How to Use Market Sentiment

Understand how the perpetual premium is computed, avoid the two most common misreads, and put it into a strategy.

The perpetual trades above spot. Does that mean the market is bullish? Market Sentiment measures how far Binance perpetuals sit above the index price, but it uses a USDT gap and compares only against the coin's own past 30 days. Those two details are where most people misread it.

What is Market Sentiment?

Market Sentiment (MS) is Blave's futures premium indicator. It compares the close of Binance USDT-margined perpetuals with the Binance index price, then converts the gap into a z-score against that coin's past 30 days. The index price is Binance's weighted blend of spot quotes from several exchanges, not the spot price of a single venue. The calculation has three steps:

Because it is a z-score, the value has no fixed bounds and every coin sits on the same scale. A positive value means the gap is wider than usual for the past 30 days: longs are willing to pay up on the perpetual. The data contains no trader counts and no positions. Studio's status badges use these thresholds:

The two critical nuances: a USDT gap, measured against itself

It is an absolute USDT gap, not a percentage

The raw input is perpetual price minus index price in USDT, and nothing in the calculation divides by price. So with the percentage premium unchanged, the USDT gap grows as the same coin's price rises. Across coins, price level does not matter, because each coin is standardized against itself.

A hypothetical example: 30 days ago a coin traded near 1 USDT and the perpetual sat 0.001 USDT above the index, a 0.1% premium. The price then climbs to 2 USDT and the premium is still 0.1%, but the gap is now 0.002 USDT. The percentage has not moved, yet the gap is above its 30-day average, so Market Sentiment rises. A sharp price drop within 30 days works the other way and pushes the value down.

High and low are relative to the coin's past 30 days, not absolute

The z-score compares the coin with its own 30-day average. If a coin's perpetual holds roughly the same premium for two months, the average rises to meet it and Market Sentiment drifts back toward 0. Here 0 means "about the same as the past 30 days", not "no premium".

The reverse also happens. When a coin falls from a large premium to a small one, the gap drops below its 30-day average and the value can turn negative while the perpetual still trades above the index. The same +2 on different coins or in different months can reflect very different actual gaps.

How do you build strategies with Market Sentiment?

No official strategy uses Market Sentiment today, so all three uses below are sketches, not backtested. The ±0.5 and 3 in the code borrow Studio's badge thresholds as a starting point. They did not come from a parameter scan, so backtest before you use them.

1. Threshold with a buffer: hold while the gap runs high

The most direct version: go long when Market Sentiment rises above an entry threshold, and go flat when it falls below an exit threshold. The buffer between the two keeps the strategy from flipping in and out when the value hovers around a single level.

ENTRY_TH = 0.5    # sketch: the Optimistic badge threshold
EXIT_TH  = -0.5   # sketch: the Pessimistic badge threshold

def compute_signals(df, entry_th=ENTRY_TH, exit_th=EXIT_TH):
    # df['MS'] = the coin's Market Sentiment value
    signal = pd.Series(np.nan, index=df.index)
    signal[df['MS'] > entry_th] = 1.0   # gap above normal -> go long
    signal[df['MS'] < exit_th]  = 0.0   # gap below normal -> go flat
    return signal

Since the value is already a 1-day median, this signal follows day-level changes in the gap and does not react to price moves within minutes. It has no backtest, so scan the thresholds and the direction (trend-following or contrarian) yourself.

2. Extremes as a risk check: no new entries at Over Optimistic or Over Pessimistic

A value of ±3 means the gap is extreme relative to the past 30 days. A conservative use is to skip it as an entry signal and add it to an existing strategy's entry rule instead, holding off new positions while it is extreme:

# sketch: no entry while |MS| >= 3 (Over Optimistic or Over Pessimistic)
too_hot = df['MS'].abs() >= 3
entry = compute_entry(df) & ~too_hot

This is not a reversal signal, and it is not backtested. Blave has not verified that price reverses after a ±3 reading, or that skipping those readings improves results; here an extreme gap simply pauses new entries. Because ±3 is relative to the past 30 days, a coin whose gap has been very stable can reach ±3 on a small change. Replace compute_entry with your strategy's existing entry condition.

3. Filter: slow Market Sentiment sets the direction, a fast indicator times the entry

Market Sentiment is a slow, day-level signal, which suits a background filter: accept long signals from a short-window indicator such as Taker Intensity only while Market Sentiment is above 0, and go flat once it drops to 0 or below. The structure below is a sketch; replace compute_fast_entry with your own condition:

ms = fetch_market_sentiment(SYMBOL, INTERVAL, START, END, hdrs)
df['MS'] = ms['alpha']
regime_long = df['MS'] > 0              # sketch: long only while the gap beats its 30-day average
fast_entry = compute_fast_entry(df)     # e.g. a short-window Taker Intensity condition
signal = pd.Series(np.nan, index=df.index)
signal[regime_long & fast_entry] = 1.0
signal[~regime_long] = 0.0

This combination is not backtested and the threshold of 0 is only illustrative. Scan the parameters and check performance after fees before you use it.

How do you choose the bar interval?

Market Sentiment has one setting: the bar interval (period), one of 5min, 15min, 1h, 4h, 8h or 1d, with 5 minutes as the minimum. There is no window parameter like Whale Hunter's timeframe. Whatever interval you pick, the value is already a 1-day median, so the interval only sets how often you read the value; it does not remove the underlying 1-day median smoothing.

Pick the interval by how often your strategy checks, not to catch short-term moves. Data starts on 2020-01-01 (or the coin's listing date). The z-score needs 30 days of samples, so meaningful values begin around 2020-01-31, and the same applies to the first 30 days of a newly listed coin.

Where do you read it in Studio?

Which indicators pair well with it?

These are ideas for combining signals, not verified results:

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

The API endpoint is GET /market_sentiment/get_alpha with symbol, period and start/end dates (start_date, end_date); in Blave Agent the helper is fetch_market_sentiment. Coverage is Binance USDT-margined perpetuals only; other exchanges are not included. Only closed bars are returned, with timestamp, alpha and stat and no price. Without a start date you get the last 365 days, and ranges longer than 365 days are truncated.

The stat object describes the latest reading, and its field names don't quite match what they compute: