Nat & Bestie’s Trading Experiments

Focus dashboard: only the experiments that still help us decide what the future bot should actually trade.

Loading…

Paper Trading + Forward Research

CURRENT GOAL

Can the frozen RN1 signal actually make money in live paper trading?

Behavioral prediction passed prospectively. The final test now paper-trades brand-new Aggressive Opposite signals using the frozen 3-minute model — no real money.

Step 2 of 6

LIVE PAPER TRADING · PRIMARY SCORECARD

Loading paper-trading results…

Frozen 3-minute model · brand-new signals only · no real money

Checking

Settled Trades

—

Wins / Losses

—

Win Rate

—

$10 Net P&L

—

ROI

—

Max Drawdown

—

$5 Paper Trades Conservative

—

ROI

—

Ending $500

—

Win Rate

—

Drawdown

—

$10 Paper Trades Primary

—

ROI

—

Ending $500

—

Win Rate

—

Drawdown

—

$25 Paper Trades Aggressive

—

ROI

—

Ending $500

—

Win Rate

—

Drawdown

—

CURRENT PAPER-TRADING GOAL

Collect the locked paper-trading cohort

0 / 30

0% complete 30 trades remaining

Nothing to analyze yet — let the paper test run.

Come back when the locked paper-trading cohort finishes or the dashboard flags a problem.

Simple rule: only do something when the top card says “COME BACK TO CHATGPT NOW.”

Otherwise, let the live paper test and Pre-Entry Scout collect automatically.

✓ Data collection
COMPLETED

✓ 3m horizon
COMPLETED

✓ Holdout
COMPLETED

✓ Frozen model
COMPLETED

✓ Prospective test
COMPLETED

Paper trading
ACTIVE

LIVE PAPER TRADING · FINAL VALIDATION

Checking Step 6 status…

No real money. Frozen 3-minute model. Brand-new signals only.

Stake	Trades	Net P&L	ROI	Ending $500 Bankroll
$5	—	—	—	—
$10 primary	—	—	—	—
$25	—	—	—	—

OVERALL RN1 STRATEGY PROGRESS

1 of 6 milestones complete · Step 2 is active

1 / 6

Completed: collect post-switch data Finish line: validated prospective shadow model

✓ Data collected ✓ Horizon selected ✓ Holdout passed ✓ Model frozen ✓ Prospective passed NOW · Shadow money

LIVE SYSTEM HEALTH

RN1 TELEMETRY

Checking…

Latest market / RN1 evidence collector.

PAPER TRADING

Checking…

Frozen 3-minute model + live shadow-money test.

PRE-ENTRY SCOUT

Checking…

Watches markets before RN1 enters.

OVERALL SYSTEM

Checking…

Summary of active collectors and current experiments.

RN1 Pre-Entry Independence Lab

Can we predict where, when, which side and how RN1 enters BEFORE RN1 trades?

Collecting

Mission: stop piggy-backing on RN1.

This research watches markets RN1 has not touched yet. RN1's first BUY is used only later as the answer key — never as an input to the prediction.

WHERE Category, liquidity, volume, market age, price zone and time-to-resolution.

WHEN Test whether entry becomes predictable 5 / 10 / 15 / 30 minutes early.

WHICH SIDE Compare the two order books, price movement, depth and imbalance before first entry.

HOW First-entry price, cash size, shares and the market conditions surrounding the entry.

Untouched Snapshots

—

Markets Observed

—

Future Entries Labeled

—

Collector

—

PRE-ENTRY MODEL READINESS

Each horizon needs both RN1-entry examples and untouched-market controls. The system will not design a model until those minimums are met.

5 minutes early

Loading…

10 minutes early

Loading…

15 minutes early

Loading…

30 minutes early

Loading…

Starting clean pre-entry collection.

The first milestone is enough untouched-market history to compare RN1 entries against markets he ignores.

Live system details

Technical diagnostics only — no action unless the dashboard tells you something is wrong

HIDDEN

These sections support the live system but do not need routine monitoring.

Decision Research · Model Diagnostics

Supporting research only · expand when you want the underlying model detail

Analyzing

CURRENT ANALYSIS

How Soon After the First Opposite BUY Can We Tell What RN1 Will Do?

This is the only research question that matters right now. The engine compares post-switch behavior at 1, 3, 5, and 10 minutes.

Post-Switch Analysis

Conditions Loaded

—

Completed

—

Switched

—

Switch Rate

—

Protect / Rebalance

—

Aggressive Opposite

—

Escalated within 10m

—

Escalation Rate

—

Loading Decision Engine…

Waiting for the consolidated lifecycle analysis.

