Portfolio Value
—
from an initial $100,000 paper balance
A fully automated, ergodicity-filtered trading system, reporting every closed position in real time. No backtests, no curve fits, no hindsight. The numbers on this page are recomputed from the database on each load.
The gambler who wins on average may still be ruined in time. We optimise the trajectory, not the ensemble.
Every signal must clear two gates before it reaches the market.
The first gate is ergodic. The expected log-return of the bet must be strictly positive — g > 0, where g is the time-average growth rate of the position:
g = p · log(1 + b) + (1 − p) · log(1 − l)
This filter rejects trades that look profitable on average but compound to zero, or worse, over a finite life. It is the central insight of Peters (2019): a bet that is attractive in expectation may still be ruinous in time.
The second gate is cognitive. The Ergonitive Risk Index measures the degree of synchronisation between AI systems facing identical market inputs. When large language models — trained on similar data, with similar architectures — converge on identical interpretations, they compress the very inefficiency the signal seeks to exploit. A trade that clears the ergodic gate may still fail the cognitive gate if the edge has already been arbitraged by synchronised AI consensus.
ERI(t) = α · CC_LLM(t) + β · NV(t) + γ · OF(t)
where CC_LLM is the cross-model cognitive convergence index, NV is narrative velocity, and OF is options flow compression. High ERI → reduced position sizing. ERI above threshold → signal rejected.
Position sizes follow a half-Kelly rule, capped at 15% of capital. Both legs of each pairs trade are sized to dollar-neutral.
Two assets. One spread. One signal.
When two structurally correlated assets temporarily diverge beyond two standard deviations of their historical ratio, the system enters simultaneously long the underperformer and short the overperformer. It does not predict direction. It trades the relationship.
z = (ratio_t − μ_60d) / σ_60d
Entry: |z| > 2.0 standard deviations.
Exit: |z| < 0.5 standard deviations.
This is the oldest documented edge in quantitative finance. It does not require the market to go up. It requires only that correlated things return to being correlated — which, structurally, they do.
The system is market-neutral by construction. A long position in the cheaper asset and a short position in the dearer asset produces approximately zero net market exposure. What remains is pure spread alpha.