AI Stock Picks — FTSE 100

Monday, September 28, 2026  ·  Five AI analysts screened this index for stocks with the best shot at gaining +5% within 30 days

Today's Market Mood

Neutral  Normal choppiness · VIX 16.1

FTSE 100 breadth is narrow and bifurcated: a cluster of energy, housebuilder and select bank/industrial names show strong, sustained momentum near their 52-week highs, while miners, gold-linked names, and several growth/services stocks have suffered sharp double-digit 30-day declines. With no fresh company-specific news pre-fetched, this call leans on price-momentum and proximity-to-high signals, favoring rotation into resilient cyclicals over broad-based buying.

Energy majors (Shell, BP) rallying on firmer oil prices, both up ~9-10% over 30 days and within single digits of their highsHousebuilder rebound (Persimmon, Barratt Redrow) on rate-cut/stimulus optimism, with double-digit 5-day and 30-day gainsSelective bank and industrial strength (Standard Chartered, NatWest, Bank of Georgia, Smiths Group, Diploma, IMI) holding near 52-week highs even as broader financials/miners diverge sharply lowerBroad weakness in miners, gold/silver names and growth-sensitive services (Fresnillo, Anglo American, Rio Tinto, Experian, Entain, Autotrader, LSEG) dragging overall index breadth down

Today's Actionable List

Which of today's picks to actually take, and at what size, given what you are already holding. Each pick's own Suggested Allocation sizes it in isolation; this table applies the portfolio limits — at most 20 open positions and 60% of the portfolio committed in total. Run with --capital 25000 to see these sizes as order amounts rather than percentages.

0 of 3
0%
58 · 254.2%
254.2% / 60%
0 of 20
RankStockActionSizeEntry → TargetNotes
#1 BTRW.L SKIP — £345.20 → £362.46
stop £311.03
 Half-Kelly sizes this at 0% — the odds carry no edge worth capital.
#2 SMIN.L SKIP — £2,827.00 → £2,968.35
stop £2,640.42
 Half-Kelly sizes this at 0% — the odds carry no edge worth capital.
#3 DPLM.L SKIP — £7,710.00 → £8,095.50
stop £7,285.95
 Position cap reached — 58 of 20 slots in use.

Already open: IMI.L (to Sep 30), HSX.L (to Oct 1), LSEG.L (to Oct 1), ADM (to Oct 2), BEZ.L (to Oct 2), BIIB (to Oct 3), DE (to Oct 3), ELV (to Oct 4), HSBA.L (to Oct 7), ITRK.L (to Oct 8), SDR.L (to Oct 9), VOD.L (to Oct 9) … and 46 more. Symbols you already hold were excluded from this run's candidates rather than re-picked.

Position-sizing guidance only — half-Kelly per pick, bounded by the limits above. Not personal investment advice.

Behind the scenes: how each AI analyst did
Aria
4 picks
Blaze
3 picks
Nova
3 picks
Orion
2 picks
Quinn
4 picks

3 Stock Picks

#1 BTRW.LBarratt Redrow plc
4/5 Agents Agree  Risk 5/10
Moderate conviction · 55%

Barratt Redrow is one of the UK's largest homebuilders, formed from the merger of Barratt Developments and Redrow. Its shares have jumped sharply over the past week and month and are clearly beating the broader market, likely on optimism around housing demand or interest-rate expectations. The main risk is that the stock is still well below its 52-week high and any wobble in the housing market outlook could quickly cool this rally.

Stop-loss£311.03-9.9%
Today's price£345.20
Target£362.46+5.0%
If you invested £1,000
▲£1,050 if the target hits (+5.0%)
▼£901 if the stop-loss is hit (-9.9%)
The AI puts the chance of the win case at 55%. Not a guarantee — always set the stop-loss.