RN1 Strategy Telemetry · Flight Recorder

Live evidence collection · market state, RN1 inventory, payoff context, and Polygon enrichment

Collecting

Parallel Evidence Collection

Are We Collecting Enough to Test RN1’s Triggers?

This does not change Strategy Model V1. It records the market and position state around RN1’s actions so we can later compare what consistently happens before first opposite-side buys, rebalances, aggressive switches, and payoff crossings.

Flight Recorder Active

Telemetry Events

—

Snapshots

—

Conditions Tracked

—

On-Chain Enriched

—

Latest Positions Fetched

—

API Errors

—

TRADE Events

—

REDEEM Events

—

SELL Events

—

SPLIT / MERGE

—

Collection Health

Loading telemetry…

Checking the latest telemetry collector run.

When Do We Analyze?

Building the sample

First trigger review target: approximately 30 clean RN1 mode-transition sequences with multiple snapshots before the transition.

Trigger-analysis rule

Snapshot count alone is not proof. We want repeated before/after sequences around RN1 transitions. Around 20 transitions is still early; ~30 is enough for the first consolidated trigger review; 50+ is a stronger sample; 100+ is substantially better. The current telemetry status endpoint does not yet directly count clean transitions, so this card reports collection maturity without pretending ordinary snapshots are transition labels.

Recent RN1 Telemetry Snapshots

Market	Latest Action	A Shares	B Shares	Weaker Payoff	Need A to $0	Need B to $0	Since Last Trade
Waiting for telemetry snapshots…

RN1 Strategy Model V1 · Forward Validation

True prospective test · predict new RN1 behavior before the lifecycle resolves

Collecting

Primary Strategy Reconstruction Test

Can We Predict What RN1 Does Next?

Historical research has been consolidated into Strategy Model V1. This validator only accepts fresh RN1 conditions and scores whether the frozen model correctly predicts Directional, Protect / Rebalance, or Aggressive Opposite behavior.

True Forward Test · No Trades

Forward Validation Progress

0 / 50 resolved 50 remaining

Predictions

—

Resolved

—

Correct

—

Incorrect

—

Accuracy

—

Status

Collecting

Directional Accuracy

—

Protect / Rebalance Accuracy

—

Aggressive Opposite Accuracy

—

Prospective Integrity

Frozen boundary loading…

Predictions must be made within 15 minutes of RN1’s first BUY and before any REDEEM exists.

Next Decision

Collecting new RN1 activity

We review the model only after 50 true prospective conditions resolve.

Recent Forward Predictions

Market	Initial	Price	Predicted	Actual	Result
Waiting for forward predictions…

Frozen Strategy Model V1

Historical reconstruction: 250 completed conditions · Directional 20.8% · Protect/Rebalance 58.4% · Aggressive Opposite 20.8%. This forward test is the gate between reverse-engineering and paper implementation.

Strategy Model V2 · Readiness Monitor

Pooled forward-chaining · event-balanced transition prediction · untouched holdout protected

Checking

Current Research Gate

When Do I Need to Come Back?

This card checks the V2 design endpoint on a slower schedule so we can see when the dataset is finally evaluable without hammering Neon every minute. The final 30% condition-level holdout stays untouched.

Waiting for Clean Data

V2 Status

Checking

Completed Conditions

—

Incomplete Conditions

—

Aggressive Completed

—

Untouched Holdout

—

Best Horizon

—

What Is Blocking V2?

Reading dataset diagnostics…

The monitor will show the smallest remaining data shortage or, once evaluable, the untouched-holdout result.

Come Back When

Wait for the monitor.

You do not need to manually inspect the JSON while this remains in a waiting state.

V2 Research Readiness

Checking readiness… —

Stage 1 · Switch Detection

—

Waiting for diagnostics.

Stage 2 · Protect vs Aggressive

—

Waiting for diagnostics.

Final Holdout

Protected

Never used to choose thresholds.

Dashboard rule

Yellow means keep collecting and do nothing. Green means V2 is finally ready for our review. Red means the dataset was sufficient but the untouched holdout says the model itself is weak. This dashboard checks V2 roughly every 15 minutes, not every 60 seconds, to reduce unnecessary Neon work.

Completed / retired research — no action needed NO ACTION NEEDED

These experiments are completed or retired. They are preserved only for reference and do not require monitoring.

COMPLETED / RETIRED RESEARCH — NO ACTION NEEDED

Older / Secondary Research — Reference Only

Candidate B · Execution Confirmation Complete

Prospective minute-30 executable-price test · simulated $5 fills only

Confirmed

Active Execution Research

Can We Actually Trade Candidate B?

