Video summary

A estratégia de alguém que está no top 1.7% dos traders | Outliers no GainClass

Main summary

Key takeaways

Finance

Finance-focused summary

The video is a live trading/automation discussion (Brazil) hosted by the GameClass (XP) program. Guests from Outliers Invest claim they rank in the top 1.7% among XP traders (for 2026) using a Copilot/ranking tool plus a control account (i.e., “robots” operated in a controlled setup).

The main technical content focuses on:

  • How they build and validate algorithmic trading robots
  • How they manage drawdown and risk at the portfolio level
  • What metrics and cost assumptions matter when moving from backtests to live / copied trading

Tickers / assets / instruments / platforms mentioned

  • XP (broker/platform context; XP group)
  • Mini-index futures (referenced for execution costs/fees)
  • Dollar chart / contracts (used as an example of liquidity/dislocation risk)
  • Rocket Trader (platform mentioned; also referenced as offering AI/real-time signals)
  • Drawdown (Draudal / drawdown) (risk metric repeatedly discussed)
  • UNTIC (XP platform for automated investments / copy trading)
  • ChatGPT (mentioned as part of Rocket Trader “AI/real-time” discussion)
  • CDI rate (performance benchmark)
  • No specific Brazilian stock tickers/ETFs were mentioned in the subtitles.

Key numbers / metrics / claims (risk, performance, validation)

Ranking claim

  • Top 1.7% of XP traders in 2026
  • Ranking based on a tool that ranks XP traders using account info (described as potentially operating via “robots” without necessarily analyzing charts manually)
  • Validation described as using a “control account” at XP

Strategy/robot operating characteristics

  • 12 automations (robots)
  • Average activity: 1220–150 trades/month across the 12, i.e. < 6 trades/day total for the full set
  • They frame frequent manual trading (e.g., “50 trades/day”) as harmful, suggesting it leads to account damage early in the day.

Backtest methodology / data windows

  • For mini-index: historical data from 2018 onward (through “today” in the discussion)
  • Minimum evaluation guidance:
    • Use an OHLC model (open/high/low/close) for faster testing (vs tick-by-tick)
    • Split data into in-sample vs out-of-sample (example: holding out “last year”)
    • Aim for ~5 years as “fair” evaluation horizon
  • Overfitting risk is emphasized throughout.

Sample-size / statistical requirements

  • Sampling-error rule:
    • To achieve ~1% error, require 400 samples
    • Therefore: require at least 400 operations for strategy evaluation (not just “5 years of data” if the strategy trades infrequently)

Drawdown / risk shutoff rules

  • “Red button” concept:
    • Example: if backtest drawdown was R$2,000, then live drawdown at R$3,000 is treated as a major red flag for overfitting
  • Historical draw example:
    • Largest draw referenced: R$3,500 lasting 121 days
    • Conceptual guideline: ~60 days drawdown + ~60 days recovery
  • Portfolio-level control:
    • They describe turning off underperforming strategies:
      • “This year” they turned off 2 robots (from 14 to 12)
    • Decisions are framed as rule-based, not emotional
  • They also reference a recovery cycle described as “selling at historical high then V-shaped recovery” (illustrative).

Costs, slippage, and net profitability assumptions

  • Execution cost note:
    • Even with zero brokerage fees, they claim there is still a minimum cost ~R$0.50 per entry/exit for mini-index futures
  • Example gross vs net:
    • Backtest: R$7,000 profit with ~26 mini-contracts (~13,000 trades)
    • Fees spent: R$6,000
    • Implies net ~R$1,000 if brokerage were zero (and they still stress real costs like fees/taxes/slippage)
  • Profit per trade targets:
    • Prefer minimum average profit per trade ≈ R$7
    • Sometimes stated goal: around R$10 to leave margin after slippage/copying losses
  • Slippage definition:
    • Parent account execution vs copied client execution at different prices
    • Example given: slippage of R$2 per trade
  • Copy-trading capacity / constraints:
    • Outliers bots linked to UNTIC
    • Limit: max ~400 clients or ~2000 contracts (whichever comes first)
    • If exchange/capacity limits aren’t met, copying can fail

Performance streak claims (portfolio)

  • “Last negative month” stated as November 2025
  • Then: 7-month streak
  • July positive, targeting an 8th consecutive non-losing month (subject to “tomorrow”)

Drawdown invalidation vs “turning off too early”

  • They argue turning off can be harmful because it changes expected probabilities.
  • They cite clients who turned off “a few days ago” and imply it missed recovery.

