decision log — the ten-agent consensus, on the record
Decision Log
Ten independent agents score every covered name from real price history and combine into one BUY / SELL / HOLD call. When that call changes on a name, the flip is logged — the prior call, the new call, the conviction behind it, and the one-line reason drawn from the agents' own votes. The log is persisted: it survives restarts and is never back-filled.
| Flip | Why — from the agents’ own votes | |||
|---|---|---|---|---|
| 09-12 19:40Z | QQQ | SELL→BUY | 7% | Now 2 buy / 1 sell / 7 hold — Trend: Price above a rising 50 over 200-day average — established uptrend. |
| 09-12 19:40Z | SPY | HOLD→BUY | 15% | Now 2 buy / 0 sell / 8 hold — Trend: Price above a rising 50 over 200-day average — established uptrend. |
| 09-12 19:40Z | WMT | SELL→HOLD | 5% | The directional votes no longer agree (2 buy / 2 sell / 6 hold) — the desk steps back to hold. |
| 09-12 19:40Z | GOOGL | SELL→HOLD | 2% | The directional votes no longer agree (1 buy / 1 sell / 8 hold) — the desk steps back to hold. |
| 09-12 19:40Z | AMZN | SELL→BUY | 15% | Now 2 buy / 0 sell / 8 hold — Trend: Price above a rising 50 over 200-day average — established uptrend. |
| 09-11 22:08Z | DIS | SELL→BUY | 12% | Now 2 buy / 0 sell / 8 hold — Relative Strength: Outperforming the market by 4.6% over ~3 months — leadership. |
| 09-11 22:08Z | QQQ | BUY→SELL | 14% | Now 1 buy / 3 sell / 6 hold — MACD: MACD histogram negative and expanding — bearish momentum building. |
| 09-11 22:08Z | SPY | BUY→HOLD | 5% | The directional votes no longer agree (2 buy / 3 sell / 5 hold) — the desk steps back to hold. |
| 09-11 22:08Z | TSLA | HOLD→SELL | 8% | Now 1 buy / 2 sell / 7 hold — Relative Strength: Lagging the market by 14.0% over ~3 months — a laggard. |
| 09-10 19:42Z | UNH | BUY→SELL | 6% | Now 1 buy / 2 sell / 7 hold — RSI Trend: RSI 44 — in the bearish 30-45 band. |
| 09-10 19:42Z | TSLA | BUY→HOLD | 4% | The directional votes no longer agree (2 buy / 1 sell / 7 hold) — the desk steps back to hold. |
| 09-10 19:42Z | AMZN | HOLD→SELL | 21% | Now 1 buy / 4 sell / 5 hold — Momentum: Trading -3.0% below its 20-day average — negative momentum. |
| 09-09 21:09Z | AMD | HOLD→BUY | 36% | Now 6 buy / 1 sell / 3 hold — Momentum: Trading 6.0% above its 20-day average — positive momentum. |
| 09-09 21:09Z | TSLA | SELL→BUY | 12% | Now 3 buy / 1 sell / 6 hold — Momentum: Trading 5.1% above its 20-day average — positive momentum. |
| 09-05 14:50Z | AMD | BUY→HOLD | 3% | The directional votes no longer agree (1 buy / 1 sell / 8 hold) — the desk steps back to hold. |
| 09-05 14:50Z | UNH | HOLD→BUY | 13% | Now 2 buy / 0 sell / 8 hold — MACD: MACD histogram positive and expanding — bullish momentum building. |
| 09-05 14:50Z | XOM | HOLD→BUY | 9% | Now 3 buy / 2 sell / 5 hold — Trend: Price above a rising 50 over 200-day average — established uptrend. |
| 09-05 14:50Z | AMZN | BUY→HOLD | 0% | The directional votes no longer agree (2 buy / 2 sell / 6 hold) — the desk steps back to hold. |
| 09-05 08:23Z | DIS | HOLD→SELL | 7% | Now 1 buy / 2 sell / 7 hold — MACD: MACD histogram negative and expanding — bearish momentum building. |
| 09-05 08:23Z | NFLX | HOLD→SELL | 16% | Now 1 buy / 3 sell / 6 hold — Relative Strength: Lagging the market by 9.2% over ~3 months — a laggard. |
| 09-05 08:23Z | AMD | SELL→BUY | 13% | Now 2 buy / 0 sell / 8 hold — MACD: MACD histogram positive and expanding — bullish momentum building. |
| 09-05 08:23Z | UNH | BUY→HOLD | 2% | The directional votes no longer agree (1 buy / 1 sell / 8 hold) — the desk steps back to hold. |
| 09-05 08:23Z | XOM | BUY→HOLD | 2% | The directional votes no longer agree (2 buy / 2 sell / 6 hold) — the desk steps back to hold. |
| 09-05 08:23Z | TSLA | BUY→SELL | 13% | Now 2 buy / 3 sell / 5 hold — Relative Strength: Lagging the market by 13.9% over ~3 months — a laggard. |
| 09-05 08:23Z | GOOGL | HOLD→SELL | 17% | Now 1 buy / 3 sell / 6 hold — Relative Strength: Lagging the market by 12.6% over ~3 months — a laggard. |
| 09-05 08:23Z | AMZN | HOLD→BUY | 8% | Now 2 buy / 1 sell / 7 hold — Trend: Price above a rising 50 over 200-day average — established uptrend. |
| 09-04 12:20Z | DIS | BUY→HOLD | 5% | The directional votes no longer agree (2 buy / 1 sell / 7 hold) — the desk steps back to hold. |
| 09-04 12:20Z | NFLX | BUY→HOLD | 1% | The directional votes no longer agree (2 buy / 2 sell / 6 hold) — the desk steps back to hold. |
| 09-04 12:20Z | QQQ | SELL→BUY | 8% | Now 2 buy / 1 sell / 7 hold — Trend: Price above a rising 50 over 200-day average — established uptrend. |
