About

In plain words

TrendOrFlat is an off-switch for trading robots by market regime. Trend strategies lose in sideways markets; mean-reversion / flat algos lose in trends — in the wrong phase the robot should stay quiet. Every few hours we say one of four things — TREND_LONG, TREND_SHORT, FLAT, or OFF — and your software follows that. If you are not a programmer, that is fine: setup takes about 15 minutes with the guide, then it runs on its own.

Why this exists

Trend bots bleed equity in sideways markets, not on “bad entries.” Mean-reversion / flat algos do the opposite: they bleed when the market actually trends. Manually switching them is exhausting. Building your own ChatGPT/Claude/Gemini ensemble is expensive and brittle. TrendOrFlat does one job: say which strategy type is in season — and which must stand down.

Under the hood (short)

We read external context (news, macro, geopolitics, flows, regulation) — no price or technicals in the LLM prompt. Several independent AIs vote. Evidence strength is scored in our code. Output: one regime that TSLab, the API, or the bot can read.

How to use it (3 steps)

You do not need to understand the models. You only need a way to feed the regime into your stack:

  1. Open the Telegram bot → start a trial and get a Regime API URL (token included).
  2. TSLab: drop in the RegimeGate cube, set SignalSource=url, paste that URL — trend agents trade only in LONG/SHORT, flat agents only in FLAT; otherwise they stay idle.
  3. TradingView: on the TradingView page, paste the ready-made Pine script with embedded regime history (visual backtest). For your own fills — upload a CSV to the Simulator and compare PnL with vs without the filter.

Available now: TSLab (RegimeGate DLL), TradingView (Pine teaser), Telegram bot, public Regime API / MCP for custom scripts and AI agents. We do not sell entry tips — only market regime.

Start 14-day free trial Setup guide TradingView

Why you can verify it

Each forecast is locked before the event. After 24 hours we grade it with an open OHLC methodology (±2.5%). The grader is open source. Misses stay on the dashboard. Download the history and recompute yourself.

Grading methodology Full history Simulator

Honest limits

A regime filter for algo systems — not buy/sell advice and not a profit guarantee. OFF and misses are part of the protocol; we do not hide them.