Signal Lab: Backtest Your Sharp Edge
The sharp signals tell you what's happening now. Signal Lab answers the harder question: does following them actually make money? Build a strategy in the console, replay it against thousands of settled signals, read the verdict, then carry what works straight to your live feed.
A note on access. Signal Lab is a Sharp-tier tool, and the backtest engine opens once the graded-signal dataset is deep enough to trust — we're settling live signals against real results every day. You can build and save strategies now; they'll be waiting, fully restored, the moment the bench unlocks. The mockups below use sample data.
What Signal Lab is
Every sharp signal we detect is quietly logged as a hypothetical bet — a shadow bet — with the price it could have been taken at. When the game finishes, that shadow bet is graded: win, loss, or push, plus its closing line value. Over months, this builds a fact table of thousands of settled signals.
Signal Lab is the workbench that interrogates that history. You set the rules of a strategy in a four-step console, run it, and see exactly how it would have performed — by record, ROI, and CLV — before you risk a dollar live.
From live signal to proven edge
Four stages turn a fleeting market move into something you can measure:
- 1DetectA steam, RLM, or drift signal fires on a live market.
- 2ShadowIt's logged as a hypothetical bet at the price available then.
- 3SettleThe game ends; the bet is graded W/L/P and CLV is computed.
- 4ProveSignal Lab replays the settled bets that match your rules.
Anatomy of the strategy console
The console is the cockpit. It's a numbered, single-open accordion — the steps read top to bottom like a build order, but only one is open at a time so the whole thing always fits on screen and the Run button never leaves the bottom. Each collapsed step shows its current setting at a glance.
Tail Steam + RLM · flat 1u · NBA
- 1The numbered steps.01 Strategy → 02 Signals → 03 Model & staking → 04 Refine. One opens at a time; the rest collapse to a single line so the console never runs off the screen.
- 2Where you start. Step 01 shows whether you're editing a fresh, Unsaved draft or a loaded Saved strategy — and lets you load one or start new.
- 3At-a-glance summaries. Every collapsed step prints its setting (Steam · RLM · Standalone, Tail · Flat 1u, 2 active) so you read your whole strategy without opening a thing.
- 4Run is always pinned. The footer stays put no matter which step is open, with a one-line summary of the strategy above it — so tweak-and-re-run is one click.
- 5Save, apply, alert — together.Save a new strategy or Update a loaded one, Apply live to filter the Sharp feed, or set an Alert — all beside Run, reachable from any step.
A strategy is just filters plus how you'd bet them. Everything you set here is saved — signals, mode, thresholds, and your bet model and staking — so loading a strategy restores the whole console, not just the matching rules. Loading never auto-runs; it sets the console up and waits for you to press Run.
Reading the verdict
Hit Run backtest and the lab grades your strategy against every matching settled signal. The headline is a plain-language verdict — but it isn't an opinion, it's the record, weighed against how much data backs it.
The verdict combines ROI with sample confidence, so a glamorous return on five bets reads as "thin sample," not "edge confirmed." Here's how the headline is graded:
| Verdict | Reads as | What it means |
|---|---|---|
| Edge confirmed | ROI ≥ +5% | A real, repeatable edge on a solid sample. Treat it as a lead, then size it with discipline. |
| Slim edge | ROI > +0.5% | Marginally profitable. There's a pulse — tighten the conditions to see if you can sharpen it. |
| Break-even | −2% … +0.5% | Treading water. No edge worth the variance — keep tuning, or flip the bet model. |
| No edge | ROI < −2% | Tailing this loses on our sample. That's a finding, not a failure — try fading it instead. |
| Thin sample | < 30 signals | Too few graded signals to trust, whatever the ROI. Loosen the filters or let more games settle. |
Below the verdict, the breakdowns are where the real work happens — ROI split by signal, by sport, and by market. In the sample above, RLM did the heavy lifting (+9.0%) while drift dragged (−1.2%): a clear cue to drop drift, re-run, and watch the edge tighten.
Walkthrough: building and proving a strategy
Here's the whole loop end to end — from a hunch to a saved, live-filtering strategy.
- 1Step 01 · Strategy
Start fresh, or load a base
Open with a New strategy (the status reads Unsaved) — or load a saved one to fork from. Either way, the console fills in and waits; nothing runs yet.
- 2Steps 02–04 · Configure
Set the rules of the bet
Pick your signals and how you'd play them:
- 02 Signals — select ◇ RLM and ◢ Steam; choose Standalone or Composite.
- 03 Model & staking — Tail the signal (or Fade it), and stake Flat 1u or Kelly.
- 04 Refine — narrow to NBA, require Pinnacle, set a 5pt magnitude floor.
The collapsed summaries keep the whole setup readable: RLM · Steam · Standalone, Tail · Flat 1u, 2 active.
- 3Footer · Run
Run the backtest
Press Run backtest. The lab grades your rules against every matching settled shadow bet and returns the verdict — here, Edge confirmed · +6.2% ROI on 814 signals, with an equity curve climbing left to right.
- 4Read the breakdown
Find what's carrying the edge
The by-signal split tells the real story — so you tighten and re-run:
RLM +9.0%Steam +4.1%Drift −1.2%RLM is the engine. You drop the dead weight, hit Run again, and the ROI ticks up — the tweak-and-re-run loop the pinned footer is built for.
- 5Footer · Save
Save it for keeps
Name it and Save. The whole console is stored — signals, mode, thresholds, bet model and staking — so when you load "NBA RLM + steam" next week it comes back exactly as you left it. Loaded strategies save in place with Update.
- 6Footer · Apply live
Take it to the live feed
Hit Apply live and the Sharp Signal feed instantly hides everything that doesn't match your tested rules — or set an Alert so the strategy buzzes you the moment a matching signal fires. The edge you proved on history now filters what you see in real time.
Why the loop matters
A proven strategy shouldn't stay in the lab. Because a strategy is just filters, the same definition drives three things at once:
- Save & re-run — your strategies live in your account and re-grade as new signals settle, so the edge you measured stays honest.
- Apply to live — filter the Sharp Signal feed down to only the signals your backtest approved.
- Alert on it — get pinged the instant a matching signal fires, with no feed-watching required.
The point of the whole loop: stop guessing which signals to trust. Test a rule on history, keep what shows an edge, and let it quietly filter your live feed — so every alert you see has already earned its place.
Explore the feed that powers it
Signal Lab grades the same steam, RLM, and drift signals you see live.