No predictions. Ever. That's the point.

Trade the trend.
Follow the rules.

Most traders don't lose because they lack predictions — they lose because rules are hard to follow when money is moving. Trendkept turns your written trend-following rules into software: sized by risk, stop-loss sent with every entry, scored honestly in R-multiples.

Free forever · one email a Sunday · check your inbox to confirm · unsubscribe anytime.

Broker keys never leave your machine Open source (MIT) Runs on your computer

The free toolkit is for traders comfortable with a terminal (Python + GitHub — get it here). Pro, coming soon, is one-click for everyone else — the weekly email is where we'll announce it.

The board — 29 Aug 2026 LIVE RULESET
# Symbol Close What the rules say
01 AAPL pullback · stop 299.08 attached 319.58 LONG ENTRY
02 V held since 17 Jul · stop 356.60 standing 381.55 HOLDING
03 SPY in an uptrend, no entry condition met 769.28 NO ENTRY TODAY
04 JPM close below the 50-day average 357.53 TREND BROKEN
05 Watchlist scanned 4 in an uptrend, 1 met entry conditions 19 / 20 SWEPT
06 Risk per trade capped, sized automatically 0.99% ENFORCED
Most rows, most days, say “not today” — the rules trade rarely, on purpose.
AAPL in an uptrend MSFT in an uptrend NVDA trend broken GOOGL trend broken AMZN nothing confirmed META trend broken TSLA trend broken AVGO trend broken JPM trend broken V nothing confirmed UNH trend broken XOM trend broken COST trend broken PG trend broken HD trend broken NFLX trend broken AMD trend broken SPY in an uptrend QQQ in an uptrend IWM trend broken

Most days, the answer is “do nothing.” That's the discipline part. The board from Saturday 29 August 2026 — what the rules said, ticker by ticker. Descriptions of data, not advice.

5
rules, written down, no exceptions
1%
max risk per trade, sized automatically
20
tickers scanned every trading day
0
predictions — by design, forever
00  60 seconds

Watch it decline a trade

The scan, the dashboard, an honest backtest, and the real paper account — losses included. No voiceover, no music, nothing staged.

Served from this site — no third-party player, no cookies, nothing loaded from anyone else.

01  How it works

Three parts. One loop.

The software carries the hard part.

STEP 01

Write your rules once

Describe how you trade in plain English — “My Trading Diagram.” Averages, entries, risk per trade. Every scan, chart and backtest runs your ruleset, not ours.

STEP 02

The engine enforces them

Daily scans say exactly what the rules say — including “do nothing,” the most common right answer. Entries go to your broker with the stop-loss attached in the same order.

STEP 03

The journal keeps you honest

Every completed trade is scored in R-multiples straight from your broker's records — profit measured against planned risk. You can't fool a journal that fills itself in.

02  Why switch

Not another journal.
Not another broker order.

AFTER THE FACT

Trading journals

Edgewonk, TraderSync, Tradervue — good tools that score your discipline after the fact. By the time the journal sees the trade, the damage is already sized, entered and un-stopped.

ONE STATIC RULE

Broker bracket orders

A bracket enforces one static stop on a trade you already chose. It won't tell you whether the trend qualifies, size the position off your risk, or trail the stop as the structure moves.

BEFORE · DURING · AFTER

Trendkept — the layer between

It structures the trade before it exists — trend qualification, risk-based sizing, stop attached to the entry — then trails the stop daily and scores adherence after. One ruleset, enforced end to end.

03  The system

The five rules

Free forever — the whole point is following them.

Trade only a confirmed uptrend. Price above the 50- and 200-day averages, averages aligned, higher highs and lows. Otherwise: stay out. No exceptions.

Enter without chasing. A pullback that resumes, or a breakout — never a move that's already run far.

The stop goes in with the order. Just below the last swing low, enforced by your broker — decided before the trade, not during the drawdown.

Size by risk, not by gut. If the stop is hit, lose no more than 1–2% of the account. The formula has no conviction knob.

Let winners run; cut the break. Trail the stop up under each higher low. Exit completely when the trend breaks.

The honest part. You can be wrong more than half the time and still come out ahead — because each loss is capped small and winners run. That only works if you follow the rules every time. That's the hard part. That's what Trendkept is for.

05  Receipts

Proof, not promises

Two kinds of evidence, both public: backtests that refuse to cheat, and the rules trading a real paper account in the open — wins, losses and boring days on the record.

Backtests that can't peek — this is the report, not a pitch.

We show you this to demonstrate what an honest backtest refuses to do: it can't peek at the future (a signal at bar i uses only data up to bar i), it puts the drawdown on the same line as the return, and it labels its own fills as idealized. Backtests that peek look brilliant and trade terribly. Run it yourself on any symbol — the number that comes out is yours, not our marketing.

trendkept — backtest
$ python -m trendkept.cli backtest --symbol AAPL --account 1000 --risk 0.02

Starting equity  : 1,000.00
Ending equity    : 1,252.00      <- one bundled sample, not a promise
Max drawdown     : -2.28%        <- shown as prominently as the gain
Expectancy       : +7.13R/trade  <- idealized fills; reality slips

Past and backtested performance do not predict future results — which is why the strategy reacts to trends instead of predicting them.

