SETHPMLABS
AUTONOMOUS TRADING STACK · COLD START
INITIALIZING · 0% RTT — ms
Independent Quant Systems Lab

Find the direction. Prove the trade. Protect the downside.

SETH PM LABS is a small autonomous trading operation built around BTC 5-minute prediction markets. BOT9 decides when a setup deserves capital. BOT3 starts watching only after a proven position exists — and buys the opposite side only when reversal evidence is strong enough and the hedge improves the economics.

See the two-engine stack → How it decides
BITCOIN · LIVE PRICE
$—
— 24h
Stack● LIVE
CoreBOT9 V444
V444
BOT9 Entry Engine
V1.10.3
BOT3 Smart Hedge
5s
Canonical Scan Grid
≤ .67
Entry Execution Ceiling
SETH PM LABS DUAL ENGINE ONLINE BOT9 V444 · PRIMARY FLOW: BINANCE FUTURES SETTLEMENT EVIDENCE · CHAINLINK TWAP60 / TWAP30 DUAL · CHAINLINK + BINANCE · PM PATH · ATR · CONFIDENCE ENTRY EXECUTION CEILING ≤ 0.67 BOT3 HEDGE · PM REPRICING · BINANCE FUTURES · BYBIT · CL60 MINIMAX SIZING · NO OVER-HEDGE MARKET-OPEN PHT · SCHEDULE-AWARE · FINAL REVALIDATION LOG · TEST · RESCAN · EXECUTE RESEARCH THE EDGE · PROTECT THE DOWNSIDE
The Approach

Not one bot. A decision engine with a risk engine behind it.

The current SETH PM LABS stack separates two jobs that should never be confused. BOT9 earns the right to enter. BOT3 earns the right to protect an already-open position. Neither gets to improvise.

01DIRECTION

Executed flow chooses the side

Binance Futures aggressive buying versus selling is the primary direction authority. The system does not ask nine indicators to vote on a side anymore; it starts from actual executed flow.

02PROOF

The setup must survive settlement-aware evidence

Chainlink TWAP60/TWAP30, DUAL price-source state, Polymarket path, ATR, confidence, token-price corridor and fresh execution checks can support, delay or veto the candidate. A hold means rescan — not force the trade.

03PROTECT

The hedge starts after the position is real

BOT3 watches the exact same market only after BOT9 truly fired or filled. It buys the opposite outcome only when persistent reversal evidence appears and fee/depth-aware math says the hedge improves the worst case.

0
Independent Engines
5s
BOT9 Scan Grid
1 market
Exact-Market Hedge
0
Blind Hedge Entries
The Creed · I
Find the signal.
The Creed · II
Prove the setup.
The Creed · III
Protect the downside.
Settlement Bridge · Chainlink TWAP60

Trade what settles — not just what moves.

BOT9 no longer treats a Brownian model as a standalone fair-value trigger. Direction begins with executed flow. The bridge is settlement-aware evidence: it asks whether the official Chainlink TWAP60 state — with TWAP30 as a faster lead — actually supports the side the flow wants to buy.

The bridge is a confirmation layer, not the steering wheel.

Bitcoin can jump on one venue while the settlement reference barely moves. That distinction matters in a five-minute binary market. BOT9 therefore separates direction authority from settlement evidence: Binance Futures CVD picks the candidate side; Chainlink TWAP60/TWAP30 tests whether that move is carrying into the reference that matters.

The uncertainty model still matters, but its job is different now. It gives the TWAP bridge a time-aware hurdle instead of assuming that being later in the window automatically makes a signal safer. Early evidence is discounted, the core window gets full weight, and late evidence decays into a hard authorization cap.

Then DUAL and Polymarket add another reality check. Explicit source opposition, repeated favorite churn, a collapsing token path or a bad fresh book can keep the system on HOLD even when the primary flow still points one way.

