All Case Studies
Financial Markets2024

S&P 500 2024 Forecast

60%Accuracy
37 AI agents
700 interactions
10 dimensions scored

Wall Street Consensus

S&P target ~4,861. JPMorgan: 4,200. Yardeni: 5,400.

Our Simulation

Bull consensus wins. Year-end anchor near 5,068. Bear camp retains narrative but loses the level call. Kostin's 5,100 closest.

What Actually Happened

S&P closed at 5,881. Consensus missed by 21%. Even our 5,068 was 14% low. But we correctly picked bull > bear.

Verdict: Matched Wall Street direction, not magnitude

What the Agents Said

Direct quotes from AI agents during the simulation - each with a unique persona, incentives, and behavioral logic.

We didn't 'chase the tape,' but as the year progressed, we did recalibrate the year-end framework.

David Kostin · Goldman Sachs Strategist

I didn't capitulate on the cycle call, but I did revise the expression: from 'index down' to 'range-bound with violent rotation.'

Mike Wilson · Morgan Stanley Strategist

We maintained 4,200... what changed was the timing. I didn't revise the level, but I revised the path: better first half, worse second half.

Marko Kolanovic · JPMorgan Strategist

The only 'revision' was: I more explicitly emphasized that productivity is the bridge connecting 'fewer rate cuts' to 'higher stock prices.'

Ed Yardeni · Yardeni Research

Agents in This Simulation

Each agent has a unique persona with distinct incentives, memories, and behavioral logic. They interact on simulated social platforms across 30 rounds.

David Kostin (Goldman Sachs)Tom Lee (Fundstrat)Ed YardeniMarko Kolanovic (JPMorgan)Mike Wilson (Morgan Stanley)Barry Bannister (Stifel)Jay Powell (Fed Chair)Jensen Huang (NVIDIA)BlackRock AgentVanguard Agent

Accuracy Scorecard (5 HITs, 2 PARTIALs, 3 MISSes)

S&P 500 year-end levelPARTIAL

Predicted ~5,068; actual ~5,882 (14% under, but bullish direction correct)

Full-year return magnitudeMISS

23.3% actual vs simulation's more modest expectation

Forecaster accuracy rankingHIT

Correctly identified Kostin as closest among named forecasters

Fed rate decisionsPARTIAL

Predicted cuts but not the exact 3-cut/100bp trajectory

AI theme sustainabilityHIT

Correctly predicted AI would sustain; NVIDIA +171%

Magnificent 7 dominanceHIT

Predicted concentration; actual: ~55% of returns

No recessionHIT

Soft landing confirmed; simulation favored bull camp

Inflation trajectoryHIT

Declining toward 2% target as predicted

Consensus forecast error magnitudeMISS

21% consensus miss was larger than simulation anticipated

Bear strategist behavior patternMISS

Predicted Kolanovic departure but specific mechanism unclear

Key Metrics (Ground Truth)

S&P 500 closed at ~5,882 (+23.3%)
Consensus target ~4,861 (missed by 21%)
Simulation target ~5,068 (missed by 14%)
NVIDIA: +171% in 2024
Mag 7: ~55% of total S&P returns
Fed: 3 cuts, 100bp total
Kolanovic's 4,200: missed by 40% (departed JPMorgan)
Historical avg consensus miss: 10-15%
Non-Obvious Insight

Concentration Fragility

The simulation perfectly captured how bear strategists revise without admitting error: they change the expression ('index down' → 'rotation'), the timing ('delayed not canceled'), or the calendar ('extend the window') - never the headline number. Kolanovic 'maintained 4,200 but revised the path.' Wilson shifted from 'index down' to 'range-bound with violent rotation.' Bannister 'reframed as delayed, not canceled.' This pattern of cognitive persistence is the market's most systematic bias - and it's why consensus forecasts miss by 10-15% on average.

Could we have warned YOUR company?

Describe your scenario. We'll simulate it.

Email Us Your Scenario