Skip to content

Strategy Methodology

Last updated 2026-11-08

This is the full technical reference for QuantAscent's strategy-testing machinery — the data foundation, signal measurement, the non-adjustable rulebook, and the walk-forward protocol behind the Adaptive Strategy Builder and the Sweep Explorer. The in-app summary (the "How is this tested?" links inside the Adaptive Builder and Sweep Explorer) covers the same machinery at a high level; this page is the in-the-weeds version — the actual algorithm, with the math.

The division of labor is fixed and worth stating up front: you author the strategy; the software executes your rules and scores the result. Nothing is pre-selected in your name: the first time you compose, the fill rule and the universe-floor setting start blank and wait for your choice; after that the Builder remembers your choices as your defaults. The tilt templates are QuantAscent starting designs — we set their cluster weights in our own research; edit any of them. What the software does control — and never lets you adjust — is the honesty machinery: point-in-time data, economic direction priors, the stability gates, the sealed holdout, and the luck accounting described below.

(The Library's ETF tab is a separate, simpler thing: passive single-ETF index trackers with historical tracking records. They carry no methodology beyond historical index tracking and none of what follows applies to them.)


1. Data foundation

Everything operates on a point-in-time metrics matrix: one row per (company, as-of date), where every metric value is computed strictly from information that had been filed and public on that date. A fundamentals row is used only on as-of dates strictly after its filing date, so a process evaluated "as of" March 2021 can only see statements companies had actually filed by March 2021. Filing dates follow four rules:

  • Statements (income statement, balance sheet, cash flow) carry the SEC filing date of the original 10-Q or 10-K that reported the period, matched on the company's SEC identifier (CIK) and the period of report; where the data vendor's date is later, the later date stands. Amendments never move a date.
  • Derived rows (ratios and per-share metrics computed from the statements) carry the latest filing date of the three statements for the same company and period.
  • No SEC match (foreign filers, which file no 10-Q, and companies whose older filings sit under a former CIK): the data vendor's filing date stands; where there is none, the row becomes usable 90 days after the period end, the 10-K deadline for the slowest filer class. A company that filed later than 90 days, on a row with no SEC match, is the one case these rules cannot see.
  • Earnings announcements count from the next day; prices are read on or before the as-of date.

Every weekly data pack is audited before publication, and a pack with any derived row usable before its statements, or any statement usable before SEC's filing of it, is not published.

  • Rebalance grid: as-of dates are spaced 90 calendar days apart. All scoring, testing, and record-keeping happens on this quarterly grid.
  • Forward returns: each row carries the company's total return over the 90 days following the as-of date, computed from dividend- and split-adjusted prices.
  • Survivorship: the database includes delisted companies (sourced from exchange delisting filings), so processes are scored against the universe as it actually existed — including the companies that later disappeared.
  • Outlier control: forward returns are clipped to ±100% per quarter before any averaging, for your process and its random benchmarks alike, so a single data artifact or extreme micro-cap move cannot dominate a record.

2. The universe you author

A process draws from a universe you define: sectors (or the whole market), a market-cap range, a share-price range (measured on the price as traded on each date, not adjusted for later splits), and any metric filters you add — hard gates on a panel metric, against a fixed value (EarningsYield > 0) or against another metric (CurrentROE > MeanReturnOnEquity). A metric filter is part of the universe, not a scoring rule: it is applied at every quarter of the record before scoring, a stock missing the metric is excluded, the random comparison portfolios of the luck accounting below are drawn from the same filtered pool, and the derived trading rules include it unchanged. Two screens apply uniformly on top of whatever you choose:

Screen Rule Rationale
Freshness Latest filing ≤ 135 days old One quarter plus normal filing lag; stale books aren't comparable books
Operating companies only Registered investment companies are excluded, by SEC's entity record A closed-end fund or ETF is a portfolio, not a business: its "revenue" is the dividends it received and its "net income" moves with its holdings, which puts it at the extremes of volatility, yield and margin screens

The freshness screen is deliberately universal and disclosed once — the algorithm is barred from learning recency-based filters, a class of screen that historically exploited data artifacts rather than economics.

There is no market-cap floor gate: universes reaching below $300M run, and display a factual liquidity disclosure (spreads and thin trading are not modeled in walk-forward records — $300M is the envelope the testing machinery was validated on).

3. Signal measurement: the Information Coefficient

The basic question — "does this metric predict returns in this universe?" — is answered with the Information Coefficient (IC): the average of per-date cross-sectional Spearman rank correlations between metric values and subsequent 90-day returns.

For each as-of date \(t\) with companies \(i\) in the universe:

\[ \mathrm{IC}_t = \rho_{\text{Spearman}}\big(\, m_{i,t},\; r_{i,\,t \to t+90d} \,\big) \qquad\qquad \mathrm{IC} = \frac{1}{T}\sum_{t=1}^{T} \mathrm{IC}_t \]

where \(m_{i,t}\) is the metric value and \(r_{i,\,t\to t+90d}\) the clipped forward return. Rank correlation is used because metric scales are arbitrary; per-date computation prevents one unusual period from dominating.

4. The rulebook: what you set, what you can't

4.1 Direction priors — labels win (never adjustable)

Every metric has one allowed direction, fixed by its economic meaning: higher return-on-capital is good (desc), heavier leverage is bad (asc), and so on. If the data says a metric "works" in the opposite direction, that metric is dropped for that quarter — never flipped. A Financial Strength slot is structurally incapable of preferring the most-levered companies, no matter what a backtest suggests.

A small set of metrics is excluded from rules entirely: short-horizon reversal instruments, economically ambiguous measures, and aliases of other metrics (so the same ratio can't enter twice under two names).

4.2 Structure — clusters and slots (you author, within the taxonomy)

A process's structure is a weighted set of slots, each naming a cluster of closely related metrics (all the margin measures form one cluster, all the leverage ratios another; the cluster taxonomy itself is fixed). You compose the slots yourself — starting from one or more tilt templates (several blend with equal mass per template, merging same-cluster slots), from the Equal Weighting design on its own (one slot per factor family; it never blends with the tilts), or as a custom composition with your weights totaling 100%.

Each slot has a fill rule you choose: strongest qualifying signal (largest mean absolute IC among the slot's qualifiers), most consistent across halves, or a pinned metric of your naming. A pinned metric always keeps its economic prior's direction, and in any quarter it fails the qualification gates its weight redistributes pro-rata for that quarter. Because each cluster contributes at most one metric, a composite can never quietly become one bet counted several times. Holdings count: 5 or more, with no upper limit; the universe is the only limit. A quarter with fewer qualifying names than the count holds every name that qualified, and random portfolios of that count cannot be drawn on it, so the record has no standing against luck.

4.3 The stability gate

A metric qualifies for a slot only if, on the training window split into two halves at its midpoint:

  1. Sign stability: its IC has the same sign in both halves,
  2. Prior agreement: that sign matches the metric's direction prior,
  3. Magnitude: the mean \(\tfrac{1}{2}|\mathrm{IC}_{H1} + \mathrm{IC}_{H2}|\) clears your minimum-signal setting (0.005–0.050; the shipped default is 0.02 — signal strength runs on different scales in different universes, so a live preview under the settings shows how many stable metrics clear your floor in the universe you chose),
  4. Coverage: at least 500 valid (metric, return) pairs in each half.

If fewer than two slots fill in a quarter, no portfolio is generated for that quarter — the process stands aside (section 6).

4.4 The remaining settings

Training window (expanding from the record's start, or trailing 12/16/20/24 quarters), stability gate (standard or strict), re-learn cadence (every quarter, or at ~2-year boundaries), the defensive universe floor on/off, a floor stability gate, and fresh-filings-only training. Every setting is bounded; nothing outside the bounds runs.

5. Universe floors

A process may carry one hard floor — a screen that removes the unfavorable half of the universe before scoring (e.g., "above-median interest coverage only"). Floors are chosen from a small curated pool of economically sensible candidates (volatility measures, balance-sheet strength, earnings quality) — never mined from the full metric library.