Why the AI likes it

What could go wrong

⚔ The Skeptic's View — Sage, the Devil's Advocate
  • RSI at 69.69 sits at the edge of overbought territory, and price (345.2) is already above the upper Bollinger Band (331.62), which historically precedes short-term mean reversion rather than continuation.
  • OBV reads -209,600,799, a negative cumulative volume figure that contradicts the 'volume confirms buying interest' narrative — price is rising while cumulative on-balance volume trends negative.
  • Price (345.2) is running 39 points above both the 50-day (306.48) and 200-day (308.55) SMAs with no golden cross confirmed, leaving significant air pocket risk if the rally stalls.
Show the detailed analysis

Calibrated probability: 55% (raw model estimate: 61%) — adjusted using the accuracy of past resolved picks.

Trade math

0.5:1
66%
-0.8%
0%

Suggested allocation is half-Kelly sizing guidance capped at 10% — not personal investment advice. Allocations are scaled to 85% of half-Kelly for the current market regime.

What each AI analyst estimated

Aria
70%
High
Blaze
56%
Medium
Nova
56%
Medium
Orion
—
—
Similar issue to Persimmon: sharp price gains aren't matched by supportive trading volume, and the stock looks overbought after its fast run-up, making the rally look fragile.
Quinn
60%
Medium

Data Sources

Yahoo FinanceFinvizMarketWatchSeeking Alpha
#2 SMIN.LSmiths Group plc
4/5 Agents Agree  Risk 4/10
Moderate conviction · 55%

Smiths Group is a diversified engineering business making detection, medical and industrial equipment. The stock has surged over the past week and month, is close to its 52-week high, and is comfortably beating the wider market, all pointing to strong investor confidence. The key risk is that the rapid pace of the recent climb (RSI above 70) means the stock is due a breather, and its earnings were just reported, which can sometimes bring volatility after the fact.

Stop-loss£2,640.42-6.6%
Today's price£2,827.00
Target£2,968.35+5.0%
If you invested £1,000
▲£1,050 if the target hits (+5.0%)
▼£934 if the stop-loss is hit (-6.6%)
The AI puts the chance of the win case at 55%. Not a guarantee — always set the stop-loss.

Why the AI likes it

What could go wrong

⚔ The Skeptic's View — Sage, the Devil's Advocate
  • RSI at 71.27 is clearly overbought, and price (2827.0) is trading essentially at the upper Bollinger Band (2828.47), a classic setup for a pullback rather than fresh upside.
  • The bull case frames being near the 52-week high as strength, but at only -1.46% from that high after a 9.53% 5-day spike, most of the anticipated 5% move may already be priced in, compressing further reward.
Show the detailed analysis

Calibrated probability: 55% (raw model estimate: 61%) — adjusted using the accuracy of past resolved picks.

Trade math

0.8:1
57%
+0.5%
0%

Suggested allocation is half-Kelly sizing guidance capped at 10% — not personal investment advice. Allocations are scaled to 85% of half-Kelly for the current market regime.

What each AI analyst estimated

Aria
65%
High
Blaze
—
—
Already trading within 1.5% of its 52-week high with RSI at 71 — little value-entry room left and closer to overbought than the selected picks
Nova
53%
Medium
Orion
66%
High
Quinn
60%
Medium

Data Sources

Yahoo FinanceFinvizMarketWatchSeeking Alpha
#3 DPLM.LDiploma PLC
3/5 Agents Agree  Risk 4/10  ⚠ Conflicting Signals
Moderate conviction · 55%

Diploma distributes specialised industrial parts like seals, controls and life-science components to other businesses. Its shares have climbed steadily over the past month, are outperforming the wider UK market, and the stock is trading above both its 50-day and 200-day average price, all signs of healthy, broad-based buying rather than a short-lived spike. The main risk is that this run-up has been strong enough that a pause or pullback would not be surprising.

Stop-loss£7,285.95-5.5%
Today's price£7,710.00
Target£8,095.50+5.0%
If you invested £1,000
▲£1,050 if the target hits (+5.0%)
▼£945 if the stop-loss is hit (-5.5%)
The AI puts the chance of the win case at 55%. Not a guarantee — always set the stop-loss.