✓Primary direction: executed Binance Futures flow — one authority, not a venue vote.
✓Settlement support: Chainlink TWAP60 state plus TWAP30 lead must be healthy enough for the current time.
✓Conflict stays a hold: DUAL, PM path, ATR, confidence, TP and execution structure can still veto or rescan.
SETTLEMENT BRIDGE · BTC 5-MIN ● LIVE ALIGNED
TWAP60 · WINDOW OPEN REFERENCE NOW · t+60s SETTLEMENT
Flow dir
UP
CL60 state
SUPPORT
DUAL
AGREE
Decision
READY
PASS · primary flow and settlement evidence agree
Illustrative state. Direction comes from executed flow; TWAP60/TWAP30 and DUAL are support/opposition evidence. Model edge remains telemetry, not fire authority.
The Decision Stack

One direction authority. Eight ways to prove it deserves capital.

The old “nine indicators vote” idea is gone. The current architecture is deliberately asymmetric: one source chooses direction; everything else measures settlement support, regime, market structure, execution quality or post-entry reversal.

01
Direction Authority

Binance Futures CVD

Aggressive buys versus aggressive sells determine the candidate side. This is executed flow, not a candle-color guess and not a multi-venue popularity contest.

Primary CVD · one direction source
02
Settlement Evidence

Chainlink TWAP60 / 30

The official-style settlement path is checked against the candidate. TWAP30 provides a faster lead; time-aware reliability prevents “later automatically means safer.”

Settlement-native bridge · time-aware hurdle
03
Cross-Source State

DUAL

Chainlink and Binance price-source behavior can support, stay neutral or materially oppose the candidate. It confirms context; it does not choose direction.

Support / neutral / explicit opposition
04
Prediction-Market Memory

PM Path & Churn

Persistent favorite flips, giveback and token-path deterioration are remembered across the window. Repeated churn raises the proof required before BOT9 can fire.

Favorite hysteresis · path persistence
05
Volatility Regime

ATR

Normalized BTC ATR and actual-dollar ATR decide whether the current regime is admissible. Scheduled ATR blocks can remove narrow loss pockets without changing direction logic.

ATR bps + actual USD ATR safety ranges
06
Candidate Quality

Confidence

Legacy confidence survives as a bounded quality check, not the steering wheel. It can have an elapsed-aware floor and schedule-specific maximum caps inside selected ATR bands.

Floor + scheduled cap · rescan on failure
07
Price Discipline

TP Corridor

The token must still be buyable inside the direction/day corridor and the hard execution ceiling. A good thesis at a bad price is still a bad order.

Fresh ask · TP floor/ceiling · hard cap
08
Final Market Structure

Persistence & Book

TP collapse, extreme OBI/PM-book opposition, slippage and fresh corridor checks are re-evaluated before capital moves. Execution cannot inherit stale permission.

Final revalidation · FAK/GTC · fill evidence
09
Post-Entry Risk Engine

BOT3 Smart Hedge

Not an entry vote. After a real BOT9 position exists, BOT3 watches Polymarket adverse repricing, Binance Futures, Bybit, Chainlink TWAP60 and the path itself for a genuine reversal.

Opposite-side protection · same exact market
Quantitative Core

The math now separates signal, execution and protection.

There is no single magic score. BOT9 creates a candidate, proves it through independent constraints, and prices a real order. BOT3 only enters the picture afterward, solving a different problem: how much opposite-side protection actually improves the binary worst case.

① Executed-Flow Direction · Primary Authority
dir = sign(aggressive buys − aggressive sells)
In English: start with who is actually crossing the futures book. If executed buying dominates, the candidate is UP; if executed selling dominates, it is DOWN. Other sources can veto or support, but they do not manufacture a direction.
② Settlement Support · TWAP Bridge
support = f(TWAP60 state, TWAP30 lead, σ, time)
In English: the move must make sense against the reference path that matters to settlement. The hurdle changes with time so the engine does not confuse “late” with “safe.”
③ Schedule Authority · Market-Open PHT
effective rule = baseline ⊕ time-of-day override
In English: entry second, last second, token-price corridor, size, ATR blocks and confidence caps can change by the exact five-minute market-open clock. Safety overlays remain tighten-only — a schedule cannot silently widen a hard bound.
④ Fresh Execution · Price Still Has To Be Real
askfresh ∈ [floor, min(TP max, 0.67)]
In English: authorization is not a blank cheque. Immediately before order placement, the live Polymarket book must still fit the corridor, ceiling, depth and slippage constraints. Otherwise BOT9 holds or recovers — it does not chase.
⑤ Hedge Minimax · Protect Only When It Helps
choose h ≤ original shares to maximize min(PnLoriginal wins, PnLhedge wins)
In English: BOT3 never hedges because a chart twitched. Once reversal proof is earned, it uses the fresh opposite book, fees, available depth and budget to size only the protection that improves the true binary worst case — never more shares than BOT9 actually owns.
Market Evidence · Four Roles

Every feed has a job. None gets to pretend it is another.

