Historical Results and Evaluation

The evaluation question

For a Buy classification, the research question is whether the observed price path reaches the defined positive 10% threshold during the following 30 trading sessions. A Sell classification uses the corresponding negative 10% threshold over the same window. The label describes this research objective; it is not an order instruction, target price, or promise.

When a result is eligible

The reference price is the official close on the publication market date. A record becomes Hit as soon as its target is touched within the next 30 trading sessions, even if the full window has not finished. With no target touch, it remains TBD until all 30 sessions are observed, then becomes Miss. A record with insufficient or invalid price data is unscorable. The resolved-record target-hit rate is Hits divided by Hits plus Misses. The backtesting chart instead shows a provisional target-hit score: 100 × (hits + 0.5 × pending) / (hits + misses + pending). Existing TBD records receive half credit without changing their classification; unscorable records remain excluded. Half credit is a scoring convention, not an estimated probability of success.

Publication history and evaluation coverage are different. Recommendation archives include editions from August 7, 2026; the historical evaluation identifies its supported history boundary and data-update date. An earlier edition's existence does not guarantee a complete, scorable price path. The previous August 21 tracking reference was not the date of the first recommendation.

Early Hits can enter the resolved-record rate before pending records can become Misses. A rate for an immature cohort can therefore overstate its eventual target-hit rate. A fully mature cohort waits until every included publication has had its full 30-session window, including records that hit early; missing data must still be reported separately. WealthGlider does not turn partial cohorts into a headline accuracy claim.

Target touch, first direction, and final return

These answer different questions. The provisional chart uses target-touch outcomes and gives pending records half credit; it does not measure first direction or final return.

• Target touch: Did the price reach the predicted 10% threshold at any point in the window? A large move in the wrong direction first does not cancel this Hit.
• First direction: Which 10% threshold was reached first: the predicted direction or the opposite direction? If both were touched in one daily bar, the order is ambiguous without finer price data. Neither touched is a separate outcome.
• Final return: Where was the split-adjusted close after session 30 relative to the publication close? A signal can touch its target early and finish on the wrong side. A Sell's underlying price change and its direction-adjusted research return have opposite signs.

None of these measures, on its own, establishes a profitable executable strategy or incremental model skill.

A worked example of an immature cohort

This is a fictional example, not WealthGlider performance. Suppose 10 records have 3 early Hits and 7 pending records. The resolved-record rate is 3 / (3 + 0) = 100%, but only 3 of 10 records have resolved. The provisional chart score is 100 × (3 + 0.5 × 7) / 10 = 65%; this is not a 65% chance of success. The other 7 have not failed. If none of them later touches its target, the fully mature score and rate become 3 / (3 + 7) = 30%. Always read the Hits, Misses, pending and unscorable counts with the rate.

What the result does not measure

• It is not the return of a customer account, model portfolio, or managed investment strategy.
• A threshold touch does not prove that an order could have been filled at that price.
• Unless expressly stated, it excludes commissions, spreads, slippage, market impact, taxes, borrowing costs, subscription fees, liquidity constraints, position sizing, and compounding.
• It does not remove survivorship, selection, look-ahead, data-snooping, overfitting, or market-regime risk.
• Past, hypothetical, or observational outcomes do not predict future results.

Inspect the available evidence

The Historical Signal Evaluation displays aggregate Buy, Sell, and combined provisional target-hit score series for an available window or date range, together with data-update information and calculation limitations. Backend corrections, source-data changes, eligibility rules, or methodology changes can revise those values.

This page does not publish a validated out-of-sample model-performance result. A public model-quality report should identify the model version, prediction dates, price-data cutoff, complete mature cohort, missing counts, first-direction outcomes, final returns and a like-for-like baseline. Repeated records for the same stock and overlapping windows must be accounted for, not treated as independent wins. Target-hit statistics alone do not establish that the model adds value over a simple baseline.

Methodology, revisions, and questions

Read the Signal Methodology and Risk Disclosure for the complete interpretation and risk framework. Material corrections should retain the original record and explain the change rather than silently rewriting history.

If a published result appears inaccurate or lacks material context, use the content correction process.