Candidate B found a promising RN1 state. This separate prospective test asks whether enough price edge remains when that state becomes knowable at minute 30. The system records the live Polymarket order book and simulates a $5 buy. No order is submitted.

Simulation Only · No Trades

Execution Confirmation Goal

0 / 50 settled simulated executions 50 remaining

Current hypothesis being watched: 65–70¢ execution VWAP

Do not change Candidate B before 50 settled execution simulations. This is a confirmation sample, not a new trading rule.

Quotes Captured

Complete $5 Fills

Settled

Record

0-0

Win Rate

—

Simulated Return

—

Average Best Ask

—

Average $5 VWAP

—

Average Capture Delay

—

Fill Rate

—

Simulated P&L

—

Simulated Cash Out

—

Incomplete / No Depth

Quote Errors

Waiting for prospective execution quotes…

This test started with a fresh prospective boundary. Candidate B's older signals are intentionally excluded.

Recent Prospective Execution Quotes

Market	Outcome	RN1 Entry	Best Ask	$5 VWAP	Delay	Settlement	Sim P&L
Waiting for the first post-boundary Candidate B execution quote.

What We Are Testing

Does Candidate B still have edge at the price available to us?

Signal accuracy alone is not enough. This test measures the actual ask, order-book depth, simulated $5 VWAP, and eventual hypothetical P&L from the moment Candidate B becomes actionable.

Next Decision

Wait for 50 settled executions

Come back when the execution progress bar reaches 50/50. Until then, keep Candidate B frozen and watch the 65–70¢ VWAP hypothesis without changing the rule.

Candidate B · 30-Minute One-Sided Forward Validation

Active RN1 state research · clean at 3 minutes + one-sided through 30 minutes

Active Forward Research

RN1 Candidate B

Research only — no paper trades and no real-money trades. Candidate B tests whether an RN1 entry that is clean at minute 3 and remains economically one-sided through minute 30 is a stronger directional state.

Research Only · No Trades

Qualified Signal Progress

Loading…—

Qualified

Target

Pending 3M

Pending 30M

Settled

Record

0-0

Win Rate

—

Observed Return

—

Total Observations

Rejected

Frozen Candidate B Rule

RN1 First BUY

60–79¢

Minute 3

Exactly 1 BUY

Opposite by 3M

Same-Side ADDs after 3M

Allowed

Opposite through 30M

Trading

None · Observer Only

Rejected · Price Range

Rejected · 3M ADD

Rejected · Opposite BUY

Forward Collector

Waiting for Candidate B status…

Candidate B is research only and does not place trades.

Recent Qualified Candidate B Signals

Market	Outcome	Initial Entry	BUYs at 30M	Cash at 30M	Later Opposite	Result	P&L
Waiting for Candidate B qualifiers.

What Is Happening

Candidate B is watching, not trading.

The automated cycle collects timestamped RN1 activity and evaluates the frozen 3-minute and 30-minute state rules. No Candidate B position is opened on Polymarket and no Candidate B paper position is created.

Next Decision

Loading…

Waiting for forward qualification progress.

Candidate A · Completed Discovery / Review Sample

Preserved forward sample that led to Candidate B · no longer actively collecting

Completed Forward Discovery

RN1 Candidate A

Candidate A is preserved as the completed forward-discovery sample that revealed the importance of continued one-sided RN1 behavior. It is not currently collecting new signals and it never placed trades.

Preserved Sample

Qualified Signal Progress

Loading… —

Qualified

Target

Settled

Record

0-0

Win Rate

—

Observed Return

—

Total Observations

Rejected

Later Switched

Later Switch Rate

Frozen Candidate A Rule

RN1 First BUY

60–79¢

Checkpoint

3 Minutes

Original BUY Count

Exactly 1

Opposite BUYs

Early ADD

Reject

Early Side Switch

Reject

Rejected · Price Range

Rejected · Early ADD

Rejected · Early Switch

Forward Collector

Waiting for Candidate A status…

Candidate A is collecting untouched future signals.

Recent Qualified Signals

Market	Outcome	Entry	Checkpoint Cash	Later Switch	Settlement	Result	P&L
Waiting for Candidate A signals.

What Happens Next

What Matters Now

Do not change Candidate A.

Candidate A is now a frozen forward experiment. New results must accumulate without changing the 60–79¢ range, the 3-minute checkpoint, or the single-BUY requirement.

Next Decision

Loading…

Waiting for forward qualification progress.

V2 · Strict 60–79¢ Forward Test

Clean price-filter benchmark

Running

Experiment V2

Filtered 60–79¢ Forward Test

Simple, strict, and currently our cleanest profitability benchmark.

