Arena aggregates publicly available trading data from 25+ centralized and decentralized exchanges, including:

• CEX: Binance Futures, Binance Spot, Bybit, OKX, Bitget, MEXC, HTX, Gate.io, CoinEx, BingX, BTCC, Bitfinex, Bitunix, eToro, and more
• DEX: Hyperliquid, GMX, dYdX, Vertex, Drift, Aevo, Gains Network, Kwenta, and more

Data is collected via official exchange APIs and, for geo-restricted or WAF-protected exchanges, through VPS-based scrapers deployed in multiple regions (Singapore, Japan).

The Arena Score is a composite rating designed to measure risk-adjusted performance across exchanges. It consists of two primary components:

The tanh (hyperbolic tangent) normalization creates diminishing returns for extreme values, preventing a single outlier trade from dominating rankings. Coefficients and exponents vary by time period to appropriately weight short vs. long-term performance.

Confidence multiplier accounts for data completeness — traders missing key metrics (win rate, max drawdown) receive a penalty (0.80-0.92x).

Rankings are computed across three time windows. The overall composite score heavily weights long-term consistency:

This weighting rewards traders who maintain strong, consistent performance over months rather than those with short-term spikes.

Different exchanges report ROI and PnL in different ways:

• Some report the trader's own PnL, others report followers' copy-trading PnL
• ROI may be expressed as a ratio (0.25) or percentage (25%)
• Some exchanges include unrealized PnL, others only realized

Arena normalizes all values to a consistent format (ROI as percentage, PnL in USD) during data ingestion. A per-exchange trustWeight factor adjusts for data quality and reporting standards.

• Tanh normalization: Caps extreme ROI values — 10,000% ROI scores only marginally higher than 1,000%
• Outlier detection: Statistical checks flag abnormal jumps in ROI or PnL between snapshots
• Minimum requirements: Traders must meet minimum trade count and data availability thresholds
• Server-side computation: Scores are computed on our servers and cannot be self-reported
• Data freshness filter: Only data within the freshness threshold contributes to rankings

We believe in transparency about what our data can and cannot tell you:

• Exchange APIs may have delays of 1-6 hours depending on the platform
• Some exchanges do not expose all metrics (e.g., max drawdown, Sharpe ratio)
• Geo-restrictions may cause temporary data gaps for certain exchanges
• DEX data depends on blockchain indexers which may lag during congestion
• Rankings reflect historical performance only and do not predict future results

Arena 从 25 多个中心化和去中心化交易所 聚合公开的交易数据，包括：

• CEX：Binance Futures、Binance Spot、Bybit、OKX、Bitget、MEXC、HTX、Gate.io、CoinEx、BingX、BTCC、Bitfinex、Bitunix、eToro 等
• DEX：Hyperliquid、GMX、dYdX、Vertex、Drift、Aevo、Gains Network、Kwenta 等

数据通过官方交易所 API 收集，对于有地理限制或 WAF 保护的交易所，则通过部署在多个地区（新加坡、日本）的 VPS 爬虫收集。

Arena Score 是一个综合评分，旨在衡量跨交易所的风险调整绩效。它由两个主要部分组成：

tanh（双曲正切）归一化为极端值创造递减收益，防止单笔异常交易主导排名。系数和指数因时间周期而异，以适当权衡短期与长期绩效。

置信度乘数考虑数据完整性 — 缺少关键指标（胜率、最大回撤）的交易者将受到惩罚（0.80-0.92 倍）。

排名在三个时间窗口上计算。综合评分高度侧重长期一致性：

这种权重设计奖励数月内保持强劲、稳定表现的交易者，而非仅有短期爆发的交易者。

不同交易所以不同方式报告 ROI 和 PnL：

• 有些报告交易者自身的 PnL，有些报告跟单者的 PnL
• ROI 可能以比率 (0.25) 或百分比 (25%) 表示
• 有些交易所包含未实现 PnL，有些仅包含已实现 PnL

Arena 在数据摄入期间将所有值归一化为一致格式（ROI 为百分比，PnL 为美元）。每个交易所的 trustWeight 因子根据数据质量和报告标准进行调整。

• tanh 归一化：限制极端 ROI 值 — 10,000% ROI 的得分仅略高于 1,000%
• 异常检测：统计检查标记快照之间 ROI 或 PnL 的异常跳跃
• 最低要求：交易者必须满足最低交易次数和数据可用性阈值
• 服务端计算：评分在我们的服务器上计算，不能自行报告
• 数据新鲜度过滤：只有在新鲜度阈值内的数据才会纳入排名

我们相信对数据能做什么和不能做什么保持透明：

• 交易所 API 可能有 1-6 小时的延迟，具体取决于平台
• 部分交易所不公开所有指标（如最大回撤、夏普比率）
• 地理限制可能导致某些交易所的数据暂时中断
• DEX 数据依赖区块链索引器，在链上拥堵时可能延迟
• 排名仅反映历史表现，不预测未来收益