The current stack is role-based instead of “more venues must be better.” Binance Futures supplies direction. Chainlink supplies settlement context. Polymarket supplies the contract path and executable book. Bybit gives the hedge engine an independent reversal check.

BINANCE FUTURES
Primary executed-flow direction

Aggressive trade flow determines the BOT9 candidate and later becomes one of BOT3's independent reversal checks. It is the directional authority, not merely an order-book imbalance.

Primary · executed CVD
CHAINLINK
Settlement-native reference

TWAP60 is the core reference path; TWAP30 acts as a faster lead. The bridge judges whether the candidate is compatible with the state that matters at settlement.

TWAP60 / TWAP30
BYBIT
Independent hedge reversal proof

BOT3 uses Bybit linear executed flow as a second independent check when deciding whether the original BOT9 thesis has genuinely reversed. Expensive protection demands stronger agreement.

Hedge evidence · executed flow
POLYMARKET
Path, price, depth & execution

The contract itself is the final reality. BOT9 watches token path and fresh execution structure; BOT3 requires adverse repricing and re-reads the opposite book immediately before a hedge.

Execution authority · same market

The last question is always: can this price actually be bought?

BOT9 can have the right direction and still refuse the order. A fresh ask above the authorized corridor, thin depth, slippage or a deteriorating token path can turn READY back into HOLD before capital moves.

After BOT9 fills, the same discipline applies in reverse. BOT3 does not buy the opposite token because it became cheap or because one venue flashed red. The opposite side must become a credible favorite, the original token must reprice adversely, multiple reversal checks must persist, and the resulting hedge must improve the worst case after fees.

Coinbase can still act as optional veto-only context in the BOT9 architecture, but a missing or neutral reading does not get to invent a direction.

→ Illustrative Polymarket book. The real stack re-reads the live book at final execution time.
BTC · 5-MIN BINARY ● LIVEHEALTHY
PriceDepthSize
spread $0.02 · mid $0.62
Bid / Ask imbalance+0%
Structure confirmed · executable depth is healthy
Market-Open PHT · Schedule Authority

The clock is part of the model.

The engine can monitor continuously while authorization changes by the exact five-minute market-open time. Entry/last seconds, TP bounds, position size, actual-dollar ATR blocks and confidence caps can all tighten a specific day or hour without creating a parallel order path.

Monday · the rulebook

Philippine Time · UTC+8
—
Representative controls armed this day
Safety block / tighter windowHold / rescan conditionSize modifierEntry / TP override
Risk & Execution Governors

The rules that can say no after the signal says yes.

The current stack is built around fail-closed boundaries where money is involved. Research can fail without stopping the bot; execution authority cannot quietly widen itself.

