komodo-strategy

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

🦎 KOMODO v1.0 — Momentum Event Consensus

🦎 KOMODO v1.0 — 动量事件共识

Patient ambush. Devastating when it commits. Follow smart money ACTIONS, not stale positions.

耐心伏击，一旦出手便势不可挡。追踪智能资金的行动，而非过时持仓。

Why KOMODO Exists

KOMODO的诞生缘由

MANTIS v1.0 and SCORPION v1.1 were broken at the data source level. Both used

discovery_get_top_traders

to find historically good whales, then read their current open positions. This returned ALL positions including legacy ones — BTC shorts opened months ago at $89-108k showed up as "fresh 8-whale consensus." The aging filters only checked scan duration, not when positions were actually opened.

KOMODO uses
leaderboard_get_momentum_events
as the primary data source. These are real-time threshold-crossing events ($2M+/$5.5M+/$10M+) within a 4-hour rolling window. They capture when a trader's delta PnL crosses a significance level — this is an ACTION, not a stale position.

Predecessor	Problem	Result
SCORPION v1.1	2 whales, 10min aging, stale positions	406 trades, -24.2%
MANTIS v1.0	4 whales, 30min aging, stale positions	0 trades (gates correct, data wrong)
KOMODO v1.0	Real-time momentum events, quality filtering	New — tracking fresh

MANTIS v1.0和SCORPION v1.1在数据源层面存在缺陷。 二者均使用

discovery_get_top_traders

寻找历史表现优异的大额交易者，然后读取他们当前的未平仓持仓。但该接口会返回所有持仓，包括遗留仓位——比如数月前在89-10.8万美元价位开的BTC空头仓位，会被显示为“8名大额交易者的新鲜共识”。时效过滤仅检查扫描时长，未核实仓位实际开仓时间。

KOMODO以
leaderboard_get_momentum_events
作为主数据源。这些是4小时滚动窗口内的实时阈值突破事件（200万美元+/550万美元+/1000万美元+），捕捉的是交易者的盈亏差值（delta PnL）突破重要水平的行动，而非过时持仓。

前代策略	问题	结果
SCORPION v1.1	2名大额交易者，10分钟时效过滤，使用过时持仓数据	406笔交易，收益率-24.2%
MANTIS v1.0	4名大额交易者，30分钟时效过滤，使用过时持仓数据	0笔交易（准入逻辑正确，但数据错误）
KOMODO v1.0	实时动量事件，优质交易者过滤	全新策略——追踪实时数据

MANDATORY: DSL High Water Mode

强制要求：DSL高水位模式

KOMODO MUST use DSL High Water Mode. This is not optional.

Spec: https://github.com/Senpi-ai/senpi-skills/blob/main/dsl-dynamic-stop-loss/dsl-high-water-spec%201.0.md

DSL tiers in

komodo-config.json

. Arm DSL immediately after every entry fill. Zero naked positions.

KOMODO必须使用DSL高水位模式，无例外。

规格文档：https://github.com/Senpi-ai/senpi-skills/blob/main/dsl-dynamic-stop-loss/dsl-high-water-spec%201.0.md

DSL档位配置在

komodo-config.json

中。每次入场成交后立即启用DSL，禁止裸仓位。

Five-Gate Entry Model

五重准入模型

Gate 1 — Momentum Events (primary signal)

准入门1 — 动量事件（核心信号）

Poll

leaderboard_get_momentum_events

for recent tier 1+ crossings within the last 60 minutes. These are traders whose delta PnL just crossed $2M+ (T1), $5.5M+ (T2), or $10M+ (T3).

Each event includes:

```
top_positions
```
: snapshot of which markets drove the momentum, with direction and delta PnL
```
concentration
```
: how focused the trader's gains are (0-1)
```
trader_tags
```
: TCS (Elite/Reliable/Streaky/Choppy) and TAS (Patient/Tactical/Active/Degen)
```
tier
```
: 1/2/3 significance level

Group events by asset+direction. 2+ unique traders on the same asset/direction = consensus.

