Skip to main content
|View as Markdown|

AI Quant Market Radar & The PhD Mathematics Behind It

Most "AI trading tools" are just chat wrappers that send a text prompt to an external server and guess direction.

TM Pro's AI Quant Regime & Portfolio Risk Engine is fundamentally different:

  • Every 5 seconds (and on every candle close), the TM Pro EA streams 120 closed OHLCV bars from your live MetaTrader chart directly into TM Pro's on-device quantitative engine.
  • It runs 100% locally on your CPU in under 2 milliseconds—requiring zero cloud API keys, zero monthly token costs, and zero network latency.
TM Pro AI tab showing the PhD Quant Regime & Portfolio Risk Engine
Live MT5 Quant telemetry on the AI tab — Regime badge, Hurst Exponent, Yang-Zhang Volatility, Efficiency Ratio, Z-Score, Quarter-Kelly recommendation, and 95% 1-Day Parametric VaR.

Part 1: The Trader's 3-Second Playbook (Plain English)​

You do not need a PhD in quantitative finance to trade with the AI Quant Radar. Every calculation is distilled into one color-coded Regime Badge and one "Apply Setup" button on your Trade tab:

How to Trade Each of the 4 Regime Badges​

Regime BadgeWhen It AppearsHow to Trade It Effectively
TREND BULL · BUY (Green)Hurst H > 0.55 and positive directional efficiencyInstitutional order-flow is persistent upward. Focus on long pullbacks and breakouts. Use a wider 3.2x ATR Take Profit and enable Trailing Stop so you don't cap big trend runners.
TREND BEAR · SELL (Red)Hurst H > 0.55 and negative directional efficiencyInstitutional order-flow is persistent downward. Focus on short pullbacks and breakdowns. Let the EA trail your stop behind lower highs.
MEAN REVERT (Amber)Hurst H < 0.45 or price is stretched (`Z
HIGH VOL CHOP · NEUTRAL (Muted/Warning)Hurst H ~ 0.50 (random walk) or extreme volatility spikePrice action has no statistical memory (50/50 coin-flip noise). Best action: sit on your hands or let Apply Setup cut your risk down to a defensive 0.25%.

What Clicking "Apply Setup" Does Automatically​

Inside the AI Quant Regime & Alpha Radar card on the Trade tab, clicking Apply Setup transfers five volatility-calibrated parameters into your Order Ticket in a single click:

  1. Risk %: Set to the Quarter-Kelly Optimal fraction (0.25% to 2.00%) based on current regime edge quality.
  2. Stop Loss (pips): Calibrated to 1.5x ATR in trends (or 1.8x ATR in mean-reversion) so normal candle noise doesn't prematurely stop you out.
  3. Take Profit (pips): Calibrated to 3.2x ATR in persistent trends (or 2.0x ATR in ranges).
  4. Break-Even Trigger (pips): Enabled at 1.0x ATR in profit.
  5. Trailing Stop (pips): Enabled at 1.2x ATR behind peak price.

As soon as you click Apply Setup, look right above your cursor at the top SELL / BUY buttons—your new lot size, SL / TP prices, and exact dollar risk are already updated and ready to arm!

Part 2: The Institutional PhD Mathematics Under the Hood​

For systematic traders, quants, and fund managers who want to inspect the exact stochastic calculus driving TM Pro's telemetry, here are the five estimators running on every 120-bar window (N = 120):

1. Rescaled-Range (R/S) Hurst Exponent (H)​

Proposed by Harold Edwin Hurst and popularized in financial fractals by Benoit Mandelbrot, the Hurst Exponent measures the long-term memory of log-returns across multiple lag windows n in {8, 16, 32, 64}:

Log-Returns:           r_t = ln(Close_t / Close_{t-1})
Mean-Adjusted Series:  Y_k = sum_{i=1..k} (r_i - mean(r))
Cumulative Range:      R(n) = max_{1<=k<=n}(Y_k) - min_{1<=k<=n}(Y_k)
Sample Std Deviation:  S(n) = sqrt( (1/n) * sum_{i=1..n} (r_i - mean(r))^2 )
Scaling Power Law:     E[ R(n) / S(n) ] = C * n^H
OLS Log-Log Slope:     H = Cov( ln(n), ln(R(n)/S(n)) ) / Var( ln(n) )
  • H = 0.50: Geometric Brownian Motion (memoryless random walk).
  • H in (0.55, 1.00): Persistent Fractal Process—positive autocorrelation where trending moves statistically follow trending moves.
  • H in (0.00, 0.45): Anti-Persistent (Mean-Reverting) Process—negative autocorrelation where up-bars are statistically followed by down-bars.