Why the AI likes it

What could go wrong

⚔ The Skeptic's View — Sage, the Devil's Advocate
  • RSI at 68.12 is approaching overbought, and price (7710.0) is closing in on the upper Bollinger Band (7789.07), leaving little technical room before resistance.
  • The 5-day momentum (5.52%) has already outpaced the 30-day return (5.18%), suggesting the move is compressing into a shorter window — a sign of acceleration that often precedes exhaustion rather than a sustainable steady uptrend.
Show the detailed analysis

Calibrated probability: 55% (raw model estimate: 67%) — adjusted using the accuracy of past resolved picks.

Trade math

0.9:1
52%
+1.5%
8.5%

Suggested allocation is half-Kelly sizing guidance capped at 10% — not personal investment advice. Allocations are scaled to 85% of half-Kelly for the current market regime.

What each AI analyst estimated

Aria
75%
Very High
Blaze
—
—
Already trading within 2% of its 52-week high with RSI at 68 — limited upside room and a weaker fallback case under degraded data than the selected picks
Nova
—
—
Positive momentum but weaker than the top picks, with today's trading volume well below average, giving less confidence the move is being actively bought.
Orion
66%
High
Quinn
60%
Medium

Data Sources

Yahoo FinanceFinvizMarketWatchSeeking Alpha

Retail Investor Execution Guide

This report identifies stocks with an AI-assessed probability of gaining +5% within 30 calendar days (target exit by October 28, 2026). Picks are ranked by how many independent AI agents agreed — more agreement means higher conviction. The guide below tells you exactly how to act on them.

Step-by-Step Action Plan

  1. Act within one trading day. Prices are freshest now. The longer you wait, the more the entry price drifts away from what the agents analysed.
  2. Verify the entry price. Check the live price before buying. If a stock has already moved more than 3% above the Current Price shown in the report, skip it or wait for a pullback — the risk/reward has shifted.
  3. Use limit orders, not market orders. Set your buy limit at or below the Current Price shown. Chasing with a market order gives brokers and algorithms an advantage over you.
  4. Set your stop-loss immediately after buying. Use the Max Downside % on each pick card as your hard exit level. If the price falls to that level, sell without hesitation — the trade thesis is broken.
  5. Set a take-profit at the Target Price. Place a limit sell order at the Target Price shown. When it fills, the trade is done — resist the urge to hold for more.
  6. Exit all positions by October 28, 2026. This is your hard deadline. If a stock has not hit its target by then, exit anyway. Holding past the window turns a short-term trade into an unsupervised long-term position.
  7. Never add to a losing position. If your stop-loss is hit, exit completely. Averaging down turns a small, controlled loss into a potentially large one.

Position Sizing by Conviction Tier

Size each position according to the coloured border on its pick card. Cap your total exposure across all picks from this report at 25% of your overall trading budget.

Agents Agreeing What It Means Max Allocation Per Pick
5/5 Highest conviction — all five independent analytical checks passed Up to 10%
4/5 Very high conviction — four of five independent checks aligned Up to 8%
3/5 High conviction — three agents independently agreed Up to 6%
2/5 Moderate conviction — two agents independently agreed Up to 4%
1 Agent Speculative — solo pick, admitted only at Very High confidence Up to 2%

The Three Exit Rules

These picks are generated by AI agents for informational and educational purposes only. They do not constitute financial advice or a recommendation to buy or sell any security. Probability estimates reflect the agents' analytical models — they are not guarantees of any outcome. Past results are not indicative of future performance. Markets can and do move against even well-researched trades. Always conduct your own research, and consider seeking advice from a licensed financial adviser before placing any orders. Never invest money you cannot afford to lose.

Understanding the Numbers

A plain-English guide to every figure shown on each pick card — what it means, how it is calculated, and what a good or bad value looks like.