Portfolio construction / risk management framework (step-by-step elements)

Robot validation (anti-overfitting) workflow

  • Use an OHLC program (faster than tick-by-tick)
  • Use a large historical dataset: mini-index data from 2018 onward
  • Split into:
    • In-sample (training/sample) to develop the strategy
    • Out-of-sample (holdout period) to validate
  • Compare performance:
    • “Metrics have to be very close” between in-sample and out-of-sample to reduce overfitting
  • Robustness methods:
    • Monte Carlo / robustness testing, with internal tooling to control parameters
  • Enforce minimum statistical sample size:
    • Require ≥ 400 operations for ~1% sampling-error target

Drawdown management in live portfolios

  • For each strategy/robot:
    • Define a drawdown threshold (“red button”)
    • Turn off when realized drawdown breaches expectations (based on out-of-sample validation + Monte Carlo averages)
  • At portfolio level:
    • Use multiple robots suited for different regimes (bull/uptrend, downtrend, reversal, sideways)
    • Ensure not all strategies lose simultaneously
  • Decorrelation (key portfolio technique):
    • Combine robots with low correlation in daily results to “smooth the curve”
    • If one robot enters drawdown, others may offset

Decorrelation measurement method

  • Construct a decorrelation matrix (correlation matrix)
  • Target correlation characteristics:
    • Ideally correlations close to 0
    • High correlation near 1 = redundant exposure
    • Near -1 = opposite behavior (conceptually could offset)
  • Use AI/automation to compute correlations from operation streams:
    • Output daily correlation using Pearson correlation (explicitly named)

Explicit recommendations / cautions

  • Avoid overfitting:
    • If backtest drawdown is small but live drawdown is much larger (e.g., 2,000 vs 3,000), treat it as a major warning
  • Do not trust gross backtest curves:
    • Backtests may understate real fees, slippage, and copying costs/taxes
  • Trade less / reduce exposure time:
    • Top performers are described as those with fewer trades and less time exposed to risk
  • Do not “chase drawdown entries”:
    • They argue turning strategies on after drawdown (“buy at the low”) is counterintuitive
    • They emphasize scaling/selling around higher curve points and being careful about “promotion”
  • Slippage matters:
    • Even small per-trade slippage (e.g., R$2) can ruin strategies with low profit per trade
  • Copy limits are real:
    • Must respect UNTIC capacity limits (e.g., 400 clients / 2000 contracts)
  • Don’t micro-manage robots emotionally:
    • Turning off should be rule-based, framed as “hire the robot to operate,” not frequently adjust it

Disclosures / disclaimers

  • No explicit “not financial advice” disclaimer appears in the subtitles provided.

Presenters / sources (mentioned people and organizations)

GameClass / XP hosts

  • Isaac (also appears as “Isaac Pisani”)
  • Mauro
  • Júnior Viana (also appears as “Junior J” / “Júnior Viana”)
  • Viana is referenced again; the clear named finance educator is Júnior Viana

Guests / Outliers Invest

  • Artur (Outliers Invest)
  • Felipe (Outliers Invest)

Other organizations / platforms

  • XP (platform/ranking context)
  • Outliers Invest
  • Rocket Trader
  • My Game (education sponsor/school; MEC-recognized mentorship)
  • UNTIC (XP automation/copy-trading platform)

Books / authors cited

  • “Outliers” by Malcolm Gladwell
  • “The Wizards of the Financial Markets” (and “The Unknown Wizards…” referenced)
  • “Think and Grow Rich” by Napoleon Hill (spelled “Napoleão Rios” in subtitles)

Original video