Replay benchmarkPlaceholder data, real results coming
KalshiBench Replay v1
Recorded NBA and college-football games streamed tick by tick (60s) inside a sandbox. Each game is one task with a fresh $100 bankroll; score is P&L after settlement, fees included.
Games
8
Leagues
NBA, NCAAF
Bankroll / game
$100
Updated
Nov 14, 2026
+$21.40
+$9.80
+$6.20
+$2.90
$0.00
−$4.70
−$9.30
−$14.10
Jev 1.14
Frontier LLM A · agent
GPT-6 Luna Decisions
Microsoft Decision-1
Hold cash
Frontier LLM B · forecast
Random trader
Liquid D1
Results table
| # | Model | Track | Total P&L | Mean / game | ± SD | Win rate | Trades | Fees | Brier |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Jev 1.14 | SYS-1System-1 · even split | +$21.40 | +$2.68 | |||||
| 2 | Frontier LLM A · agent | LLMLLM · agent with tools | +$9.80 | +$1.23 | |||||
| 3 | GPT-6 Luna Decisions | SYS-1System-1 · even split | +$6.20 | +$0.78 | |||||
| 4 | Microsoft Decision-1 | SYS-1System-1 · self-sized | +$2.90 | +$0.36 | |||||
| 5 | Hold cash | BASEBaseline · hold cash | $0.00 | $0.00 | |||||
| 6 | Frontier LLM B · forecast | LLMLLM · forecast only | −$4.70 | −$0.59 | |||||
| 7 | Random trader | BASEBaseline · random | −$9.30 | −$1.16 | |||||
| 8 | Liquid D1 | SYS-1System-1 · self-sized | −$14.10 | −$1.76 |
P&L per game
green = profit, red = loss| Game | Jev 1.14 | Frontier LLM A · agent | GPT-6 Luna Decisions | Microsoft Decision-1 | Hold cash | Frontier LLM B · forecast | Random trader | Liquid D1 |
|---|---|---|---|---|---|---|---|---|
| DEN @ NOP nba | +$4.10 | −$2.20 | +$1.40 | +$6.30 | $0.00 | −$1.20 | −$0.90 | −$5.50 |
| DAL @ CLE nba | +$2.00 | +$5.10 | −$0.80 | −$3.90 | $0.00 | +$0.60 | −$1.40 | +$2.20 |
| DET @ MIA nba | +$7.90 | +$3.30 | +$4.40 | +$1.10 | $0.00 | −$3.00 | −$2.00 | −$6.80 |
| BKN @ MEM nba | −$1.60 | −$6.00 | +$2.20 | −$2.50 | $0.00 | +$2.40 | −$0.60 | +$1.00 |
| CHI @ MIN nba | +$3.30 | +$8.80 | −$1.90 | +$4.00 | $0.00 | −$0.90 | −$1.70 | −$3.10 |
| ORST @ BSU ncaaf | +$5.20 | +$1.90 | +$3.00 | +$7.70 | $0.00 | +$1.10 | −$0.80 | +$4.40 |
| UNLV @ UNM ncaaf | −$3.40 | +$2.40 | −$4.20 | −$9.30 | $0.00 | −$2.60 | −$1.10 | −$2.90 |
| SJSU @ AFA ncaaf | +$3.90 | −$3.50 | +$2.10 | −$0.50 | $0.00 | −$1.10 | −$0.80 | −$3.40 |
Source: web/content/results.json · Kalshi order books + ESPN, recorded 2026-11-14