How SquareRadar verifies every signal you see

Throughout the day, we read what every active trader on our platform publishes to Binance Square. The list of tracked traders grows whenever users submit new ones through the Track Trader form and we confirm the account actually publishes trading calls. We cover the four markets traders actually operate in — SPOT, FUTURES, ALPHA, and STOCK — so a trader's activity across different markets all lands in one place.

The collector reads Binance Square's public feed in a fast, lightweight mode for posts that are easy to parse. When a post is image-heavy, exotically formatted, or otherwise tricky, it falls back to a fully-rendered read — the same view a real user would see — so nothing important slips through.

Nothing about this stage is opinionated. We're reading what traders chose to publish in public. No DMs, no private channels, nothing scraped from behind a paywall.

Every post that survives the screen goes through AI extraction. The model reads the post and pulls out the structured pieces — the asset, the side (long or short), the entry price or zone, up to three take-profit levels, the stop loss, and the leverage if it's a futures call.

Two cleanup passes run on top of the raw extraction. First, a normalisation step rewrites ticker shorthand to match the actual listings on Binance — a trader writing one ticker name is automatically mapped to the listed name, so leveraged tokens, tokenised stocks, and renamed assets all line up correctly. Second, an auto-fix pass catches obvious slip-ups — for example, a stop-loss written as 0.0765 when context clearly means 0.765 gets corrected and the signal is flagged for human review.

Every AI extraction is logged. If we ever correct or override a trader's call, you can see that on the signal's detail view — we don't quietly rewrite history.

We don't always see a signal the moment it's posted. A few minutes can pass between the trader hitting publish and us reading the post — long enough that a fast move could already have hit the entry, run to a target, and started reversing before we ever got there.

So for any signal we see after a delay, we replay the actual price history. We pull minute-by-minute prices covering the gap and simulate the trade chronologically — was the entry zone reached first, then which target hit first, then did the stop ever trigger. The simulation respects order: a target can only count after the entry is filled.

That way, a fast-moving signal that closed before we ever saw it still gets recorded honestly, with the right outcome and the right ROI. We mark these signals so you can filter them out if you only want signals tracked from the moment we caught them onward.

One honesty note: replay isn't available for ALPHA and STOCK markets, because Binance doesn't publish historical price data for them. Those signals are tracked from the moment we extract them onward, with entry assumed to have happened at the trader's stated price.

We don't manually curate winners. Every signal a tracked trader posts goes through the same pipeline. We don't pick favourites, we don't hide flops, and we don't boost any trader's metrics by quietly dropping losing calls. If a trader has a 12% win rate, that's what you'll see.

We don't simulate trades. Every ROI you see on this platform is the actual price movement against the trader's stated entry, TP, and SL — not what would have happened with some hypothetical position size or risk model. Slippage and fees are not modelled.

We don't pretend to know intent. If a post looks like it might be a signal but is genuinely ambiguous, the AI flags it with a low confidence score and we don't use it in any trader-level metrics. Those signals still show up so you can judge them yourself.