I've seen that and wholeheartedly concur with the many detractors.

Earlier I alerted all to a graphic Kelly Simulator. That showed traumatic volatility over its limit of ~450 runs.

The 650 runs mentioned, even though over 5 years are clearly contrived, or fallaciously backfitted if you prefer.

Extend that sample 10-fold and the percolating POT will evaporate.