We don't ask you to trust the signals. We prove them.
Every signal the engine fires is logged the instant it's issued and graded against live market prices — immutable, no hindsight, no cherry-picking. Then we put the whole record through the same four-stage statistical validation a quant fund uses on its own strategies. Here are the live results.
Avg return / signal
—
direction-adjusted
Statistical edge
—
significance grade
Resolved signals
—
closed & graded
Universe
S&P 500
liquid, large-cap
Window
The validation cycle
Four questions. One honest answer each.
A backtest that looks great can still be a calibration mirage, a statistical fluke, a ranker no better than a coin, or a fragile edge that dies out of sample. Each stage below interrogates the live record from a different angle — and reports a letter grade you can hold us to.
Stage 1 · Calibration
Does the score mean what it claims?
"When the engine says 70, does it win about 70% of the time?"
—
Reliability of the score-to-outcome mapping (Expected Calibration Error + Brier score).
Stage 2 · Significance
Is the edge real, or could it be luck?
"Would a coin-flipper produce a record this good by chance?"
—
Win-rate confidence interval + return t-test against the no-edge null.
Stage 3 · Ranking power
Does a higher score pick more winners?
"Can you actually size by conviction and earn the spread?"
—
ROC-AUC: the probability a random winner outranks a random loser by conviction.
Stage 4 · Robustness
Will the edge last, or is it a fluke?
"Does it survive fat tails, time, and out-of-sample?"
—
Probabilistic Sharpe (fat-tail-corrected) + in-sample vs out-of-sample hold.
The record, drawn
Cumulative realized return.
Every step is one closed, pre-logged signal — graded against live prices, in the order it was issued. A smooth climb is a durable edge; one jump that carries the whole curve is fragility (Stage 4 measures exactly that).
Loading the equity curve…
Where the edge lives
Per-sector accuracy. Wins and losses, by sector.
The same resolved S&P 500 record, cut by GICS sector — so you can see which sectors the engine actually predicts well, and which it doesn't. Win-rate is direction-adjusted; sectors with fewer than 20 resolved signals are omitted as too thin to quote.
The engine, self-tested
How each factor scored. And which ones we stopped trusting.
A nightly walk-forward validator measures each of the five factors by its Information Coefficient — the rank correlation between what the factor said at issuance and what the market subsequently did — across the full resolved record, then re-measures it on the most recent 60 days. These are historical-validation readings, not forecasts. When a factor's coefficient decays or flips sign on the record, we surface it here and review its weighting in the live composite rather than leaving it unexamined. This is our research methodology, in public.
Does the edge survive out-of-sample?
Each strategy horizon is trained on a rolling window and tested on the next, unseen window. We report the train→test Sharpe ratio — how much of the edge survives once the engine sees data it wasn't fit on. A horizon that fits its training window but fades out-of-sample is labelled OVERFIT, not buried.
How it's verified
No hindsight. No survivorship. No edits.
Immutable logging
Logged at issuance
Every signal is written to a permanent, append-only record the moment it fires — with its score, factor breakdown, entry price, target and stop. The fill columns can't be edited after the fact. You're seeing what the engine actually called, not a curated highlight reel.
Automated outcome
Graded against live prices
Outcomes resolve on a fixed schedule against live market data — target hit, stop hit, or horizon expiry — using the same target/stop the signal shipped with. Long-horizon calls resolve on closes, not intraday wicks, the institutional convention.
Honest universe
S&P 500, return-adjusted
The measured cohort is scoped to liquid S&P 500 names so the headline isn't propped up by thin micro-caps. Returns are direction-adjusted: a bearish call that drops is a win with a positive return, so the engine's directional judgment is what's being scored.
The engine
5 factors + an ML blend
Technical, momentum, fundamental, options, and microstructure factors combine into a composite, nudged by a macro overlay and a learned market-regime model, then blended with a continuously-retrained ML layer. The grades above measure that whole stack — not a single indicator.
⚠ What we do NOT claim
Past performance does not guarantee future results. A strong grade today is evidence the engine has an edge on the record so far — not a promise about tomorrow's market.
This is a signal track record, not a live brokerage statement. Signals are graded against live market prices; they are not real-money fills, and they exclude slippage, commissions, and your own execution.
Young samples are labelled, not hidden. Where a stage doesn't yet have enough resolved signals to be conclusive, it says so ("seasoning") instead of showing a flattering number. The grades only firm up as the record grows.
Grades can move against us. The validation runs on the live record and updates automatically. If the edge decays, these numbers will show it before we ever say a word.
QuantLogix is not a registered investment advisor, broker-dealer, or financial planner. All content, signals, scores, and analysis provided on this platform are for informational and educational purposes only and do not constitute financial advice, investment recommendations, or solicitations to buy or sell securities. Past performance does not guarantee future results. Trading stocks, ETFs, options, and other financial instruments involves substantial risk of loss and is not suitable for every investor. You should consult with a qualified financial advisor before making any investment decisions. By using this platform, you acknowledge that you are solely responsible for your own investment decisions and that QuantLogix bears no liability for any losses incurred.