EXTRACTOR

A real fruit-fly brain, wired to stock-market data, making trades, and compared against brains with scrambled wiring.

Live decision · latest recorded day

Decisions appear once an experiment has a locked replay.

Market → neurons: why did the fly decide?

Select a day with a recorded decision.

Seeded mapping, not biology. Which visual neurons receive which stock feature, and which descending neurons count as “buy” or “sell”, was assigned at random from a fixed seed before the experiment ran. This chain shows the computation the model performed. It is not evidence that flies perceive momentum or volume.

Watch the brain think

Recorded firing around the selected day

BUY motor groupSELL motor group dopaminevisual inputsselected day

Results: official locked test

NO LOCKED TEST HAS FINISHED. Results appear here once one does.

Equity curves

Equity curves appear once an official locked test has finished.

Trade ledger

Trades appear once an experiment has a locked replay.

Does the wiring matter?

The comparison appears once an official locked test has finished.

Real vs experimental

REAL · FROM BIOLOGY
  • Neuron locations (soma positions)
  • The connectome: who connects to whom
  • Cell types and classes
  • Connection structure and synapse counts
  • Neurotransmitter predictions (sign of each connection)
EXPERIMENTAL · DESIGNED BY US
  • Stock features (1/5/20-day returns, volume ratio)
  • Feature → neuron assignment (seeded, arbitrary)
  • BUY / SELL interpretation of motor neurons
  • Leaky integrate-and-fire simplification
  • Reward pulse into dopamine neurons
  • Trading, sizing and risk rules

Reproduce this experiment

Reproducibility details appear once an experiment is locked.

Methodology

Marketdaily bar at the close
Features1-, 5-, 20-day return; volume vs 20-day average
EyesON / OFF groups of visual projection neurons
Brain40 leaky integrate-and-fire steps through the connectome
Motortwo descending-neuron groups: BUY and SELL
Decisionact only when one group clearly outfires the other
Risklong only, $10k cap, daily-loss and trade caps, kill switch, no margin
Fillnext open, with slippage and fees

Locked before testing

Parameters are chosen only on the training period, then written to a hashed lock together with the code, data and connectome hashes and every threshold. The test period runs once. Any change voids the lock.

Edge criterion

A predefined experimental criterion: Extractor must outperform every control brain, exceed the 95th percentile of 200 coin-flip runs, and outperform momentum and buy & hold. It is a success rule set in advance, not a claim of formal statistical significance.

Limitations

Limitations are listed with each official report.

Want the same test on your own trading strategy? Free preliminary backtest on your rules, with your fees and slippage. Written in plain English or built from dropdowns.

Test your strategy free →