2. Yang-Zhang Minimum-Variance OHLC Volatility (sigma_YZ)​

Standard close-to-close volatility ignores intrabar highs/lows and overnight opening gaps. TM Pro implements Yang & Zhang's (2000) unbiased, minimum-variance estimator, which combines overnight jump variance (sigma_o^2), open-to-close drift variance (sigma_c^2), and Rogers-Satchell intrabar extreme variance (sigma_RS^2):

Overnight Log-Jump:    o_i = ln(Open_i / Close_{i-1})
Open-to-Close Drift:   c_i = ln(Close_i / Open_i)
Rogers-Satchell Term:  rs_i = ln(High_i / Close_i)*ln(High_i / Open_i) + ln(Low_i / Close_i)*ln(Low_i / Open_i)

Optimal Weighting:     k = 0.34 / ( 1.34 + (N + 1) / (N - 1) )
Per-Bar YZ Variance:   sigma_YZ^2 = Var(o) + k * Var(c) + (1 - k) * Mean(rs)
Annualized Vol (%):    Vol_Annual = sqrt( sigma_YZ^2 * BarsPerYear ) * 100

This makes TM Pro's volatility reading up to 14x more statistically efficient than standard close-to-close standard deviation on small sample windows.

3. Kaufman Efficiency Ratio (ER), Wilder ATR & Z-Score Stretch (Z)​

To measure directional signal-to-noise alongside mean-reversion stretch over a 20-bar lookback (L = 20):

True Range (TR_t):     max( High_t - Low_t, |High_t - Close_{t-1}|, |Low_t - Close_{t-1}| )
ATR (14-bar):          (1 / 14) * sum_{t=N-13..N} TR_t   (converted to Pips)
Efficiency Ratio (ER): (Close_N - Close_{N-20}) / sum_{t=N-19..N} |Close_t - Close_{t-1}|   in [-1, +1]
Z-Score Stretch (Z):   (Close_N - SMA_20) / StdDev_20
  • When |ER| > 0.35 and H > 0.55, directional trend velocity is clean and low-noise.
  • When |Z| > 2.0, price is stretched more than 2 standard deviations from its 20-bar equilibrium—warning trend traders not to chase late entries.

4. Fractional Quarter-Kelly Optimal Sizing (f*)​

The classic Kelly Criterion (John L. Kelly Jr., 1956; Edward O. Thorp) maximizes the expected geometric growth rate of capital E[ln(1 + f * R)]. Because full Kelly exhibits high drawdown variance under parameter estimation error, institutional quant desks use Fractional Quarter-Kelly (0.25 * f_full):

Estimated Win Prob:    p = clamp( 0.50 + 0.18 * Confidence * |ER|, 0.42, 0.68 )
Loss Probability:      q = 1 - p
Payoff Odds Ratio:     b = Suggested_TP_Pips / Suggested_SL_Pips
Full Kelly Fraction:   f_full = (p * b - q) / b
Quarter-Kelly Risk %:  f* = clamp( 0.25 * f_full * 100, 0.25%, 2.00% )

5. Portfolio 95% 1-Day Parametric Value-at-Risk (VaR_95) & Stop-Locked Risk Audit​

In the AI tab under Portfolio Risk Audit, TM Pro continuously evaluates all open positions on your account against two institutional risk metrics:

1. Stop-Locked Cash Risk ($):
   Sum across all open positions of: |OpenPrice - StopLoss| * TickValue * Lots
   (Flags any unprotected position with StopLoss == 0 immediately in amber/red)

2. 95% 1-Day Parametric VaR ($):
   DailySigma   = (Vol_Annual / 100) / sqrt(252)
   VaR_95_1Day  = 1.64485 * DailySigma * GrossNotionalExposureUSD

This tells you—in real dollars—the maximum 1-day statistical loss expected on 95% of trading days given current market volatility, alongside your hard stop-loss floor.