AI Conviction Meter
Estimated probability of hitting the target, shown as a 5-bar gauge
The filled bars and label translate the AI's probability estimate into plain language: 75%+ is Very strong, 60–74% Strong, 50–59% Moderate, below 50% Cautious. Where enough past picks have been resolved, the probability is calibrated against the AI's real track record.
e.g. "Moderate conviction · 58%" — the AI thinks this trade works slightly more often than a coin flip.
✓ More filled bars = more confidence — but even Very strong picks fail sometimes; always size positions accordingly.
If You Invested $1,000
$1,000 × (1 + Target Gain ÷ 100)  /  $1,000 × (1 + Max Downside ÷ 100)
The same Target Gain and Max Downside percentages expressed in concrete money terms for a $1,000 position — what you'd have if the target hits versus if your stop-loss is triggered.
e.g. +10% target and −12.3% stop → $1,100 on a win, $877 on a stopped-out loss.
✓ Scales linearly — a $500 position sees half these amounts, a $2,000 position double.
Current Price
Market price at time of analysis
The stock's price when this report was generated. All other figures on the card are calculated from this reference point.
e.g. $215.20 — check the live price before placing an order; if it has moved more than ~3%, re-evaluate the risk/reward.
✓ No positive/negative distinction — it is simply your cost basis.
Target Price
Current Price × (1 + Target Gain ÷ 100)
The price the agents project the stock could reach within the time window. This is your take-profit level — place a limit sell order here after buying.
e.g. $215.20 × 1.085 = $233.49 target on an 8.5% pick.
✓ Always above Current Price — this tool only selects upside candidates.
Target Gain
(Target Price − Current Price) ÷ Current Price × 100
The percentage return you would earn if the stock hits its Target Price from your entry. Set by the run parameters — every pick in this report targets the same percentage.
e.g. +8.5% means a $100 stock has a $108.50 target; a $500 stock has a $542.50 target.
✓ Always positive — picks are upside-only. A larger % means a bigger move is required.
Max Downside
Estimated stop-loss level, as % below entry
The worst-case loss the agents estimated if the thesis fails — based on key support levels and recent volatility. Use this as your hard stop-loss. If the price falls to this level, exit immediately; the trade thesis is broken.
e.g. −12.0% means: if you bought at $100, sell if it drops to $88.00.
⚠ Always negative. A smaller magnitude (e.g. −5%) = tighter risk. A larger magnitude (e.g. −20%) = more room to fall before you exit — riskier.
Reward : Risk
Target Gain ÷ |Max Downside|
How much you could gain for every dollar you are risking. A ratio of 2.0:1 means you stand to make $2 for every $1 at risk. Higher is better.
e.g. 8.5% gain ÷ 12% downside = 0.7:1 (unfavourable). 10% gain ÷ 5% downside = 2.0:1 (good).
✓ ≥ 2.0:1 is strong. ⚠ < 1.0:1 means you are risking more than you could gain — approach with caution.
Breakeven Win Rate
|Max Downside| ÷ (Target Gain + |Max Downside|) × 100
The minimum percentage of trades at this reward:risk that must succeed for you to break even over time — regardless of how good your stock picking is.
e.g. 12 ÷ (8.5 + 12) = 59% — you need roughly 6 in 10 trades to win just to avoid losing money.
✓ Lower is better — ≤ 35% means even a poor win rate is survivable. ⚠ ≥ 55% means you need most trades to work out.
Expected Value
(Probability% × Target Gain%) + ((1 − Probability%) × Max Downside%)
The average return per trade if the AI's probability estimate is accurate over many similar trades. A positive Expected Value means the trade has a mathematical edge; a negative value means the math works against you on average, even if you sometimes win.
e.g. 65% × +8.5% + 35% × −12% = +1.3% EV. Over 100 similar trades you would expect an average gain of 1.3% per trade.
✓ Positive = edge in your favour over many trades. ⚠ Negative = the math loses money on average — even if individual wins feel good. Note: EV is only as reliable as the probability estimate.
Suggested Allocation
half-Kelly: (Probability ÷ |Max Downside| − (1 − Probability) ÷ Target Gain) ÷ 2, capped at 10%
The Kelly criterion computes the portfolio fraction that maximises long-run growth for a bet with these odds; half-Kelly is the standard hedge against the probability estimate being too optimistic. Uses the calibrated probability where available. 0% means the math sees no edge worth sizing; — means it couldn't be computed.
e.g. 57% probability, +10% target, −13% downside → Kelly ≈ 8.5% of portfolio, half-Kelly ≈ 4.2%.
Generic sizing guidance only — not personal investment advice. Position sizing must reflect your own portfolio, risk tolerance and circumstances.
Every stock Scout considered — and why each was discarded

Scout scanned 100 FTSE 100 constituents and short-listed 24 candidates for deep analysis by the five analysts. Here is what happened to each one.

