变化显著性追踪器

并非所有排名变化都有意义，有些只是随机噪音。此页面使用统计分析来判断哪些模型评分变动是真实趋势，哪些只是正常波动。

已分析模型

290

显著变化

噪音 (不显著)

272

双时间维度

这意味着什么

18 / 290 个模型的评分变化具有统计显著性--这些是真实的性能变化，不是随机噪音。
272 个模型的评分变化在正常波动范围内--不要过度解读这些模型的小幅排名变化。
2 个模型在每日和每周两个时间维度上都显示显著变化--这些是最强、最可靠的真实性能变化信号。
使用方法：当模型排名发生变化时，先查看此页面。如果未标记为显著，则变化很可能是暂时的噪音。如果确实显著，则模型正在真正地改善或下降。

最显著变化 (按|Z分数|)

LMMarketCap.com

真实性能变化

18 个模型的近期评分偏离其历史平均值足以被认为是真实变化（非噪音）。按变化极端程度排序。Z分数衡量变化的异常程度--超过±1.96表示有95%的概率该变化是真实的。

模型	提供商	当前	基准	偏差	Z分数	方向	置信度
Qwen3 30B A3B Instruct 2507Alibaba	Alibaba	64.1	39.7	+24.39	5.04	提升	±9.48
Qwen3.5 Plus 2026-04-20Alibaba	Alibaba	74.8	41.6	+33.22	4.58	提升	±14.21
GLM 5.1Zhipu AI	Zhipu AI	78.0	75.9	+2.08	4.29	提升	±0.95
Qwen3.7 MaxAlibaba	Alibaba	74.8	41.8	+32.97	4.24	提升	±15.23
Claude Sonnet 5Anthropic	Anthropic	85.2	43.5	+41.72	3.46	提升	±23.61
gpt-oss-120bOpenAI	OpenAI	61.6	42.0	+19.65	3.02	提升	±12.77
Claude Opus 5Anthropic	Anthropic	95.1	45.5	+49.59	3.00	提升	±32.40
Kimi K2.5Moonshot AI	Moonshot AI	47.5	57.7	-10.17	-2.99	下降	±6.68
Claude Fable 5Anthropic	Anthropic	95.9	96.8	-0.92	-2.68	下降	±0.67
Grok 4.3xAI	xAI	88.8	82.0	+6.83	2.65	提升	±5.06
Muse Spark 1.2meta	meta	80.9	45.1	+35.79	2.65	提升	±26.51
Grok 4.5xAI	xAI	88.8	79.2	+9.60	2.64	提升	±7.11
Grok 4.6xAI	xAI	88.8	88.4	+0.43	2.45	提升	±0.34
Claude Fable 5 (batch)Anthropic	Anthropic	95.9	96.9	-1.03	-2.45	下降	±0.82
GPT-4o (2024-11-20)OpenAI	OpenAI	71.2	48.9	+22.31	2.41	提升	±18.11
GPT-5 MiniOpenAI	OpenAI	53.8	61.4	-7.62	-2.10	下降	±7.12
GPT-5 NanoOpenAI	OpenAI	61.9	50.2	+11.75	2.01	提升	±11.47
Kimi K2.7 CodeMoonshot AI	Moonshot AI	40.0	51.1	-11.12	-2.00	下降	±10.90

短期变化 vs. 持续变化

24小时内的排名变化可能只是暂时的。但如果7天内也在变动，那就是真正的趋势。在两个时间维度上都被标记的模型最值得关注--它们代表已确认的持续性能变化。

在两个时间维度上均显著(最强信号)

模型	提供商	评分	24小时变化	7天变化
Claude Sonnet 4Anthropic	Anthropic	73.9	-5	-15
Claude Opus 4.1Anthropic	Anthropic	74.4	-61	-11

仅每日(可能是噪音)

没有仅在每日维度上显著的模型。

仅每周(正在形成趋势)

