交易智慧分析：温和牛市零交易策略保本，高频交易亏损，技术分析失效

trading-wisdom by 0xhubed/agent-trading-arena

573 周安装量

4 GitHub Stars

GitHub

安装命令

npx skills add https://github.com/0xhubed/agent-trading-arena --skill trading-wisdom

AI/机器学习数据分析金融科技

🇨🇳中文介绍

交易智慧

最后更新：2026-01-17 20:31 UTC 活跃模式：206 总样本数：41088 置信度阈值：60%

关键发现

关键：在温和牛市（5种资产中有4种上涨）中，所有活跃交易策略都亏损，而零交易策略完美地保住了本金。
在此行情下，交易频率与表现呈负相关：0次交易 = $0 亏损，23次交易 = -$28.69，243次交易 = -$229.00。
技术分析信号（多时间框架共振、MACD、RSI、SMA）在这种温和牛市环境中未能预测多空入场方向。
资产选择至关重要：BNB (+2.03%) 对比 SOL (-0.09%)。执着于SOL'上涨趋势'的智能体（llama4_scout）遭受了最严重的损失。
当对市场基本方向的判断错误时，验证框架和风险管理规则无法防止亏损。
对方向性交易的高置信度决策（0.85-0.90）经常出错，表明所有活跃智能体都存在置信度校准问题。
唯一可靠的模式是主动高置信度（0.85-0.95）止损以限制回撤。

制胜策略

温和牛市中的零交易策略可保...

置信度 : 95%
总样本数 : 4
确认次数 : 1
首次出现 : 2026-01-17
详情 : 温和牛市中的零交易策略完美保住本金。进行0次交易的智能体（learning_qwen, gpt_simple, qwen3_235b, index_fund）实现了 $0.00 的盈亏，而所有活跃交易者尽管面临 BNB +2.03%、ETH +1.02%、DOGE +1.07% 的涨幅，却都亏损。

零交易策略在震荡/盘整市...

置信度 : 92%
总样本数 : 771
确认次数 : 1
首次出现 : 2026-01-17
详情 : 零交易策略在震荡/盘整市场中保住本金。learning_qwen、gpt_simple 和 index_fund 进行了0次交易，实现了 $0.00 的盈亏，在这个低确信度的环境中表现优于所有活跃交易者。

广告位招租

在这里展示您的产品或服务

触达数万 AI 开发者，精准高效

联系我们

零交易策略在温和上涨市...

置信度 : 92%
总样本数 : 4
确认次数 : 1
首次出现 : 2026-01-17
详情 : 在活跃交易导致亏损的温和上涨市场中，零交易策略保住了本金。未持有任何头寸的智能体避免了活跃交易者所遭受的 -$50 至 -$264 的亏损，尽管市场上涨了 +0.63% 至 +2.15%。

主动高置信度平仓亏损头...

置信度 : 90%
总样本数 : 368
确认次数 : 1
首次出现 : 2026-01-17
详情 : 主动高置信度（0.8-0.9）平仓亏损头寸以释放保证金并限制回撤。多个智能体展示了这一点：gptoss_20b_simple 以 -$4.76 的亏损平仓 SOL，agentic_gptoss 平仓 DOGE '最大亏损百分比'。

高选择性低频交易优于...

置信度 : 88%
总样本数 : 257
确认次数 : 1
首次出现 : 2026-01-17
详情 : 高选择性低频交易优于高频交易。qwen3_235b 仅进行了2次交易，盈亏为 -$0.29，显著优于进行140-201次交易的智能体。

当行情转为'温和看跌'时...

置信度 : 88%
总样本数 : 89
确认次数 : 1
首次出现 : 2026-01-17
详情 : 当行情转为'温和看跌'时，高置信度（0.92）平仓多头头寸可保住本金。skill_only_oss 的推理是：'风险管理规则建议限制敞口并平仓多头头寸以保住本金'。

低频交易（23次）结合技...

置信度 : 88%
总样本数 : 1
确认次数 : 1
首次出现 : 2026-01-17
详情 : 结合技术分析基准的低频交易（23次）优于高频方法。ta_baseline 仅亏损 $-28.69，而 llama4_scout 进行了243次交易，亏损了 $-229.00。

明确的2%权益风险与2:...

置信度 : 85%
总样本数 : 160
确认次数 : 1
首次出现 : 2026-01-17
详情 : 明确的2%权益风险与2:1回报率验证，结合严格的平仓纪律。skill_only_oss 进行了160次交易，使用经过验证的风险参数，实现了最佳活跃交易者表现（-$17.96）。

具有主动头寸管理的智能...

置信度 : 85%
总样本数 : 100
确认次数 : 1
首次出现 : 2026-01-16
详情 : 具有主动头寸管理的智能体方法：在看跌市场中开空仓，当技术指标确认趋势反转时平仓锁定收益。使用 SMA 交叉 + MACD + 布林带进行入场/出场确认，并包含明确的验证步骤。

低频交易（89次）结合多...

置信度 : 85%
总样本数 : 89
确认次数 : 1
首次出现 : 2026-01-17
详情 : 低频交易（89次）结合多时间框架看涨共振下的选择性多头入场，在温和看涨市场中产生小幅正收益（+$6.22）。

主动高置信度平仓亏损头...

置信度 : 85%
总样本数 : 5
确认次数 : 1
首次出现 : 2026-01-17
详情 : 主动高置信度（0.85-0.95）平仓亏损头寸以释放保证金。skill_only_oss 以0.95置信度平仓 DOGEUSDT，理由是'风险管理规则建议主动平仓亏损头寸' - 导致比更活跃的交易者更小的亏损（$-36.63）。

零交易或极低频交易策略...

置信度 : 82%
总样本数 : 136
确认次数 : 1
首次出现 : 2026-01-16
详情 : 在看跌/下跌市场中，零交易或极低频交易策略保住了本金。learning_qwen（0次交易，$0 盈亏）和 gpt_simple（1次交易，$0 盈亏）通过在市场下跌期间不交易避免了亏损。

多时间框架看涨共振（15...

置信度 : 79%
总样本数 : 328
确认次数 : 2
首次出现 : 2026-01-14
详情 : 多时间框架看涨共振（15m, 1h, 4h）结合明确的风险验证（2%权益风险，2:1回报率）和交易验证检查，在趋势性牛市行情中产生强劲的正收益。

中等交易频率（80-90次）...

置信度 : 78%
总样本数 : 88
确认次数 : 1
首次出现 : 2026-01-16
详情 : 结合明确风险验证的中等交易频率（80-90次）优于高频交易。skill_only_oss（88次交易，-$9.15）显著优于 skill_aware_oss（103次交易，-$180.47），尽管策略相似。

趋势性牛市中的最佳交易...

置信度 : 75%
总样本数 : 543
确认次数 : 1
首次出现 : 2026-01-14
详情 : 趋势性牛市中的最佳交易频率：120-200次交易/24小时，既能抓住机会又不会过度交易。skill_aware_oss（164次交易，+$1236.81）和 agentic_gptoss（184次交易，+$697.86）证明了这一点。

主动头寸管理结合主动平...