Selection uses per-date median-split lift. For each training date, the universe is split at the metric's median; the lift is the difference in mean forward return between halves, averaged across dates:

\[ \mathrm{lift} = \frac{1}{D}\sum_{d=1}^{D}\Big( \bar r_{\text{above-median},\,d} - \bar r_{\text{below-median},\,d}\Big) \]

A date only counts if it has ≥ 40 valid pairs and ≥ 15 companies in each half; at least 8 such dates are required. The learned favorable side must match the metric's direction prior (the drop-don't-flip rule extends to floors), and the qualifying metric with the largest absolute lift is selected. A floor is never forced: if no candidate clears the gates — its learned direction contradicts its prior, or the pool's metrics are too sparse in that universe — the process simply runs without one. With the floor-stability setting on, the floor must additionally keep its prior-consistent sign in both halves of the training window.

6. Portfolio construction

At each rebalance date the process scores its screened universe with a weighted percentile composite. Each slot metric contributes its cross-sectional percentile rank (direction-adjusted, so "better" is always higher), weighted by its slot weight:

\[ \mathrm{score}_i = \sum_{k} w_k \cdot \mathrm{pctrank}_k(i) \]

The process holds the top-scored names at your holdings count, at the amounts your Amount per stock setting gives them (§6.2; equal amounts unless you set otherwise), for the following 90 days, then re-derives and re-scores.

A quarter where fewer than two slots qualify — or the screened universe is empty — is recorded as stood aside: the process holds cash at 0% while the benchmark return still compounds. Standing aside can never improve the record; hiding those quarters would understate the benchmark a real adopter was trailing.

6.1 Stock selection: the ranking peer group

The percentile rank each slot metric contributes is taken within a peer group you choose: the whole screened universe (the default), the stock's own sector, or its own sector and market-cap band. Within-group ranks change which names score highest — ranked against the whole universe, the picks can all come from the sector that happens to score best; ranked within sectors, every sector competes with itself and the picks spread. The qualification of the slot metrics (§3, §4.3) is measured on the whole universe either way; the peer group applies at scoring time, exactly as it does for a deployed strategy's construction_mode.

6.2 Amount per stock: how the money is split

The picks never change; the setting only changes how much of the money each one gets on the buy date. With N picks whose market values are \(v_i\) on that date and a share \(s\) (0 to 1) of the money that follows size,

\[ \mathrm{base}_i = v_i \;\text{(toward larger)}\quad\text{or}\quad \frac{1}{v_i}\;\text{(toward smaller)}, \qquad w_i = \frac{1-s}{N} + s\,\frac{\mathrm{base}_i}{\sum_j \mathrm{base}_j} \]

so \(s = 0\) is equal amounts, \(s = 1\) toward larger holds each pick in proportion to its market value (the S&P 500's rule applied to the picks; the panel carries total market value, not the index's float-adjusted value), and every pick always holds at least \((1-s)/N\) of the money. A pick with no market value on its buy date takes the median value of the picks that have one; a quarter where no pick has a value is split equally. The market value itself is the share count the company filed with the SEC, as of its latest filing before the date, times the share price as traded that day (never a restated count; where no filed count applies the data vendor's count is used and the row says so), checked against the value of non-affiliate shares the company states on its own 10-K cover (a value that fails the check takes the vendor's if that one passes, and has no row if neither does), so one wrong data row cannot take a whole quarter. The record stores each quarter's weights with its picks, the largest weight and the effective number of holdings (\(1/\sum_i w_i^2\)).

The random portfolios a by-size record is graded against (§8) split their money the same way, from the same drawn picks: an equal-amounts random field against a by-size record would not be a like-for-like standing. The universe comparator stays the equal-weight average of the universe: it is the base rate of picking, and the benchmark already covers the market by size.

7. The walk-forward protocol

The record on every card is produced by walk-forward testing with periodic re-learning — the strictest of the standard backtest designs, because it tests the process, not one frozen set of rules. At each 90-day date past the minimum history, the process re-derives its slot picks by your rules using only data available before that date, holds one quarter, and chains the returns. Single-sector universes benchmark against their sector ETF (XLK for Technology, XLV for Healthcare, and so on); anything broader against SPY; a universe-equal-weight comparator is computed alongside.

