Aperture Analytics is the statistics and data science arm of The Camera Group. We build projection models, decision systems, and live-data pipelines for people who have to commit before the clock runs out. Every figure we hand over is sourced, time-stamped, and reproducible.
We work in domains where a wrong number has a cost. So we keep the scope narrow and the discipline high.
Ensemble models that project performance from live data: player stats, matchups, market movement. Each model is reweighted against how accurate it actually has been, not how confident it sounds.
Integer-linear programming over real constraints: budget, positions, stacking, eligibility. The optimizer finds the best feasible answer, not a plausible one, and it is deterministic and auditable.
Week-scoped ingestion from independent sources, cross-referenced and tracked by manifest. If a number is not in the week's manifest, the honest answer is "no data." We do not fill the gap.
We deliberately separate the two. A small ensemble of models makes suggestions; a deterministic optimizer makes the decision; a feedback loop corrects the models when they are wrong. No single opinion can steer the result.
$ aperture run --week 1 ingest fantasypros, sleeper, odds, weather ok fetched 2026-08-31T18:04:12Z models quant · adversarial · news ok blend reweighted by realized error ok optimize integer program, 9 feasible lines ok guard no look-ahead, all figures sourced pass
This is a real run, not a demo. The full methodology — the model roles, the data rules, and the feedback loop — is on the methodology page.
The same engine powers a live draft coprocess. Between rounds it pre-computes a sourced, ranked shortlist, so when your pick clock starts, the best pick and two alternates are already on screen. You confirm; nothing waits on a model mid-round.
Aperture Analytics is the statistics and data science division of The Camera Group. The group builds the systems; we build the numbers behind them.