Welcome to the lab. Nobody's selling you a sleeper.
SignalTuned is a fantasy football rankings engine. It computes player tiers from real NFL production data, age-curve analysis, and your league's exact scoring settings. It does not defer to FantasyPros, KeepTradeCut, or any other expert-aggregator that exists primarily to chase clicks. It does the math itself, every week, against your roster.
Most rankings tools are consensus machines. They aggregate expert opinion, average it out, feed the result back into the pool, and call it a ranking. The problem is that process doesn't produce signal. It produces mean-seeking behavior. Perception biased by aggregation, fed back in, reinforcing the bias. Your group chat runs on the same loop. The engine doesn't.
The Lab is the open notebook for that engine. Hypotheses tested. Consensus claims audited. The occasional methodology deep-dive when the math is interesting enough to deserve one. If the engine learns something worth shipping, you read about it here first.
What lives in here
- Hypothesis writeups. Every week the research agent picks a candidate signal (a metric, an age-curve tweak, a positional weighting) and tests it against historical outcomes. Survives, gets shipped. Fails, gets buried with honors. Negative results matter.
- Calibration notes. When a model parameter shifts (say, the RB age-decline slope), the change goes out paired with the validation data behind it. You see exactly what moved, why, and how much.
- Methodology deep-dives. Occasional long-form pieces on how a specific layer of the engine works. Speed Score derivation, the LASSO residual ensemble, the Bayesian rookie projection cohort baselines. The kind of thing the analyst on TV won't show you because it isn't a hot take.
What doesn't
- Hot takes. If the engine disagrees with consensus on a specific player, the rankings page already tells you that. The Lab is for why the engine is built the way it is, not roster advice you can screenshot.
- Marketing. No subscriber funnels, no upsell prompts, no email-gate before you read the methodology. Just the work.
More soon.