Two spans keep the test honest:

  • The design span — everything you tune against.
  • The sealed holdout — roughly the most recent two years, computed but never displayed during design. Locking a process is permanent: the holdout is computed once, drawn onto the record, and stamped. Editing a locked process seeds a new, unlocked process — the stamped record never changes.

The record books each quarter from the data panel's forward return; it has no daily curve. For the day-by-day view — drawdown between quarters, monthly returns, allocation, holdings — the record's saved picks are replayed at daily prices with the Backtester's standard trading costs and shown there as the record's own card. The replay prices the design span until the process is locked and the held-out quarters are appended then; it never prices the sealed span early. The stamped record remains the evidence; the replay reads below it by the trading costs.

8. Luck accounting

Every candidate reports, factually and without a pass/fail badge:

  • its record's percentile against 500 random portfolios drawn from the same universe at the same holdings count and the same dates, scored identically (stood-aside quarters excluded on both sides) — in a universe where everything went up, beating the benchmark is not evidence of skill;
  • weight robustness — the record recomputed with slot weights deformed to their square roots and their squares (renormalized); results that collapse under small weight changes are curve-fit artifacts;
  • the luck line — the fixed threshold of 99%: a record stands above it when it beat at least 495 of its 500 random portfolios. Because a record with no skill is exchangeable with its random portfolios, its rank among the 501 is uniform and it stands above the line with probability 6/501, about 1.2%. A sweep of N records therefore states N × 6/501 as the number luck alone would put above the line (about 1.4 of 120), beside the number that did; the user compares the two counts. A record you authored by hand was selected out of nothing and carries no such count; its percentile is reported as a plain fact. "None beat the benchmark" is a designed, normal outcome.

Why these are records, not forecasts

Before release, a pre-registered study asked: do candidates that stand above the luck line keep outperforming on data they never saw? The answer was no — standing above the luck line did not predict holdout performance, while risk characteristics (volatility, drawdown) did persist. That finding shapes the product: the luck accounting can reject a result as explainable by chance, it never endorses one; the default orderings are risk-first; and every record sentence ends the same way — a historical record, not a prediction of future performance.

9. The Sweep Explorer

The Sweep Explorer runs the same machinery at scale. You author a design space instead of a single process — one fixed universe, the structures you allow, and ranges for the settings and holdings — and the engine fills it with draws sampled mechanically from that space. It never optimizes, never steers toward good-looking results, and never uses values you didn't allow: a search that chased winners would quietly hand authorship back to the software and corrupt the luck accounting.

Every draw is a complete walk-forward record (design span only — the holdout stays sealed), and every draw is accounted for: completed records, draws your own rules refused, and duplicates collapsed — nothing is dropped silently.

Every draw is graded against its own random portfolios at the same luck line (§8): bright when it beat at least 99% of them, dimmed otherwise once Dim records below the line is on; it starts off, because standing above the line has not picked better strategies out of sample. The sweep's size enters as a count, stated before you run for the planned draw count and on the results page for the completed count: how many of these records luck alone would put above the line (N × 6/501), beside how many did. Refused and errored draws produced no record and are outside both counts; a sweep you stop early counts what completed, and resuming it grows both counts without changing any record's standing. Opening a candidate's full record computes it, saves it to the Records tab — tagged with how many records it was picked from — and opens it there, where the ordinary Lock & Reveal and adoption path applies unchanged.

The distribution. One dot's standing is one record against its own random portfolios. The sweep page's Distribution view asks the group question: every record's CAGR, max drawdown, volatility or Sharpe over a span's quarters (a quarter it stood aside counts as cash at 0%), and random portfolios of the same holdings counts over the same quarters, as two histograms on one axis with the two medians and the sweep's benchmark marked. The random side is weighted by the holdings mix of the records drawn, so a subset is compared with random portfolios of its own sizes. Two numbers carry it: the middle (the records' median minus the random median, the lift) and the width (the 10th to 90th percentile of the records, the range any one record landed in). The universe and the scoring move the middle; the holdings count, the factor families and how much the universe's companies differ from each other set the width, and a wider range means bigger best cases and bigger losses, not a higher middle. Until the sweep is revealed the view covers the design span; after the reveal it also shows the sealed two years alone, quarters no record was designed on, where "the top X% by backtest CAGR" shows what picking by backtest got. The random portfolios are the ones the sweep stored when it ran, drawn on its own data pack over every quarter it traded, so a later pack, with another grid and another universe, does not change the comparison; a sweep run before that file existed draws them from the installed matrix on the nearest grid date, and the view says so. This is a description of what the sweep produced on past quarters, not a persistence test and not a forecast.

