KPI Detail & Verification Hub

Home is summary; this page explains metric meaning and fixed formulas as a static verification page.

What is this page?

This page is a KPI/proof page to validate NoahAI performance and trust using real operation data.

For general users, focus on these 3 items

  • Win Rate: how often decisions are correct
  • Max Drawdown (MDD): how deep the worst decline was
  • PnL trend: overall profit/loss structure

Core one-liner

NoahAI is designed around loss control and repeatable stability, not only headline win rate.

Top 3 KPI summary

Win Rate

-

Max Drawdown (MDD)

-

Total PnL

-

Cohort values are primary; if a cohort field is unavailable, a public-safe operating aggregate is shown as a fallback.

How to interpret these numbers (important)

Current public figures are conservative rolling-30-day cohort aggregates. Do not judge by a single period win rate/PnL alone; interpret together with loss control (MDD), early exits, and repeatable stability.

How to read Data Health

Data Health is a status indicator combining data freshness and activity.

In short

  • NoahAI is validated on real usage logs.
  • Decisions and outcomes are recorded and published as aggregates.
  • Risk control is prioritized over headline win rate.
  • Learning/feedback loops improve repeatable stability.

Where to verify detailed evidence

The KPI hub is intentionally summary-first. Use the pages below for improvement evidence and mechanism details.

AlphaArena KPI (engine/meta-AI validation)

Technical metrics for engine behavior, throughput, and operational status. This is different from user cohort performance.

AlphaArena Public KPI Snapshot

Public-safe operating metrics only. Internal thresholds and raw session-level data are excluded; values are shown by public JSON snapshot timestamp, not real-time.

Real-trade PoC aggregate period: 2026-01-01 ~ present | Purpose: Transparency of AlphaArena technology performance | Engine cumulative PnL is anonymized within public-safe scope

Data Health

-

-

Trade Records

-

latest -

Model Chats

-

latest -

Balance History

-

latest -

Engine cumulative realized PnL (public summary): Loading data.These metrics are based on public-safe aggregates from AlphaArena real-trade PoC logs and do not constitute investment advice or guaranteed returns.

Data Freshness (elapsed time from last update)

  • Trade: -
  • ModelChat: -
  • Balance: -

Activity

  • Trades 24h: 0
  • Trades 7d: 0
  • Chats 24h: 0
  • Chats 7d: 0

Engine detail metrics

EngineTradesWin RateAvg PnL/Trade

How should I read these numbers?

Data Health is a composite state score from freshness and activity signals. Engine-level cumulative/average PnL and win rate are benchmark comparison references, not investment recommendations or guaranteed returns.

How to read zero activity

A 24h/7d activity value of 0 in this public snapshot does not necessarily mean service shutdown. This page is a static public snapshot at a disclosure timestamp, not a real-time operations console, and live operation may continue in a limited beta cohort.

NoahAI User KPI (real-user cohort aggregate)

Performance/risk metrics from real-user cohorts. Prioritize this block for service-use decisions.

NoahAI User Trading KPI Snapshot

Public-safe summary metrics derived from NoahAI PoC user logs (trading.db). Public representative figures follow the rolling-30-day cohort aggregate basis; account identifiers, raw prompts, and sensitive session traces are excluded.

PoC operation period: 2024-11-01 ~ present | Purpose: Transparency of NoahAI user trades and AI decisions | Data scope: Real-account test-user transaction logs and AI decision records (public KPI basis: rolling-30-day cohort aggregate)

Sample notice

This screen shows a random 1-user snapshot from a 30-user cohort. It is a case sample, not a representative performance metric.

For marketing/investor messaging, use cohort aggregates (median, quantiles, standard deviation) rather than a single-user value.

How to read: cohort metrics are representative values, while Random 1/N is a case sample. Judge with win rate, PnL, and max drawdown (MDD) together.

Cohort aggregate KPI

Representativeness should be evaluated on cohort aggregates. (fixed recent-30d window)

Cohort Win Rate

-

Cohort Total PnL

-

Cohort Max Drawdown

-

Cohort MDD is not shown because it is not present in the current public source dataset.

Data Health

-

-

Closed Trades

-

AI Decisions

-

Sample Win Rate

-

Random 1/30

Sample Total PnL

-

Random 1/30

Top Exit Reasons

  • Loading data

Top Decision Types

  • Loading data

These metrics are for operating transparency and do not constitute investment advice or guaranteed returns.

Note: This snapshot is a public-safe summary across crypto and securities PoC operation logs. Account-level details, raw prompts, and sensitive session traces are excluded.

How to interpret cohort PnL

The public recent-30-day cohort PnL is a validation aggregate under conservative risk rules and early-exit policy. NoahAI prioritizes risk limits and long-run repeatability over short-term PnL maximization.

Exchange / channel summary

ExchangeTradesWin RateSample PnL (1/N)
Loading data
AI Learning Insights

NoahAI Blockchain Learning Insights

AI learning performance and core mechanisms based on PoC operation data. Provided for operational transparency; certain technical parameters are proprietary and not disclosed.

Items marked "Proprietary" are withheld to protect trade secrets and prevent competitive intelligence disclosure.

AI Staircase Learning Curve

Based on public monthly reports (rolling 30-day windows), NoahAI has shown phased staircase-like improvements after the initial operation period.

