Video summary

STEAL This Trading Champion’s Exact Strategy - Math Based Models for Prop Firms

Main summary

Key takeaways

Finance

Finance/Trading focus extracted

  • The video is about statistical/mechanical trading for prop firms, using data-driven trade rules and advanced risk models.
  • Topics mentioned include:
    • autocorrelation,
    • Kelly-based sizing,
    • regime/volatility modeling (e.g., GARCH-class tools),
    • drawdown shielding,
    • variance “switching” concepts.
  • No specific broad macro assets (like rates or inflation) are emphasized; the focus is mainly on execution rules, backtesting methodology, and trade/risk metrics.

Tickers / Assets / Instruments mentioned

  • Forex: strategy said to work across many forex pairs.
  • Indices / Dow Jones: mentions a centralized volume profile “for the Dow Jones.”
  • Metals: described as weaker / non-strong results (compared to other instruments).
  • Total coverage: 33 assets (strategy claimed to work across 33 assets: forex + indices).
  • CFDs and futures: mentioned in a prop-firm context.
  • Volatility regimes: referenced via GARCH / garch-class tools.
  • Specific tickers: not named.

Key presenters / sources (end of transcript)

  • Gian Luca Brun / Gian Luca Rooney
    • Host intro contains both names; the main guest is consistently referenced as Gian Luca Brun.
  • “Char Fanatics”
    • Host/interviewer name not provided in subtitles.
  • Hamilton James Hamilton
    • Credited for an autocorrelation concept described as “market of switching autoaggression” / “mark of switching” (paper name mentioned).
  • “Law E.M.”
    • Credited for discovery related to autocorrelation; paper referenced: “Stock market prices do not follow a random walk.”

Step-by-step / methodology frameworks shared

A) Convert discretionary concepts into a testable strategy (objectification)

  • Pure “concepts” (e.g., subjective ICT/FVG/harmonic/Fibonacci ideas) are not sufficient by themselves.
  • Convert concepts into objective/statistical rules.
  • Validate across large samples to reduce the chance that results are coincidence.

B) Market structure objectification (single timeframe)

Reads market structure in layers on one fixed timeframe:

  1. Peripheral structure (larger picture)
  2. Actual structure (latest breakout defining bias)
    • Also references internal structure for additional cases.

Rules:

  • Use one fixed timeframe (example given: 15-minute).
  • Do not switch timeframes for breakout confirmation.
  • Use “latest two breakouts” to decide which structure zone(s) to follow.
  • A breakout must be validated with a candle body close (not wick-only).
  • Directional potential handling:
    • Follow actual structure until a boundary is hit.
    • After directional potential is exhausted, switch to peripheral structure logic.

C) Directional potential boundary using discount/premium or volume profile

Boundary options described:

  • Discount/Premium zone (via Fibonacci)
    • Mentions checking around a 75 level (described imperfectly; likely a ~75% retracement / 0.75 area).
  • More precise method: Volume Profile
    • Use Value Area High/Low as targets/limits.
  • Claimed improvement:
    • The statistical approach using value area boundaries provides better edge than the simplistic “75” rule.

D) Internal structure handling

  • If peripheral + actual are aligned, trade can proceed.
  • If not aligned:
    • Check internal structure for the latest breakout alignment.
  • Internal structure can redefine P (peripheral) and A (actual) within the same timeframe.
  • If conditions produce a third breakout:
    • Older zones are discarded and the “latest two breakouts” rule continues.

E) “Triple Print” to objectify demand/supply zones

Candle validity rule:

  • A candle is valid if body length > longer week length (a numerical ratio method is implied).

Zone construction:

  • Require three valid candles (a “triple print”) near the zone to qualify demand/supply.
  • You may include invalid candles as long as you count the required three valid ones.

Zone placement:

  • Use the latest three candles before the impulse (triple print is used to qualify the candles defining the zone).
  • If no triple print exists near the expected area:
    • “Go down” (look for the next qualifying structure).

