Aligrithm is an independent research publication on systematic trading, quantitative research, market microstructure, and adaptive systems. Long-form essays, code notebooks, and architecture breakdowns across eight pillars, built for traders who care more about how markets behave than about hype.
More about the publication →Read in order. Each pillar walks you out of one trap with one new ability. The newest article is the worst place to begin.
Decentralized prediction markets, cut into eight swappable stages. Seven are engineering trade-offs. The eighth, resolution, is where a handful of token wallets can settle a market against the truth.
CryptoMamba, a compact Mamba SSM, forecasts Bitcoin's next close and trades $100 into $262. But it never runs the naive or buy-and-hold baselines, on one bull year of daily bars.
A 14B model trained with outcome-only RL matched o1 and beat it on calibration. The lesson: drop GRPO's variance scaling, block leakage, and the win is honest probabilities, not beating the market.
When the game is against you, betting small is slow suicide. Bold play, swinging your whole stack at the target, maximizes your chance of hitting it. Optimal bet size flips with the sign of your edge.
Hyperliquid runs fully on-chain, so a node captures every order event: wallet IDs, counterparty inventory, and the ~89% of orders that are rejected and invisible in LOBSTER-style data.
Two economists shocked 817 prediction markets by 5 points each. Sixty days later the shove was still there. Prices revert, but slowly, partly, and cheaply beaten in thin markets.
A prediction-market price is not a probability. It is where capital-weighted Kelly bets cancel. You bet the gap, not the belief, and getting the probability wrong costs more than mis-sizing.
Prediction markets have no Black-Scholes. A recent paper builds one: model log-odds as a jump-diffusion, force the price to be a martingale, trade what's left. Clean theory, thin evidence.
A two-regime MS-GARCH turned 7% buy-and-hold lumber into 158%. The edge was all in the asymmetric variance model. Adding market and behavioral factors made the good versions worse.
The random walk still wins for developed FX. But a stock-return signal beats it for emerging currencies, netting about 7% a year, when it works. The edge is real, and it comes and goes.
The exact P&L of a European trend-follower is a weighted sum of return autocorrelations plus drift squared. Positive long-horizon autocorrelation pays; the story is optional, the sign is not.
A year of Polymarket data, $40M in arbitrage. But almost all of it is plain single-market rebalancing harvested by a few bots during volatility, not the exotic cross-market kind.