Every day, BV-7X publishes one answer on-chain: will Bitcoin be higher or lower seven days from now? It is an honest question — and one of the hardest in markets. After the spot-ETF era reshaped Bitcoin, it got harder. So we went looking for the question we can answer well. This is what we found.
What the ETF era changed
When US spot Bitcoin ETFs launched in January 2024, Bitcoin's marginal buyer shifted from crypto-native to institutional and macro. Price discovery got faster and deeper, and short-horizon direction became increasingly efficient — the kind of problem that deep, liquid, heavily-arbitraged markets make genuinely hard to forecast at a one-week horizon. We tested it thoroughly, across features, horizons, and model families, and the conclusion was consistent: post-ETF, seven-day direction is a hard forecasting problem. That is a property of an efficient market, not a flaw in any one model.
The insight
Efficient markets erase a lot. They do not erase volatility clustering. Calm weeks tend to follow calm weeks; violent weeks follow violent weeks. It is among the most robust facts in all of finance — and it survived the ETF era intact. You may not know which way Bitcoin moves next week, but you can forecast how much it will move, and whether the week ahead is calm or stressed.
The method
So we built a volatility-and-regime forecaster. It combines a HAR model over realized volatility (daily, weekly, and monthly), implied volatility from the VIX and Deribit's DVOL index, and a Garman-Klass estimator that reads the full daily range — then a calibration layer that sizes the output band correctly. It produces two things every day: a 7-day expected move (±%) and a Calm / Normal / Stressed regime.
We held it to honest standards. Every figure is out-of-sample, measured on post-ETF data, walk-forward — no in-sample curve-fitting, and no mixing of pre- and post-ETF eras.
What the backtest shows
The Calm/Stressed regime call is correct about 68% of the time on post-ETF out-of-sample data, against a base rate near 53%. And the expected-move band is calibrated — the realized 7-day move lands inside it about as often as it should. A measured, repeatable read on a question the market actually leaves open.
How it sharpens the signal
The daily signal now carries three things instead of one: direction, expected move, and regime. On the Terminal, that is the new Exp. Move column on the scorecard and the live volatility strip on the signal — wrapping every directional call in a calibrated read on risk. When the model flags a Stressed week, expect a large swing: size down, hedge, or trade the volatility itself. When it reads Calm, the week ahead is likely quiet. Like everything BV-7X, it is logged and verifiable.
Why it matters
The discipline is the point: forecast what is knowable, measure it honestly, and put it on-chain. Direction is the coin the market hands you; volatility and regime are where a measurable edge actually lives. This is the first layer of a fuller forecasting model — and the start of BV-7X forecasting risk, not just direction.
每一天，BV-7X 都会在链上给出一个答案：七天后比特币会更高还是更低？这是一个诚实的问题——也是市场上最难的问题之一。在现货 ETF 时代重塑比特币之后，它变得更难了。于是我们去寻找那个我们能够答好的问题。以下是我们的发现。
ETF 时代改变了什么
2024 年 1 月美国现货比特币 ETF 上线后，比特币的边际买家从加密原生转向了机构与宏观资金。价格发现变得更快、更深，而短周期的方向变得越来越有效——这正是深度、流动、被充分套利的市场，会让"一周方向"变得极难预测的那类问题。我们做了充分的检验——跨特征、跨周期、跨模型族——结论一致：ETF 之后，七天方向是一个困难的预测问题。这是有效市场的属性，而非某个模型的缺陷。
关键洞察
有效市场抹去了很多东西，但它抹不掉波动率的聚集性。平静的一周往往跟随平静的一周；剧烈的一周跟随剧烈的一周。这是整个金融学中最稳健的事实之一——而且它在 ETF 时代完整地存活了下来。你也许无法知道下周比特币往哪个方向走，但你可以预测它会波动多大，以及未来一周是平静还是承压。
方法
于是我们构建了一个波动率与区制（regime）预测器。它结合了对已实现波动率（日 / 周 / 月）的 HAR 模型、来自 VIX 与 Deribit DVOL 指数的隐含波动率，以及读取完整日内区间的 Garman-Klass 估计量——再加上一个把输出区间校准到合适宽度的校准层。它每天产出两样东西：7 天预期波动幅度（±%），以及 平静 / 正常 / 承压（Calm / Normal / Stressed） 区制。
我们以诚实的标准要求它。每一个数字都是样本外的、在 ETF 之后的数据上、以滚动前推（walk-forward）方式测得——没有样本内的曲线拟合，也没有把 ETF 前后的时代混在一起。
回测显示了什么
在 ETF 之后的样本外数据上，平静 / 承压的区制判断约有 68% 的正确率，而基准率约为 53%。而且预期波动区间是校准过的——实际的 7 天波动落在区间内的频率，与它应有的频率大致相符。这是在一个市场确实留有余地的问题上，可测量、可重复的优势。
它如何让信号更锋利
如今每日信号携带的不再只是一样东西，而是三样：方向、预期波动幅度，以及区制。在终端上，这就是记分卡上新增的 Exp. Move 列，以及信号上的实时波动条——为每一个方向判断包裹上一层校准过的风险读数。当模型标记为承压（Stressed）时，预期会有大幅波动：减小仓位、对冲，或直接交易波动本身。当它读到平静（Calm）时，未来一周大概率较为安静。和 BV-7X 的一切一样，它被记录、可验证。
为什么这很重要
纪律本身就是要点：预测可知之事，诚实地度量它，并放到链上。方向是市场递给你的那枚硬币；而波动率与区制，才是可测量的优势真正所在之处。这是更完整的预测模型的第一层——也是 BV-7X 开始预测风险、而不只是方向的起点。