F) “Wick Gap” entry filtration (liquidity/trigger quality by wick distance)

  • Setup: inefficiency using wick structure around entry/stop.
  • Define two closest wicks to the stop loss.

  • Filter threshold uses pip distance between those two wicks:

    • ≤ 0.2 pips: not profitable
    • ≥ 0.3 pips: becomes “profitable/break-even” (improved)
    • Explored examples:
      • 0.4+: first profitable result mentioned
      • Higher values improve until a maximum threshold, after which performance worsens

  • Process:

    • Test incrementally (0.1, 0.2, 0.3, …)
    • Increase until performance degrades again
  • Claimed: applies similarly for long and short setups.

G) Backtesting framework to reduce overfitting

Core goals:

  • Define backtest as validation of edge, not simply bar replay.

Bias and overfitting controls:

  • Avoid lookahead bias (example given: trading the same asset with later-known direction).
  • Guard against overfitting using chunk optimization:
    • Split ~10,000 trades (example) into 10 chunks (100 trades each).
    • Performance should remain stable across chunks.
    • Overfit warning: one chunk looks great while others collapse.

Realism:

  • Include realistic costs:
    • spread,
    • commission,
    • slippage,
    • swap.

Sample size guidance:

  • Minimum ~1,000 trades suggested to reduce variance misguiding conclusions.
  • More samples → less variance.

H) Performance metrics used (order)

  1. Expectancy (primary “truth” metric)

    • Defined as return per $ risk.
    • Example interpretation:
      • Expectancy 0.10 cents (as stated) means about +10% per $1 risk (interpreted as 1 + 0.10 return).
    • Ranges described:
      • 0.10–0.30: low / modest “healthy lower” (break-even to modest)
      • 0.40–0.50: very strong
      • >0.50: amazing
    • Claim: win rate is “vanity”; expectancy matters most.

  2. Profit factor (with caution)

    • Profit factor = total profit / total loss

    • 1 profitable; <1 losing

    • Caution: can be inflated by a single lucky outlier trade.

  3. Sharpe ratio / “Share pressure”

    • A consistency measure (Sharpe-like), emphasizing smoother equity curve behavior.

  4. “Optimal target” (take-profit selection)

    • When using fixed risk-reward strategies, test which R multiple is best.
    • Example approach:
      • evaluate ranges (e.g., 1R to 6R) to find break-even/limit behavior.

I) Risk model: Autocorrelation-based variable risk (positive/negative)

Core premise:

  • Coin flips are IID (50/50), but market returns are autocorrelated by regimes.

Steps:

  • Compute:
    • Unconditional win rate (UC)
    • Conditional win rate (CO) after a losing trade (described as win rate following a losing trade)
  • Rule:
    • If UC > CO ⇒ positive autocorrelation
    • If UC < CO ⇒ negative autocorrelation
    • If equal ⇒ zero autocorrelation

Autocorrelation strength scaling (0–10 bands):

  • 0–2: none
  • 3–4: weak
  • 5–6: solid
  • 7–8: stronger
  • 9–10: extreme

Risk application:

  • Positive autocorrelation
    • After wins: next trade more likely wins → risk can be higher
    • After losses: next trade more likely losses → risk lower or skip
  • Negative autocorrelation
    • Expect alternation → opposite risk adjustments
  • Caution:
    • If you vary risk without knowing autocorrelation, you may be “gambling.”
    • Suggested safety baseline: fixed position sizing when autocorrelation is ~0 or unknown.
  • Caution on risk/reward:
    • Effects can be diluted by high R, because outcomes cluster less reliably.

J) Kelly criterion-based sizing (fractional Kelly)

Concept:

  • Kelly uses win% and loss% relative to average loss/error.

Cautions and rule:

  • Full Kelly is often too aggressive → use fractional Kelly.

Sizing example concept:

  • Use different Kelly “fractions” depending on prop/broker constraints.
  • Mentions “one point Kelly scaled” (e.g., Kelly/5, /4, /3) and “one 0.1 Kelly…” (example stated unclearly).
  • Main idea: fractional Kelly with scaling.

