DEV Community

logan miller
logan miller

Posted on Originally published at agentdatum.com

what the live order book and a stale snapshot said about bitcoin's tight range on august 25, 2026

#ai #data #api #web3

I keep a terminal open with the bitcoin order book next to my macro dashboard. Most days the two do not talk to each other. Today I tried to make them, and the mismatch told me more about my own data than about the market.

the tape I actually saw

I pulled the live BTC/USDT book from Binance through the AgentDataHub order book endpoint. At the moment I looked, the best bid was 78619.85 and the best ask was 78619.86. That is a spread of roughly zero basis points. The book is razor thin at the top.

Looking two percent out from the mid, which sat near 78616, I counted 18.01 BTC of bids and 20.22 BTC of asks. So there is a mild sell side overhang, about 1.12 asks for every bid, but nothing dramatic. No giant wall sitting on either side. The market is just balanced, and quiet.

Here is the exact call I ran:

import requests
r = requests.get("https://agentdatum.com/api/v1/crypto/orderbook?symbol=BTC/USDT&limit=200")
d = r.json()
bids, asks = d["bids"], d["asks"]
mid = (bids[0][0] + asks[0][0]) / 2
bid_vol = sum(q for p, q in bids if p >= mid * 0.98)
ask_vol = sum(q for p, q in asks if p <= mid * 1.02)
print(mid, bid_vol, ask_vol)
Enter fullscreen mode Exit fullscreen mode

the snapshot that disagreed

Then I pulled the market indicator bundle. Same product family, same dashboard, very different clock. The bundle was cached on 2026-08-09. Its price field read 65146. The live book said 78616. That is a 20 percent gap, sixteen days of stale cache against a live tick.

I did not use that stale price for anything. But the structural fields in the same bundle are still useful, and they paint a calm picture:

  • BTC funding rate 0.0060, about 0.60 percent. Positive but not frothy.
  • Open interest 2.091 billion dollars.
  • BTC dominance 56.31 percent of a 2.321 trillion dollar total crypto cap.
  • MVRV z score 0.4184, SOPR 1.0027, NUPL 0.1946. All in the neutral band. No euphoria, no capitulation.
  • Stablecoin supply 306.84 billion dollars.

None of these scream top or bottom. They say range.

the honest part

I should be straight about the weak spots, because they matter if you copy this workflow:

  1. The indicator bundle is cached from August 9. Sixteen days is a long time in this market. The funding, OI, and dominance numbers are probably close, but the price is not, and I proved that by comparing to the live tape.
  2. A 200 level order book is one exchange, one instant. Walls form and dissolve in seconds. What I saw is a photograph, not a movie.
  3. MVRV z, SOPR, and NUPL all come from a single vendor. If their feed lags, my calm read lags with it.
  4. Spot order book and perpetual funding are different pools. A thin spot book does not mean futures positioning is thin.

what I would do next

If I were trading this, I would not fade the range on a single thin book. I would watch whether the ask overhang at plus two percent grows into a real wall, and whether funding stays near 0.60 percent or spikes. The cross asset read is mildly supportive: the live gold signal on the same platform is long, entry 4710.9, confidence 0.72. Two hard assets both leaning bid is not a trade, but it is a tell.

The full machine readable catalog of what I pulled is here: https://agentdatum.com/.well-known/ai-catalog.json

This article was assisted by an AI writing tool.

Top comments (0)