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.
Your backtest's "spread" is probably a continuous-time estimate that reads 0.04% on a real 1.00% cost when trading is thin. EDGE fixes the discreteness bias from OHLC alone and beats Roll and CS.
The theory under the ML arc: why fitting the past predicts the future, what each parameter costs in variance, and why complexity's tax shrinks only as one over the square root of your sample.
A tick-level DRL market maker with real submit and cancel latency: PPO beats DQN and Avellaneda-Stoikov, an alpha signal nearly doubles it, and slower cancels raising profit is a risk trap.
Fourteen popular intraday MNQ setups, 947 days, one honest test. Zero cleared costs. The gross edge tops out near 1.5 points and two points of friction eats it. The signal ceiling is real.
Fair value is a low-pass filter with a dial, not a moving average. LAFO makes the cutoff explicit and neural filters turn corners faster than an EMA, but the headline Sharpe of 11 is pure in-sample.
Cut Bitcoin's day at US market hours and only the overnight leg predicts anything. It forecasts the next VIX move (beta minus 0.262) while trading-hour Bitcoin is noise.
Everyone reads the VIX to time the S&P. Rob Hanna ran it both ways: SPX oversold readings predict VIX futures far better than the reverse, and the real trade is a filtered short-VX book.
Pair a Deribit inverse call with Polymarket binaries and you build a payoff that never loses: 24 trades, zero losses, 20.7% each. The catch: it is a directional bet with a floor, and it barely ever fires.
A prop account is a discretionary payout contract on a simulated book, not capital. Read the framing here; download the full book as PDF.
Cross-sectional momentum only uses the order, so train on the order: learn-to-rank beats regress-then-rank in equities. But the 3.40 Sharpe is frictionless daily demo; weekly it is 0.54.
Crypto time series momentum only pays when the market return is weighted by trading volume, not manipulable market cap: 0.94%/day, Sharpe 2.17. Equal-weight the same signal and it loses 1.19%/day.
In low-signal economics and finance, the do-nothing zero forecast is Bayes-optimal. Ridge can beat it, Lasso cannot for any penalty, and signal weakness, not sparsity, is why.