LAB ONLINE·00:00:00 UTC

MARKET CONTROL ROOM / BTC

Signal, evidence, outcome.

DATA FRESHNESS 12 SECSource consensus 3/3
MARKET REGIMEHIGH VOLATILITY
--/ 100
PATHFINDER SCORE

CONNECTING

Collecting independent market sources and calculating the first verified snapshot.

BEARISHNEUTRALBULLISH
MARKET PULSELIVE
BTC

$78,240

−2.31%
FUNDING--OPEN INTEREST--SOURCES0/3
AGENT CONSENSUS4 / 5 aligned
MMarket AgentBEARISH72%
DDerivativesBEARISH63%
NNews AgentNEUTRAL54%
RRisk AgentHIGH RISK81%
FFlow AgentBEARISH63%
WHY THIS SCORE?
Technical structure−18
Derivatives positioning−9
Liquidation pressure−11
Funding balance+3
Macro conditions−8
COMPOSITE CONTRIBUTION−43
PAPER BOT / ALPHA 01SIMULATION
EQUITY

$104,380

+$4,380 +4.38%
WIN RATE56.2%MAX DD−3.82%TRADES48
LIVE MARKET · RESEARCH V1가격·지표·Pathfinder 값은 서버가 조회한 실제 공개 시장 데이터입니다. 뉴스·연구·모델 성과 화면은 아직 설계 샘플이며, 실거래 주문은 실행하지 않습니다.

MARKET / BTC

Raw inputs, clearly exposed.

FEED QUALITY HEALTHY3 sources reporting
RSI · 14

--

Connecting
MACD

--

Connecting
VOLATILITY · 24H

--

Realized
EMA 20 / 50

--

Connecting
PRICE + EMA STRUCTURE1H
Illustrative trendLive EMA values above
SOURCE CONSENSUSCONNECTING

-- median

Bybit----
Binance----
OKX----

Cross-source spread --

EVENT ENGINE

News becomes measurable events.

RECENT EVENT CLUSTERSUTC
MARKET REACTIONREGULATION

Regulatory guidance updated

EVENT PRICE$78,420
+5m−0.18%
+15m−0.42%
+1h−1.07%
+4h−2.18%
+24hPending

The 24-hour window updates automatically when complete.

AGENT NETWORK

Specialists before consensus.

MMARKET AGENT

--

CONNECTING

Computed from live RSI, MACD and EMA structure.

DDERIVATIVES AGENT

--

CONNECTING

Funding and open-interest snapshot from the derivatives market.

NNEWS AGENT

--

PENDING CONNECTION

Will use the existing GraphFriends News database in the next integration step.

RRISK AGENT

--

CONNECTING

Calculated from live volatility and source availability.

FFLOW AGENT

--

PENDING DATA

Volume-flow model stays disabled until normalized historical volume is stored.

PPATHFINDER

--

CONNECTING

Waiting for verified sources.

PATHFINDER / VERSION 1.0

Every conclusion keeps its receipt.

REPRODUCIBLE
LIVE WEIGHTED SCOREWaiting for data
Technical70%--
Derivatives30%--
NewsPending--
MacroPending--
FlowPending--
EventsPending--
CURRENT INTERPRETATIONCONNECTING
Collecting live data. No Pathfinder conclusion is created until at least one verified source is available.
CONFIDENCE--
Source quality and indicator completeness are evaluated separately from direction.
V1 score = technical × 0.70 + derivatives × 0.30. News, events and flow remain disabled until their production datasets are connected.

PAPER BOT / ALPHA 01

Claims meet the simulated market.

NO REAL ORDERS
CAPITAL

$100,000

Initial
EQUITY

$104,380

+4.38%
MAX DRAWDOWN

−3.82%

48 decisions
WIN RATE

56.2%

27 / 48
IMMUTABLE DECISION LOG
TIMEASSETACTIONENTRYSCORECONF.MODEL
Loading stored decisions…

PERFORMANCE / VERSIONED

The record includes the misses.

EQUITY CURVE90 DAYS
RETURN+4.38%PROFIT FACTOR1.31AVG. HOLD6.4h
STRATEGY VERSIONSAUDITABLE
v1.0Weighted scoring baselineCURRENT
v0.3Calibration testARCHIVED
v0.2Technical-only benchmarkARCHIVED

Results are never reassigned after an algorithm change.