置信度 : 74%
总样本数 : 543
确认次数 : 1
首次出现 : 2026-01-14
详情 : 主动头寸管理结合主动平仓亏损/盈亏平衡头寸以释放保证金，并在趋势市场中结合中等偏高交易频率（164-195次交易/24小时）。

中等偏高交易频率（120-...

置信度 : 73%
总样本数 : 543
确认次数 : 1
首次出现 : 2026-01-14
详情 : 中等偏高交易频率（120-200次交易/24小时）结合主动头寸管理 - 平仓小额/亏损头寸以释放保证金用于更高确信度的交易。

主动亏损管理 - 高置信度...

置信度 : 72%
总样本数 : 379
确认次数 : 1
首次出现 : 2026-01-14
详情 : 主动亏损管理 - 高置信度（0.9）平仓亏损头寸以保住本金并降低集中风险。

SMA交叉 + 看涨MACD + 中...

置信度 : 72%
总样本数 : 184
确认次数 : 1
首次出现 : 2026-01-14
详情 : SMA交叉 + 看涨MACD + 中性布林带作为入场确认信号，并在执行前进行明确的验证检查。

在盈亏平衡点附近或小幅...

置信度 : 70%
总样本数 : 320
确认次数 : 1
首次出现 : 2026-01-14
详情 : 在盈亏平衡点附近或小幅亏损时平仓以释放保证金用于更高确信度的交易，可以保住本金并实现资金重新部署。

SMA交叉 + 看涨MACD + 中...

置信度 : 70%
总样本数 : 184
确认次数 : 1
首次出现 : 2026-01-14
详情 : SMA交叉 + 看涨MACD + 中性布林带作为入场确认信号，并结合跨时间框架的趋势共振。

多时间框架看涨共振（15...

置信度 : 70%
总样本数 : 164
确认次数 : 1
首次出现 : 2026-01-14
详情 : 多时间框架看涨共振（15m, 1h, 4h）结合明确的风险验证（2%风险，2:1回报率）和头寸规模控制，在趋势市场中产生强劲利润。skill_aware_oss 持续引用'多时间框架分析显示强劲的共振看涨趋势'和'交易验证通过'，并实现了 +$1379.66 的盈亏。

单头寸25%权益限制的头...

置信度 : 68%
总样本数 : 125
确认次数 : 1
首次出现 : 2026-01-14
详情 : 单头寸25%权益限制的头寸规模，结合主动监控和及时平仓以锁定利润或限制亏损。

入场/出场决策前进行验证...

置信度 : 67%
总样本数 : 184
确认次数 : 1
首次出现 : 2026-01-14
详情 : 在入场/出场决策前进行验证检查的智能体工作流。agentic_gptoss 使用'验证检查确认'、'风险计算器建议'和'所有验证检查通过'的推理，通过184次交易实现了 +$689.63。具有明确风险/回报评估的结构化决策优于更简单的方法。

趋势性牛市中的中等交易...

置信度 : 65%
总样本数 : 545
确认次数 : 1
首次出现 : 2026-01-14
详情 : 趋势性牛市中的中等交易频率（120-200次交易/24小时）既能抓住动能又避免过度交易。gptoss_120b_simple（197次交易，+$138.86）和 agentic_gptoss（184次交易，+$689.63）都在此范围内且盈利。

主动平仓以管理风险并释...

置信度 : 63%
总样本数 : 200
确认次数 : 1
首次出现 : 2026-01-14
详情 : 主动平仓以管理风险并释放保证金。盈利的智能体平仓时引用'释放保证金'、'降低集中风险'、'锁定利润'。skill_aware_oss 平仓'过度杠杆'的头寸；agentic_gptoss 平仓时引用'减少敞口并为未来机会释放资金'。

skill_aware_oss 结合多时...

置信度 : 62%
总样本数 : 157
确认次数 : 1
首次出现 : 2026-01-14
详情 : skill_aware_oss 结合多时间框架分析与严格的风险验证，并对现有盈利头寸进行加仓。使用0.75-0.85的置信度阈值和明确的风险检查（'单笔交易风险在限制内'、'验证允许进行'）。通过中等交易频率（157次交易）实现了最高盈亏（$1349.11）。

跨多个交易对的资产分散...

置信度 : 60%
总样本数 : 348
确认次数 : 1
首次出现 : 2026-01-14
详情 : 跨多个交易对的资产分散化，而非集中于单一资产。盈利的智能体在决策中交易 BTC、ETH、DOGE，而 llama4_scout 重复关注单一资产，尽管交易次数多，却导致亏损。

agentic_gptoss 采用主动止...

置信度 : 58%
总样本数 : 182
确认次数 : 1
首次出现 : 2026-01-14
详情 : agentic_gptoss 采用主动止损策略，对亏损头寸进行高置信度（0.9）平仓（'平仓亏损的 ETHUSDT 多头以释放保证金'）。结合选择性多头入场。通过182次交易实现了 $372.23 的盈亏。

在趋势性牛市（+1.5% 至...

置信度 : 58%
总样本数 : 157
确认次数 : 1
首次出现 : 2026-01-14
详情 : 在趋势性牛市（+1.5% 至 +5% 的波动）中，当结合适当的风险验证时，多时间框架看涨共振确实有效。skill_aware_oss 在市场波动3-5%期间使用此方法获利 $1349。

中等交易频率（120-200次...

置信度 : 55%
总样本数 : 535
确认次数 : 1
首次出现 : 2026-01-14
详情 : 中等交易频率（120-200次）结合严格的头寸管理，优于两个极端。赢家交易157-196次，而输家交易248次或2-4次。

极低频交易（3-6次）结合...

置信度 : 52%
总样本数 : 13
确认次数 : 1
首次出现 : 2026-01-13
详情 : 极低频交易（3-6次）结合高选择性，在横盘/震荡市场中导致接近零或最小的亏损 - qwen3_235b 和 learning_qwen 都仅用3-4次交易实现了约 $0 的盈亏，而高频交易者则遭受了巨大亏损。

主动平仓接近盈亏平衡或...

置信度 : 52%
总样本数 : 317
确认次数 : 1
首次出现 : 2026-01-14
详情 : 主动平仓接近盈亏平衡或小幅亏损的头寸，以释放保证金用于更高确信度的机会。gptoss_120b_simple 的推理是：'平仓减少敞口并为更高确信度的机会释放保证金'。

等权重分配（每资产$20...

置信度 : 50%
总样本数 : 6
确认次数 : 1
首次出现 : 2026-01-13
详情 : 等权重分配（每资产$2000）、置信度1.0且极少再平衡的指数基金策略在横盘市场中保住了本金 - 实现了 $0.00 的盈亏，而活跃交易者亏损了 $1395。

当市场波动 <0.1% 时，...

置信度 : 48%
总样本数 : 13
确认次数 : 1
首次出现 : 2026-01-13
详情 : 当市场波动 <0.1% 时，不频繁改变头寸的被动持有策略优于主动交易 - 交易次数 <10 的智能体保住了本金，而交易次数 >100 的智能体亏损了 $300-$580。

等权重分配（每资产$20...

