A FORWARD-VERIFIED REGIME LAYER FOR TRADING SYSTEMS

Your bot knows how to enter. TrendOrFlat knows when it should stay out.

TrendOrFlat reads news and macro, classifies the BTC market regime, and gives your strategies one simple answer: TREND_LONG, TREND_SHORT, FLAT, or OFF. It does not place trades or access your funds. Every regime is committed before the outcome and remains in the public track record.

News and macro in the LLM layer, no price or TA · evidence computed in code · verified 24 hours later

Current public regime

Why this call
forecasts in the public record
Проверка через 24ч
Методика v2.1
Ошибки не удаляем

A good strategy can still trade at the wrong time

A trend strategy loses its edge in a range. A mean-reversion or grid strategy becomes fragile in a strong move. TrendOrFlat does not rewrite entries: it adds a regime check in front of them.

Trend strategy in FLAT

Mean-reversion / grid in TREND

Not another strategy. A control layer above yours.

Read the external backdrop

News, macro, calendar, flows, geopolitics, and regulation — without BTC price or TA in the 24h LLM prompt.

Commit the regime

Models vote; evidence and the final call are computed in code.

Pass permission

Your bot receives TREND_LONG, TREND_SHORT, FLAT, or OFF and applies its own rules.

The record comes before the claim.

Every forecast stays in the journal with time, version, market move, and grade. Mistakes stay as visible as hits.

Date Forecast BTCUSDT 24h Fact Grade

Test the idea on your past trades

Upload an export of completed trades. The calculation runs entirely in your browser and shows a retrospective "with / without the filter" scenario. It is a simulation, not your account's actual performance or a promise of future returns.

Test it on my trades

Works with the stack you already use

Pay for automation, not another dashboard

Basic

Monitor the regime

Max

Scale regime control

Pricing

What TrendOrFlat does — and what it cannot know

The service does not place trades or receive withdrawal/trading access to your exchange. It does not know your strategy or actual P&L. The regime is probabilistic and can be wrong. OFF is a normal protective outcome. Price is used in the grader and simulator after the fact, but not in the 24h LLM prompt.

Test it on my trades Try the automatic gate for 14 days

How the control layer works

Current public regime
Model Vote Strength
Why (Claude)

Next forecast in
Outcome strength average of models that agreed with the outcome (−14…+14)
Consensus accuracy how often the shared AI decision was right
Forecasts evaluated how many forecasts have already been checked against the market
Best model which model is currently the most accurate

Loading…

7-dimension radar
Filter vs reference strategy

Exact reference rules →

Model leaderboard

Model Accuracy % Evaluated total Correct Partial Wrong Uncertainty (was: Refusals)

Recent forecast checks

What we called → what BTC did in 24h → grade. Trend band ±2.5%. LONG = up, SHORT = down, FLAT = sideways, OFF = refuse.

Date Forecast BTCUSDT 24h Fact Grade