Trading desk

What it looks at

Live candles from Binance for crypto and Yahoo for US stocks, down to one-minute bars. Code computes everything: RSI and stochastic for momentum, MACD, moving averages and ADX for trend and its strength, Bollinger position and width, average true range and where today’s volatility sits in this asset’s own history, on-balance volume for whether volume confirms the move, the named candle pattern, and the nearest swing high and low as real levels rather than drawn lines.

Jev is handed those results in words and asked only what needs judgement: is this setup worth acting on, has it already run too far to chase, do the statistics agree with each other, and is there enough room up to justify the room down. It is never asked to calculate, because it cannot.

Intraday, on US stocks

Switch to US stocks on a minute or five-minute bar and the desk reads the session rather than the chart: where price sits against VWAP — the reference an intraday desk actually trades against — the overnight gap, whether it has been given back, the break of the opening half-hour range, where price sits in today’s range, and today’s volume against a normal day. Jev is then asked which way the day is leaning, as a separate question from the chart.

The universe is chosen to be worth day-trading: NVIDIA, Tesla, Sandisk, Strategy, Coinbase and AMD. Measured over one session their daily ranges ran from 6.5% to 27%, against roughly 1% for a mega-cap index — a quiet stock gives an intraday system nothing to work with.

Intraday means the day is not carried overnight. Nothing new opens in the last half hour, everything is closed out in the last ten minutes, and the opening half hour is skipped entirely: it prints the widest bars of the day and a stop placed there is usually taken out by noise rather than by being wrong.

Which candle to watch

Hourly by default, because that is where trends currently live. Measured across six pairs: trend strength sits in the 50th to 80th percentile of its own history on the hourly, and in the bottom third on every shorter bar. The shorter frames therefore spend almost all their time correctly refusing to trade — honest, but a dead board to watch. Entries are decided on the hourly; stops and targets are still checked on every pass, because the current hour’s candle keeps moving.

When it refuses to trade

Most of the time, and deliberately. A system that trades whenever it has an opinion loses money to its own opinions. Before any judgement is acted on, code checks the market:

• Trend weaker than usual for this asset, below the 40th percentile of its own ADX history — the market is chopping, and indicator signals in chop are noise.
• Volatility above its own 95th percentile — stops get run in conditions like that.
• Trend pointing down — this desk does not go short, so it stands aside.
• Already holding 4 names — concentration is a risk of its own.
• Down 15% from the peak — new positions halt entirely until it recovers.
• The statistics disagree, or there is more room down than up — no edge worth the risk.

Every refusal names itself on screen, so a quiet desk is explicable rather than broken.

Where losses are actually limited

Not by prediction. Every position gets a stop 2 average-true-ranges below entry and a target 3 above, written down the moment it opens and checked in code before anything is asked of a model — a stop is arithmetic, not an opinion. Once a trade is halfway to its target the stop trails up past break-even, so a winner cannot quietly become a loser.

Position size comes from the stop rather than from conviction: each trade risks 1% of equity between entry and stop, unless the 25% position cap binds first, which means a volatile asset automatically gets a smaller position. On intraday stocks the stop is tight enough that the cap usually does bind, and the desk’s risk panel shows the figure actually risked. That single rule flattens a drawdown more than any signal does.

What this does not do

It does not predict prices, and none of the above makes it profitable. Technical statistics on short candles, after fees, very often lose to simply buying and holding — which is precisely why the benchmark sits beside the equity line and why that comparison is the only number worth reading. Better risk management narrows losses; it does not manufacture a gain. If the figure against doing nothing is negative, the demo shows it.

Simulated throughout. There is no exchange and no account behind this, fills are imagined at the last candle close, and none of it is advice.