K) Drawdown Shield model (prop-firm payout stability)

Observation:

  • Drawdowns do not occur monotonically; winners can appear even during losing months.

Mechanism (account splitting):

  • Instead of one large account (example: $200k),
    • use multiple accounts (example implies 450k total capacity).
  • Risk ~1% per account, but rotate which account is actively traded.

During drawdown weeks:

  • “Shield” equity by switching to other accounts so you can still qualify for prop firm payouts.

Outcome and tradeoff:

  • In losing months, the model aims for some profits to keep payouts alive.
  • Tradeoff: lower payout size in winning months, but smoother payout consistency.
  • Prop-firm mechanics mentioned:
    • Breaking payout/drawdown thresholds can cause payout denial → keep payout/sizing safer to maintain consistency.

L) “Variance Accelerator Model” (VAM) / market switching autocorrelation

  • Uses the idea that strategy autocorrelation can shift between positive and negative “states” during a drawdown phase.
  • Applied concept:
    • After sequences like 7–8 losing trades, autocorrelation shifts strongly → reversal probability rises.
  • Combines:
    • drawdown state transitions + Kelly-like sizing.

M) “GARCH / GARC H” volatility regime tool (volatility clustering)

Purpose:

  • Forecast the “weather” of the next session: volatility level, not direction.

Regime buckets:

  • calm / middle / extreme (three levels)

Numerical claims and rules:

  • In “calm” regimes:
    • strategy baseline: win rate shifts to about 75% loss rate
    • elsewhere baseline win rate described as about 65%
  • Controversial rule (reversal):
    • If model predicts poor performance (e.g., 75% loss rate),
    • reverse the side (e.g., if model says long → take short),
    • based on data indicating improved win rate in that reversed condition.
  • Emphasis: follow what the data indicates for the regime, not the explanation.

Key numbers & explicit recommendations/cautions

Performance / trade statistics

  • Claimed championship performance:
    • over 200% return in 9 months.
  • Win rate / reward horizon references:
    • Guest mentions ~60% win rate with 1 to 4 week reward (intro claim).
  • Regime example:
    • Baseline win rate around 65%, but in calm regimes described as ~75% loss rate.
  • Expectancy targets:
    • 0.10–0.30: low/modest healthy lower bound
    • 0.40–0.50: very strong
    • >0.50: amazing
  • Backtest sample size:
    • minimum ~1,000 trades
  • Chunk optimization:
    • example: 10,000 trades split into 10 chunks
  • Wick gap filter:
    • ≤ 0.2 pips not profitable
    • first profitable at ~0.4 pips or above
  • Worst-case loss streak (Monte Carlo claim):
    • up to 13 losing trades in a row

Risk control / model rules

  • No “guessing”: execute rules only after validation.
  • Breakout definition requires candle body close.
  • Overfitting rejection:
    • if equity collapses in chunks or improves only in lucky periods → reject.
  • Include all costs in backtests:
    • spread, commission, slippage, swap.
  • Autocorrelation:
    • if autocorrelation near 0 → don’t vary risk (fixed sizing).
  • Kelly:
    • must be fractional due to regime shifts.
  • Drawdown Shield:
    • aim for consistent prop payouts, even if performance dips.

Disclosures / disclaimers

  • Sponsor disclosures in transcript:
    • NinjaTrader,
    • Apex Trader Funding,
    • Tradzella,
    • Chart Academy.
  • No explicit “not financial advice” line appears in subtitles, but the video repeatedly emphasizes data validation over trust.

Mentioned tools/platforms (non-asset but finance-relevant)

  • NinjaTrader: futures platform; simulator mode.
  • Apex Trader Funding: prop firm; payout rules; claims about $600M+ paid.
  • TradingView: volume profile used; manual data constraints discussed.
  • Tradezella: AI trading journal/sync/backtest tooling.
  • Chart Academy: education platform.

Original video