Running

Equity

$500

P&L

Return

Open Exposure

Open

Closed

Wins

Losses

Win Rate

Status

Benchmark

Frozen Rules

RN1 Entry

60–79¢

US Entry

60–79¢

Max Premium

+3¢

Per Trade

Max Event

$15

Max Total

$100

V6 · 60–79¢ + Averaging-Up Confirmation

Clean $5 challenger to V2

Collecting Data

Experiment V6

Price Edge + RN1 Conviction

Same $5 sizing and 60–79¢ execution discipline as V2, but entry requires RN1 averaging-up confirmation.

Active Challenger

Equity

$500

P&L

Return

Open Exposure

Open

Closed

Wins

Losses

Win Rate

Stage

Collecting

Frozen Rules

RN1 Price

60–79¢

US Entry

60–79¢

Max Premium

+3¢

Per Trade

Averaging Up

Required

Exit

Resolution

Open V6 Trades

Market	Outcome	RN1 Confirm	Our Entry	Amount
Waiting for the first V6 signal.

Recent Closed V6 Trades

Market	Outcome	Entry	Result	P&L
No closed V6 trades yet.

V5 · Completed Behavior Baseline

Broad RN1 mirroring completed; legacy positions winding down

Completed Baseline

V5 · Broad RN1 Behavior Mirror

Broad RN1 mirroring did not produce positive account growth. V5 is preserved as a comparison baseline while legacy positions wind down.

Winding Down

Equity

$500

P&L

Return

Open Exposure

Open

Closed

Wins

Losses

Win Rate

Mirrored ADDs

Recent Behavior Mirror Actions

Market	RN1 Action	RN1 Price	V5 Action	V5 Price	Dollars	Status
Waiting for V5 actions.

V5.1 · Completed Discovery Research

53+ trade discovery experiment that identified the V6 hypothesis

Discovery Research

RN1 Confirmed Conviction Mirror

Selective strategy: ADD confirmation at 40¢+, separate 60–79¢ cohort, averaging-up tag, maximum three mirrored adds.

Supporting Research

Equity

$500

P&L

Return

Open Exposure

Open

Closed

Wins

Losses

Win Rate

Qualified

Candidate Quality

Watching

Core 60–79¢

Averaging Up

Core + Avg Up

Frozen V5.1 Rules

Minimum Candidate

40¢

Core Zone

60–79¢

Requires ADD

Yes

Max Adds

Initial Entry

Max Event

$20

Performance by Qualification Tag

Tag	Total	Open	Closed	W / L	Win Rate	P&L
Waiting for V5.1 cohorts.

Recent V5.1 Actions

Market	Tag	RN1 Action	RN1 Price	V5.1 Action	V5.1 Price	Dollars	Status
Waiting for V5.1 actions.

Research & validation details Collapsed

WHAT STEP 2 STILL NEEDS

These are the exact gates before we can move on.

✓

Enough lifecycle data

Loading current sample…

Done

✓

Enough opposite-side switches

Loading current switch sample…

Done

Score 1m / 3m / 5m / 10m escalation horizons

This is the missing evaluator that tells us the earliest reliable point after the first opposite BUY.

Active

Choose earliest horizon that passes

We need a real out-of-sample score, not simply more conditions.

Pending

1 minute

Loading…

Running forward validation

3 minutes

Loading…

Running forward validation

5 minutes

Loading…

Running forward validation

10 minutes

Loading…

Running forward validation

STEP 4 / STEP 5

Checking frozen model status…

The winning horizon must be frozen before brand-new prospective cases count.

PROGRESS

Collect enough post-switch data

Decision Engine has a large completed lifecycle sample and clean transition history.

Done

Find the earliest reliable escalation horizon

Compare 1m, 3m, 5m, and 10m after RN1’s first opposite BUY.

Now

Pass untouched holdout

The chosen escalation rule must work on unseen conditions.

Freeze the escalation model

No changing thresholds during the true prospective test.

Later

Run prospective validation

Test brand-new RN1 transitions only.

Later

Turn on shadow-money tracking

Show what $5 / $10 / $25 would have done only after the move logic is validated.

Later

WHAT DO I DO?

Data is collecting. The horizon scorer is the remaining blocker.

Watch the live movement cards above. If counts rise, collection is working. Step 2 finishes only after the 1m / 3m / 5m / 10m horizons receive real holdout scores.

Come back when

One of the 1m / 3m / 5m / 10m horizons has a real holdout score and passes the review gate.

Still pending: 4 major milestones after the current step.

Decision Engine Conditions

—

Completed Lifecycles

—

RN1 Switch Rate

—

Post-Switch Escalation

—