairs (EUR/USD, GBP/USD, etc.)
Finnhub — 12 stock tickers (AAPL, MSFT, TSLA, etc.)

Signal Pipeline 20 ALGOS

Prices → live_signals.php → lm_signals table → live_trade.php → lm_trades table. Each algorithm scores independently. Regime gate filters out counter-trend signals.

Risk Management ACTIVE

5% position sizing, 10 max concurrent positions, auto stop-loss & take-profit, max-hold time limit, sector concentration cap (max 3), circuit breakers for drawdown protection.

The 20 Algorithms

1. Ichimoku Cloud

Trend + momentum

2. StochRSI Crossover

Momentum oscillator

3. MACD Divergence

Trend reversal

4. RSI Reversal

Mean reversion

5. Bollinger Squeeze

Volatility breakout

6. Volume Breakout

Volume + price

7. EMA Crossover

Trend following

8. Support/Resistance

Key price levels

9. VWAP Deviation

Institutional flow

10. OBV Trend

Volume confirmation

11. ADX Trend

Trend strength

12. Parabolic SAR

Trend reversal

13. Keltner Channel

Volatility bands

14. Fibonacci Retrace

Key levels

15. Williams %R

Overbought/oversold

16. CCI Divergence

Commodity channel

17. Stochastic Momentum

SMI oscillator

18. Alpha Predator

Multi-signal fusion

19. VAM

Volume-adjusted momentum

20. Challenger Bot

Smart money consensus

World-Class Comparison

Metric	Renaissance	Two Sigma	Citadel	Ours
Sharpe Ratio	3-6	1.5-2.5	2-4	TBD (12 trades)
Annual Return	66%	15-30%	20-40%	TBD
Algorithms	100+	50-200	100+	20
ML Pipeline	Full	Full	Full	Grid search only
Risk Management	EGARCH+VaR	EGARCH+VaR	Dynamic	Fixed 5%

Target Benchmarks

Win Rate

58.33%

Target: 50-65%

Profit Factor

2.61

Target: >1.5

Sharpe Ratio

TBD

Target: >1.5

Max Drawdown

<20%

Professional std

Min Trades

385

95% confidence

Research & Future ROADMAP

Academic Foundations

Systematic Trading

Narang (2013): “Inside the Black Box” — systematic trading requires alpha models, risk models, transaction cost models, and execution algorithms working together.
Key insight: No single alpha model is sufficient. Combining uncorrelated signals via ensemble methods dramatically improves risk-adjusted returns.

Regime Detection

Hamilton (1989): Markov-switching models identify bull/bear/sideways regimes with high accuracy. Most strategies fail when deployed in the wrong regime.
Our status: Regime gating is active on 19/20 algos — this is a strong foundation, but currently uses simple SMA-based detection rather than HMM.

Multi-Algorithm Ensembles

Dietterich (2000): Ensemble methods reduce variance and improve generalization. Combining diverse models outperforms any single model, provided individual models have some predictive edge.
Key insight: 3-5 diverse strategy families beats 20 correlated momentum variants.

Position Sizing & Risk

Kelly Criterion: Optimal bet sizing maximizes long-term growth rate. Quarter-Kelly is recommended for noisy estimates.
EGARCH (Nelson 1991): Volatility forecasting model that captures leverage effects — used by Renaissance, Two Sigma, and Citadel for dynamic position sizing.

Strategy Families Deployed

Trend Following

Ichimoku, EMA Crossover, Parabolic SAR, ADX — ride momentum in the direction of the prevailing trend. Best in bull/bear regimes.

Mean Reversion

RSI Reversal, Bollinger Squeeze, Williams %R, CCI Divergence — trade price snapping back to the mean. Best in sideways regimes.

Momentum Oscillators

StochRSI, MACD, Stochastic Momentum — detect acceleration and deceleration in price movement. Works across regimes with filters.

Volume-Based

Volume Breakout, OBV Trend, VWAP Deviation, VAM — confirm price moves with volume. High-confidence signals when aligned with price.

Multi-Signal Fusion

Alpha Predator, Consensus, Challenger Bot — combine multiple indicators into a single score. Highest conviction, lowest frequency.

Key Level Trading

Support/Resistance, Fibonacci Retracement, Keltner Channel — trade bounces and breakouts at structurally significant price levels.

Walk-Forward Validation (Missing)

Pardo (2008): The Evaluation and Optimization of Trading Strategies

Walk-forward analysis trains on in-sample data, then tests on out-of-sample data, rolling the window forward. This is the gold standard for avoiding overfitting. Our system currently lacks this — the grid search parameter learning is a step in the right direction, but it does not partition data into train/test windows. Priority P2 on the roadmap.