  • SHEL.LShell plcAlready held
    Scout: Up 9.3% over 30 days and 4.8% over 5 days, only ~2.8% off its 52-week high — strong sustained momentum tracking the recent oil price rally.
    Already held: a position opened 2026-09-14 runs until 2026-10-14 (16 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • BP.LBP p.l.c.Dropped at merge
    Scout: 30-day return of 9.5% with continuing 5-day strength (+3.8%), benefiting from the same energy tailwind as Shell.
    Only Blaze selected this stock — single-agent picks require Very High confidence and at least 70% probability (this pick: Medium, 56% raw (48% after calibration)).
  • SMIN.LSmiths Group plcPicked #2
    Scout: 8.2% 30-day and 9.2% 5-day gains, just 1.5% from its 52-week high — accelerating momentum near highs.
    Made the final report at #2.
  • DPLM.LDiploma PLCPicked #3
    Scout: Steady compounder up 5.2% over 30 days and 3.0% over 5 days, within 2% of its 52-week high.
    Made the final report at #3.
  • STAN.LStandard Chartered PLCAlready held
    Scout: Up 6.7% over 30 days and trading just 1.5% below its 52-week high, showing durable re-rating versus peer banks.
    Already held: a position opened 2026-09-18 runs until 2026-10-18 (20 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • BGEO.LLion Finance Group PLCLiquidity filter
    Scout: 8.8% 30-day gain and 3.8% 5-day gain, close to its 52-week high — strong emerging-market bank momentum.
    20-day average daily volume 77,797 shares is below the 500,000-share liquidity floor.
  • SMT.LScottish Mortgage OrdAlready held
    Scout: Up 4.9% over 30 days, only 1.8% from its 52-week high, indicating a sustained recovery in growth-trust sentiment.
    Already held: a position opened 2026-09-21 runs until 2026-10-21 (23 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • VOD.LVodafone Group Public Limited CompanyAlready held
    Scout: 5.9% 30-day gain and just 5.0% off its 52-week high despite a modest 5-day pullback — underlying trend still positive.
    Already held: a position opened 2026-09-09 runs until 2026-10-09 (11 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • PSN.LPersimmon PlcDropped at merge
    Scout: Sharp 12.4% 30-day and 13.3% 5-day gains signal a strong housebuilder rebound, likely on rate-cut expectations.
    Average probability across the 4 agent(s) that selected it was 58% raw (48% after calibration), below the 52% minimum to qualify.
  • BTRW.LBarratt Redrow plcPicked #1
    Scout: 8.5% 30-day and 13.6% 5-day gains mirror the housebuilder rally seen in Persimmon.
    Made the final report at #1.
  • KGF.LKingfisher plcAlready held
    Scout: 5.4% 30-day and 8.8% 5-day gains — home-improvement retail benefiting from the same housing-sector optimism.
    Already held: a position opened 2026-09-23 runs until 2026-10-23 (25 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • TSCO.LTesco PLCAlready held
    Scout: Steady 4.7% 30-day gain, only 6.2% off its 52-week high, reflecting resilient grocery demand.
    Already held: a position opened 2026-09-14 runs until 2026-10-14 (16 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • AZN.LAstraZeneca PLCAlready held
    Scout: 4.4% 30-day gain shows recovering momentum after a deeper prior drawdown.
    Already held: a position opened 2026-09-15 runs until 2026-10-15 (17 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • PCT.LPolar Capital Technology OrdAlready held
    Scout: 4.3% 30-day gain with a positive 5-day move, tracking a rebound in tech-sector sentiment.
    Already held: a position opened 2026-09-21 runs until 2026-10-21 (23 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • LGEN.LLegal & General Group PlcAlready held
    Scout: Modest but positive 1.2% 30-day gain, only 6.7% off its 52-week high — steady insurer holding up better than peers.
    Already held: a position opened 2026-09-21 runs until 2026-10-21 (23 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • NWG.LNatWest Group plcAlready held
    Scout: 1.6% 30-day gain and just 5.0% from its 52-week high despite a recent 5-day dip — underlying bank strength intact.
    Already held: a position opened 2026-09-17 runs until 2026-10-17 (19 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • DCC.LDCC plcAlready held
    Scout: Small positive 30-day return and only 4.5% from its 52-week high — stable energy-distribution momentum.
    Already held: a position opened 2026-09-25 runs until 2026-10-25 (27 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • ITRK.LIntertek Group plcAlready held
    Scout: Essentially at its 52-week high (−0.1%), with flat-to-positive 30-day and 5-day returns — quiet but persistent strength.
    Already held: a position opened 2026-09-08 runs until 2026-10-08 (10 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • IMI.LIMI plcAlready held
    Scout: Trading just 1.4% off its 52-week high with modest positive 30-day and 5-day returns — steady industrial momentum.
    Already held: a position opened 2026-08-31 runs until 2026-09-30 (2 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • BEZ.LBeazley plcAlready held
    Scout: Within 0.2% of its 52-week high with a small positive 30-day return — resilient specialty insurer.
    Already held: a position opened 2026-09-02 runs until 2026-10-02 (4 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • SDR.LSchroders plcAlready held
    Scout: Only 1.8% off its 52-week high with a positive 30-day return, bucking the wider asset-manager weakness.
    Already held: a position opened 2026-09-09 runs until 2026-10-09 (11 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • IHG.LInterContinental Hotels Group PLCLiquidity filter
    Scout: 5-day return of 3.0% signals a fresh bounce, with the stock only 9.8% off its 52-week high.
    20-day average daily volume 293,673 shares is below the 500,000-share liquidity floor.
  • HSX.LHiscox LtdAlready held
    Scout: Positive 2.1% 30-day return and just 3.6% from its 52-week high, showing relative insurer strength.
    Already held: a position opened 2026-09-01 runs until 2026-10-01 (3 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.
  • IAG.LInternational Consolidated Airlines Group S.A.Already held
    Scout: Small positive 5-day (+1.2%) move with the stock recovering ground, aided by resilient travel demand.
    Already held: a position opened 2026-09-22 runs until 2026-10-22 (24 day(s) left). Re-picking it would concentrate the portfolio, not diversify it.