What the count does not do. It accounts for the candidates inside one sweep, not for how many sweeps you have run. Twenty separate 25-draw sweeps over the same universe each state the count for 25 candidates (about 0.3), even though you have in truth looked at 500 designs (about 6 above the line by luck alone) and are free to keep the best one you saw anywhere. No number in the app prices that, deliberately: the alternative we tried — a running per-universe tally — was defeated by changing a single universe filter, and it made every sweep's standing drift as unrelated work happened elsewhere. Your Records instead lists every sweep you have run, so the size of your real search is visible in the record rather than compressed into one dot's brightness. Reading it is your job, and it is the strongest reason to treat a bright dot as a reason to test rather than a result.

The complete record. Every draw keeps, for every quarter it traded, the companies it held, the rules in force (each slot's metric and weight), the floor, the universe's equal-weight return, its benchmark's return and SPY's; and it keeps its standing against the random portfolios on four measures — CAGR, max drawdown, volatility and Sharpe — not CAGR alone. Nothing on the Map view changed: the dot is still the CAGR standing. The rest is there so the record can be analyzed outside the app. Export… on a sweep (on its page or its Records card), on a record or on a saved backtest writes a zip of spreadsheets with a README that names every column, and every table on the Research page has an Export… link of its own. Company names are the strategy's own output and are in the file; prices, per-company returns and statement values are licensed data and never are.

Cuts and persistence. The question the luck line cannot answer is whether a record's standing carried over: whether the records that ranked high on one stretch of history ranked high on the next. A cut splits every record's quarters at a date into a before side and an after side. The sweep page's Persistence view takes the rank correlation, across the sweep's records, of each measure before the cut with the same measure after it, with a 95% range from resampling the records (1,000 resamples). +1 means the same order on both sides, 0 no relation, −1 the order turned over; a range that excludes zero is the only line the page draws, and it draws no verdict. Cuts are free: any date inside the visible span with at least eight quarters before it and four after, as many as you like, chosen at any time, because a cut inside the span the Map view already shows is a check on data you have already seen. Three sit there by default, every eight quarters back from the end of the span. Two cautions the page carries: cuts share most of their history, so several cuts are not several independent samples; and the standing against random portfolios at a cut uses the portfolios the sweep stored when it ran, drawn on its own data pack over every quarter it traded. A sweep run before that file existed draws them from the installed matrix instead, on the matrix date nearest each of its quarter dates (a weekly pack moves the whole grid a week), and the page says so: the comparison then sits on the installed pack's grid and universe, and running the sweep again makes it exact.

Windows. The other slicing sits on the same view. Windows of N quarters splits the span into consecutive blocks that do not overlap, aligned to the end of the span, and asks whether the records' rank in one block carried over to the next, one row per adjacent pair. Each quarter belongs to one block, so the pairs are separate tests, and the last row pools them: the records ranked within each pair, one correlation over every record's pairs, with a range from resampling the records, so that a record's pairs stay together. A rank across the pairs would compare one period's CAGRs with another period's, and the difference between the periods' levels would be counted as carry-over or turnover. Shorter windows give more pairs and noisier ranks, since a single quarter can swing a four-quarter CAGR; only lengths that give at least two blocks are offered.

The sweep reveal. The last two years of every record stay sealed, the same span Lock & Reveal seals on a single record, and they are the one cut the page will not make until you ask. Reveal the last two years… on the sweep page is permanent and dated: every record's sealed span becomes visible on the Persistence view (the design end becomes a cut) and in the export, and the Map view does not change. The cost is stated before you confirm. After the reveal no candidate from that sweep can earn a holdout verdict, because the holdout was visible when it was picked: records already promoted from the sweep and not yet locked are marked selected after the reveal, as is every candidate promoted later. They can still be locked and adopted, and their held-out span is then shown as a stated fact, not a verdict. Records locked before the reveal keep their verdicts.