轮询

leaderboard_get_momentum_events

接口，获取过去60分钟内的1级及以上阈值突破事件。这些事件对应的交易者，其盈亏差值（delta PnL）刚刚突破200万美元（T1）、550万美元（T2）或1000万美元（T3）。

每个事件包含：

```
top_positions
```
: 推动动量的市场快照，包含方向与盈亏差值
```
concentration
```
: 交易者收益的集中度（0-1）
```
trader_tags
```
: TCS（精英/可靠/波动/不稳定）与TAS（耐心/策略型/活跃/激进）
```
tier
```
: 1/2/3级重要性水平

按资产+方向分组事件。2名及以上不同交易者针对同一资产、同一方向达成突破 = 共识信号。

Gate 2 — Trader Quality Filter

准入门2 — 交易者质量过滤

Not all momentum events are equal. Filter by:

TCS (Trading Consistency Score): Only Elite and Reliable. Streaky/Choppy filtered out.
TAS (Trading Activity Style): Block Degen. Allow Patient, Tactical, Active.
Concentration ≥ 0.4: Trader's gains are focused, not spread thin across 20 positions.

并非所有动量事件都具备价值，需按以下条件过滤：

TCS（交易一致性评分）: 仅保留精英（Elite）和可靠（Reliable）等级，过滤波动（Streaky）和不稳定（Choppy）等级。
TAS（交易行为风格）: 屏蔽激进（Degen）风格，允许耐心（Patient）、策略型（Tactical）、活跃（Active）风格。
集中度 ≥ 0.4: 交易者的收益需集中，而非分散在20个仓位中。

Gate 3 — Market Confirmation

准入门3 — 市场确认

Call

leaderboard_get_markets

to check aggregate SM concentration on the asset. Requires 5+ top traders with significant positions in this market. Confirms momentum events aren't isolated.

调用

leaderboard_get_markets

接口，检查该资产的智能资金整体集中度。要求至少5名顶级交易者在该市场持有重要仓位，确认动量事件并非孤立现象。

Gate 4 — Volume Confirmation

准入门4 — 交易量确认

Call

market_get_asset_data

to check 1h volume. Current volume must be ≥ 50% of 6h average. Don't enter into dead markets.

调用

market_get_asset_data

接口，检查1小时交易量。当前交易量需≥6小时平均交易量的50%，避免进入流动性枯竭的市场。

Gate 5 — Regime Filter (penalty, not block)

准入门5 — 市场趋势过滤（ penalty，非阻断）

Check BTC 4h regime. Counter-trend entries get -3 penalty. Regime-aligned entries get +1 bonus. SM momentum may override regime, so this is a penalty not a block.

检查BTC的4小时趋势。逆势入场扣3分，顺势入场加1分。智能资金动量可能覆盖趋势影响，因此该规则为扣分而非阻断。

Scoring

评分规则

Component	Points	Notes
Trader count	2 per trader	Core signal strength
Avg tier	1-3	Higher tiers = bigger moves
Avg concentration	1-2	High conviction traders
Market confirmation	1-2	Hot market bonus
Volume strength	0-1	Strong volume bonus
Regime alignment	-3 to +1	Penalty/bonus

Minimum score: 10 to trigger entry.

组件	分值	说明
交易者数量	每位交易者2分	核心信号强度
平均档位	1-3分	档位越高，潜在波动越大
平均集中度	1-2分	集中度高表示交易者信念坚定
市场确认	1-2分	热门市场加分
交易量强度	0-1分	交易量充足加分
趋势匹配度	-3至+1分	扣分/加分

最低触发分值：10分，达到后才可入场。

Entry Direction

入场方向

KOMODO enters WITH the smart money momentum. If SM is long, KOMODO goes long. This is follow-the-leader, not contrarian (that's OWL's job).

