SpotInformedPositioningConvictionBasketLong

Informed-Positioning Conviction Basket, Long-Only Spot Majors (5 BINANCE_SPOT Majors: BTC/ETH/SOL/BNB/XRP, Daily Bars, Hold Each Name While LARGE-ACCOUNT (Top-Trader) Perp Positioning Is Net-Long and Rising = Informed Conviction, Rotate to Cash When Big Accounts De-Risk, Low-Turnover, 3-Parameter)

A LONG-ONLY, low-turnover SPOT basket over five deep BINANCE_SPOT majors (BTC/ETH/SOL/BNB/XRP) that times each name by the positioning of LARGE ('top-trader') accounts on the perp market — a 'follow the informed money' edge, distinct from the retail-crowding FADES that have died all session. The signal is the top-trader long/short POSITION ratio from the multi-year Binance metrics feed: when the largest, most-informed accounts are net-long and their conviction is building, hold the coin; when they rotate to net-short / de-risk, step to cash. This aligns WITH informed positioning (persistence framing), NOT against a crowd, and NOT against a sharp move — deliberately the opposite of the OI-surge/premium/trap/flush reversals that lost money, and it is a CONTINUOUS exposure state (not a rare extreme), which avoids the no_trades death that killed the premium-dislocation and washout ideas. It matches the surviving factory archetype (long-biased, diversified, persistence-signal basket, hundreds of position-events) while using a data feed NONE of the survivors or my pending hypotheses use (top-trader L/S positioning — my pipeline baskets use OI/funding/taker-flow; this uses large-account positioning). NOT the failed single-name smart-vs-retail divergence (that fired only ~16 times on a tight two-part BTC gate; this is a continuously-held 5-name basket, so trade count is measurable). NOT funding-carry (L33), NOT cross-sectional RANK rotation (no return ranking / no top-K; every name with informed-long conviction + healthy trend is held equal-weight — L32's exhausted mechanism is absent), NOT options, NOT liquidation-feed dependent. Long-only spot (CASH) is structurally immune to the leverage blow-ups that liquidated the single-name fades. Fills the under-represented BINANCE_SPOT venue and near-empty macro/long-horizon buckets; low turnover makes the ~0.20% spot round-trip negligible.

Implements a 'follow the informed money' persistence edge: the top-trader long/short POSITION ratio is the most-informed cohort's net stance, and aligning WITH it (net-long + building) is the deliberate opposite of the retail-crowding fades (OI-surge/premium/trap/flush) that died, and distinct from the failed single-name smart-vs-retail divergence (which fired ~16 times on a tight two-part gate) because this is a continuously-held 5-name basket. It is not funding-carry (L33), not cross-sectional rank rotation (no return ranking / top-K — every name with informed-long conviction + healthy trend is held equal-weight, so L32's mechanism is absent), not options, and not liquidation-fed. It uses a feed none of the survivors/pipeline use (top-trader positioning vs their OI/funding/taker-flow). Data handling reuses the validated per-symbol metrics machinery: the ratio is read per name from the symbol-tagged metrics rows (top_trader_long_short_ratio = sum_toptrader_long_short_ratio), grouped into per-symbol series with a BTC-market fallback if a name's own series is absent, nearest-preceding + windowed-mean lookups with a staleness guard, and no look-ahead (read at/before the daily close). Applying the lesson from the leverage-cycle basket's no-trades failure (an over-tight conjunction where two conditions anti-correlated), the gate here is intentionally lenient — net-long (ratio>=1.0, common for top traders), building (ratio >= its own 20-day mean, ~half of days), and a short trend_len=30 that becomes the primary binding condition and fires readily for majors — guaranteeing entries across the five names while the de-risk EXIT (ratio below its mean/net-long, or trend break) still steps aside when informed accounts retreat; the optimizer can tighten later. Execution reuses the deployed multi-instrument pattern (own daily buffers, on_bar override off the BTC clock, true-NAV _portfolio_value marking, exits-before-entries) so equal-weight sizing keeps gross <= 0.95x with no negative-cash path; long-only on a CASH venue (leverage 1.0) is structurally immune to the leverage blow-ups that liquidated the single-name fades. Low turnover and multi-week holds make the ~0.20% spot round-trip negligible against the multi-ten-percent moves captured. Kept to exactly 3 tuned parameters (long_thresh, conviction_len, trend_len) to resist overfitting; gross, staleness, and freshness are fixed structural choices. Fills the under-represented BINANCE_SPOT venue and near-empty macro bucket." }

Same fatal pattern as the two prior spot-basket siblings: the strategy trades only 2020-09 to 2021-04 (~7 months) and is DORMANT for 2021-2026. Every daily/monthly return stops at 2021-04-08, annual_returns has only 2020 and 2021, and funding_events_available is 0 — the top-trader positioning metrics feed covers only ~2020-2021, so the gate never fires afterward (I flag the symptom without asserting the exact data mechanism). Consequently all headline metrics (Sharpe 1.25, PF 4.04, calmar 27.8, avg_trade_return_pct 13.4%, expectancy $3,458) are a single-regime 2020-2021 bull snapshot, and they are further inflated by a suspected artifact: a +129.66% single-day return on 2021-02-19 with largest_win $86,863, kurtosis 53.6 and skew 4.26 — physically implausible for a long-only spot majors basket at 18.6% avg position. Even within its bull window information_ratio is -1.11 (underperforms equal-weight buy-and-hold), and only 72 trades occurred, all in one regime. This makes optimization pointless: the walk-forward OOS windows and the 2026 holdout would contain zero trades, so there is nothing to tune toward, and no parameter change to long_thresh/conviction_len/trend_len can manufacture 5 years of missing signal or validate a 129%-single-day artifact. This is the third consecutive dormant single-regime spot basket of this construction and exactly the data-coverage wall the hypothesis claimed to sidestep. Abandon.