What the study found, in plain words. On this engine, the backtest tells you the risk you will take, not the return you will get: across 1,500 strategies, drawdown and volatility rank carried over to data the strategies had never seen, and return rank did not. The Persistence view lets you run that check on your own sweeps, cut by cut, and the export lets you run any other.

10. Adoption and deployment

Only a locked process can be adopted. Adoption creates a scoring system that runs your process's derivation against current data at every rebalance. A derivation matching the rules you already run simply confirms them; a changed derivation becomes a pending proposal — the rebalance keeps trading the rules you last accepted, and nothing changes until you review the proposal (accept or decline) in the Strategy Manager. Accepting each proposal as it arrives keeps the deployed behavior identical to the recorded process; declining departs from it, and the record no longer describes what you run. The rebalance preview shows the rules in force (each slot's metric and weight, the floor, and the derivation date) and flags any pending proposal, and every trade list is preview-and-confirm. If the process cannot derive picks (wrong data basis, missing or stale matrix), list generation stops and says why; a one-time explicit override can deploy the last-derived picks with the stale date disclosed.

11. Limitations

Honest reading of the records requires knowing what they do not include:

  • Trading costs. Walk-forward records are computed from market returns gross of commissions, spreads, and slippage. Quarterly-rebalanced portfolios trade little, but the drag is not zero. Each record's replay in the Backtester (§7) is its after-cost version: the same picks, priced daily with the Backtester's default commission ($1.00 for every stock bought or sold) and slippage (0.10% of the value of every stock bought or sold).
  • Hypothetical performance. No capital tracked these records when they were generated. Fills at the rebalance price are assumed.
  • Sample size. Even the longest records span ~10 years of quarterly data — ~40 observations. The random-portfolio comparison exists precisely because records this short are easy to luck into.
  • Narrow universes. Your universes are your own, so a record's sample can be far narrower than a sector's — the luck accounting exists precisely because short records over narrow universes are easy to luck into.
  • Search across sweeps. The count of records luck alone would put above the line accounts for the candidates in one sweep, never for your search history as a whole. If you run sweep after sweep and keep the best result you have ever seen, no figure in the app corrects for that — the honest accounting for it is the list of sweeps on the Records tab, and your own memory of how hard you looked.
  • Benchmark choice. Sector ETFs are investable, dividend-inclusive comparators, but any single benchmark is a simplification.

The process end to end — a replication checklist

Everything above, restated as the sequence of steps the machinery takes for one process. Someone with a point-in-time dataset could follow this list and reproduce the test; each step names the section that specifies it fully.

  1. Take the authored process (§2, §4): the universe, the slot composition with weights and fill rules, and the settings.
  2. Walk the quarterly grid (§7). At each 90-day date past the minimum history, restrict to data available before that date.
  3. Screen the universe (§2, §5): the authored sector/cap/price bounds, the 135-day freshness screen, and the floor if one qualifies.
  4. Qualify and fill each slot (§3, §4.3): split the training window in half; keep metrics whose IC carries the same prior-consistent sign in both halves, clears the signal floor, and has at least 500 valid pairs per half; fill each slot by its fill rule.
  5. Score and hold (§6): if at least two slots filled, rank within the chosen peer group, hold the top-scored names for 90 days at the amounts the Amount per stock setting gives them (equal amounts unless set) and record the clipped return; otherwise record a stood-aside quarter at 0% while the benchmark compounds.
  6. Seal the holdout (§7): grade and display only the design span; compute the final ~two years once, at lock, and stamp it.
  7. Account for luck (§8): report the record's percentile against 500 random same-universe portfolios over the same dates, its weight robustness, and — when it was selected out of a sweep — how many records it was picked from and how many of those luck alone would put above the luck line.

QuantAscent provides research and analysis tools, not investment recommendations. Backtested and walk-forward results are hypothetical, have inherent limitations, and are not a guarantee of future performance. See the full Disclosures.