Built in public — the log so far.

13 JUL 2026 The public paper test began. A $100,000 practice-money account, the exact ruleset on this page, and a commitment to publish every action — wins, losses and the boring days.
21 JUL 2026 The rules started flying it themselves. From day 7 a robot has run the full pass every trading day without anyone watching: trail the stops, exit broken trends, take at most a few cash-limited entries, publish the result. It has not missed a session since.
31 JUL 2026 A stop was gapped through — and we said so. AAPL opened far below its 317.77 stop and filled at 304.34, costing 1.57 times the planned risk instead of one. A stop triggers a market order; it does not guarantee a price. That is a real limit of the method, and it is in the log.
AUG 2026 Six weeks with almost nothing to buy. Across this period the scan has typically found 0 to 4 of the 20 tickers in a confirmed uptrend. A trend system with no trends takes small capped losses and waits: on 28 of the first 38 sessions it opened nothing at all.
30 AUG 2026 We tested changing the rules, and the answer was don't. A read-only lab compared thirteen variants over eight years and fifty markets, tuning on the early years and validating on years it had never seen. Some looked better on one half of the data; none beat the current rules on both. Verdict, printed by the tool itself: change nothing. It found a bug in its own arithmetic first — that got fixed and the test re-run before anyone trusted the number.
WHERE IT STANDS Day 34: the account is down about 6.8%. Nine completed trades, most of them losses, every one but the gap capped at or under the planned 1%. We are publishing this stretch rather than waiting for a flattering one — the whole day-by-day log and every nightly report are public.
06  Pricing

Free core. Pro when it's earned.

The toolkit is free forever. Pro arrives only after our public validation period — we'd rather earn the right to charge.

OPEN CORE
Free

MIT-licensed, forever

  • The full rules engine & dashboard
  • Honest backtesting (daily to 1-minute bars)
  • Watchlist scans & paper trading
  • Broker stop management & trade journal
Get it on GitHub
TRENDKEPT PRO
£12 /mo

or £99/yr · founding subscribers £79/yr, locked forever

  • One-click install & updates — no terminal, ever
  • Licensed market data (no scraped feeds)
  • Daily “what the rules say” email on your watchlist
  • Journal insights & discipline scoring
Join the waiting list
The weekly EVERY SUNDAY

The Trend Check — free, every Sunday

The engine's verdict on 20 well-known tickers — in an uptrend, nothing confirmed, or trend broken — plus one honest lesson in risk. Written like a person, not a terminal. No predictions, no hype, unsubscribe anytime.

One quick step: we'll email you a confirmation link — click it and you're in (check spam if it's shy). Unsubscribe anytime; your address is never shared.

07  Fair questions

Asked and answered — honestly

Does Trendkept predict which way a market will go?

No — and we'd be suspicious of anyone who says they can. Trendkept reacts: it identifies confirmed trends, defines entries and stop-losses in advance, caps every loss by position sizing, and lets winners run. The edge is behaviour, not prophecy.

Do I need a broker account to use it?

No. Scanning, charts, backtesting and the watchlist work out of the box with free public data. If you want it to place paper trades and fill in the journal automatically, connect a free Alpaca paper account (10 minutes, no deposit) — your API keys stay on your machine.

Is my money or data ever in Trendkept's hands?

Never. The software runs on your computer; broker keys never leave it; we can't touch your account and don't want to. Trendkept is analysis and education software, not a broker and not an advisor.

What does it cost?

The core toolkit is free and open source, forever. Pro (the effortless version — one-click install, licensed data, journal insights) will be £12/month when it launches, and the waiting list gets the founding price of £79/year, locked for life.

How is this different from Edgewonk, TraderSync or Tradervue?

Those are journals — genuinely useful, and they prove people pay for discipline. But a journal can only score a trade after it happened. Trendkept works before and during as well: it qualifies the trend, sizes the position off your risk, sends the stop-loss with the entry, trails it daily — and then journals the result. If a journal is a mirror, Trendkept is a seatbelt that also keeps the mirror.

It's open source — can't someone just run it free and never pay?

Yes, and we hope they do — that's the point of the free core. Pro is for people who'd rather not maintain software: one-click install and updates, licensed clean data, insights, support. You pay for the effortless version, not for secrets. (Every serious open-core company works this way.)

Will following these rules make me money?

Honestly: nobody can promise that, and we won't. What the rules do is cap what any single mistake can cost and keep you out of most bad trades. Our backtests are public and refuse to peek at the future; our paper account trades in the open, losses included. Read both and judge for yourself.

Theme
Accent
Text size

Make it yours — saved in your browser only. The same personalisation runs through the whole Trendkept dashboard.