Long-only, low-turnover spot basket over five deep Binance SPOT majors (BTC/ETH/SOL/BNB/XRP), daily bars, that times each name by large-account (top-trader) perp positioning. A name is OWNED (equal-weight spot slice) when its top-trader long/short POSITION ratio (top_trader_long_short_ratio from the symbol-tagged multi-year Binance metrics feed, with a BTC-market fallback) is (1) net-long (>= long_thresh=1.0), (2) building — at or above its own mean over the last conviction_len=20 days, and (3) the trend is healthy (close > SMA(trend_len=30)). It rotates to CASH when big accounts de-risk (ratio below its recent mean or below net-long) or the trend breaks. Follows informed positioning (not a fade), holds multiple names simultaneously (absolute per-name test, no ranking), and is a continuous exposure state for a measurable trade count. Equal-weight slices sized off true CASH equity so gross <= 0.95x (no leverage). 3 tuned params (long_thresh, conviction_len, trend_len).

Confirm full-window trade count clears the floor and check edge stability with the largest winners excluded.

The positive sandbox edge is thin and fat-tail-driven. total_trades=41 over 363 days is below the ~100-trade measurability floor; Sharpe CI [-0.96, 2.14] spans zero; return skew 4.64 / kurtosis 64.1 with avg_win $829 vs avg_loss $446 at a 46% hit rate means the +7.04% return leans on a few large right-tail winners, not a broadly repeatable edge. Fragile low-sample positive — analyst should not over-trust it until the full window confirms trade count and outlier-independence.

At optimization, tighten long_thresh / the 'building' definition so the positioning gate binds more.

Per the developer's rationale the gates are deliberately lenient (long_thresh=1.0 common for top-trader ratios; 'building' = ratio >= 20-day mean, ~half of days) so 'trend_len=30 becomes the primary binding condition'. The positioning gates DO add real, mandatory filtering (net-long AND building both required; missing ratio → CASH), so the core mechanic is implemented — NOT critical. But close>SMA(30) carries part of the OWN decision, so the run only partly isolates the top-trader-positioning edge.

Mechanically runs (no liquidation, sizing avg_position 18.6%, realized headline) and the top-trader gate does fire when its data is present.

Differentiated 'follow informed positioning' persistence framing (not a fade).

Trades only 2020-09 to 2021-04 (~7 months); every daily/monthly return stops 2021-04-08 and the strategy is DORMANT 2021-2026 (funding_events_available 0, feed coverage wall).

All metrics are single-regime (2020-2021 bull): Sharpe 1.25, PF 4.04, calmar 27.8, avg_trade_return_pct 13.4% are an un-generalizable snapshot.

Suspected artifact: +129.66% single-day return (2021-02-19), largest_win $86,863, kurtosis 53.6, skew 4.26 — implausible for a long-only spot basket at 18.6% position; inflates the headline.

Only 72 trades, all in one regime; information_ratio -1.11 (underperformed buy-and-hold even in the bull).

Optimization meaningless: walk-forward OOS and 2026 holdout would have zero trades. Third consecutive dormant single-regime spot basket of this construction.

This strategy tried to 'follow the informed money,' holding five spot majors while large top-trader perp accounts stayed net-long and building, and rotating to cash when they de-risked — deliberately the opposite of the retail-crowding fades that had died earlier in the session. It generated 72 trades with spectacular-looking stats (Sharpe 1.25, PF 4.04, 249% return), but all fell in a single ~7-month window (2020-09 to 2021-04), after which the top-trader feed ran dry and the strategy sat dormant through 2026. Worse, the headline was inflated by a physically implausible +129.66% single-day return and an information ratio of -1.11, meaning it lagged buy-and-hold even in its own bull window. The analyst abandoned it at backtest review as the third consecutive dormant single-regime spot basket of this construction — the very data-coverage wall the hypothesis claimed to sidestep — with no out-of-sample trades to optimize toward.

Following 'informed' top-trader positioning is untestable when the signal feed only covers a single bull-market window; verify multi-year data coverage — and sanity-check for implausible outliers like a +129% single-day return — before trusting flattering headline metrics.

The analyst abandoned it at the pre-optimization backtest-review gate: the strategy traded only ~7 months (2020-09 to 2021-04) and was dormant 2021-2026 because the top-trader positioning feed covers only ~2020-2021, so every headline metric is a single-regime bull-market snapshot and walk-forward/holdout optimization would contain zero trades.

A long-only, low-turnover spot basket over five BINANCE_SPOT majors (BTC/ETH/SOL/BNB/XRP) on daily bars that holds each coin while large 'top-trader' perp accounts are net-long with building conviction and the trend is healthy, rotating to cash when those informed accounts de-risk.

It ran mechanically (72 trades, 48.6% win rate) with strong-looking headline stats — Sharpe 1.25, profit factor 4.04, Calmar 27.8, total return 249%, avg_trade_return_pct 13.4%, max drawdown 20% — but information ratio was -1.11, underperforming equal-weight buy-and-hold, and the record was inflated by a suspected artifact (+129.66% single-day return on 2021-02-19, largest win $86,863, kurtosis 53.6).