Future Priorities

Priority	Improvement	Expected Impact	Status
P1	Deploy top backtested algos via bridge	Close backtest-forward gap	PLANNED
P2	Walk-forward validation pipeline	Eliminate overfitting	PLANNED
P3	EGARCH volatility-based position sizing	Dynamic risk adjustment	PLANNED
P4	Ensemble weighting (IC-based)	+16% return improvement	PLANNED
P5	Portfolio-level risk controller	VaR/CVaR limits	PLANNED

Benchmark Gap Analysis

Dimension	World-Class	Ours	Gap
Sharpe Ratio	2-6	TBD (12 trades)	HIGH
Active Algos	50-200	20	MED
ML Pipeline	Full pipeline	Grid search only	HIGH
Risk Sizing	EGARCH + VaR	Fixed 5%	HIGH
Validation	Walk-forward	None	HIGH

Sources: Narang (2013) “Inside the Black Box”, Hamilton (1989) Markov Regime Switching, Dietterich (2000) Ensemble Methods, Nelson (1991) EGARCH, Pardo (2008) Walk-Forward Analysis, Kelly (1956) Optimal Sizing, FinRL ACM ICAIF 2024 Competition.

Glossary of Trading Terms iPlain-language explanations of every term used on this dashboard.

▼

Paper Trading

Simulated trading with fake money ($10,000 starting balance). No real money is risked. The system pretends to buy and sell at real market prices to test strategies.

Why it matters: Lets you evaluate strategy performance without financial risk.

Signal

An alert generated when one of the 13 active algorithms detects a trading opportunity. A signal says "this asset might go up (or down) based on these technical patterns." Not all signals become trades. (7 stock-only algorithms are currently paused.)

Signals are suggestions, not guarantees. The system filters for the strongest ones.

Win Rate

The percentage of closed trades that made money. Example: 60% win rate means 6 out of 10 trades were profitable. A win rate above 50% is good; above 60% is excellent.

Win rate alone doesn't tell the whole story — you also need to know HOW MUCH each win/loss was.

Take Profit (TP)

The price target where a trade automatically closes to lock in profit. Example: TP of +2% means the system sells when the price goes up 2% from entry.

Locks in gains before the price reverses. Prevents greed from turning a winner into a loser.

Stop Loss (SL)

The price level where a trade automatically closes to limit damage. Example: SL of -1.5% means the system sells if the price drops 1.5% from entry.

The most important risk management tool. Limits the maximum you can lose on any single trade.

Drawdown

The percentage drop from the portfolio's peak value. If the portfolio hit $10,500 then dropped to $10,200, the drawdown is ~2.9%. Measures "how much pain" the strategy experiences.

High drawdown = the strategy goes through rough patches. Even profitable systems have drawdowns.

Circuit Breaker

Safety limits that automatically stop trading when losses get too large. Like a fuse box for your portfolio. If too many trades lose in a row, or daily losses exceed a threshold, trading pauses.

Prevents a bad day from becoming a catastrophe. The system protects itself from runaway losses.

Position Sizing (5%)

Each trade uses 5% of the total portfolio. With a $10,000 portfolio, each trade risks ~$500. This limits exposure — even if a trade goes wrong, only 5% is affected.

Professional risk management. Never bet everything on one trade.

P&L (Profit & Loss)

The net gain or loss on a trade or the entire portfolio. Shown as both a dollar amount ($+50) and a percentage (+0.5%). Green = profit, red = loss.

Equity Curve

A chart showing the portfolio's total value over time. An upward-sloping equity curve means the strategy is making money overall. Dips are normal but the trend should be up.

Understanding the 20 Algorithms (12 Active, 7 Stock Algos Paused + Challenger Bot) iEach algorithm looks for different patterns. Multiple algorithms agreeing = stronger signal. 7 stock-only algorithms are currently paused.

▼

The system has 20 algorithms total. 13 are active (12 original + Challenger Bot). 7 stock-only algorithms are paused (ETF Masters, Sector Rotation, Sector Momentum, Blue Chip Growth, Technical Momentum, Composite Rating, Cursor Genius) due to poor TP/SL execution despite strong directional accuracy. They fall into these categories:

Core Algorithms (8)

Momentum Burst — detects sudden price acceleration
RSI Reversal — finds oversold bounces (RSI below 30 then rising)
Breakout 24h — price breaking above its 24-hour high
DCA Dip — identifies dips in an uptrend worth buying
Bollinger Squeeze — low volatility about to expand (like a spring)
MACD Crossover — momentum shifting from bearish to bullish
Consensus — only fires when 3+ other algorithms agree
Volatility Breakout — price breaking out of a tight range

Science-Backed (5)

Trend Sniper — multi-timeframe trend alignment (ADX + SMA)
Dip Recovery — buying dips confirmed by volume recovery
Volume Spike — abnormal volume preceding price moves
VAM (Vol-Adjusted Momentum) — momentum weighted by volume conviction
Mean Reversion Sniper — extreme deviation from average, likely to snap back

Advanced Technical (6)

ADX Trend Strength — measures how strong a trend is (not direction)
StochRSI Crossover — momentum oscillator with faster response
Awesome Oscillator — market momentum using median price
RSI(2) Scalp — ultra-short 2-period RSI for quick reversals
Ichimoku Cloud — Japanese system using support/resistance clouds
Alpha Predator — multi-indicator composite for highest-conviction trades

Asset Classes Explained iWe trade 3 different types of assets, each with different characteristics.