模型	提供商	评分	24小时变化	7天变化
Claude Opus 5Anthropic	Anthropic	95.1	0	+273
GPT-5.5 ProOpenAI	OpenAI	92.7	0	+9
GPT-5.3-CodexOpenAI	OpenAI	90.5	0	+19
GPT-5.2 ChatOpenAI	OpenAI	90.5	0	+81
Grok 4.5xAI	xAI	88.8	-1	+57
Grok 4.3xAI	xAI	88.8	-1	+31
Grok 4.20xAI	xAI	88.8	-1	-7
GPT-5 ProOpenAI	OpenAI	88.7	-1	-9
GPT-5 Pro (batch)OpenAI	OpenAI	88.7	-1	-9
GPT-5OpenAI	OpenAI	88.7	-1	-9
GPT-5 (batch)OpenAI	OpenAI	88.7	-1	-9
Gemini 3 Flash PreviewGoogle	Google	88.4	-1	-9
Gemini 3 Flash Preview (batch)Google	Google	88.4	-1	-9
Grok 4.20 Multi-AgentxAI	xAI	87.9	-1	-8
GPT-5.1 (batch)OpenAI	OpenAI	87.8	-1	-6
Claude Sonnet 5Anthropic	Anthropic	85.2	-1	+232
Claude Sonnet 4.6Anthropic	Anthropic	85.2	-1	-6
Claude Sonnet 4.6 (batch)Anthropic	Anthropic	85.2	-1	-6
Claude Opus 4.5Anthropic	Anthropic	85.1	-1	-6
Claude Opus 4.5 (batch)Anthropic	Anthropic	85.1	-1	-6
Gemini 2.5 ProGoogle	Google	83.5	-1	-6
Gemini 2.5 Pro (batch)Google	Google	83.5	-1	-6
Gemini 2.5 Pro Preview 06-05Google	Google	83.5	-1	-6
DeepSeek V3.2DeepSeek	DeepSeek	83.4	-1	-6
Claude Sonnet 4.5Anthropic	Anthropic	82.4	-1	-6
Claude Sonnet 4.5 (batch)Anthropic	Anthropic	82.4	-1	-6
Muse Spark 1.3meta	meta	80.9	+1	+162
Muse Spark 1.2meta	meta	80.9	+1	+190
Muse Spark 1.1meta	meta	80.9	+1	-6
Gemma 4 31BGoogle	Google	80.5	+1	-6
Gemma 4 31B (free)Google	Google	80.5	+1	-6
Qwen3.5 397B A17BAlibaba	Alibaba	79.4	+1	-6
R1 0528DeepSeek	DeepSeek	79.4	+1	-6
GPT-5.4 NanoOpenAI	OpenAI	79.3	+1	-6
GPT-5.4 Nano (batch)OpenAI	OpenAI	79.3	+1	-6
GPT-5.4 MiniOpenAI	OpenAI	79.3	+1	-6
GPT-5.4 Mini (batch)OpenAI	OpenAI	79.3	+1	-6
Gemini 3.1 Flash Lite PreviewGoogle	Google	79.3	+1	-6
Gemini 2.5 Flash LiteGoogle	Google	79.1	+1	-6
Gemini 2.5 Flash Lite (batch)Google	Google	79.1	+1	-6
Gemini 2.5 FlashGoogle	Google	79.1	+1	-6
Gemini 2.5 Flash (batch)Google	Google	79.1	+1	-6
Gemini 3.5 FlashGoogle	Google	79.0	+1	-6
Gemini 3.5 Flash (batch)Google	Google	79.0	+1	-6
GLM 5.3 (batch)Zhipu AI	Zhipu AI	78.6	+1	-7
GLM 5.2Zhipu AI	Zhipu AI	78.6	+1	-7
GLM 5.2 (batch)Zhipu AI	Zhipu AI	78.6	+1	-7
GLM 5.3 FlashZhipu AI	Zhipu AI	78.0	+1	-7
GLM 5.1Zhipu AI	Zhipu AI	78.0	+1	+11
GLM 5 TurboZhipu AI	Zhipu AI	78.0	+1	+43
GLM 5Zhipu AI	Zhipu AI	78.0	+1	-9
GLM 5.3 Flash (batch)Zhipu AI	Zhipu AI	77.9	+1	-9
Qwen3.5-122B-A10BAlibaba	Alibaba	77.7	+1	-8
