Tested to fail. Built to last.
We build bots that trade the markets, then spend most of our time trying to prove they're wrong. The few that survive trade live, on a short leash. Right now we're just making our own desk work. If you're curious how, have a look around.
What we do, in plain words
We build trading bots
Automated strategies for any market we can test: indices, currencies, energy, metals and rates. No gut feelings, no 3 a.m. hunches.
We try hard to break them
Most ideas look brilliant for about an hour. A team of AI agents tests, attacks and retests until the weak ones sink.
The survivors trade, carefully
Every fill is checked, loss limits act on their own, and one human signs off. The bots do not get the car keys.
Below is the voyage every strategy takes. Nothing here is for sale. If one day we open it up, the list hears first.
Diver goes looking
Each night a research agent tests new ideas on years of 1-minute data, with realistic costs, neighbouring settings and walk-forward windows. Every run is logged, failures included. Especially the failures.
Undertow pushes back and manages the storms
A separate red team reproduces the result and attacks it: look-ahead, unrealistic fills, overfitting, one lucky year. Most ideas stop here. The survivors get risk management: how they fit in a portfolio, daily loss limits and the worst-case drawdown for each account. Undertow is not fun at parties.
Keel builds the boat
Survivors are coded in NinjaTrader 8 with two reviews, timing and order safety, then matched trade by trade against the research engine. If they disagree, nobody sails.
Bosun runs the sea trial
Each strategy runs on its own simulation account. After every close, every fill is checked against the plan and the backtest. Bosun has never missed a close.
Harbor brings it to the decision
The desk manager puts the evidence and the question in one brief. One person decides the account, the size and the loss limits, in writing. The agents advise; the human signs.
Lighthouse keeps watch
Live strategies are checked every day against their own history. A daily loss limit closes the trade and stops the account, and no account ever holds long and short at once.
Strategies
Real market, real time, simulated money for now. Our top 3 by live result plus the portfolio account, updated by Bosun after every close. The rest of the fleet keeps sailing, just not on this shelf.
Top 3 + portfolio, last 2 trading days
NINJATRADER 8 SIM · MNQ · NOV 12 – 13, 2026 · 1 CONTRACT · AFTER COSTSStrategy 1
- Trades
- 1
- Account
- Sim102
- Days
- 2
Strategy 2
- Trades
- 1
- Account
- Sim107
- Days
- 2
Strategy 3
- Trades
- 1
- Account
- Sim108
- Days
- 2
Portfolio account
- Trades
- 12
- Account
- Sim110
- Last day
- +$74
Simulated accounts in NinjaTrader 8 on live market data, 1 contract, after costs. Two days is far too little to judge anything: this shelf is here so you can watch the record grow, bad days included. Simulated results are not real trading and do not predict future results.
The strategies we run, each tested in NinjaTrader 8 on real market data before it was allowed to trade. Live results join this shelf once there are enough of them to mean something.
Real strategies, backtested over 5 years
NINJATRADER 8 BACKTEST · MNQ · JUN 2021 – NOV 2026 · 1 CONTRACT · AFTER $2 A TRADEStrategy A
- Net
- +$34,930
- Max drawdown
- −$1,524
- Trades
- 663
- Positive years
- 6 of 6
Strategy B
- Net
- +$29,000
- Max drawdown
- −$2,260
- Trades
- 2,593
- Positive years
- 6 of 6
Strategy C
- Net
- +$21,535
- Max drawdown
- −$1,836
- Trades
- 488
- Positive years
- 5 of 6
A + B + C
- Net
- +$85,464
- Max drawdown
- −$3,430
- Trades
- 3,744
- Positive years
- 6 of 6
Backtests are hypothetical and were run on data that also helped choose the settings. They show how each strategy behaved in the past, not what it will do. Live trading is the real test, and those results will be shown here once there are enough of them. Backtest and live rankings are kept separately, so Strategy A is not necessarily Strategy 1.
What never changes
Loss limits act on their own
At the limit, positions close and the day ends.
Drawdown first
Profit is always judged against the worst drawdown in dollars.
Simple wins ties
Between two similar results, fewer rules stay.
Meet the crew
Six AI agents and one very patient boat. They find, test, build and guard every strategy; the human brings the coffee and the final yes. Not for sale: this is how our own desk runs.
Diver · Strategy Builder
Backtests your idea with costs, nearby settings, walk-forward and a trial log.
Undertow · Risk Management
Red team, risk management, account portfolios, daily loss limits and prop-firm odds.
Keel · Build and Operations
NinjaTrader 8 code with safety reviews and backtest matching.
Bosun · Sim Monitor
Every fill checked after the close; forward-test gates.
Harbor · Desk Brief
Results and every decision waiting, one click each.
Lighthouse · Dashboard
Private web dashboard with its own login.
Helm
NinjaTrader add-on where every crew member's work lands: strategies set once per account, a daily loss limit that closes the trade, and no opposite positions.
How we sail it
The method behind the desk, the mistakes we made (plenty) and the checks that now catch them. Maybe a course one day; for now, it's our playbook.
- Testing an idea without fooling yourself
- Red-teaming your own results
- Building it in NinjaTrader 8, safely
- Matching live trades to the backtest
- Loss limits that really stop you
- Running prop and funded accounts
- Your agent team and dashboard
Curious? Leave your email
Nothing to buy. If we ever share the playbook, you'll hear first. No spam: we're too busy arguing with backtests.