▼

Crypto (32 pairs)

Cryptocurrencies like Bitcoin, Ethereum, Solana. Traded 24/7, never closes. Most volatile — can move 5-20% in a day. Prices via Kraken/FreeCryptoAPI (2-5 second delay). Fee: 0.20% per trade (NDAX rate).

Forex (10 pairs)

Currency pairs like EUR/USD, GBP/JPY. Trades Mon-Fri, 24 hours. Less volatile than crypto — typical daily moves of 0.5-2%. Prices via TwelveData (~15s delay). Fee: 0 (cost is in the spread).

Stocks (12 symbols)

Major US stocks: AAPL, MSFT, GOOGL, AMZN, NVDA, META, JPM, WMT, XOM, NFLX, JNJ, BAC. NYSE hours only (Mon-Fri 9:30-4:00 ET). Prices via Finnhub (real-time). Fee: $0.0099/share, min $1.99 (Moomoo rate).

What We Don't Have (Yet) iFull transparency about what this system can't do today. We believe listing our gaps is as important as listing our features.

▼

No system is complete. Here's what's honestly missing and what we're working toward:

Current Gaps

⚠No auto-execution. The system generates 26+ signals per scan every 30 minutes, but signals must be manually executed via the Execute button. Signals expire after 30 minutes — meaning the page often shows 0 active signals between scans. Auto-execution of high-conviction signals is planned.

⚠No historical stress tests. Our algorithms haven't been backtested against the 2008 crash, COVID crash, or other black-swan events. Self-learning only trains on recent trades, so extreme tail-risk scenarios are uncharted territory.

⚠No statistical significance testing. When self-learning says "win rate improved 3%", we don't yet test whether that improvement is statistically meaningful or just noise. Small sample sizes can create illusions of improvement that vanish with more data.

⚠No real-money track record. Everything is paper-traded or backtested. We have no audited live-money performance. Paper trading doesn't account for slippage, partial fills, or the psychological pressure of real capital.

⚠Recency bias in learning. The self-learning system optimizes against recent closed trades. If market regime shifts dramatically (bull → bear, low vol → high vol), learned parameters may become stale before the system adapts. There's no mechanism yet to detect regime changes proactively.

⚠Limited trade history. The system launched Feb 9, 2026. With only a handful of closed trades, performance metrics (win rate, profit factor, max drawdown) are not statistically reliable yet. Hundreds of trades are needed before the numbers stabilize.

Recently Implemented

✔Statistical significance testing. Binomial confidence intervals and p-value testing now applied to win rates. Reports whether performance is statistically better than coin flip or just noise. Wilson confidence intervals on all tier breakdowns.

✔Regime detection. Market classified as bull/bear/sideways/volatile for crypto (BTC SMA + volatility), forex (USDJPY trend), and stocks (individual SMA). Regime data shown in System Health panel and used to gate signal generation.

✔Discord alerts. Strong signals (strength 80+) now trigger Discord webhook notifications with symbol, algorithm, and trade parameters. No need to watch the page constantly.

Still Planned

✓Auto-execute strong signals. Automatically enter trades when consensus signals score above a threshold, with proper position sizing and risk management already in place.

✓Historical backtesting engine. Replay algorithms against historical data covering multiple market regimes (crashes, rallies, sideways) to stress-test each strategy before deploying it live.

System Health

Equity Curve

Trades by Asset Class

Trades by Algorithm

Trades by Time of Day (UTC)

Current Status

Issues Identified

What's Working

Algorithm Performance Breakdown (Forward)

ML Parameter Learning Status

System Architecture

The 20 Algorithms

World-Class Comparison

Target Benchmarks

Academic Foundations

Systematic Trading

Regime Detection

Multi-Algorithm Ensembles

Position Sizing & Risk

Strategy Families Deployed

Walk-Forward Validation (Missing)

Future Priorities

Benchmark Gap Analysis

Glossary of Trading Terms iPlain-language explanations of every term used on this dashboard.

Understanding the 20 Algorithms (12 Active, 7 Stock Algos Paused + Challenger Bot) iEach algorithm looks for different patterns. Multiple algorithms agreeing = stronger signal. 7 stock-only algorithms are currently paused.

Asset Classes Explained iWe trade 3 different types of assets, each with different characteristics.

What We Don't Have (Yet) iFull transparency about what this system can't do today. We believe listing our gaps is as important as listing our features.

System Analysis DIAGNOSIS

Data Sources & Transparency LIVE