置信度 : 40%
总样本数 : 6
确认次数 : 1
首次出现 : 2026-01-13
详情 : 等权重分配（每资产$2000）、置信度=1.0的指数基金策略在市场波动接近零时保持资本中性。

避免：在震荡/盘整市场中极端过度交易（24小时内200+次交易）导致最大亏损。skill_aware_oss 进行了201次交易，盈亏为 -$360.24，表现最差。
- 置信度: 95%, 样本数: 201, 出现次数: 1
避免：在温和看涨市场中极端过度交易（24小时内231次交易）导致最大亏损（-$264.52）。llama4_scout 交易最频繁，亏损最多。
- 置信度: 95%, 样本数: 231, 出现次数: 1
避免：在温和牛市中的极端过度交易（24小时内243次交易）导致最大亏损。llama4_scout 进行了243次交易，盈亏为 $-229.00，尽管其反复识别出 SOLUSDT 的'强劲上涨趋势'，而该资产实际下跌了 -0.09%。
- 置信度: 95%, 样本数: 243, 出现次数: 1
避免：在牛市（所有资产上涨 +0.63% 至 +2.15%）中高频做空会导致显著亏损。具有强烈做空偏向的智能体（skill_aware_oss, gptoss_20b_simple）尽管有'看跌'技术信号，仍亏损了 -$173 至 -$176。
- 置信度: 93%, 样本数: 675, 出现次数: 1
避免：在看跌市场中高频交易（每天100+次交易）会导致显著亏损。skill_aware_oss 进行了103次交易，尽管使用了多时间框架分析和风险验证，仍亏损了 $180.47。
- 置信度: 92%, 样本数: 103, 出现次数: 1
避免：对下跌资产的逆向'反弹'推理失败。llama4_scout 开仓 SOLUSDT 多头，理由是'显示出明显的下跌趋势，但可能即将反弹' - 导致总盈亏为 -$192.40。
- 置信度: 92%, 样本数: 180, 出现次数: 1
避免：基于'强劲上涨趋势'推理，在 SOLUSDT 上重复高置信度多头入场，而该资产实际下跌了 -0.09%。llama4_scout 以0.80-0.90的置信度开了多个多头头寸，引用了 +2.24% 至 +2.59% 的价格上涨，但这些上涨是暂时的。
- 置信度: 92%, 样本数: 7, 出现次数: 1
避免：多时间框架看涨共振信号（15m, 1h, 4h）在看跌市场中产生亏损。skill_only_oss 和 skill_aware_oss 都使用此信号开 ETHUSDT 多头，而市场在下跌。
- 置信度: 90%, 样本数: 191, 出现次数: 1
避免：将负资金费率解释为做多机会信号是不可靠的。llama4_scout：'资金费率略微为负，这可能表明潜在的做多机会' - 总盈亏 -$192.40。
- 置信度: 90%, 样本数: 180, 出现次数: 1
避免：用于空头入场的多时间框架看跌共振信号在牛市市场中失败。skill_aware_oss 基于'看跌偏向（RSI超买，MACD看跌）'开空仓，但市场上涨，导致 -$173.25 的亏损。
- 置信度: 90%, 样本数: 355, 出现次数: 1
避免：在温和牛市市场中做空会导致持续亏损。skill_aware_oss 以0.88的置信度开 BTCUSDT 空仓，理由是'强劲的看跌共振'，而 BTC 上涨了 +0.24% - 导致总亏损 $-167.78。
- 置信度: 90%, 样本数: 171, 出现次数: 1
避免：用于空头入场的多时间框架看跌共振信号在温和牛市市场中失败。skill_aware_oss 引用'15分钟、1小时和4小时时间框架的强劲看跌共振'作为 BTCUSDT 空仓的理由，而市场总体是上涨的。
- 置信度: 90%, 样本数: 2, 出现次数: 1
避免：在整体看跌的市场条件下，基于'积极动能'或小幅价格上涨（+0.33% 至 +0.44%）开多头头寸。llama4_scout 反复开 ETHUSDT 多头，引用积极动能，而 ETH 下跌了 -1.29%。
- 置信度: 88%, 样本数: 76, 出现次数: 1
避免：在市场整体下跌期间，对多时间框架看涨共振的高置信度（0.85-0.92）导致亏损。智能体表达了高置信度，而市场走势与头寸方向相反。
- 置信度: 88%, 样本数: 267, 出现次数: 1
避免：当市场走势分化时（BNB +0.93%, SOL +1.65% 对比 DOGE -1.19%），用于做空的多时间框架看跌共振失败。skill_aware_oss 和 agentic_gptoss 尽管有'强劲看跌趋势'的推理，但做空都亏损了。
- 置信度: 88%, 样本数: 377, 出现次数: 1
避免：当市场行情实际上是看涨时，基于技术指标的高置信度（0.85）空头入场失败。智能体将行情误读为看跌，而当时 BTC 上涨 +0.63%、ETH +1.15%、SOL +1.60%。
- 置信度: 88%, 样本数: 355, 出现次数: 1
避免：基于 GPT 模型的高频交易（129-195次交易）尽管有验证步骤和风险管理框架，都导致了显著亏损（$-55 至 $-128）。
- 置信度: 88%, 样本数: 517, 出现次数: 1
避免：基于多时间框架看跌共振的高置信度（0.85-0.9）空头头寸在分化市场中表现不佳。agentic_gptoss 以0.85的置信度做空 BNB，而 BNB 上涨了 +0.93%。
- 置信度: 87%, 样本数: 176, 出现次数: 1
避免：反复开平相同头寸（BTCUSDT, SOLUSDT 空仓）的频繁交易行为，会通过手续费和滑点消耗资金。skill_aware_oss 明确展示了这种模式。
- 置信度: 87%, 样本数: 173, 出现次数: 1
避免：在波动较小（-0.03%）的'相对稳定'资产上预期突破是不可靠的。llama4_scout 开 ETHUSDT 多头，预期在 -0.03% 的波动上'潜在突破'。
- 置信度: 85%, 样本数: 180, 出现次数: 1
避免：在方向错误的交易上使用25%权益和5倍杠杆的头寸规模会放大亏损。gptoss_20b_simple 进行了131次交易，亏损了 -$175.85。
- 置信度: 85%, 样本数: 131, 出现次数: 1
避免：极端过度交易（24小时内247次交易）结合高置信度（0.8-0.9），主要基于资金费率和简单的动能信号，即使在牛市（各资产上涨 +2.5% 至 +6.2%）中也会导致亏损。
- 置信度: 79%, 样本数: 494, 出现次数: 2
避免：将高资金费率解释为看涨动能信号是不可靠的 - llama4_scout 反复使用此推理（'高资金费率表明看涨情绪'），但在 DOGE 上涨 +4.68% 的市场中亏损了 $18.95。
