{"id":"6d55be7d-b3d6-4077-aa66-909156e6c7c8","arxiv_id":"2606.09025","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":3.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"Two continuous filters combined with a max-cash rule improve CAGR from 16.62% to 20.45% and cut max drawdown from 33.59% to 16.77% on 2017-2026 data for a static risky sleeve.","lead":"This paper develops two continuous cash-overlay filters for a static 50/50 growth-defensive ETF sleeve to decide daily cash allocation. A smart generalist might read it to understand if simple rules can improve returns and reduce drawdowns in a fixed risky portfolio.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.3","headline":"Unadjusted multiple-testing risk in filter parameters and max-cash rule on 2017-2026 window","rationale":"The reader's weakest_assumption pinpoints exactly the data-snooping exposure that keeps the verdict UNVERDICTED. The walk-forward OOS numbers provide partial mitigation, but the leading common-window results remain exposed until the multiple-testing issue is quantified.","tokens_in":1922,"tokens_out":309,"duration_ms":11511,"concrete_test":"Re-run the walk-forward procedure while enumerating a modest grid around the reported filter parameters (e.g., ±20% on each of the three parameters) and apply a Bonferroni or FDR correction to the t-test on the CAGR difference; if the adjusted p-value for 20.45% vs 16.62% rises above 0.05, the central claim weakens.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline performance figures (20.45% CAGR / -16.77% max DD on the full 2017-2026 window; 18.05% CAGR / -22.05% max DD in the expanding OOS version) rest on two continuous filters plus a fixed max-cash combination, each with parameters (reader notes parameter_count=3). The paper itself flags the need for multiple-testing-adjusted inference as future work, so the reported edge over the static 50/50 sleeve could reflect selection on the evaluation period rather than genuine out-of-sample robustness.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript develops two continuous cash-overlay filters (slow-tail compensation targeting persistent deterioration in risky-sleeve compensation and V-shape crash-brake targeting fast drawdowns) for a fixed 50/50 growth-defensive ETF sleeve, combined via a max-cash rule. It reports performance gains on the 2017-2026 window, including 20.45% CAGR and -16.77% max DD for the combination versus 16.62% CAGR and -33.59% max DD for the static sleeve, with similar improvements shown in walk-forward out-of-sample variants.","tokens_in":2053,"tokens_out":462,"duration_ms":17559,"significance":"If the filters prove robust after addressing multiple-testing concerns, the modular cash-overlay framework could offer a practical, separable tool for drawdown control in portfolio management. The explicit use of walk-forward validation is a methodological strength that partially mitigates overfitting risks.","major_comments":[{"comment":"Abstract: The mathematical definitions of the slow-tail compensation filter and V-shape crash-brake filter (including the three free parameters: slow-tail threshold, V-shape parameters, and max-cash rule) are not supplied, preventing verification of whether the reported performance gains depend on in-sample fitting to the 2017-2026 characteristics.","section":"Abstract"},{"comment":"Abstract: The headline metrics (20.45% CAGR / -16.77% max DD on the full window; 18.05% CAGR / -22.05% max DD in expanding OOS) are presented without statistical significance tests, transaction-cost adjustments, or multiple-testing corrections, even though the manuscript explicitly defers the latter to future work; this is load-bearing for the claim that the filters provide genuine improvement over the static sleeve.","section":"Abstract"}],"minor_comments":[{"comment":"The abstract would be clearer if it briefly noted the specific ETF tickers or data sources used for the growth and defensive baskets.","section":null}],"recommendation":"major_revision","confidential_remarks":"The self-acknowledgment of multiple-testing needs is appropriate and reduces the risk of overclaiming, but the parameter count and period overlap still warrant caution in interpreting the unadjusted results as robust evidence."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the detailed and constructive report. We address each major comment below and will revise the manuscript to improve clarity and robustness where feasible.","responses":[{"response":"The full manuscript provides the exact mathematical definitions, including the slow-tail threshold, V-shape parameters, and max-cash rule, in the methodology sections. To address the concern directly in the abstract and facilitate verification without requiring the full text, we will add concise equations and parameter descriptions to the revised abstract.","revision_made":"yes","referee_comment":"[Abstract] Abstract: The mathematical definitions of the slow-tail compensation filter and V-shape crash-brake filter (including the three free parameters: slow-tail threshold, V-shape parameters, and max-cash rule) are not supplied, preventing verification of whether the reported performance gains depend on in-sample fitting to the 2017-2026 characteristics."},{"response":"We agree that statistical significance tests and transaction-cost adjustments would strengthen the claims. We will add bootstrap-based significance tests for the performance differentials and incorporate realistic transaction-cost adjustments in the revised results. Multiple-testing corrections are explicitly deferred to future work as they require screening a larger universe of candidate filters; the current analysis relies on walk-forward validation as the primary control for overfitting.","revision_made":"partial","referee_comment":"[Abstract] Abstract: The headline metrics (20.45% CAGR / -16.77% max DD on the full window; 18.05% CAGR / -22.05% max DD in expanding OOS) are presented without statistical significance tests, transaction-cost adjustments, or multiple-testing corrections, even though the manuscript explicitly defers the latter to future work; this is load-bearing for the claim that the filters provide genuine improvement over the static sleeve."