Gemma 2 27BGoogle	Google	77.4	+1	-8
Qwen3.5-27BAlibaba	Alibaba	77.0	+1	-6
GPT-5 Mini (batch)OpenAI	OpenAI	76.8	+1	-6
GPT-5 Nano (batch)OpenAI	OpenAI	76.8	+1	-6
Gemini 3.5 Flash LiteGoogle	Google	76.5	+1	-6
Gemini 3.5 Flash Lite (batch)Google	Google	76.5	+1	-6
MiMo-V2.5-ProXiaomi	Xiaomi	76.2	0	-7
Qwen3.5-35B-A3BAlibaba	Alibaba	76.0	0	-7
GLM 5.2 (free)Zhipu AI	Zhipu AI	75.7	0	-7
Kimi K2.6Moonshot AI	Moonshot AI	75.7	0	-7
Qwen3.8 Max (0902)Alibaba	Alibaba	75.6	0	+120
Qwen3.7 PlusAlibaba	Alibaba	75.6	0	-8
o3 MiniOpenAI	OpenAI	75.3	0	-7
Claude Opus 4.1 (batch)Anthropic	Anthropic	75.2	0	-7
GLM 4.7Zhipu AI	Zhipu AI	75.1	0	+13
GLM 4.6Zhipu AI	Zhipu AI	75.1	0	+27
GLM 4.5Zhipu AI	Zhipu AI	75.1	0	-9
Qwen3.7 MaxAlibaba	Alibaba	74.8	0	+176
Qwen3.5 Plus 2026-04-20Alibaba	Alibaba	74.8	0	+187
Qwen3.6 Max PreviewAlibaba	Alibaba	74.8	0	-11
Hy3Tencent	Tencent	74.4	0	-11
o1OpenAI	OpenAI	74.4	0	-10
o3 Mini (batch)OpenAI	OpenAI	74.3	0	-10
Qwen3.6 PlusAlibaba	Alibaba	74.1	0	-10
MiniMax M3MiniMax	MiniMax	73.9	0	-10
R1DeepSeek	DeepSeek	73.8	-1	-11
o1-proOpenAI	OpenAI	73.6	0	-10
Qwen3.8 27BAlibaba	Alibaba	73.4	0	-10
MiMo-V2.5Xiaomi	Xiaomi	73.0	0	-10
Gemma 4 26B A4B Google	Google	73.0	0	-10
Gemma 4 26B A4B (free)Google	Google	73.0	0	-10
Inklingthinkingmachines	thinkingmachines	72.9	0	-10
Inkling (free)thinkingmachines	thinkingmachines	72.9	0	-10
GLM 5V TurboZhipu AI	Zhipu AI	72.3	+1	-8
o4 Mini HighOpenAI	OpenAI	72.1	+1	-8
GPT-4o (batch)OpenAI	OpenAI	72.0	+1	-8
DeepSeek V3.1DeepSeek	DeepSeek	71.8	+1	+7
DeepSeek V3 0324DeepSeek	DeepSeek	71.8	+1	-10
Mistral Medium 3.5Mistral AI	Mistral AI	71.6	+1	-10
MiniMax M2.7MiniMax	MiniMax	71.6	+1	+6
MiniMax M2.5MiniMax	MiniMax	71.6	+1	-11
GPT-4o (2024-11-20)OpenAI	OpenAI	71.2	+1	+60
GPT-4o (2024-08-06)OpenAI	OpenAI	71.2	+1	-12
GPT-4oOpenAI	OpenAI	71.2	+1	-12
GPT-4o (2024-05-13)OpenAI	OpenAI	71.2	+1	-12
MiniMax M2MiniMax	MiniMax	71.0	+1	-12
MiniMax M1MiniMax	MiniMax	70.8	+1	-12
GLM 4.5 AirZhipu AI	Zhipu AI	70.7	+1	-12
Llama 4 MaverickMeta	Meta	70.7	+1	-12
Claude Haiku 4.5Anthropic	Anthropic	69.9	+1	-9
Claude Haiku 4.5 (batch)Anthropic	Anthropic	69.9	+1	-9
DeepSeek V3DeepSeek	DeepSeek	69.5	+1	-8
Qwen3 VL 235B A22B InstructAlibaba	Alibaba	69.3	+1	-8
DeepSeek V3.1 TerminusDeepSeek	DeepSeek	69.3	+1	-8