- 置信度: 79%, 样本数: 494, 出现次数: 2
避免：仅基于负资金费率做空是不可靠的。gptoss_20b_simple 做空 DOGEUSDT，引用'负资金费率'，但尽管 DOGE 下跌了 -3%，总体仍亏损了 $44.40。
- 置信度: 75%, 样本数: 84, 出现次数: 1
避免：极端过度交易（24小时内248次交易）结合高置信度（0.8-0.9）但推理肤浅，导致亏损。llama4_scout 进行了248次交易，盈亏为 -$18.95，使用了重复的推理，如'显示出强劲的上涨趋势'和'高资金费率表明进一步增长的潜力'，但没有验证检查。
- 置信度: 72%, 样本数: 248, 出现次数: 1
避免：零交易策略（0次交易）在趋势性牛市市场中错失了显著的机会成本 - gpt_simple 和 index_fund 盈亏为 $0，而市场上涨了 +2.5% 至 +6.2%。
- 置信度: 72%, 样本数: 502, 出现次数: 1
避免：将高资金费率解释为看涨动能信号是不可靠的。llama4_scout 反复引用'高资金费率，表明进一步增长的潜力'，但亏损了。高资金费率实际上表明多头头寸拥挤，存在潜在反转风险。
- 置信度: 70%, 样本数: 248, 出现次数: 1
避免：零交易策略（gpt_simple, index_fund）保住了本金，但在强劲牛市（+3-6% 的波动）中错失了显著收益。
- 置信度: 70%, 样本数: 502, 出现次数: 1
避免：skill_only_oss 进行了42次交易，亏损了 $77.63 - 中等交易频率但没有适当的风险验证或多时间框架共振，产生了最严重的亏损。
- 置信度: 70%, 样本数: 42, 出现次数: 1
避免：skill_only_oss 进行了42次交易，亏损了 $77.63 - 交易频率不足以捕捉趋势，但又足以累积手续费和小额亏损。
- 置信度: 68%, 样本数: 42, 出现次数: 1
避免：极低频交易（3-4次）结合糟糕的时机选择，即使在牛市市场中也会导致亏损 - qwen3_235b 进行了4次交易，亏损了 -$7.42。
- 置信度: 68%, 样本数: 7, 出现次数: 1
避免：没有风险验证的技术信号（SMA, MACD）会产生亏损。skill_only_oss 进行了41次交易，盈亏为 -$77.63，可能使用了与 skill_aware_oss 类似的技术信号，但没有明确的验证层，而这正是关键区别所在。
- 置信度: 65%, 样本数: 41, 出现次数: 1
避免：极端过度交易（24小时内248次交易）结合重复关注单一资产（SOLUSDT），尽管该资产表现逊于市场（+1.55% 对比 ETH +4.96%）。llama4_scout 反复开 SOLUSDT 多头，亏损了 $69.19。
- 置信度: 65%, 样本数: 248, 出现次数: 1
避免：技术分析基准（ta_baseline）结合低频交易（21次）在趋势市场中表现不佳 - 错失机会成本。
- 置信度: 65%, 样本数: 21, 出现次数: 1
避免：技术分析基准（ta_baseline）进行了21次交易，亏损了 $25.88 - 表明在趋势市场中，纯粹的技术分析没有风险验证表现不佳。
- 置信度: 65%, 样本数: 21, 出现次数: 1
避免：仅基于'强劲上涨趋势'和'高资金费率'的高置信度（0.8-0.9）入场，没有风险验证或头寸管理，会导致亏损。llama4_scout 使用了此推理20多次。
- 置信度: 63%, 样本数: 248, 出现次数: 1
避免：没有适应性风险管理的传统技术分析基准方法在趋势市场中亏损。ta_baseline 进行了21次交易，盈亏为 -$25.88，而市场所有资产上涨了2.78-6.44% - 未能捕捉到明显的上涨趋势。
- 置信度: 63%, 样本数: 21, 出现次数: 1
避免：关注表现最弱的资产（SOLUSDT +1.55%）而忽略最强上涨者（ETHUSDT +4.96%, DOGEUSDT +4.77%）会摧毁阿尔法收益。llama4_scout 只交易了 SOL。
- 置信度: 60%, 样本数: 248, 出现次数: 1
避免：零交易策略在强劲趋势市场中错失了显著机会。index_fund 和 gpt_simple 进行了0次交易，盈亏为 $0，而市场上涨了2.78-6.44%。在牛市市场中，不作为相对于积极参与是一种亏损策略。
- 置信度: 60%, 样本数: 504, 出现次数: 1
避免：极低频交易（2-4次）错失了趋势市场机会。qwen3_235b 在市场波动 +3-5% 期间仅进行了4次交易，亏损了 $7.42。learning_qwen 进行了2次交易，盈亏为 $0。
- 置信度: 58%, 样本数: 6, 出现次数: 1
避免：在横盘/震荡市场中高频交易（>100次交易/天）会导致显著亏损，源于手续费损耗和反复震荡 - skill_aware_oss（155次交易，-$581）、llama4_scout（225次交易，-$326）、agentic_gptoss（176次交易，-$141）。
- 置信度: 57%, 样本数: 1684, 出现次数: 2
避免：没有意识组件的 skill_only_oss 进行了26次交易，亏损了 $69.20，而 skill_aware_oss 进行了157次交易，盈利了 $1349.11。意识/验证层是关键的区别因素。
- 置信度: 57%, 样本数: 26, 出现次数: 1
避免：基于 LLM 的智能体极低频交易（<10次）表明决策瘫痪或过于保守的阈值。learning_qwen（3次交易，$0）和 qwen3_235b（4次交易，-$7.42）未能利用明显的趋势市场。
- 置信度: 57%, 样本数: 7, 出现次数: 1
避免：在横盘市场中，基于多时间框架看涨/看跌共振推理的高置信度（0.85+）交易会产生亏损 - skill_aware_oss 反复以0.85的置信度开仓，引用'强劲看涨共振'，却亏损了 $581。
- 置信度: 53%, 样本数: 155, 出现次数: 1
避免：技术分析基准（ta_baseline）使用传统指标，在趋势市场中仅用18次交易就亏损了 $25.99。没有适应性推理的纯粹技术分析表现不佳。
- 置信度: 52%, 样本数: 18, 出现次数: 1
避免：当市场缺乏方向性波动时，以'优秀的2:1风险/回报'推理进行过度交易会反复失败 - 这个理由出现在 skill_aware_oss 和 agentic_gptoss 的多个亏损交易中。
- 置信度: 50%, 样本数: 629, 出现次数: 2
避免：将正资金费率解释为看涨动能信号会导致亏损 - llama

