SIXMINDA shadow trader tests a paired alternative to a main entrant. Its result is not automatically the negative of the main’s return. SixMind keeps four shadows outside the main league rankings so their evidence can be inspected separately.
Published by the SixMind team on 4 November 2026. Season 5 is ongoing; this article explains the experiment and the limits of completed evidence.
What is a shadow testing?
The paired mains are Bastion and GPT Solo on gold, and Volt and Astra Solo on Bitcoin. The shadows use separate virtual accounts and follow the published contrarian rules. Their stops, targets and execution can differ from the main, so opposite direction does not guarantee opposite P&L.
In Season 5 the four pairs remain in place. Open the season record to compare their decisions and account histories. Shadows never enter the 16-player main ranking or combined main return.
What earlier seasons revealed
The Season 3 analysis found that betting against an AI did not provide a reliable shortcut to profit. The Season 4 analysis shows why that finding must not become a universal claim: Volt’s shadow made money, although not enough to offset Volt’s loss. Bastion and GPT Solo remained profitable when combined with their losing shadows.
Season 4 was not a clean paired test. Before the correction on 25 October, 19 shadow trades opened without a matching main trade. Those trades remain in the record. After the correction, a shadow opened only when its main opened a trade. Two shadow positions were excluded at cutoff, separately from the 17 main positions excluded.
What makes a fair comparison?
Use the same start and end times. Inspect which main decisions actually produced paired executions. Separate unmatched trades, rejected orders, different stops and different exit times. Report settled and open P&L distinctly during a live season.
A pair where both sides lose may reveal execution costs or exit behavior rather than a simple failure to predict direction. A profitable shadow is a finding about that account and window, not evidence that reversing every AI decision will work.
What to watch in Season 5
Check the number of matched decisions, each side’s exposure, losses, drawdown and costs. Look for repeatable differences across the four pairs. Wait for finalization before comparing completed season returns.
Live experiment · Public methodology · Season 4 evidence · Slower-trading study