Hard entry price ceiling
BOT9's fresh execution path cannot silently pay beyond the configured directional ceiling. The current architecture keeps the absolute execution boundary at or below 0.67.
≤ 0.67
Fresh CLOB corridor
The order book is re-read at execution. Floor, TP ceiling, depth and slippage are checked again so a stale decision cannot chase a changed market.
Final recheck
Settlement / DUAL opposition
Material opposition from the settlement bridge or the Chainlink/Binance DUAL state can hold the candidate even when primary CVD still points to a direction.
HOLD → rescan
PM path & churn memory
Repeated favorite flips and token-path deterioration raise the proof required. Late, choppy windows can trigger stronger reconfirmation instead of getting a free pass.
Persistent path
ATR safety layers
Normalized ATR controls the broad BTC regime; actual-dollar ATR rules can block narrow scheduled ranges that forward evidence identifies as poor.
bps + USD ATR
Confidence floor / cap
Confidence is a bounded quality check. It can require a minimum after a configured elapsed second and can also cap overconfident candidates inside selected time/ATR pockets.
Elapsed + scheduled
FAK / GTC / fill evidence
An attempted order is not treated as a position. Fill evidence, bounded recovery and ghost protection keep accounting tied to what actually happened on the exchange.
Proof before position
Hedge requires a real BOT9 position
BOT3 ignores READY and ordinary order attempts. It arms from proven fire/fill authority, and live hedge spending can require independent wallet proof.
No phantom hedge
Evidence gets stricter with price
Standard, strong and emergency hedge tiers require progressively stronger persistence and flow evidence. Confirmation fixes the maximum tier; the final executor cannot upgrade it.
Tier-capped
No over-hedge
Cumulative opposite-side shares can never exceed the proven original BOT9 shares. The hedge budget, fees, depth and minimum risk improvement remain binding.
h ≤ original shares
Execution Corridor

A good thesis still needs a buyable price.

Direction, settlement support and timing can all be correct — and the engine can still refuse to pay. Price authority gets tighter as the order approaches the exchange.

Zone
Relative Range
Action
Authority
NOT AUTHORIZED
Wait
Timing / proof incomplete
BOT9 TP CORRIDOR
Eligible
Day + direction + schedule
FRESH CLOB CHECK
Revalidate
Floor · depth · slippage
HARD ENTRY CEILING
Never chase
≤ 0.67
HEDGE TIERS
Stronger proof
Standard → Strong → Emergency
BOT3 uses a different corridor because it is solving a different problem. The opposite side must first become a credible favorite; standard protection is capped around the cheaper tier, while more expensive protection requires stronger reversal proof. The final fresh-book read can move inside an earned tier, but it cannot promote itself into a more expensive tier.
Two-Engine Pipeline

From flow to capital — then from capital to protection.

Six stages. BOT9 owns the entry. BOT3 owns post-fill protection. The handoff happens only after a real position exists.

BOT9 · 01
Wake

Warm feeds · exact market · 5s grid

BOT9 · 02
Direction

Binance Futures executed CVD

BOT9 · 03
Prove

CL60/30 · DUAL · PM path · ATR · confidence

BOT9 · 04
Execute

TP corridor · final recheck · FAK/GTC · fill proof

BOT3 · 05
Monitor

Same market · adverse PM repricing · BNF · Bybit · CL60

BOT3 · 06
Hedge or Hold

Tier proof · minimax sizing · no over-hedge

Continuous Monitor · PHT (UTC+8)

The engine can be awake all day. Authorization does not have to be.

The old website showed a cron-limited strategy. The current architecture is better described as continuous monitoring with schedule-aware authorization: the process can observe every five-minute market while individual windows tighten entry time, TP, size, ATR or confidence rules.

Monitoring Map

Philippine Time · full week

Current PHT: --:--:--
← swipe to see the full week →
Monitoring online Partial / controlled window Process offline Current hour
Market-open clock: timing rules attach to the exact five-minute market open in Philippine time, not to whatever second the scanner happens to wake up.
Independent modifiers: entry/last timing, TP bounds, live size, USD-ATR blocks and confidence caps can each have their own scheduled rule.
Shadow research stays separate: selected windows can be observed with zero capital so a hypothesis earns forward evidence before it is allowed near live execution.
Hedge budget can be scheduled too: BOT3 freezes the applicable market-open budget rule for the entire position, while hard multiplier and dollar ceilings remain fail-closed.
The Founder

A small lab with a serious operating standard.

SETH PM LABS
BOT9 · ENTRYBOT3 · HEDGE
Founder · Engineer · Systems Researcher
Cedie
Autonomous trading systems & prediction-market research
BOT9 V444 · HEDGE V1.10.3 · OPERATIONAL
EngineerQuantResearcher

A $32 server. Claude and GPT. A bedroom. An idea that refused to stay small.

SETH PM LABS did not begin on a trading floor. It began with a cheap cloud server, AI tools open in another tab, and one practical question: could one person build a digital trading operation disciplined enough to research, execute and protect capital even while he slept?