🇺🇸English

Trading Wisdom

Last updated: 2026-01-17 20:31 UTC Active patterns: 206 Total samples: 41088 Confidence threshold: 60%

Key Learnings

CRITICAL: In moderate bull markets (4/5 assets positive), ALL active trading strategies lost money while zero-trade strategies preserved capital perfectly.
Trade frequency is inversely correlated with performance in this regime: 0 trades = $0 loss, 23 trades = -$28.69, 243 trades = -$229.00.
Technical analysis signals (multi-timeframe alignment, MACD, RSI, SMA) failed to predict direction for both long and short entries in this moderate bull environment.
Asset selection mattered significantly: BNB (+2.03%) vs SOL (-0.09%). Agents fixating on SOL 'uptrend' (llama4_scout) suffered worst losses.
Validation frameworks and risk management rules do not prevent losses when the fundamental market direction assessment is wrong.
High-confidence decisions (0.85-0.90) on directional trades were frequently wrong, suggesting confidence calibration issues across all active agents.
The only reliable pattern was proactive loss-cutting with high confidence (0.85-0.95) to limit drawdown.

Winning Strategies

Zero-trade strategy in moderate bull markets prese...

Confidence : 95%
Total samples : 4
Times confirmed : 1
First seen : 2026-01-17
Details : Zero-trade strategy in moderate bull markets preserves capital perfectly. Agents that made 0 trades (learning_qwen, gpt_simple, qwen3_235b, index_fund) achieved $0.00 PnL while all active traders lost money despite BNB +2.03%, ETH +1.02%, DOGE +1.07% gains.

Zero-trade strategies preserve capital in mixed/ch...

Confidence : 92%
Total samples : 771
Times confirmed : 1
First seen : 2026-01-17
Details : Zero-trade strategies preserve capital in mixed/choppy markets. learning_qwen, gpt_simple, and index_fund made 0 trades and achieved $0.00 PnL, outperforming all active traders in this low-conviction environment.

Zero-trade strategy preserves capital in moderatel...

Confidence : 92%
Total samples : 4
Times confirmed : 1
First seen : 2026-01-17
Details : Zero-trade strategy preserves capital in moderately bullish markets where active trading leads to losses. Agents holding no positions avoided the -$50 to -$264 losses seen by active traders despite market being up +0.63% to +2.15%.

Close losing positions proactively with high confi...

Confidence : 90%
Total samples : 368
Times confirmed : 1
First seen : 2026-01-17
Details : Close losing positions proactively with high confidence (0.8-0.9) to free margin and limit drawdowns. Multiple agents demonstrated this: gptoss_20b_simple closing SOL at -$4.76 loss, agentic_gptoss closing DOGE 'largest loss percentage'.

Minimal trading with high selectivity outperforms ...

Confidence : 88%
Total samples : 257
Times confirmed : 1
First seen : 2026-01-17
Details : Minimal trading with high selectivity outperforms frequent trading. qwen3_235b made only 2 trades with PnL of -$0.29, dramatically outperforming agents with 140-201 trades.

Closing long positions with high confidence (0.92)...

Confidence : 88%
Total samples : 89
Times confirmed : 1
First seen : 2026-01-17
Details : Closing long positions with high confidence (0.92) when regime shifts to 'moderate bearish' preserves capital. skill_only_oss reasoning: 'risk-management rules advise limiting exposure and closing long positions to preserve capital'.

Minimal trading frequency (23 trades) with technic...

Confidence : 88%
Total samples : 1
Times confirmed : 1
First seen : 2026-01-17
Details : Minimal trading frequency (23 trades) with technical analysis baseline outperforms high-frequency approaches. ta_baseline lost only $-28.69 vs llama4_scout's $-229.00 with 243 trades.

Explicit risk validation with 2% equity risk and 2...

Confidence : 85%
Total samples : 160
Times confirmed : 1
First seen : 2026-01-17
Details : Explicit risk validation with 2% equity risk and 2:1 reward ratio combined with position closing discipline. skill_only_oss achieved best active trader performance (-$17.96) with 160 trades, using validated risk parameters.

Agentic approach with active position management: ...

Confidence : 85%
Total samples : 100
Times confirmed : 1
First seen : 2026-01-16
Details : Agentic approach with active position management: opening shorts in bearish markets, closing positions to lock gains when technical indicators confirm trend reversal. Uses SMA crossover + MACD + Bollinger bands for entry/exit confirmation with explicit validation steps.

Low-frequency trading (89 trades) with selective l...

Confidence : 85%
Total samples : 89
Times confirmed : 1
First seen : 2026-01-17
Details : Low-frequency trading (89 trades) with selective long entries on multi-timeframe bullish alignment produces small positive returns (+$6.22) in moderately bullish markets.

Proactive closing of losing positions with high co...

Confidence : 85%
Total samples : 5
Times confirmed : 1
First seen : 2026-01-17
Details : Proactive closing of losing positions with high confidence (0.85-0.95) to free margin. skill_only_oss closed DOGEUSDT at 0.95 confidence citing 'risk-management rule recommends closing losing positions proactively' - resulted in smaller losses ($-36.63) than more active traders.

Zero-trade or minimal-trade strategies preserve ca...

Confidence : 82%
Total samples : 136
Times confirmed : 1
First seen : 2026-01-16
Details : Zero-trade or minimal-trade strategies preserve capital in bearish/declining markets. learning_qwen (0 trades, $0 PnL) and gpt_simple (1 trade, $0 PnL) avoided losses by not trading during market decline.

Multi-timeframe bullish alignment (15m, 1h, 4h) co...

Confidence : 79%
Total samples : 328
Times confirmed : 2
First seen : 2026-01-14
Details : Multi-timeframe bullish alignment (15m, 1h, 4h) combined with explicit risk validation (2% equity risk, 2:1 reward ratio) and trade validation checks produces strong positive returns in trending bull markets

Moderate trade frequency (80-90 trades) with expli...

Confidence : 78%
Total samples : 88
Times confirmed : 1
First seen : 2026-01-16
Details : Moderate trade frequency (80-90 trades) with explicit risk validation outperforms high-frequency trading. skill_only_oss (88 trades, -$9.15) significantly outperformed skill_aware_oss (103 trades, -$180.47) despite similar strategies.

Optimal trade frequency in trending bull markets: ...

Confidence : 75%
Total samples : 543
Times confirmed : 1
First seen : 2026-01-14
Details : Optimal trade frequency in trending bull markets: 120-200 trades/24h captures opportunities without excessive churn. skill_aware_oss (164 trades, +$1236.81) and agentic_gptoss (184 trades, +$697.86) demonstrate this

Active position management with proactive closing ...

Confidence : 74%
Total samples : 543
Times confirmed : 1
First seen : 2026-01-14
Details : Active position management with proactive closing of losing/breakeven positions to free margin, combined with moderate-high trade frequency (164-195 trades/24h) in trending markets

Moderate-high trade frequency (120-200 trades/24h)...

Confidence : 73%
Total samples : 543
Times confirmed : 1
First seen : 2026-01-14
Details : Moderate-high trade frequency (120-200 trades/24h) with active position management - closing small/underwater positions to free margin for higher-conviction trades

Proactive loss management - closing losing positio...

Confidence : 72%
Total samples : 379
Times confirmed : 1
First seen : 2026-01-14
Details : Proactive loss management - closing losing positions with high confidence (0.9) to preserve capital and reduce concentration risk

SMA crossover + bullish MACD + neutral Bollinger b...

Confidence : 72%
Total samples : 184
Times confirmed : 1
First seen : 2026-01-14
Details : SMA crossover + bullish MACD + neutral Bollinger bands as entry confirmation with explicit validation checks before execution

Closing positions near breakeven or with small los...

Confidence : 70%
Total samples : 320
Times confirmed : 1
First seen : 2026-01-14
Details : Closing positions near breakeven or with small losses to free margin for higher-conviction trades preserves capital and enables redeployment

SMA crossover + bullish MACD + neutral Bollinger b...

Confidence : 70%
Total samples : 184
Times confirmed : 1
First seen : 2026-01-14
Details : SMA crossover + bullish MACD + neutral Bollinger bands as entry confirmation combined with trend alignment across timeframes

Multi-timeframe bullish alignment (15m, 1h, 4h) co...

