The Filters Analysis tab answers a simple but fundamental question: why is the portfolio shaped this way? It shows, filter by filter, how the rule chain narrowed the strategy pool in a rotation — how many entered, how many each filter discarded, and which ones remained at the end.
In simple terms: it’s the selection funnel’s “X-ray”. Where the Equity tab shows you the portfolio’s result and the Comparison tab lets you compare alternatives, Filters Analysis shows you the process: the filter sequence, the cascading effect (each filter only works on what survived the previous one), and for each filter the percentage of strategies passing it. It’s the first-choice tool when a portfolio comes out too crowded, too thin, or even empty, and you want to understand which rule cut what.
A key aspect: the analysis is done per single rotation (by default the last one, the one going live) and reconstructs the funnel exactly as it was on that date — with the same lookback window and the same parameters the walk-forward generation actually used. It isn’t a “as of today” recomputation: it’s a faithful snapshot of the decision made in that rotation.
An important premise. QANTHOS is an analysis, validation and discovery tool, not a financial advisor. This tab is purely diagnostic: it explains how the filters you’ve set acted on a strategy pool at a given moment. It doesn’t judge whether those filters are “right”: that stays in your hands.
1. What it’s for
When you generate a rotational portfolio, the Portfolio Rules rule chain kicks in at every rotation: filters narrow the strategies, the correlation stage decorrelates, the rebalancer rebalances. You see the final result in the Equity tab — but how you got there stays, without this tab, a black box.
Filters Analysis opens the box. It’s mainly useful for three things:
Diagnosing an unexpected funnel. Is the portfolio empty in certain rotations? Does it have too few strategies? Here you immediately see which filter is the bottleneck: a row with Pass = 0 or a very low pass rate jumps out.
Understanding each filter’s effect. For each filter you see how many strategies come in (input), how many pass, and how many are excluded. Pair this reading with filter order and understand whether a quality gate is too strict, or whether a downstream filter is effectively useless because the upstream one already narrowed everything.
Validating filter order. Since the effect is cascading, order matters. Filters Analysis makes it evident how one filter’s output becomes the next one’s input, helping you place filters in the most sensible sequence.
2. Selecting the rotation
At the top, the Rotation dropdown chooses which rotation to analyze. By default the last one is selected — marked 🔴 (LIVE) — because it’s the one the portfolio is about to trade; the menu shows, for each rotation, the number, the date range and the number of strategies. You can also select past rotations to see what the funnel looked like at those moments. The Refresh button reloads the list and re-analyzes the selected rotation.
When you change rotation, QANTHOS recalculates the filter metrics on that rotation’s lookback window: the rules only look at data available up to that date, consistent with the portfolio’s walk-forward nature (each rotation’s decisions are made without looking at the future). This is the same logic that produces the Equity tab’s equity, seen here from the selection side.
Shows the real funnel, not an “as of today” simulation. The tab uses the same lookback window and the same Min trades with which generation actually built that rotation (values frozen at generation time). This is a deliberate choice: if it recalculated with the filters you have now in the Portfolio Rules tab — maybe changed without regenerating — the shown funnel would diverge from the one that really produced the portfolio. Here you see what happened, not what would happen.
3. The summary table
The tab’s core is a table with one row per filter, in the order the filters are applied:
| Column | Meaning |
|---|---|
| # | The filter’s position in the sequence. |
| Filter | The filter’s name. |
| Period | The lookback window used (e.g. “Last N months” or “Full history”). A ⏳ icon flags that for one or more strategies the window was extended (see below). |
| Pass | How many strategies pass the filter. |
| Fail | How many are excluded. |
| Total | How many strategies entered the filter (the input). |
| % Pass | The percentage of strategies passing. |
Rows are colour-coded to draw the eye where needed: red if no strategy passes (Pass = 0), green if the pass rate is high, orange if it’s very low. It’s the fastest way to spot a filter zeroing out the pool or one cutting drastically.
The cascading effect reads by scrolling down the Total column: each filter’s input is the number of strategies that survived the previous filter. So the table tells, row by row, how the pool narrows down the chain.
The ⏳ icon and lookback extension. Some filters require a minimum number of trades in the period (the Min trades field in Portfolio Rules). If a strategy’s nominal window contains fewer trades than the minimum, QANTHOS extends the window backward for that strategy, so metrics are computed on enough trades. The ⏳ icon in the Period column flags how many strategies had this extension: useful information to know that some values come from a wider window than the nominal one.