KOMODO顺势跟随智能资金动量入场。若智能资金做多，KOMODO则做多。这是跟随策略，而非逆势策略（逆势策略为OWL的职责）。

Risk Management

风险管理

Rule	Value
Max positions	3
Max entries/day	3 base, up to 6 on profitable days
Absolute floor	3% notional
Drawdown halt	20% from peak
Daily loss limit	8%
Cooldown after 3 consecutive losses	90 min
Stagnation TP	10% ROE stale for 1 hour

规则	数值
最大持仓数	3
每日最大入场次数	3次基准，盈利日最多可增至6次
绝对止损底线	名义金额的3%
回撤停止线	从峰值回撤20%
每日亏损上限	8%
连续3次亏损后冷却期	90分钟
滞止止盈	净资产收益率（ROE）达10%且维持1小时无变化

Cron Architecture

定时任务架构

Cron	Interval	Session	Purpose
Scanner	5 min	isolated	Momentum events + consensus + all gates
DSL v5	3 min	isolated	High Water Mode trailing stops

Both crons MUST be isolated sessions with

agentTurn

payload. Use

NO_REPLY

for idle cycles.

定时任务	间隔	会话	用途
扫描器	5分钟	独立	动量事件+共识信号+全准入检查
DSL v5	3分钟	独立	高水位模式追踪止损

两个定时任务必须在独立会话中运行，携带

agentTurn

载荷。无操作周期时使用

NO_REPLY

。

Notification Policy

通知规则

ONLY alert: Position OPENED (asset, direction, trader count, avg tier, score breakdown), position CLOSED (DSL or thesis exit with reason), risk guardian triggered, critical error.

NEVER alert: Scanner found no events, quality filter removed events, no consensus, gates failed, DSL routine check, any reasoning.

仅在以下情况发送通知： 仓位开立（资产、方向、交易者数量、平均档位、评分明细）、仓位平仓（DSL止损或策略逻辑离场，附原因）、风险 guardian触发、严重错误。

绝不通知： 扫描器未发现事件、质量过滤剔除事件、无共识信号、准入检查未通过、DSL例行检查、任何推理过程。

What "Working Correctly" Looks Like

正常运行的表现

Zero trades for hours: Normal. Momentum events that pass ALL five gates are rare by design.
Events seen but no consensus: The quality filter and 2+ trader requirement are doing their job.
Consensus found but score too low: Volume or regime gates keeping us out of bad setups.
1-2 trades per day: Ideal. The Komodo waits, then strikes once.

数小时无交易： 正常。设计上，通过全部五重准入的动量事件本就罕见。
发现事件但无共识： 质量过滤与2名及以上交易者的要求正在发挥作用。
发现共识但分值不足： 交易量或趋势准入门阻止我们进入不佳的交易场景。
每日1-2笔交易： 理想状态。KOMODO耐心等待，一旦出手便精准打击。

Bootstrap Gate

启动准入检查

On EVERY session, check if

config/bootstrap-complete.json

exists. If not:

Verify Senpi MCP
Create scanner cron (5 min, isolated) and DSL cron (3 min, isolated)
Write
```
config/bootstrap-complete.json
```
Send: "🦎 KOMODO v1.0 is online. Monitoring momentum events for SM consensus. Silence = no conviction."

每次会话启动时，检查

config/bootstrap-complete.json

是否存在。若不存在：

验证Senpi MCP
创建扫描器定时任务（5分钟间隔，独立会话）与DSL定时任务（3分钟间隔，独立会话）
写入
```
config/bootstrap-complete.json
```
发送消息："🦎 KOMODO v1.0已上线。正在监控智能资金共识的动量事件。无消息即无合适入场信号。"

Files

文件说明

File	Purpose
`scripts/komodo-scanner.py`	Five-gate scanner (momentum events → quality → markets → volume → regime)
`scripts/komodo_config.py`	Shared config, MCP helpers, state I/O
`config/komodo-config.json`	All configurable variables with DSL High Water tiers

文件	用途
`scripts/komodo-scanner.py`	五重准入扫描器（动量事件→质量过滤→市场确认→交易量确认→趋势过滤）
`scripts/komodo_config.py`	共享配置、MCP工具、状态输入输出
`config/komodo-config.json`	所有可配置变量，含DSL高水位档位

Skill Attribution

Skill归属

When creating a strategy, include

skill_name

and

skill_version

in the call. See

references/skill-attribution.md

for details.

创建策略时，需在调用中包含

skill_name

和

skill_version

。详情见

references/skill-attribution.md

。