GPT-4o-miniOpenAI	OpenAI	69.3	+1	-8
Inkling Smallthinkingmachines	thinkingmachines	68.7	+1	-8
Inkling Small (free)thinkingmachines	thinkingmachines	68.7	+1	-8
Qwen3.5-FlashAlibaba	Alibaba	68.6	+1	-8
Hy3 previewTencent	Tencent	68.4	+1	-8
Qwen3.5 Plus 2026-02-15Alibaba	Alibaba	68.2	+1	+159
Qwen3 Max ThinkingAlibaba	Alibaba	68.2	+1	-9
MiniMax M2-herMiniMax	MiniMax	68.1	+1	-9
GPT-4.1OpenAI	OpenAI	67.7	+1	-9
GPT-4.1 (batch)OpenAI	OpenAI	67.7	+1	-9
Qwen3 MaxAlibaba	Alibaba	67.4	+1	-8
Mistral Large 3 2512 (batch)Mistral AI	Mistral AI	67.0	+1	-8
Llama 3.3 70B InstructMeta	Meta	66.8	+1	-8
Qwen3 Next 80B A3B ThinkingAlibaba	Alibaba	66.7	+1	+6
Qwen3 Next 80B A3B InstructAlibaba	Alibaba	66.7	+1	-9
GPT-4 TurboOpenAI	OpenAI	66.7	+1	-9
Qwen3.5-9BAlibaba	Alibaba	66.5	+1	-9
Step 3.5 FlashStepFun	StepFun	66.1	+1	-9
Mistral Large 2407Mistral AI	Mistral AI	65.9	+1	+23
Mistral LargeMistral AI	Mistral AI	65.9	+1	-10
Composer 2Cursor	Cursor	65.7	+1	-10
Composer 2 FastCursor	Cursor	65.7	+1	-10
GLM 4.6VZhipu AI	Zhipu AI	65.6	+1	-10
Qwen3 235B A22B Thinking 2507Alibaba	Alibaba	65.3	+1	-10
Llama 3.1 70B InstructMeta	Meta	65.3	+1	-10
GPT-4OpenAI	OpenAI	64.9	+1	-10
Qwen3 235B A22B Instruct 2507Alibaba	Alibaba	64.7	+1	-10
Qwen3 30B A3B Thinking 2507Alibaba	Alibaba	64.1	+1	-9
Qwen3 30B A3B Instruct 2507Alibaba	Alibaba	64.1	+1	+170
Qwen3 30B A3BAlibaba	Alibaba	64.1	+1	-12
o3 Mini HighOpenAI	OpenAI	63.9	+1	-10
GLM 4.7 FlashZhipu AI	Zhipu AI	63.5	+1	-10
Mixtral 8x22B InstructMistral AI	Mistral AI	63.4	+1	-10
Trinity Large Thinkingarcee-ai	arcee-ai	63.1	+1	-10
GPT-4o-mini (batch)OpenAI	OpenAI	62.5	+1	-10
GLM 4.5VZhipu AI	Zhipu AI	62.3	+1	-10
GPT-5 NanoOpenAI	OpenAI	61.9	+1	-10
gpt-oss-120bOpenAI	OpenAI	61.6	+1	-10
GPT-4 Turbo (batch)OpenAI	OpenAI	61.5	+2	-9
Mercury 2.5Inception	Inception	61.2	+2	-9
Qwen3 8BAlibaba	Alibaba	61.0	+2	-9
Mercury 2Inception	Inception	60.9	+2	-9
Nova 2 LiteAmazon	Amazon	60.4	+2	-9
Llama 4 ScoutMeta	Meta	60.2	+2	-9
Phi 4Microsoft	Microsoft	60.2	+2	-9
GPT-4.1 Mini (batch)OpenAI	OpenAI	58.9	+2	-9
GPT-4.1 Nano (batch)OpenAI	OpenAI	58.9	+2	-9
gpt-oss-20bOpenAI	OpenAI	57.4	+2	-9
GPT-4o-mini (2024-07-18)OpenAI	OpenAI	56.5	+2	-9
GPT-4.1 MiniOpenAI	OpenAI	56.2	+2	-9
Qwen3 235B A22BAlibaba	Alibaba	54.0	+2	-8
