How to Use Blave Top Traders

How "top" is actually defined, what the exposure number measures, and the four readings that trip people up.

When you read "top traders are long", the first question is how "top" was defined. At Blave the answer is not the intuitive one — performance plays no part in it.

What are Blave Top Traders?

Blave Top Traders is a set of accounts drawn from Blave's own users: after someone links an exchange API key in Studio, the platform periodically reads that account's wallet balances and positions, then ranks accounts by asset size and keeps the top 10%. Their positions, aggregated, are what the exposure, position and watchlist pages show. It is not an exchange leaderboard and not an on-chain address list.

Selection has three steps:

Nothing in that path looks at performance. Large assets may just mean large capital, or capital that was never withdrawn — a different thing from trading well.

The three pages answer different questions:

Four things worth knowing

1. "Top" is asset size, not performance

This is the most common misreading, and our own older copy got it wrong too. The sort key is summed wallet balance; returns, Sharpe and drawdown do not exist anywhere in the calculation. So the indicator tells you how the largest balances are positioned, not how winners are positioned.

2. It is market-wide, not BTC

Exposure sums the group's perpetual positions across every coin into a single number, with no symbol dimension. The exposure page lets you pick a coin, but that only changes the price line underneath — the bars stay identical. For per-coin detail, use the position page.

3. It is a raw ratio, not a z-score

Market sentiment and whale hunter are standardized against each coin's own last 30 days, which is what makes ±2 and ±3 meaningful there. Exposure is not standardized at all. Its distribution over the past year (2025-09-15 to 2026-09-15; hourly points before 2025-10, 5-minute after):

The group is structurally net long, so "exposure above zero is bullish" is nearly always true and carries no information. What does carry information is the relative position: exposure falling into the low end of its recent range usually means this group is cutting longs.

4. The watchlist is not positions

Watchlist data comes from lists these traders built inside Blave Studio, not from their holdings. Watching a coin is not holding it.

What it covers, and what it does not

A few boundaries change how you should read it, so they go here rather than in a footnote:

• The denominator is total assets; the numerator is only part of the positions. Net exposure is perpetual net notional ÷ all wallet assets, so spot and earn balances sit in the denominator without ever appearing in the numerator. A trader heavy in spot reads as lower exposure.
• Some position types are excluded. Only certain perpetual position types enter the calculation; other contract types and spot positions do not. This is one slice of the group's book, not the whole book.
• The sample is small and it moves. The top 10% is a share, not a fixed roster, so the count falls when active accounts fall, and the median is taken only among those currently holding countable positions. Values from different periods are not strictly comparable.
• Five-minute bars are false precision. The series is written every five minutes, but the underlying balances and positions refresh hourly. Intra-hour moves mostly come from prices repricing the same notional, not from anyone trading.
• History starts 2025-03-03, and the early part is hourly before it becomes 5-minute. Account for that in any backtest.

Where do you read it in Studio?

Treat the position page's entry price carefully: it is a plain average of every position row in that coin, regardless of direction or size, so an entry on the long table can carry short-side cost in it. It is not the group's cost basis.

What should you know when pulling it via API or Blave Agent?

Exposure has a public endpoint, GET /blave_top_trader/get_exposure (API plan), taking a period and a date range and no symbol — the series is market-wide by construction. Position and watchlist have no public endpoint today; they live on the Studio pages.

Inside Blave Agent, fetch it with fetch_top_trader_exposure(interval, start, end, headers); the returned alpha is the ratio described above.

# no symbol argument: this is a single market-wide series
tt = fetch_top_trader_exposure("1h", START, END, hdrs)
print(tt["alpha"].describe())

How to put it into a strategy

None of the three sketches below is backed by an official strategy or a backtest. They are starting points to validate yourself.

1. Use the relative level, not zero

Because the series is structurally positive, using zero as the dividing line means being long almost always. Compare it to its own recent distribution instead:

tt = fetch_top_trader_exposure("1h", START, END, hdrs)
# 90-day quantiles as levels; more meaningful than an absolute zero
lo = tt["alpha"].rolling("90D").quantile(0.1)
hi = tt["alpha"].rolling("90D").quantile(0.9)
crowded = tt["alpha"] > hi      # this group is more loaded than usual
flush   = tt["alpha"] < lo      # usually means this group is cutting longs

2. Watch it diverge from price

Price making highs while exposure slides means the group is selling into strength. That is an observation, not a rule — it can also mean positions moved into types the calculation does not count.

px = fetch_kline(SYMBOL, "1h", START, END, hdrs)["close"]
d = px.to_frame("px").join(tt["alpha"]).dropna()
# over 20 bars: price up, exposure down
div = (d["px"] > d["px"].shift(20)) & (d["alpha"] < d["alpha"].shift(20))

3. Treat it as context, not an entry

The indicator updates slowly and rests on a small sample, which makes it background — useful for sizing or for deciding whether to take a signal, not for timing entries on its own. To check whether unusual flow accompanied a move, read it alongside whale hunter.

Which indicators pair well with it?