| 09-04 12:20Z | TSLA | SELL→BUY | 24% | Now 5 buy / 1 sell / 4 hold — Momentum: Trading 8.4% above its 20-day average — positive momentum. |
| 09-04 12:20Z | GOOGL | SELL→HOLD | 5% | The directional votes no longer agree (1 buy / 1 sell / 8 hold) — the desk steps back to hold. |
| 09-04 12:20Z | AMZN | SELL→HOLD | 0% | The directional votes no longer agree (2 buy / 2 sell / 6 hold) — the desk steps back to hold. |
| 09-03 08:20Z | DIS | HOLD→BUY | 11% | Now 3 buy / 1 sell / 6 hold — Momentum: Trading 1.4% above its 20-day average — positive momentum. |
| 09-03 08:20Z | TSLA | HOLD→SELL | 17% | Now 1 buy / 3 sell / 6 hold — Relative Strength: Lagging the market by 19.6% over ~3 months — a laggard. |
| 09-03 08:20Z | META | HOLD→BUY | 13% | Now 3 buy / 1 sell / 6 hold — Momentum: Trading 3.2% above its 20-day average — positive momentum. |
| 09-02 12:07Z | DIS | BUY→HOLD | 5% | The directional votes no longer agree (2 buy / 1 sell / 7 hold) — the desk steps back to hold. |
| 09-02 12:07Z | AMD | HOLD→SELL | 26% | Now 1 buy / 4 sell / 5 hold — Relative Strength: Lagging the market by 12.2% over ~3 months — a laggard. |
| 09-02 12:07Z | QQQ | BUY→SELL | 15% | Now 1 buy / 3 sell / 6 hold — Relative Strength: Lagging the market by 5.5% over ~3 months — a laggard. |
| 09-02 12:07Z | UNH | SELL→BUY | 19% | Now 3 buy / 0 sell / 7 hold — MACD: MACD histogram positive and expanding — bullish momentum building. |
| 09-02 12:07Z | TSLA | BUY→HOLD | 5% | The directional votes no longer agree (2 buy / 2 sell / 6 hold) — the desk steps back to hold. |
How the agent weighting adapts
learning — weights graded from real outcomesThe ten agents' voting weights move with how well each agent's past calls scored against real outcomes, within a published range. Hit-rates carry their 95% confidence range and the count of distinct graded calls behind them; excess edge is measured vs the market benchmark over the same window. A figure not yet earned from real outcomes shows as “—”.
trend1.00×hit 42% · 95% CI 26–59%30 gradededge -1.05% vs markettrend: hit 42% of excess-return calls on 30 distinct bar-dates (95% CI 26–59%) — the interval still spans the coin-flip line, equal weight held.
momentum1.00×hit 44% · 95% CI 28–61%30 gradededge -0.82% vs marketmomentum: hit 44% of excess-return calls on 30 distinct bar-dates (95% CI 28–61%) — the interval still spans the coin-flip line, equal weight held.
meanRevert1.00×hit 52% · 95% CI 35–69%30 gradededge -0.56% vs marketmeanRevert: hit 52% of excess-return calls on 30 distinct bar-dates (95% CI 35–68%) — the interval still spans the coin-flip line, equal weight held.
macd1.00×hit 45% · 95% CI 29–63%30 gradededge -0.67% vs marketmacd: hit 45% of excess-return calls on 30 distinct bar-dates (95% CI 29–63%) — the interval still spans the coin-flip line, equal weight held.
breakout1.00×hit 46% · 95% CI 29–63%30 gradededge -1.16% vs marketbreakout: hit 46% of excess-return calls on 30 distinct bar-dates (95% CI 29–63%) — the interval still spans the coin-flip line, equal weight held.
volume1.00×hit 47% · 95% CI 31–65%30 gradededge -1.05% vs marketvolume: hit 47% of excess-return calls on 30 distinct bar-dates (95% CI 31–65%) — the interval still spans the coin-flip line, equal weight held.
volatility1.00×hit —— gradededge — vs marketvolatility: regime commentary only — excluded from reweighting by design; weight pinned at 1.00x.
rsiTrend1.00×hit 42% · 95% CI 27–60%30 gradededge -1.02% vs marketrsiTrend: hit 42% of excess-return calls on 30 distinct bar-dates (95% CI 27–60%) — the interval still spans the coin-flip line, equal weight held.
levels1.00×hit 49% · 95% CI 32–66%30 gradededge +0.35% vs marketlevels: hit 49% of excess-return calls on 30 distinct bar-dates (95% CI 32–66%) — the interval still spans the coin-flip line, equal weight held.
relStrength1.00×hit 44% · 95% CI 28–61%30 gradededge -0.88% vs marketrelStrength: hit 44% of excess-return calls on 30 distinct bar-dates (95% CI 28–61%) — the interval still spans the coin-flip line, equal weight held.
window 180 days · 5960 graded calls · effective n 30
Why effective n is so much smaller than the graded-call count: calls are graded on overlapping multi-day windows and cluster on correlated names, so they are far from independent observations. Effective n is the equivalent count of independent samples — the confidence ranges above are built on it, not on the raw call count.
Method: bounded, disclosed online reweighting — every change is graded against real outcomes and replayable · Equal weight (every agent at 1.0x) is the permanent unlearned baseline every disclosed weight is read against.
A steady desk adds no rows by design.