Prediction markets react to information before mainstream news covers it. This is not a theory -- it is an observable, repeatable pattern in the data. When insiders, researchers, or well-connected traders learn something material, they do not wait for a journalist to publish a story. They trade on it. That trade activity -- volume spikes, price shifts, smart money movements -- is visible on-chain before any news organization runs a headline. Depthy's signal engine detects these patterns in real time across 50,000+ active Polymarket markets, giving anyone with an API key access to the same early warning system that informed traders rely on.

The Information Cascade

Every market-moving event follows the same information cascade. Understanding the timeline is the key to understanding where the edge lives.

  1. Information emerges. A leak surfaces, an insider learns something, early data from a report circulates in private channels. The public does not know yet.
  2. Informed traders act. Volume spikes as those with information take positions. Depthy fires a VOLUME_ANOMALY signal within seconds of the spike crossing the detection threshold.
  3. More traders notice. The price begins moving significantly as the initial activity attracts attention from other sophisticated participants watching for unusual flow.
  4. News organizations publish. Reporters confirm the underlying event. The story hits Reuters, Bloomberg, or crypto news desks. This is when most people first hear about it.
  5. Retail traders react. The general public piles in after reading the headline. By this point, the price has already moved substantially. The best entry is long gone.

The gap between steps 2 and 4 is where the edge lives. It can be minutes. It can be hours. In some cases, it is a full day. Depthy detects step 2 -- the earliest observable footprint of new information entering the market.

Real-World Pattern: Anatomy of a Volume Spike

Here is a realistic example based on the patterns we observe daily in the data. The specifics are constructed, but the dynamics are taken directly from actual signal behavior across thousands of markets.

Consider a market titled "Will [geopolitical event] happen by March 31?" trading at 35 cents. Baseline activity has been approximately 30 trades per hour over the prior 24 hours -- normal, unremarkable flow for a mid-tier market.

  • 2:14 PM -- Depthy detects a VOLUME_ANOMALY signal. Trade count in the last hour: 247. That is an 8.2x spike against the 30/hr baseline. Severity score: 89 out of 100.
  • 2:14 - 2:45 PM -- Price moves from 35c to 52c, a 49% increase. The volume anomaly is attracting attention from other participants who are now front-running the expected price discovery.
  • 3:30 PM -- Reuters publishes a report about the underlying geopolitical event. This is the first mainstream coverage.
  • 4:00 PM -- Market reaches 68c as retail traders read the news and buy in. The move from 52c to 68c is driven primarily by late arrivals.

The volume anomaly signal provided a 76-minute head start before any news coverage. For an automated agent, that is an eternity. Even for a human trader checking alerts on their phone, it is more than enough time to evaluate the market and take a position before the crowd arrives.

Here is the actual signal data structure your agent would receive: 

{
  "signal_type": "VOLUME_ANOMALY",
  "market_id": "0x3488...",
  "severity": 89.2,
  "title": "Volume spike 8.2x in Will [event]...",
  "metadata": {
    "last_hour_count": 247,
    "avg_hourly_24h": 30.1,
    "ratio": 8.2
  },
  "created_at": "2026-02-17T14:14:00Z"
}

The severity field is a normalized 0-100 score that accounts for the volume ratio, the absolute trade count, and the market's liquidity profile. An 89.2 is in the top 2% of all volume anomaly signals. Your agent can threshold on this field to filter out low-conviction spikes and only act on the signals most likely to represent genuine information flow.

Setting Up Automated Detection

The fastest way to catch these signals is through Depthy's SSE (Server-Sent Events) stream. Open a persistent connection and receive every signal as it fires, with zero polling delay. Here is a Python script that monitors specifically for high-severity volume anomalies: 

import requests
import json

API_KEY = "dpth_free_your_key_here"
SEVERITY_THRESHOLD = 75

with requests.get(
    "https://depthy.io/v1/pm/signals/stream",
    headers={"Authorization": f"Bearer {API_KEY}"},
    stream=True
) as resp:
    for line in resp.iter_lines():
        if line and line.startswith(b"data: "):
            signal = json.loads(line[6:])
            if (signal["signal_type"] == "VOLUME_ANOMALY" and
                signal.get("severity", 0) > SEVERITY_THRESHOLD):
                ratio = signal.get("metadata", {}).get("ratio", 0)
                print(f"HIGH SEVERITY ALERT: {signal['title']}")
                print(f"  Volume ratio: {ratio}x | Severity: {signal['severity']}")
                # Alert via Telegram, Slack, or execute trade

If SSE is not practical for your setup, polling works too. Hit GET /v1/pm/signals/volume-anomalies every 60 seconds and filter for new signals since your last request using the created_at timestamp. You will miss signals by up to 60 seconds compared to the stream, but for strategies with longer time horizons this is perfectly acceptable.

Both approaches return the same signal structure. The stream is better for latency-sensitive agents; polling is better for simpler architectures that do not need to maintain persistent connections.

Confirming Signals: Volume + Depth

Volume alone can be noise. A single whale placing a large order, a bot running a test sequence, or a brief burst of speculative activity can all trigger a volume anomaly without any underlying information driving the move. This is why the highest-conviction setup in the dataset is not a single signal -- it is a dual signal.

When volume spikes AND bid-side depth surges on the same market within a short window, the move has conviction. Real money is not just trading -- it is stacking liquidity behind the direction, signaling that informed participants expect the price to hold and continue moving. Depthy's DEPTH_IMBALANCE signals detect exactly this kind of structural shift in the order book.

The pattern looks like this:

  • VOLUME_ANOMALY fires -- unusual trade activity detected, severity above 75
  • DEPTH_IMBALANCE fires on the same market within 30 minutes -- bid-side liquidity stacking confirms the direction of the move

When both signals fire on the same market, you have two independent data sources agreeing: trade flow says something is happening, and order book structure says the market believes it will continue. This dual-signal pattern is rare. Depthy generates 23,463 signals per day across 3,430 markets, but only a handful of markets on any given day produce both a volume anomaly and a depth imbalance within a tight window. When they do, the signal-to-noise ratio is the highest we observe in the entire dataset.

Limitations and Honest Caveats

This approach is detection, not prediction. It is important to be clear about what volume anomaly signals can and cannot tell you.

  • Not all volume spikes lead to news. Sometimes the spike is a whale repositioning, a liquidation cascade, or simply noise. A volume anomaly tells you something unusual is happening. It does not tell you why.
  • Free tier has a 5-minute delay. The SSE stream on the free plan delivers signals with a 5-minute lag. For the pattern described above -- where 76 minutes of lead time matters -- this still leaves significant edge. But for fast-moving markets where the window is narrower, that delay can reduce or eliminate the advantage. Metered and above tiers get real-time delivery.
  • Prediction markets can be thin. Low-liquidity markets are susceptible to reflexivity: a single large order moves the price, which triggers a volume signal, which attracts more traders, which moves the price further. The signal is real, but the cause may be self-reinforcing rather than information-driven. Check market liquidity before acting.
  • Interpretation is still required. Depthy tells you that volume spiked 8x on a geopolitical market. Your agent -- or you -- still needs to decide what that means and whether to act. The signal is the starting point, not the conclusion.

The best trade you can make on a prediction market is the one you enter before the crowd knows why the price is moving. Volume anomaly detection does not guarantee that edge on every signal, but it consistently surfaces the moments where new information is entering the market. Combined with depth imbalance confirmation, it is the closest thing to a systematic early warning system that exists for prediction markets today.

The data is streaming. The signals are firing. The only question is whether you are listening.