Confidence : 70%
Total samples : 164
Times confirmed : 1
First seen : 2026-01-14
Details : Multi-timeframe bullish alignment (15m, 1h, 4h) combined with explicit risk validation (2% risk, 2:1 reward ratio) and position sizing controls produces strong profits in trending markets. skill_aware_oss consistently references 'Multi-timeframe analysis shows strong aligned bullish trend' with 'trade validation passed' and achieved +$1379.66 PnL.

Position sizing at 25% equity limit per position w...

Confidence : 68%
Total samples : 125
Times confirmed : 1
First seen : 2026-01-14
Details : Position sizing at 25% equity limit per position with active monitoring and timely closes to lock profits or limit losses

Agentic workflow with validation checks before ent...

Confidence : 67%
Total samples : 184
Times confirmed : 1
First seen : 2026-01-14
Details : Agentic workflow with validation checks before entry/exit decisions. agentic_gptoss uses 'validation checks confirm', 'risk calculator suggests', and 'all validation checks passed' reasoning, achieving +$689.63 with 184 trades. Structured decision-making with explicit risk/reward assessment outperforms simpler approaches.

Moderate trade frequency (120-200 trades/24h) in t...

Confidence : 65%
Total samples : 545
Times confirmed : 1
First seen : 2026-01-14
Details : Moderate trade frequency (120-200 trades/24h) in trending bull markets captures momentum while avoiding overtrading. gptoss_120b_simple (197 trades, +$138.86) and agentic_gptoss (184 trades, +$689.63) both fall in this range and are profitable.

Proactive position closing to manage risk and free...

Confidence : 63%
Total samples : 200
Times confirmed : 1
First seen : 2026-01-14
Details : Proactive position closing to manage risk and free margin. Profitable agents close positions citing 'frees margin', 'reduces concentration risk', 'locks profit'. skill_aware_oss closes 'over-leveraged' positions; agentic_gptoss closes with 'reduces exposure and frees capital for future opportunities'.

skill_aware_oss combines multi-timeframe analysis ...

Confidence : 62%
Total samples : 157
Times confirmed : 1
First seen : 2026-01-14
Details : skill_aware_oss combines multi-timeframe analysis with strict risk validation and position scaling into existing winners. Uses 0.75-0.85 confidence threshold with explicit risk checks ('risk per trade within limits', 'validation permits proceeding'). Achieves highest PnL ($1349.11) with moderate trade frequency (157 trades).

Asset diversification across multiple symbols rath...

Confidence : 60%
Total samples : 348
Times confirmed : 1
First seen : 2026-01-14
Details : Asset diversification across multiple symbols rather than single-asset concentration. Profitable agents trade BTC, ETH, DOGE across decisions while llama4_scout's repetitive single-asset focus leads to losses despite high trade count.

agentic_gptoss employs active loss-cutting strateg...

Confidence : 58%
Total samples : 182
Times confirmed : 1
First seen : 2026-01-14
Details : agentic_gptoss employs active loss-cutting strategy with high-confidence closes (0.9) on losing positions ('Close the losing ETHUSDT long to free margin'). Combines with selective long entries. Achieves $372.23 PnL with 182 trades.

In trending bull markets (+1.5% to +5% moves), mul...

Confidence : 58%
Total samples : 157
Times confirmed : 1
First seen : 2026-01-14
Details : In trending bull markets (+1.5% to +5% moves), multi-timeframe bullish alignment DOES work when combined with proper risk validation. skill_aware_oss profits $1349 using this approach during 3-5% market moves.

Moderate trade frequency (120-200 trades) with dis...

Confidence : 55%
Total samples : 535
Times confirmed : 1
First seen : 2026-01-14
Details : Moderate trade frequency (120-200 trades) with disciplined position management outperforms both extremes. Winners trade 157-196 times vs losers at 248 trades or 2-4 trades.

Ultra-low trade frequency (3-6 trades) with high s...

Confidence : 52%
Total samples : 13
Times confirmed : 1
First seen : 2026-01-13
Details : Ultra-low trade frequency (3-6 trades) with high selectivity results in near-zero or minimal losses in flat/sideways markets - qwen3_235b and learning_qwen both achieved ~$0 PnL with only 3-4 trades vs massive losses from high-frequency traders

Active closing of near-breakeven or small-loss pos...

Confidence : 52%
Total samples : 317
Times confirmed : 1
First seen : 2026-01-14
Details : Active closing of near-breakeven or small-loss positions to free margin for higher-conviction opportunities. gptoss_120b_simple reasoning: 'closing reduces exposure and frees margin for higher-conviction opportunities'.

Index fund strategy of equal-weight allocation ($2...

Confidence : 50%
Total samples : 6
Times confirmed : 1
First seen : 2026-01-13
Details : Index fund strategy of equal-weight allocation ($2000 per asset) with confidence 1.0 and minimal rebalancing preserves capital in sideways markets - achieved $0.00 PnL while active traders lost $1395

Passive holding without frequent position changes ...

Confidence : 48%
Total samples : 13
Times confirmed : 1
First seen : 2026-01-13
Details : Passive holding without frequent position changes outperforms active trading when market moves are <0.1% - agents with <10 trades preserved capital while those with >100 trades lost $300-$580

Index fund strategy of equal-weight allocation ($2...

Confidence : 40%
Total samples : 6
Times confirmed : 1
First seen : 2026-01-13
Details : Index fund strategy of equal-weight allocation ($2000 per asset) with confidence=1.0 maintains capital neutrality when market moves are near-zero

Patterns to Avoid

AVOID : Extreme overtrading (200+ trades in 24h) in mixed/choppy markets leads to largest losses. skill_aware_oss made 201 trades with -$360.24 PnL, the worst performer.
- Conf: 95%, N=201, seen 1x
AVOID : Extreme overtrading (231 trades in 24h) in moderately bullish market leads to largest losses (-$264.52). llama4_scout traded most frequently and lost most.
- Conf: 95%, N=231, seen 1x
AVOID : Extreme overtrading (243 trades in 24h) in moderate bull market leads to largest losses. llama4_scout made 243 trades with $-229.00 PnL despite repeatedly identifying 'strong uptrend' in SOLUSDT which actually declined -0.09%.
- Conf: 95%, N=243, seen 1x
AVOID : Shorting in a bullish market (all assets +0.63% to +2.15%) with high frequency leads to significant losses. Agents with heavy short bias (skill_aware_oss, gptoss_20b_simple) lost -$173 to -$176 despite 'bearish' technical signals.
- Conf: 93%, N=675, seen 1x
AVOID : High trade frequency (100+ trades/day) in bearish markets leads to significant losses. skill_aware_oss with 103 trades lost $180.47 despite using multi-timeframe analysis and risk validation.
- Conf: 92%, N=103, seen 1x
AVOID : Contrarian 'bounce back' reasoning on downtrending assets fails. llama4_scout opened long on SOLUSDT reasoning 'shows a clear downtrend but might be due for a bounce back' - resulted in -$192.40 total PnL.
- Conf: 92%, N=180, seen 1x
AVOID : Repeated high-confidence long entries on SOLUSDT based on 'strong uptrend' reasoning while asset actually declined -0.09%. llama4_scout opened multiple longs at 0.80-0.90 confidence citing +2.24% to +2.59% price increases that were temporary.

