Terabyte-scale market memory for trading agents. Propose a trade and see how similar trades performed throughout history under comparable market conditions.
Start with a trading idea. Replay it. Compare it. Add context. Then let the model decide whether it still deserves attention.
The point is knowing which ideas survived contact with the data.
01
Ask the clean question first: what would this strategy have done on this ticker over this window?
Signals
1 Active
Research Engine
Inputs are scored, compared, and resolved into evidence.
Result Evidence
Return
Drawdown
Trades
The output stays tied to replayable results: return, risk, trade count, and ledger evidence.
02
Then measure whether two or more strategies agreeing actually improves the outcome.
Signals
2 Active
Research Engine
Inputs are scored, compared, and resolved into evidence.
Result Evidence
Lift
Win Rate
Trades
The output stays tied to replayable results: return, risk, trade count, and ledger evidence.
03
The same signal means different things in different markets. The lab measures the conditions around it.
Signals
2 Active
Context Layer
Inputs are scored, compared, and resolved into evidence.
Result Evidence
Matches
Hit Rate
Edge
The output stays tied to replayable results: return, risk, trade count, and ledger evidence.
04
Finally, the model scores what survived and separates trades worth attention from trades to block or shadow-track.
Signals
2 Active
Model Router
Inputs are scored, compared, and resolved into evidence.
Result Evidence
Approved
Blocked
Shadow
The output stays tied to replayable results: return, risk, trade count, and ledger evidence.
At this decision threshold, 14 trades would have executed and marked +0.40% after 15 minutes. Slide the threshold to compare all protectiveness levels.
Decision Threshold
Higher settings require stronger evidence, so fewer signals pass.
46%
SPY
+0.27%
VTI
+0.31%
Executed Trades
14
Wins v. Losses
8/4
15m Mark
+0.40%
Model Result
+0.40%
Executed Trades
14
8/4 Wins v. Losses
The model beat SPY on 5 of 5 verified sessions.
Model (P48)
+0.60%
SPY
−0.45%
Model (P28)
+2.16%
SPY
−0.06%
Model (P41)
+0.72%
SPY
−0.13%
Model (P50)
+1.17%
SPY
−0.05%
Model (P39)
+0.38%
SPY
+0.16%
Research Center
The paid research center turns the archive into searchable replays, model comparisons, and proof ledgers. Subscribers see the exact strategy names, scores, decision paths, and trade-by-trade results.
Market Archive
Terabyte-Class
Cleaned, stored, indexed, and turned into repeatable simulations.
Minute Data
Since 2010
Verified one-minute stock bars powering repeatable simulations.
Strategy Types
Rules to Models
Single signals, blended signals, context filters, and adaptive models compared side by side.
Trade Ledgers
Line by Line
Every headline result can be checked against the modeled buys and sells behind it.
Last updated at —.
The Paper-Account Leaderboard
See the strongest results so far today. Strategy names are hidden here.
Each account starts with $100k, monitors 200 symbols, and deploys cash only when a buy signal triggers. Sale proceeds return to the shared account pool; symbols without signals remain cash. These are research simulations, not live broker returns or investment advice.
Create a free account to see them all.Give your AI historical context before it makes a decision. Make smarter trades with strategy replays; member-only model performance and paper-trading results remain in the Research Center.