GRAPHFRIENDS DATASET / RESEARCH

Questions that become evidence.

6 STUDIES
FEATURED · MACRO

RESEARCH QUESTION

How does Bitcoin react after FOMC decisions?

Event clusters are matched with market prices at +5m, +15m, +1h, +4h and +24h. The study separates hawkish, neutral and dovish outcomes.

SAMPLES23WINDOW2024—2026LAST UPDATEDSEP 11
MEDIAN BTC REACTION · HAWKISH
+5m−0.18%
+1h−0.74%
+4h−1.42%
+24h−0.83%
MARKET REGIME

Bitcoin behavior in risk-off regimes

Volatility, drawdown and recovery statistics by regime.

48 samples · updated daily
DERIVATIVES

Funding-rate extremes and next-day returns

Distribution after crowded long and short positioning.

1,284 observations
OPEN INTEREST

Falling OI during price declines

Separating deleveraging from new short participation.

816 observations
NEWS IMPACT

High-importance regulation events

BTC and ETH response across five evaluation windows.

37 event clusters
FIXED RESEARCH URLS각 연구는 질문, 요약, 데이터셋, 역사 통계, 방법론, 출처, 최근 갱신일을 가진 독립 페이지로 확장하는 구조입니다.

MODEL REGISTRY / GOVERNANCE

Improve the model, preserve the record.

1 ACTIVE · 2 CANDIDATES
CURRENTRules V1Weighted scoring
+
NEXTXGBoost V1Candidate ensemble
+
LATERNeural V1After sufficient data
MODEL VERSIONSTIME-SERIES VALIDATION
MODELF1PROFIT FACTORMAX DDSAMPLESSTATUS
Pathfinder Rules v1.00.571.31−3.82%4,820ACTIVE
XGBoost candidate v0.40.611.38−5.10%4,820CANDIDATE
Logistic baseline v0.20.521.08−7.44%4,820ARCHIVED
PROMOTION GATEBLOCKED
F1 exceeds currentPASS
Max drawdown ≤ 12%PASS
Sample count ≥ 5,000FAIL · 4,820

Candidate models never replace production automatically until every validation gate passes.

REAL TIMEPredictionDAILYOutcome evaluationWEEKLYModel retrainingMONTHLYDeep review

OPERATIONS / SOURCE ROUTER

Bad data stops the analysis.

OVERALL STATUSOPERATIONAL

Pathfinder may run. Market consensus is within freshness and source-count thresholds.

ANALYSIS ENABLED
MARKET FEEDS3 SOURCES
BybitLIVE146 ms · 4s ago
BinanceLIVE182 ms · 7s ago
OKXDEGRADED940 ms · 42s ago
CoinbaseSTANDBYFallback route
NEWS SOURCES4 / 4 LIVE
BLSLIVELast success 08:30
Federal ReserveLIVELast success 11:42
SECLIVELast success 11:31
ReutersLIVELast success 11:58
AGENT RUNSV1.0
Market AgentOK42s ago
News AgentOK54s ago
Risk AgentOK48s ago
PathfinderOK58s ago
FALLBACK ORDERNO SINGLE-SOURCE DEPENDENCY
01Fresh source consensus→02Cached consensus→03Last valid data→04Pause analysis

OPEN METHODOLOGY

How Pathfinder works.

01

Collect

Market, derivatives, news and macro inputs enter through independent source routes.

02

Validate

Freshness, source count and dispersion gates decide whether analysis may run.

03

Calculate

Indicators and features are computed from stored, timestamped inputs.

04

Analyze

Specialized agents produce deterministic scores with confidence and reasons.

05

Combine

Pathfinder weights agent outputs according to the detected market regime.

06

Simulate

Paper Bot records the decision before the outcome is known. No real trade executes.

07

Evaluate

Returns and errors attach to the original decision without rewriting history.

08

Train

Time-ordered datasets retrain candidate models on a controlled schedule.

09

Govern

Only validated versions enter the ensemble; older models remain auditable.

COLLECT→DB→FEATURES→AGENTS→PATHFINDER→RESULT→TRAIN↻
LIMITATIONSModels can be wrong. Historical simulation does not guarantee future performance. This research environment is educational and does not provide personalized investment advice.