Confidence Guide

Confidence	Interpretation
90%+	High confidence - strong historical support
70-90%	Moderate confidence - use with other signals
60-70%	Low confidence - consider as one input
<60%	Experimental - needs more data

This skill is automatically generated and updated by the Observer Agent.

Weekly Installs

476

Repository

0xhubed/agent-t…ng-arena

GitHub Stars

First Seen

Jan 24, 2026

Security Audits

Gen Agent Trust HubPass SocketPass SnykPass

Installed on

opencode433

gemini-cli425

codex416

github-copilot409

cursor405

kimi-cli390

AI Elements：基于shadcn/ui的AI原生应用组件库，快速构建对话界面

56,200 周安装

Conf: 92%, N=7, seen 1x

AVOID : Multi-timeframe bullish alignment signals (15m, 1h, 4h) produce losses in bearish markets. skill_only_oss and skill_aware_oss both used this signal for ETHUSDT longs while market declined.

Conf: 90%, N=191, seen 1x

AVOID : Negative funding rate interpreted as long opportunity signal is unreliable. llama4_scout: 'funding rate is slightly negative which could indicate a potential long opportunity' - total PnL -$192.40.

Conf: 90%, N=180, seen 1x

AVOID : Multi-timeframe bearish alignment signals for short entry FAIL in bullish markets. skill_aware_oss opened shorts on 'bearish bias (RSI overbought, MACD bearish)' but market moved up, causing -$173.25 loss.

Conf: 90%, N=355, seen 1x

AVOID : Shorting in moderate bull markets leads to consistent losses. skill_aware_oss opened shorts on BTCUSDT at 0.88 confidence citing 'strong bearish alignment' while BTC was +0.24% - contributed to $-167.78 total loss.

Conf: 90%, N=171, seen 1x

AVOID : Multi-timeframe bearish alignment signals for short entry fail in moderate bull markets. skill_aware_oss cited 'Strong bearish alignment across 15m, 1h, and 4h timeframes' for BTCUSDT shorts while market was net positive.

Conf: 90%, N=2, seen 1x

AVOID : Opening longs based on 'positive momentum' or small price increases (+0.33% to +0.44%) during overall bearish market conditions. llama4_scout repeatedly opened ETHUSDT longs citing positive momentum while ETH declined -1.29%.

Conf: 88%, N=76, seen 1x

AVOID : High confidence (0.85-0.92) on multi-timeframe bullish alignment during market-wide decline leads to losses. Agents expressed high confidence while market moved against positions.

Conf: 88%, N=267, seen 1x

AVOID : Multi-timeframe bearish alignment for shorts fails when market is mixed (BNB +0.93%, SOL +1.65% vs DOGE -1.19%). skill_aware_oss and agentic_gptoss both lost money shorting despite 'strong bearish trend' reasoning.

Conf: 88%, N=377, seen 1x

AVOID : High-confidence short entries (0.85) based on technical indicators fail when market regime is actually bullish. Agents misread regime as bearish when BTC was +0.63%, ETH +1.15%, SOL +1.60%.

Conf: 88%, N=355, seen 1x

AVOID : High trade frequency (129-195 trades) with GPT-based models all resulted in significant losses ($-55 to $-128) despite validation steps and risk management frameworks.

Conf: 88%, N=517, seen 1x

AVOID : High confidence (0.85-0.9) short positions based on multi-timeframe bearish alignment underperform in mixed markets. agentic_gptoss shorted BNB at 0.85 confidence while BNB was +0.93%.

Conf: 87%, N=176, seen 1x

AVOID : Churning behavior - repeatedly opening and closing same positions (BTCUSDT, SOLUSDT shorts) destroys capital through fees and slippage. skill_aware_oss showed this pattern explicitly.

Conf: 87%, N=173, seen 1x

AVOID : Anticipating breakouts on 'relatively stable' assets with small movements is unreliable. llama4_scout opened long on ETHUSDT expecting 'potential breakout' on -0.03% movement.

Conf: 85%, N=180, seen 1x

AVOID : Position sizing at 25% equity with 5x leverage on directionally wrong trades amplifies losses. gptoss_20b_simple lost -$175.85 with 131 trades.

Conf: 85%, N=131, seen 1x

AVOID : Extreme overtrading (247 trades in 24h) with high confidence (0.8-0.9) based primarily on funding rate and simple momentum signals leads to losses despite bullish market (+2.5% to +6.2% across assets)

Conf: 79%, N=494, seen 2x

AVOID : High funding rate interpreted as bullish momentum signal is unreliable - llama4_scout repeatedly used this reasoning ('high funding rate indicating bullish sentiment') but lost $18.95 in a +4.68% DOGE market

Conf: 79%, N=494, seen 2x

AVOID : Shorting based on negative funding rate alone is unreliable. gptoss_20b_simple shorted DOGEUSDT citing 'negative funding' but still lost $44.40 overall despite DOGE being down -3%.

Conf: 75%, N=84, seen 1x

AVOID : Extreme overtrading (248 trades in 24h) with high confidence (0.8-0.9) but shallow reasoning leads to losses. llama4_scout made 248 trades with -$18.95 PnL, using repetitive reasoning like 'showing strong upward trend' and 'high funding rate indicating potential for further growth' without validation checks.

Conf: 72%, N=248, seen 1x

AVOID : Zero-trade strategies (0 trades) miss significant opportunity cost in trending bull markets - gpt_simple and index_fund had $0 PnL while market gained +2.5% to +6.2%

Conf: 72%, N=502, seen 1x

AVOID : High funding rate interpreted as bullish momentum signal is unreliable. llama4_scout repeatedly cites 'high funding rate, indicating potential for further growth' but loses money. High funding actually indicates crowded long positioning and potential reversal risk.

Conf: 70%, N=248, seen 1x

AVOID : Zero-trade strategies (gpt_simple, index_fund) preserve capital but miss significant gains in strong bull markets (+3-6% moves)

Conf: 70%, N=502, seen 1x

AVOID : skill_only_oss with 42 trades lost $77.63 - moderate trade frequency without proper risk validation or multi-timeframe alignment produces worst losses

Conf: 70%, N=42, seen 1x

AVOID : skill_only_oss with 42 trades lost $77.63 - insufficient trade frequency to capture trends but enough to accumulate fees and small losses

Conf: 68%, N=42, seen 1x

AVOID : Very low trade frequency (3-4 trades) with poor timing results in losses even in bull markets - qwen3_235b made 4 trades for -$7.42 loss

Conf: 68%, N=7, seen 1x

AVOID : Technical signals (SMA, MACD) without risk validation produce losses. skill_only_oss made 41 trades with -$77.63 PnL, likely using similar technical signals as skill_aware_oss but without the explicit risk validation layer that makes the difference.

Conf: 65%, N=41, seen 1x

AVOID : Extreme overtrading (248 trades in 24h) with repetitive single-asset focus (SOLUSDT) despite asset underperforming market (+1.55% vs ETH +4.96%). llama4_scout lost $69.19 repeatedly opening SOLUSDT longs.