Granite 4.2 8BIBM	IBM	53.8	+2	-8
GPT-5 MiniOpenAI	OpenAI	53.8	+2	-8
Claude 3 HaikuAnthropic	Anthropic	51.3	+2	-7
Command ACohere	Cohere	50.8	+2	-7
Command R (08-2024)Cohere	Cohere	48.7	+3	-6
Command R+ (08-2024)Cohere	Cohere	48.7	+2	-7
Kimi K2.5Moonshot AI	Moonshot AI	47.5	+2	-7
Llama 3.1 8B InstructMeta	Meta	44.5	+2	-7
GPT-4.1 NanoOpenAI	OpenAI	42.1	+2	-7
R1 Distill Llama 70BDeepSeek	DeepSeek	40.7	+2	-7
Schematron V2 Turboinference-net	inference-net	40.0	0	-9
Schematron V2 Smallinference-net	inference-net	40.0	0	-9
GPT Astra Latest~openai	~openai	40.0	0	-9
GPT Sol Latest~openai	~openai	40.0	0	-9
GPT Terra Latest~openai	~openai	40.0	0	-9
GPT Luna Latest~openai	~openai	40.0	0	-9
Fugu Ultra v2sakana	sakana	40.0	0	-9
Fugu Maxsakana	sakana	40.0	0	-9
Ling 3.0 Flash VLinclusionai	inclusionai	40.0	0	-9
Ling 3.0 Flash VL (free)inclusionai	inclusionai	40.0	0	-9
DeepSeek V4.1 FlashDeepSeek	DeepSeek	40.0	0	-9
Ling 3.0 Flash Sante (free)inclusionai	inclusionai	40.0	0	-9
Muse Spark 1.3 Contributormeta	meta	40.0	0	-8
Hy4 previewTencent	Tencent	40.0	0	-7
Ling 3.0 Flash Fininclusionai	inclusionai	40.0	0	-7
Ling 3.0 Flash Fin (free)inclusionai	inclusionai	40.0	0	-7
GLM Flash Latest~z-ai	~z-ai	40.0	0	-7
Qwen3.8 FlashAlibaba	Alibaba	40.0	0	-7
Muse Spark 1.2 Contributormeta	meta	40.0	0	-7
DeepSeek V4 Flash Vision ExpDeepSeek	DeepSeek	40.0	0	-7
DeepSeek V4 Flash Vision Exp (batch)DeepSeek	DeepSeek	40.0	0	-7
Hy-MT2-1.8BTencent	Tencent	40.0	0	-7
Hy-MT2-30B-A3BTencent	Tencent	40.0	0	-7
GLM Latest~z-ai	~z-ai	40.0	0	-7
Hy-MT2-7BTencent	Tencent	40.0	0	-7
Dots3-Note Preview (free)dots-studio	dots-studio	40.0	0	-7
Gemini 3.7 FlashGoogle	Google	40.0	0	-7
Gemini 3.7 Flash (batch)Google	Google	40.0	0	-7
Seed 2.1 TurboByteDance	ByteDance	40.0	0	-7
Qwen3.8 2.4T A95BAlibaba	Alibaba	40.0	0	-7
Seed-2.0-CodeByteDance	ByteDance	40.0	+1	-6
DeepSeek V4 Pro 0813DeepSeek	DeepSeek	40.0	+1	-6
DeepSeek V4 Pro 0813 (batch)DeepSeek	DeepSeek	40.0	+1	-6
LFM2.5-2.6B (free)Liquid AI	Liquid AI	40.0	+1	-6
Nemotron 3.5 LightningNVIDIA	NVIDIA	40.0	+1	-6
Nemotron 3.5 Lightning (free)NVIDIA	NVIDIA	40.0	+1	-6
Sakana Namazusakana	sakana	40.0	+1	-6
Solar Pro 4Upstage	Upstage	40.0	+1	-6
Muse Glimmer 30Bmeta	meta	40.0	+1	-6
Muse Glimmer 30B (batch)meta	meta	40.0	+1	-6

哪些模型嘈杂、哪些模型稳定?

有些模型天然评分稳定--即使小幅排名变化也有意义。其他模型评分波动较大--需要更大的变化才值得关注。CV%（变异系数）告诉你每个模型的波动程度。越高 = 越嘈杂。

最不稳定模型(最高CV% - 最宽显著性阈值)

模型	提供商	评分	CV%	标准差	显著性阈值
Gemma 2 27BGoogle	Google	77.4	41.4%	26.17	±51.28
Phi 4Microsoft	Microsoft	60.2	37.7%	18.81	±36.87
Claude Opus 5Anthropic	Anthropic	95.1	36.3%	16.53	±32.40
Gemini 3.6 FlashGoogle	Google	40.0	35.3%	18.81	±36.87
Muse Spark 1.3meta	meta	80.9	35.3%	17.71	±34.71
R1DeepSeek	DeepSeek	73.8	34.0%	21.15	±41.45
Kimi K3Moonshot AI	Moonshot AI	40.0	32.7%	15.62	±30.62
Command ACohere	Cohere	50.8	31.7%	13.70	±26.85
Qwen3.8 Max (0902)Alibaba	Alibaba	75.6	31.5%	15.42	±30.21
GPT-4OpenAI	OpenAI	64.9	31.5%	17.41	±34.13
Mixtral 8x22B InstructMistral AI	Mistral AI	63.4	31.0%	16.81	±32.95
Llama 3.1 70B InstructMeta	Meta	65.3	31.0%	17.29	±33.89
Llama 3.3 70B InstructMeta	Meta	66.8	30.8%	17.58	±34.45
Mistral LargeMistral AI	Mistral AI	65.9	30.7%	17.30	±33.91
Gemini 3.6 Flash (batch)Google	Google	40.0	30.6%	14.00	±27.43
MiniMax M2-herMiniMax	MiniMax	68.1	30.5%	18.05	±35.38
Muse Spark 1.2meta	meta	80.9	30.0%	13.53	±26.51
o3 MiniOpenAI	OpenAI	75.3	29.8%	19.24	±37.72
DeepSeek V3DeepSeek	DeepSeek	69.5	29.4%	17.60	±34.49
GPT-6 AstraOpenAI	OpenAI	81.8	29.0%	19.70	±38.62

最稳定模型(最低CV% - 最窄显著性阈值)

模型	提供商	评分	CV%	显著性阈值
GPT-5.5 Pro (batch)OpenAI	OpenAI	92.7	0.0%	±0.00
GPT-5.5 (batch)OpenAI	OpenAI	92.7	0.0%	±0.00
GPT-5.2 Pro (batch)OpenAI	OpenAI	90.5	0.0%	±0.00
GPT-5.2 (batch)OpenAI	OpenAI	90.5	0.0%	±0.00
GPT-5.6 Luna Pro (batch)OpenAI	OpenAI	89.0	0.0%	±0.00
GPT-5.6 Luna (batch)OpenAI	OpenAI	89.0	0.0%	±0.00
GPT-5.6 Terra Pro (batch)OpenAI	OpenAI	89.0	0.0%	±0.00
GPT-5.6 Terra (batch)OpenAI	OpenAI	89.0	0.0%	±0.00
GPT-5.6 Sol Pro (batch)OpenAI	OpenAI	89.0	0.0%	±0.00
GPT-5.6 Sol (batch)OpenAI	OpenAI	89.0	0.0%	±0.00
GPT-5 Pro (batch)OpenAI	OpenAI	88.7	0.0%	±0.00
GPT-5 (batch)OpenAI	OpenAI	88.7	0.0%	±0.00
GPT-5.1 (batch)OpenAI	OpenAI	87.8	0.0%	±0.00
o3 (batch)OpenAI	OpenAI	86.7	0.0%	±0.00
Claude Sonnet 4.6 (batch)Anthropic	Anthropic	85.2	0.0%	±0.00
Gemini 2.5 Pro (batch)Google	Google	83.5	0.0%	±0.00
Grok 4.3 (batch)xAI	xAI	81.0	0.0%	±0.00
GPT-5.4 Nano (batch)OpenAI	OpenAI	79.3	0.0%	±0.00
GPT-5.4 Mini (batch)OpenAI	OpenAI	79.3	0.0%	±0.00
Gemini 3.5 Flash (batch)Google	Google	79.0	0.0%	±0.00

如何计算显著性

了解我们显著性分析背后的统计方法，帮助您区分真实的性能变化和随机波动。

统计显著性

我们使用95%置信度阈值(|z| > 1.96)的z分数。z分数衡量模型当前评分偏离其历史基准的标准差倍数。只有超过1.96个标准差的变化才被标记为统计显著。

基准评分

基准值是根据每个模型14天波动曲线数据的算术平均值计算的。该滚动平均值平滑了每日波动,提供了检测有意义偏差的稳定参考点。

置信区间

每个模型的95%置信区间计算公式为：基准值 +/- 1.96 x 标准差。落在此范围之外的评分表示统计上有意义的变化。"置信度"列显示 +/- 阈值。

多时间维度分析

每日(24小时)和每周(7天)排名变化分别分析。每日显著性要求排名移动超过3位,每周要求超过5位。在两个时间维度上都显著的模型代表最强、最可靠的信号。

噪音与信号

变异系数(CV%)衡量相对波动性。高CV模型天然评分嘈杂,需要更大的绝对变化才能达到显著性。低CV模型更可预测,因此即使小偏差也可能代表真实变化。