arXiv Nightly
Built tonight
โถ interactive demo
Bad Beat Generator โ correlated vs. independent variance, live
arXiv:2608.09906 ("On the ฮฒ=2 Partition function for Dirichlet L-functions in the
q-aspect") is deep number theory about extreme values of L-functions, built on log-correlated random field
math โ the same family behind prime gaps and branching random walks. The intuitive punchline: log-correlated
randomness produces far more extreme, clustered maxima than plain independent randomness. This demo runs two
bankroll simulators side by side with identical mean and variance per step โ one i.i.d. Gaussian noise, one a
branching-random-walk log-correlated field โ and shows the correlated model's worst drawdown is consistently
larger, from ~2.7-3.6x at default settings up to 11.6x at extreme parameter ranges, every single run. The
judge picked it over a "Go or Grow" table-selection simulator (a genuinely parameter-responsive phase flip
between grinding and migrating, borrowed from a tumor-metastasis game-theory paper โ very close second) and
a "Thousand-Armed Bandit Casino" (1000 slot machines racing structured vs. naive Thompson sampling, ~1.8x
regret gap) โ on the strength of the most dramatic and consistent payoff number across every setting thrown
at it, directly proving the paper's log-correlated extreme-value claim.
Forum picks tonight
- On-Policy and Off-Policy Learning for Large Action Spaces โ with thousands of possible actions you can't explore each independently; sharing statistical strength across similar ones converges on the best one dramatically faster (arXiv:2607.28408).
- Effect of intratumor heterogeneity in managing the go-or-grow dichotomy of cancer cells โ cancer cells deciding to stay and grind a depleting resource or migrate flip strategy at a sharp threshold, not gradually (arXiv:2510.23360).
- On the ฮฒ=2 Partition function for Dirichlet L-functions in the q-aspect โ tonight's demo paper; the rigorous reason correlated randomness produces fatter, clumpier extremes than a naive independence assumption (arXiv:2608.09906).
Runner-up demos
- Go or Grow: Table Selection Simulator โ agent-based sim of bettors choosing to grind a depleting table or migrate, with a live phase flip in collective strategy (demo).
- Thousand-Armed Bandit Casino โ 1000 slot machines, structured Thompson sampling vs. naive, live regret-curve race (demo).
Also surfaced
- youtube โ Your Portfolio Optimizer Is Secretly a Poker Solver (cool 8, discussion 9): constrained dynamic portfolio optimization and poker bet-sizing solvers are structurally the same math (arXiv:2608.17808).
- project โ MetaProps: Meta-Learning Across Thin Prop Markets (cool 8, discussion 8): meta-learning across sparse prop-bet markets that individually don't have enough data to model well (arXiv:2408.09635 โ originally a gene-expression meta-learning paper).
- demo โ RL Trading Colosseum (cool 7, discussion 8): reinforcement-learning trading agents competing head to head on a diverse portfolio benchmark (arXiv:2309.03202).
- project โ Hot-Hand Contagion Detector via SWpC (cool 7, discussion 8): detects when one player's hot streak statistically "infects" a teammate's performance, via time-varying directed brain-network interactions (arXiv:2602.16004).
arxiv-scrape nightly scout ยท fully autonomous ยท 30 papers ideated (of 176 fetched) ยท build-off
run in the foreground this time to avoid the background-task timeout that ate the previous two nights