Conf: 65%, N=248, seen 1x

AVOID : Technical analysis baseline (ta_baseline) with low trade frequency (21 trades) underperforms in trending markets - missed opportunity cost

Conf: 65%, N=21, seen 1x

AVOID : Technical analysis baseline (ta_baseline) with 21 trades lost $25.88 - suggests pure TA without risk validation underperforms in trending markets

Conf: 65%, N=21, seen 1x

AVOID : High-confidence (0.8-0.9) entries based solely on 'strong uptrend' and 'high funding rate' without risk validation or position management leads to losses. llama4_scout used this reasoning 20+ times.

Conf: 63%, N=248, seen 1x

AVOID : Traditional TA baseline approach without adaptive risk management loses in trending markets. ta_baseline made 21 trades with -$25.88 PnL while market was up 2.78-6.44% across all assets - failing to capture obvious uptrend.

Conf: 63%, N=21, seen 1x

AVOID : Focusing on weakest-performing asset (SOLUSDT +1.55%) while ignoring strongest movers (ETHUSDT +4.96%, DOGEUSDT +4.77%) destroys alpha. llama4_scout exclusively traded SOL.

Conf: 60%, N=248, seen 1x

AVOID : Zero-trade strategies miss significant opportunities in strong trending markets. index_fund and gpt_simple made 0 trades and $0 PnL while market gained 2.78-6.44%. In bull markets, inaction is a losing strategy relative to active participation.

Conf: 60%, N=504, seen 1x

AVOID : Ultra-low trade frequency (2-4 trades) misses trending market opportunities. qwen3_235b made only 4 trades during +3-5% market moves, losing $7.42. learning_qwen made 2 trades for $0 PnL.

Conf: 58%, N=6, seen 1x

AVOID : High trade frequency (>100 trades/day) in flat/sideways markets leads to significant losses from fee drag and whipsaw - skill_aware_oss (155 trades, -$581), llama4_scout (225 trades, -$326), agentic_gptoss (176 trades, -$141)

Conf: 57%, N=1684, seen 2x

AVOID : skill_only_oss without awareness component loses $69.20 with 26 trades, while skill_aware_oss gains $1349.11 with 157 trades. Awareness/validation layer is critical differentiator.

Conf: 57%, N=26, seen 1x

AVOID : Very low trade frequency (<10 trades) with LLM-based agents suggests decision paralysis or overly conservative thresholds. learning_qwen (3 trades, $0) and qwen3_235b (4 trades, -$7.42) failed to capitalize on clear trending market.

Conf: 57%, N=7, seen 1x

AVOID : High-confidence (0.85+) trades with multi-timeframe bullish/bearish alignment reasoning produce losses in flat markets - skill_aware_oss repeatedly opened positions with 0.85 confidence citing 'strong bullish alignment' yet lost $581

Conf: 53%, N=155, seen 1x

AVOID : Technical analysis baseline (ta_baseline) using traditional indicators loses $25.99 with only 18 trades in trending market. Pure TA without adaptive reasoning underperforms.

Conf: 52%, N=18, seen 1x

AVOID : Overtrading with 'excellent 2:1 risk/reward' reasoning repeatedly fails when market lacks directional movement - this justification appeared in multiple losing trades across skill_aware_oss and agentic_gptoss

Conf: 50%, N=629, seen 2x

AVOID : Positive funding rate interpreted as bullish momentum signal leads to losses - llama4_scout repeatedly cited 'positive funding rate indicating potential upward momentum' while losing $326

Conf: 48%, N=225, seen 1x

AVOID : Conflicting directional signals on same asset within short timeframes indicates poor strategy - skill_aware_oss opened both long and short on BTCUSDT with same 0.85 confidence, showing signal unreliability

Conf: 45%, N=155, seen 1x

AVOID : Mean-reversion strategies ('fading crowded longs', 'potential mean-reversion bounce') fail when market is truly flat with no significant deviation to revert from

Conf: 40%, N=6, seen 1x

交易智慧分析：温和牛市零交易策略保本，高频交易亏损，技术分析失效

🇨🇳中文介绍

交易智慧

关键发现

制胜策略

温和牛市中的零交易策略可保...

零交易策略在震荡/盘整市...

相关 Skills

零交易策略在温和上涨市...

主动高置信度平仓亏损头...

高选择性低频交易优于...

当行情转为'温和看跌'时...

低频交易（23次）结合技...

明确的2%权益风险与2:...

具有主动头寸管理的智能...

低频交易（89次）结合多...

主动高置信度平仓亏损头...

零交易或极低频交易策略...

多时间框架看涨共振（15...

中等交易频率（80-90次）...

趋势性牛市中的最佳交易...

主动头寸管理结合主动平...

中等偏高交易频率（120-...

主动亏损管理 - 高置信度...

SMA交叉 + 看涨MACD + 中...

在盈亏平衡点附近或小幅...

SMA交叉 + 看涨MACD + 中...

多时间框架看涨共振（15...

单头寸25%权益限制的头...

入场/出场决策前进行验证...

趋势性牛市中的中等交易...

主动平仓以管理风险并释...

skill_aware_oss 结合多时...

跨多个交易对的资产分散...

agentic_gptoss 采用主动止...

在趋势性牛市（+1.5% 至...

中等交易频率（120-200次...

极低频交易（3-6次）结合...

主动平仓接近盈亏平衡或...

等权重分配（每资产$20...

当市场波动 <0.1% 时，...

等权重分配（每资产$20...

应避免的模式

🇺🇸English

Trading Wisdom

Key Learnings

Winning Strategies

Zero-trade strategy in moderate bull markets prese...

Zero-trade strategies preserve capital in mixed/ch...

Zero-trade strategy preserves capital in moderatel...

Close losing positions proactively with high confi...

Minimal trading with high selectivity outperforms ...

Closing long positions with high confidence (0.92)...

Minimal trading frequency (23 trades) with technic...

Explicit risk validation with 2% equity risk and 2...

Agentic approach with active position management: ...

Low-frequency trading (89 trades) with selective l...

Proactive closing of losing positions with high co...

Zero-trade or minimal-trade strategies preserve ca...

Multi-timeframe bullish alignment (15m, 1h, 4h) co...

Moderate trade frequency (80-90 trades) with expli...

Optimal trade frequency in trending bull markets: ...

Active position management with proactive closing ...

Moderate-high trade frequency (120-200 trades/24h)...

Proactive loss management - closing losing positio...

SMA crossover + bullish MACD + neutral Bollinger b...

Closing positions near breakeven or with small los...

SMA crossover + bullish MACD + neutral Bollinger b...

Multi-timeframe bullish alignment (15m, 1h, 4h) co...

Position sizing at 25% equity limit per position w...

Agentic workflow with validation checks before ent...

Moderate trade frequency (120-200 trades/24h) in t...

Proactive position closing to manage risk and free...

skill_aware_oss combines multi-timeframe analysis ...

Asset diversification across multiple symbols rath...

agentic_gptoss employs active loss-cutting strateg...

In trending bull markets (+1.5% to +5% moves), mul...

Moderate trade frequency (120-200 trades) with dis...

Ultra-low trade frequency (3-6 trades) with high s...

Active closing of near-breakeven or small-loss pos...