Video summary

This GPT-5.6 Trading Bot Is CRUSHING Hyperliquid 24/7 (so far)

Main summary

Key takeaways

Finance

Finance-focused summary (markets, strategy, performance, risk)

  • The creator describes an AI-built fully automated trading system running 24/7 on Hyperliquid (crypto perpetuals implied by the mention of “leverage” and fast polling, though no specific perpetual contract is named).
  • The system is rule-scored and then automated:
    • It scans the market every 5 minutes (background process).
    • For any scanned instrument, if its score ≥ 70, it opens a trade.
    • Once opened, a spawned “manager” controls the position and closes it when exit/parameter rules are met.

“Evolutionary mode” (how the strategy evolves)

  • The approach is framed as evolutionary mode:
    1. Collect results/data
    2. Return to GPT-5.6 to analyze and improve the strategy over time
  • After initial deployment, the creator:
    • Allows the system to run,
    • Implements non-strategy changes first (robustness and risk controls),
    • Then continues with longer backtests and more live data.

Instruments / tickers / assets mentioned

  • SpaceX (used as an example instrument):
    • Move cited: declined 1.6% across 48 price checks
    • RSI value shown: 47.3, with RSI “range” 42 to 65
    • Example short trade details:
      • Entry: 139
      • Exit: 137.88
      • Gross profit: $25
      • Trading costs: $0.50
      • Net: $24
      • Monitoring window: every 5 seconds for 43 minutes
  • Bitcoin
    • Referenced as a separate fully automated 5-minute “up and down Bitcoin setup” used on prior days.
  • France–Spain game (Polymarket betting example)
    • Not a finance market ticker, but treated as an automated wagering experiment in the subtitles.
  • No explicit ETF/bond/commodity tickers are provided.

Key numbers & performance claims

System performance (Hyperliquid experiment)

  • Started “yesterday” and is up $170 at the time of recording.
  • A 7-day view shows a spike (exact date range not specified).

Example live trade (SpaceX)

  • $25 profit
  • $0.50 trading cost
  • Net $24
  • Monitored for 43 minutes

Account / leverage sizing (risk implications)

  • The system uses leverage for a “small account” approach (higher risk).
  • Mentions: “300 times 10” and $3,000 position with $3,000 context; the exact leverage math is unclear in the subtitles.
  • The creator explicitly states: leverage increases risk.
  • Example starting balance: $385 free, 1 slot open.

Broader results / attribution

  • Over the last week: $389 in profit, mostly attributed to the fully automated 5-minute Bitcoin setup.

Scoring framework (entry methodology)

The creator uses a point-based scoring system. Entry occurs when the total score meets a threshold.

Explicit scoring components mentioned

  • +20: “easy to trade” / high liquidity
  • +15: longer move condition
    • Example: SpaceX -1.6% across 48 price checks
  • +15: RSI balance within a specified range
    • Example: RSI 47.3 within 42–65
  • +20: “trend is wider”
    • Example logic: fast average below slow average implies broad falling
    • Example gap mentioned: 0.60% (also stated as “.6% or .60%”)
  • +18: weak bounce
    • “Bounce” defined as a brief rise during a fall; weak bounce scored over the last three checks
  • Mentioned “small bounce” scoring as well.

Entry rule

  • Enter if score ≥ 70
  • Example: a total score of 96
    • This is 26 points above the threshold.

Risk management / execution controls mentioned

  • Leverage is used intentionally for training/small-account sizing, but the creator warns this is “much more risky.”
  • After improvements, the focus is described as robustness rather than pure strategy edge, including:
    • Retries
    • Emergency exit setups
    • Updated data collection/testing infrastructure
  • Exits are governed by “meeting parameters”, though not all exit parameters are specified.

Backtesting / validation timeline

  • After “Sol Max” ran for 25 minutes, the creator:
    • Decided not to make big strategy changes
    • Implemented improvements (robustness and risk controls)
  • Claims include “44 pad tests passed” (benchmark; “pad” likely a transcription/autocorrect error).
  • Data collection expanded to enable testing over 200+ trades (exact count not specified beyond “200-plus”).

Explicit workflow (step-by-step as described)

  1. Check account balance
  2. Scan the market
  3. For each candidate, compute a score using factors such as:
    • liquidity/ease-to-trade
    • multi-check move magnitude
    • RSI positioning
    • trend relationship (fast vs slow average)
    • weak bounce / small bounce characteristics
  4. If score passes entry parameters (≥ 70):
    • Open trade
    • Spawn a trade manager that:
      • monitors continuously (example: every 5 seconds)
      • closes when “parameters” are met
  5. Run continuously in the background with a 5-minute scan cadence.

Disclosures / cautions

  • No explicit “not financial advice” disclaimer is shown in the provided subtitles.
  • The creator explicitly highlights risk:
    • Using leverage is “much more risky.”
  • Performance uncertainty is acknowledged:
    • The system “might go to zero”
    • It could revert toward the mean (“reverse to the mean”) despite current outperformance.

Presenters / sources

  • Presenter/creator: the person speaking throughout the video (name not given in the subtitles)
  • System/model referenced: GPT-5.6
    • Also briefly references GPT 5.5
  • Trading venue referenced: Hyperliquid

Original video