IDFX Trading
Trading ideas,
tested with code.
Free algorithmic studies of the setups traders talk about. Each idea is written down as a strategy, backtested in the IDFX engine, and judged only on data the optimizer never saw. Whatever the answer is, it gets published.
- Out-of-sample only
- Before trading costs, always labelled
- Data published as JSON
Spec lines quoted from the Silver Bullet study. The animation shows the process, not a result.
The newest result
Tori Trades' trend line strategy on five markets, 4H
One rule on 5 markets, with three-touch lines and with any two-touch line. 5 of 10 averaged above zero out of sample. Every test is listed with equal weight.
| Market | 3+ touches | 2+ touches |
|---|---|---|
| Platinum | −0.012R | −0.139R |
| WTI oil | −0.021R | +0.126R |
| Gold | +0.108R | +0.260R |
| US30 | −0.099R | −0.078R |
| NAS100 | +0.198R | +0.091R |
ICT killzones: when does the market actually move?
Five years of 5-minute bars on NAS100, gold and EURUSD, measured on the New York clock: when the big moves happen, when the day's high and low are made, and what happens after Asian, London and previous-day levels are swept.
CRT (Candle Range Theory) across four timeframes, NAS100 and gold
One rule on 4 candle sizes, 2 markets and three reward-to-risk floors. 12 of 24 averaged above zero out of sample. Every test is listed with equal weight.
| Candle | NAS100 | Gold |
|---|---|---|
| Daily | −0.210R | −0.134R |
| 4H | +0.185R | +0.065R |
| 1H | −0.022R | +0.035R |
| 15m | −0.012R | +0.007R |
IVB (Initial Volume Breakout) on NAS100 and US500: three readings, price only
3 readings of the rules on 2 markets. 6 of 6 averaged above zero out of sample. Every test is listed with equal weight.
| Reading | NAS100 | US500 |
|---|---|---|
| Protocol | +0.085R | +0.032R |
| Breakout | +0.057R | +0.037R |
| Retest | +0.041R | +0.014R |
How every idea is tested
The same four steps for every study, in the same engine, so results can be compared.
- 01
Describe
The idea is written in plain English, then turned into a declarative strategy spec: indicators, entry conditions, exits.
- 02
Compile twice
One spec becomes a Python backtest and a Pine Script v6 file for TradingView. Both come from the same source, so they cannot drift.
- 03
Walk forward
Parameters are fitted on one window of history and then traded, frozen, on the next. Only those out-of-sample trades count.
- 04
Verdict
The engine attacks the result: sample size, overfitting, shuffled prices, trade order. It grades the idea and says why.
What you can rely on
Out-of-sample only
The numbers come from walk-forward testing: rules fitted on one stretch of history, then traded on the next stretch, untouched. Optimized in-sample results are never presented as performance.
Before trading costs, said every time
Studies test whether a signal has an edge at all. Spread, slippage and commission depend on your broker and account, so they are left out, and every result is labelled that way.
Failures get published
Most ideas do not survive a fair test. When one fails, the study says “not this time” and shows why. That is as useful to know as a pass.
Nothing invented
Every number and chart comes from a recorded engine run, and its data is published as JSON. No testimonials, no screenshots of P&L, no claims about live or prop-firm money.
Watch them being built
Studies are built and tested on camera, in the engine's dashboard, on the IDFX Trading channel.