The goal was never a passive-income fantasy. It was repeatability — turn a thesis into code, turn code into logged evidence, and turn evidence into a system that can operate without fear, excitement or fatigue changing the rules halfway through a trade.

What started as one trading bot became a two-engine stack. BOT9 decides when a BTC five-minute setup deserves capital. After a real position exists, BOT3 watches what happens next and buys protection only when the original thesis is failing hard enough, the reversal is independently proven, and the hedge mathematically improves the downside.

The bigger lesson is not that everyone should build a trading bot. It is that serious work can start small. A bedroom, a low-cost server and modern AI do not replace skill or discipline — they lower the price of beginning. Start before the setup looks impressive. Keep the losses as data. Change one thing at a time. Let persistence compound into capability.

SETH PM LABS is the business built around that process: independent quantitative research, autonomous execution infrastructure, risk engineering and the belief that a small team — even one person at the beginning — can build systems that deserve to be taken seriously.

"Start small. Measure everything. Keep building until the system can carry the discipline for you."
For The Curious · For The New

If you're new, start here.

Plain answers about what SETH PM LABS is building, what the two engines do, and what they deliberately refuse to promise.

What is SETH PM LABS — a bot, a website, or a business?
It is an independent quantitative systems lab. The website documents the research and engineering; the trading stack is one operating system inside the lab. The current focus is autonomous BTC five-minute prediction-market execution, post-entry risk protection, telemetry and forward research.
What does BOT9 actually do?
BOT9 is the entry engine. Binance Futures executed flow chooses a candidate direction. Chainlink TWAP60/TWAP30, DUAL state, Polymarket path, ATR, confidence, timing, token-price corridor and execution checks decide whether that candidate deserves a real order. A failed gate normally means HOLD and rescan, not “buy anyway.”
What happened to the old Brownian-bridge fair-value strategy?
The math was not thrown away; its role changed. The current architecture uses a settlement-native TWAP60 bridge as support evidence rather than treating a model probability or a fixed “fair value minus ask” edge as the direct fire authority. Direction now comes from executed flow. Edge remains useful telemetry, but it is not the gate that makes the trade.
What does BOT3 hedge — and why doesn't it hedge every losing-looking move?
BOT3 only watches after BOT9 has a proven position. It evaluates the opposite outcome of that exact same five-minute market. A hedge requires persistent adverse Polymarket repricing, independent flow reversal evidence, a healthy path, fresh executable depth and an economic improvement after fees. A temporary wiggle is intentionally not enough, because unnecessary protection can turn a winning BOT9 trade into a worse result.
Can the hedge eliminate every BOT9 loss?
No. No causal signal can guarantee that. A market can jump straight into settlement, liquidity can disappear, or a losing position can fail without producing the configured reversal proof. The engineering target is more defensible: monitor every proven BOT9 position and only spend on protection when the evidence and worst-case math justify it.
What actually stops the system from chasing?
Multiple hard boundaries: market-open timing, TP floors/ceilings, a fresh CLOB corridor, an execution ceiling, final revalidation, fill evidence and bounded recovery. On the hedge side, price tiers, budget caps, favorite proof, depth, minimum risk improvement and the no-over-hedge rule remain binding through the final fresh-book read.
Is the system fully automatic?
The software is built to scan, authorize, execute, audit and protect positions autonomously, but the research process remains human-supervised. New ideas can run in shadow / research-only mode before they are allowed to affect live capital. Automation is used to enforce discipline, not to remove accountability.
Is this financial advice or a promise of consistent returns?
No. SETH PM LABS is documenting its own research and engineering process. It is not a signal service, managed fund, return guarantee or financial recommendation. The goal is repeatable process and controlled risk; real markets can still produce losses, failed fills, regime changes and unexpected behavior.
Build · Research · Collaborate

Research the edge.
Automate the discipline. Protect the downside.

For systems research, AI automation, prediction-market engineering, partnerships, or a conversation about building small teams with serious infrastructure.

Start a Conversation →Back to Top
Copied
Press J/K to navigate sections