}],"tokens_in":1495,"tokens_out":391,"duration_ms":15438,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper develops two continuous filters for cash overlays on top of a fixed 50/50 growth-defensive ETF sleeve. The slow-tail filter watches for persistent drops in compensation when cash yields rise, while the V-shape filter catches sharp drawdowns and re-entry points. These feed into a max-cash rule that takes the larger cash weight each day.\n\nThe modular setup is useful because it isolates the overlay from any style timing in the risky sleeve. The walk-forward expanding-window version reports 18.05% CAGR and -22.05% max drawdown against the sleeve's 16.09% and -33.59% in the main out-of-sample window. That is concrete and the author is upfront about leaving multiple-testing adjustments for later work.\n\nThe main softness is that the filters have free parameters and the combination rule is also chosen, all evaluated on the same 2017-2026 period that produces the headline numbers. Even with walk-forward weights, the selection process can still capitalize on period-specific features. Transaction costs and statistical significance are not addressed in the available description.\n\nThis is for quant portfolio managers who run static sleeves and want simple add-on cash rules. It is not a broad theoretical result but could serve as a practical template if the exact filter definitions hold up under scrutiny.\n\nSend it to referees. The walk-forward evidence is worth checking in detail, and the author's own note on future statistical work gives a clear path for revision.","headline":"The paper shows backtested gains from two custom cash filters on a fixed sleeve with walk-forward checks, but the unadjusted parameter search on 2017-2026 data is the clear limitation.","tokens_in":2562,"tokens_out":382,"would_cite":false,"duration_ms":12607,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Two continuous cash-overlay filters combined by max-cash rule improve CAGR and reduce drawdown on a static risky sleeve.","keywords":["cash overlay","portfolio allocation","drawdown control","walk-forward validation","max-cash combination","risky sleeve","continuous filters","growth-defensive sleeve"],"falsifier":"Performance of the same filters applied to data after 2026 or to a different set of assets would show whether the reported improvements hold.","tokens_in":2784,"feed_emoji":"📈","tokens_out":559,"duration_ms":12739,"temperature":0.7,"pith_summary":"The paper develops two continuous filters for overlaying cash on a fixed 50/50 growth-defensive ETF sleeve: one for slow-tail compensation when risky returns deteriorate relative to cash, and one for V-shape crash brakes during fast drawdowns. These are combined daily by taking the larger cash weight from each filter. On 2017-2026 data the combination delivers 20.45 percent CAGR versus 16.62 percent for the static sleeve while cutting maximum drawdown from 33.59 percent to 16.77 percent, with walk-forward out-of-sample validation confirming gains. A sympathetic reader cares because the approach separates the cash decision from any style-timing policy, offering a modular way to manage tail risk without altering the underlying risky allocation.","feed_headline":"Cash overlays boost CAGR to 20.45% and halve drawdown on risky sleeve","feed_subtitle":"Slow-tail compensation and V-shape crash-brake filters on a static 50/50 growth-defensive sleeve deliver higher returns with lower maximum l","key_machinery":"The max-cash combination rule, under which the portfolio uses the larger of the two cash weights from the slow-tail compensation filter and the V-shape crash-brake filter each day.","core_discovery":"The selected-weight max-cash combination of the slow-tail compensation filter and the V-shape crash-brake filter earns a 20.45 percent CAGR versus 16.62 percent for the static risky sleeve on the 2017-2026 window, and improves maximum drawdown from -33.59 percent to -16.77 percent. A stricter walk-forward version in the main out-of-sample window earns 18.05 percent versus 16.09 percent with maximum drawdown of -22.05 percent versus -33.59 percent. The evidence supports modular continuous cash overlays as drawdown-control tools.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Max-cash filters post 20.45% CAGR versus 16.62% on risky sleeve","Slow-tail and crash-brake filters cut drawdown to -16.77% from -33.59%","Walk-forward max-cash earns 18.05% CAGR with -22.05% max drawdown","Continuous cash overlays show 20.45% CAGR and lower drawdown on sleeve"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The filter parameters and combination rule were not optimized in a way that capitalizes on the specific characteristics of the 2017-2026 period.","fun_headline_variants_meta":{"raw":{"variants":["Max-cash filters post 20.45% CAGR versus 16.62% on risky sleeve","Slow-tail and crash-brake filters cut drawdown to -16.77% from -33.59%","Walk-forward max-cash earns 18.05% CAGR with -22.05% max drawdown","Continuous cash overlays show 20.45% CAGR and lower drawdown on sleeve"]},"model":"grok-4.3","cost_usd":0.004472,"raw_usage":{"total_tokens":2303,"prompt_tokens":813,"num_sources_used":0,"completion_tokens":97,"cost_in_usd_ticks":44724500,"prompt_tokens_details":{"text_tokens":813,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1393,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":813,"tokens_out":97,"duration_ms":8649,"temperature":1.0,"reasoning_tokens":1393,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-27T14:18:32.668307+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Performance of the same filters applied to data after 2026 or to a different set of assets would show whether the reported improvements hold.","supporting_citations":[],"review_version":1}