Agent Analytical Approaches

Aria

Momentum & Technical

Checks 30-day price history for every candidate before making any selection. Requires positive period return and positive 5-day momentum as mandatory gates. Ranks picks by proximity to their period high and adjusts probability based on volume. Uses news only to rule out major negative catalysts.

Blaze

Fundamental & Earnings

Searches earnings results, analyst upgrades, and revenue trends before checking price data. Requires at least one verifiable fundamental catalyst — earnings beat, analyst upgrade, or revenue acceleration — within the last 60 days. Favours stocks with strong fundamentals trading below recent highs (value entry).

Nova

News Catalyst & Event-Driven

Hunts for specific events within the last 21 days: FDA approvals, major contract wins, product launches, earnings surprises, or significant analyst upgrades. Requires a positive price reaction confirming the market is recognising the catalyst. Also scans for upcoming events that could drive further gains.

Orion

Macro & Sector Rotation

Maps the macroeconomic regime (risk-on / risk-off / neutral) and identifies sectors benefiting from current conditions before looking at individual stocks. Only selects stocks from macro-aligned sectors with confirmed sector tailwinds. Adjusts for interest rate sensitivity and geopolitical factors.

Sage

Devil's Advocate

Reviews the final picks after all four analysts agree and actively challenges each bull thesis. Searches for bearish technical signals, negative news, insider selling, and elevated short interest that the agreeing agents may have underweighted. Produces evidence-based counter-arguments shown on each pick card.

Quinn

Quantitative / Statistical

Pure-quant analyst: no news, no narrative, only numbers. Identifies mean-reversion setups from RSI extremes (<35) with momentum turning, Bollinger squeeze breakouts, and multi-factor quant scores using RSI, MACD, OBV, short interest, and insider transaction data. Counterbalances narrative bias across the four analyst agents.