4. The summary bar
Below the table, a bar summarizes the rotation’s overall outcome: the analyzed rotation with its date range, the number of filters, and above all the Strategies: initial → final balance — how many you start with vs how many remain after the whole chain.
The bar also shows how many strategies were actually Selected into the portfolio. If this number diverges from the funnel’s survivors, a ⚠️ warning appears with the detail of “extra” strategies (present in the analysis but not selected) or “missing” ones (selected but not appearing in the analysis). It’s a consistency check between what the analysis reconstructs and what the portfolio actually chose: a discrepancy signals something worth investigating (typically tied to stages acting after the list filters, such as correlation or rebalancer).
5. A filter’s detail
A double-click on a table row opens that filter’s detail dialog: the full ranking of strategies the filter evaluated. The table has columns # / Status / Strategy / Symbol / Value, with rows coloured green (✅ pass) or red (❌ excluded). It’s sortable by clicking headers — useful for seeing strategies ordered by the metric’s value and precisely spotting the “cutoff line” between what’s in and what’s out. Strategies with an extended lookback for Min trades are marked with ⏳ and have an explanatory tooltip.
From here you can go one level deeper: double-clicking a strategy opens the detail of its trades within the lookback window — exactly the trades on which the filter’s metrics for that rotation were computed. Maximum transparency: from the overall funnel, to a single filter, down to a single strategy’s trades in the evaluated period.
6. The “few trades in lookback” warning
When a selected strategy has few or zero trades in the lookback window — and the Min trades field isn’t set — QANTHOS shows a warning. The reason is substantive: with too few trades, metrics like Sharpe, PSR or DSR look valid but aren’t statistically reliable. The warning lists the affected strategies and offers a few choices: exclude them from the rotation (the default, more cautious option, because QANTHOS never silently includes a low-reliability strategy), include them accepting the low reliability, or set the Min trades field in Portfolio Rules to extend the window until reaching the trade minimum — the solution that eliminates the problem at the root.
This warning is an example of QANTHOS’s philosophy: better an honest, cautious figure than an apparently precise but fragile number. The final decision remains yours, but the tool puts the risk in front of you instead of hiding it.
7. Common issues and notes
The tab says to configure filters and generate. Filters Analysis populates from an already generated portfolio. Set the filters in Portfolio Rules, generate the rotational portfolio, then select a rotation here.
A row is red with Pass at zero. That filter, in that rotation, excluded all input strategies: from there on the funnel is empty. Often a quality gate too strict for that period, or a badly placed filter in the sequence. Open the detail to see the values and threshold.
The summary bar shows ⚠️ with “missing” or “extra” strategies. The selected strategies don’t match the reconstructed funnel’s survivors. Worth investigating: usually the difference comes from stages acting after the list filters (correlation, rebalancer), which can remove already-passed strategies.
The numbers don’t match what I expect from the current filters. Normal and intentional: the tab reconstructs the funnel with the parameters used at that rotation’s generation, not the filters you currently have in the tab. If you changed filters without regenerating, the two funnels diverge. Regenerate to align them.
Many ⏳ in the Period column. Many strategies had their window extended to reach the trade minimum: a signal that the nominal lookback is short relative to those strategies’ trade frequency. Consider a wider window or a consistent Min trades.
8. An example workflow
A typical diagnosis path, not a prescription.
- In Portfolio Rules set the filters and GENERATE the rotational portfolio.
- Open Filters Analysis: by default you see the last rotation (LIVE).
- Scan the summary table: spot filters with a very low % Pass or Pass = 0 (orange/red rows).
- Double-click the suspect filter to see the full ranking and where the cutoff line falls.
- If needed, double-click a strategy to inspect its trades in the lookback window.
- Check the summary bar: the initial → final balance and any discrepancy warning.
- Go back to Portfolio Rules, adjust thresholds or filter order, regenerate, and recheck here. Then compare variants in the Comparison tab.
Filters Analysis turns selection from a black box into a readable process. But “understanding why the funnel looks this way” isn’t the same as “the funnel is right”: validating the soundness of the choices remains entrusted to Walk-Forward Analysis, Lock-box and observation over time.
9. To learn more
- Portfolio Rules — the filter chain, the order, the Min trades field.
- Percentile Rank — the ranking filters, clearly visible in Filters Analysis with their cutoff line.
- Equity — the funnel’s result.
- Portfolio Comparison — comparing portfolios against each other.
- Correlations and Portfolio Rebalancer — the stages acting after the list filters, which can explain discrepancies.