Phase 1

~18%

Early Learning Signal collection and pattern recognition

Phase 2

~30%

Maturation Stable strategy convergence

Phase 3

~33%

Plateau + Adjustment Preparing next-level optimization

NOW

Phase 4

34%+

Re-acceleration New mechanisms applied

3 Core Learning Mechanisms

Why NoahAI is different from conventional automated trading

01Patent pending

Dynamic Risk Adjustment System

Automatically switches between aggressive and defensive modes

When win rate drops, AI immediately reduces trade size. When performance improves, it gradually scales up. Thresholds and scaling factors are proprietary.

02

Intelligent Coin Selection Learning

Learns "which coin wins now" from 140k+ signals

Accumulates and learns from 140,000+ AI signals during PoC operations to identify optimal entry timing per coin. Selection algorithm is proprietary.

03β Verified

AI Early-Exit Logic

"Small certain gains" outperform "waiting for full TP"

Real data confirms AI early-exit achieves higher effective win rate than traditional TP-targeting. The exit judgment model is proprietary.

Technical Differentiation: NoahAI vs Conventional Bots

Core technology advantages over competitors

FeatureConventional BotNoahAI
Risk AdjustmentFixed (no change)✅ Dynamic (performance-reactive)
Coin SelectionPre-defined (static)✅ Real-time learning (evolving)
Exit StrategyMechanical TP/SL✅ AI judgment-based (adaptive)
Learning DataManual tuning windows✅ Real-operation cumulative (140k+)
Market Sentiment❌ Not supported✅ Fear · Neutral · Greed
Adjustment ModelManual setting✅ Real-time execution adjustment + 7-day policy review

Technical Specifications (Public Scope)

⚠️ Some parameters are not disclosed to prevent competitive intelligence leakage.

Cumulative AI Signals140,000+ (PoC operation cumulative)
Supported ExchangesBinance · Bybit · OKX · Bitget operational, Upbit · Bithumb spot integration available
Technical IndicatorsRSI · MACD · Bollinger Bands · Volume · Trend
Portfolio Coins14–20 (dynamic selection)
Adjustment CycleReal-time execution adjustment + 7-day sliding review
Market Sentiment ClassesFear · Neutral · Greed (3 levels)
Dynamic Risk ThresholdsProprietary
Exit Model ParametersProprietary
Coin Scoring AlgorithmProprietary
Base LeverageUp to 10× (adjustable per setting)

Technology Roadmap

Completed

PoC Learning System

  • Staircase learning curve confirmed
  • Dynamic risk adjustment live
  • AI early-exit logic validated
  • 140,000+ signals accumulated
In Progress

Judgment Quality Optimization

  • Balance TP/SL to reduce avoidable loss
  • Increase AI early-exit share
  • Use a user-specific 7-day data adjustment loop
Planned (2026 Q3)

Multi-exchange Operations Hardening

  • Exchange-specific execution quality/slippage optimization
  • Cross-market learning hardening (staged enablement)
  • User risk profile policy presets
Long-term (2026 Q4+)

Advanced Strategies

  • Spot strategy quality upgrades
  • Community signal pool concept validation
  • Options strategy feasibility review

Frequently Asked Questions

Q. Does the AI actually learn, or is it just luck?

It actually learns. Random performance would fluctuate erratically, but NoahAI has shown staircase improvement across operation phases. The 7-day sliding architecture and 140k+ accumulated signals support statistical validity.

Q. What is the basis for Phase 1 (~18%)?

It is a rounded reference from the 2025-10 monthly report value (17.7%). This is not a calendar-month fixed value, but a rolling 30-day window (2025-09-20 to 2025-10-19).

Q. How long did it take from 17.7% to 34.3%?

Under the public comparison rule, it is 211 days (about 7 months): from the 2025-10 report window (2025-09-20 to 2025-10-19) to the 2026-05 report window (2026-04-19 to 2026-05-18).

Q. Why is win rate below 50%?

NoahAI executes only a rigorously filtered subset of all signals. The effective win rate on those filtered signals is higher than the published figure. Systems sustaining 50%+ consistently risk overfitting, so stability is intentionally prioritized.

Q. Can I see the specific thresholds or parameters?

Core parameters are proprietary to prevent competitive replication. We publicly disclose the concepts and principles of each mechanism, but not specific numerical values.

Q. These are PoC-user results — will commercial users see the same?

PoC operations run on real accounts under strict risk controls. Public figures are published with verifiable aggregate rules, and commercial rollout may include additional optimization.

Q. Is risk managed during market crashes?

Yes. The dynamic risk adjustment system detects performance degradation and automatically reduces positions. Real-time market sentiment (fear/neutral/greed) is also incorporated.

This section is for educational purposes about AI system operation. Results are not guaranteed. This is not investment advice.

Why these metrics matter

  • Win Rate: not just the percentage, but whether stability is maintained across regimes.
  • MDD: a core metric for real downside risk experienced by users.
  • Early exits: execution quality for limiting downside before losses expand.
  • Data Health: an operational signal for freshness and activity level.

KPI Definition (fixed public)

Data Health Score is a weighted composite from freshness and activity conditions. AlphaArena uses public-safe anonymized aggregates from real-trade PoC logs, while NoahAI user KPI uses a rolling-30-day cohort aggregate. Public pages expose anonymized aggregate metrics only.

XAI KPI Draft (public scope)

XAI remains governed by the technology/XAI page. The KPI hub will add only public-safe aggregate indicators in phases.

  • Explanation log completeness: required-field coverage (%)
  • Reason-to-outcome linkage: rationale/result match rate (%)
  • Replayability: decision consistency under same conditions (%)
  • Risk alert timeliness: median response time after alert
Open XAI architecture

Engine PnL (bar view)

No engine data