Altora Trading
Three systematic books across uncorrelated markets, live on demo, building the out-of-sample record.
Risk-parity blend of the three sleeves (each scaled to the same volatility), net of measured spreads and the overnight financing of every day held. The sleeves are genuinely orthogonal (monthly correlations of −0.10 to +0.17 between them), which is why the combined book beats every sleeve on its own. Run at 10% annual volatility, its worst peak-to-trough was −26%. Before 2023 the blend's Sharpe was 0.97; the held-out figure is flattered by an unusually good recent stretch for index and gold trends.
Revised 3 October 2026. A full re-test after we found and fixed several bugs: three strategy engines compared different instruments when our price files were sorted by date; the live bots acted on stop and target exits one candle late; financing was mis-modelled (one tool over-charged FX six-fold, others applied today's rates to all of history), and is now charged at historical rates for every day held; and the commodities bot could not trade markets that were shut at its run time for about a month. Every number on this page now uses the corrected engines, measured spreads and full financing, on OANDA's costs (on IG's lower costs the blend is 1.41). The construction is simpler than August's (equal risk per trade in each book), so the figures are not directly comparable. The live record below is what actually matters.
Earlier: corrected 1 August 2026, downwards. The previous figures here (1.52 and 2.10) came from a script that no longer exists and used an indices input backtested with almost no trading cost. We rebuilt the whole blend to mirror what the three live bots actually do: their real engine rosters, their own rebalance rules, real measured spreads, and commodity carry. The rebuilt figures are what you see above, with twelve of sixteen years positive and a 21% maximum drawdown at the blend level. Full detail on the corrections page.
11 major pairs (EUR/USD, GBP/USD, USD/JPY, the yen & euro crosses…). Six rule-locked edges: RSI exhaustion, price-action structure breaks, pullback quality, volatility breakout, weekly CFTC positioning and cross-sectional currency rotation. (Currency-strength divergence was removed on 21 September 2026 after it lost money in both halves of the history.)
Two-sided · checks every 4 hours · fixed risk per trade · thin edge per trade, so it is the most cost-sensitive book
S&P 500, Nasdaq 100, DAX, Nikkei 225, gold and silver. Seven trend/dip engines (Donchian breakout, momentum burst, RSI trend-dip, turn-of-month…), long only, exactly as the live bot trades. The two event overlays once listed here (pre-FOMC drift and a COT tilt) were switched off in August 2026 after failing their re-tests. Revised 3 October 2026 (see note above); the history of this book's earlier figures is below.
Corrected 31 July 2026, downwards. This book previously showed 1.19. That figure came from a script we can no longer find, and the surviving code it was built on charged index CFDs a foreign-exchange spread of 0.0001 of price, which is thousands of times too small. We rebuilt the backtest to mirror what the live bot actually does and charged real measured spreads. Rebuilt: 0.77 over 2011–26 and 1.80 over the held-out era from 2023, at a 29% maximum drawdown. Costs explain only part of the gap; the rest we cannot account for against a script that no longer exists, so we publish the number we can defend rather than the one we liked.
Long-only · rebalances daily · volatility-targeted sizing
Brent, WTI, natural gas, corn, soybeans, wheat and sugar. Four trend engines (volatility expansion, higher-timeframe volatility expansion, multi-timeframe pullback, trend pullback) with a 3×ATR trailing-stop exit. Copper, platinum and palladium were removed after losing in both halves of the history; most of the edge comes from oil.
Two-sided · rebalances daily · fixed risk per trade (halved to 0.5% in October 2026)
| Book | Started with | Net P&L | NAV now |
|---|---|---|---|
| FX | £4,000 | −£135.32 (−3.38%) | £3,864.68 |
| Indices & metals | £10,500 + £4,633 added 1 Oct | −£253.78 (−2.27%) | £14,879.22 |
| Commodities | £10,000 | +£130.02 (+1.3%) | £10,130.02 |
| Commodities — previous book, closed 2 Jul – 5 Oct 2026, restarted at £10,000 after the October bug fixes | £5,000 | −£1005.18 (−20.10%) | closed at £3,994.82 |
| Portfolio | −£259.08 (−0.89%) | £29133.0 deployed | |
| Portfolio including closed books | −£1,264.26 | ||
| Under evaluation — not in the portfolio figure | |||
| simple_trader | £25,000 | −£627.75 (−2.51%) | £24,372.25 |
| Explorer v2 | £10,000 | +£2,612.99 (+26.13%) | £12,612.99 |
simple_trader is a from-scratch lab running classic indicators, and Explorer v2 is a candidate bench whose rules were frozen before it started, with a verdict due 15 December 2026. Neither is a validated book yet, so neither counts toward the portfolio return above — but both are published here for the same reason everything else is.
Demo accounts, verified against the broker's own transaction history and marked to live NAV. Live since mid-June 2026 — a sample this young is mostly noise, which is why it updates weekly whether it flatters us or not. Where capital was added to a running book it is shown under its starting figure and is not counted as profit. Updated 19 Oct 2026.
Execution-timing disclosure: live results before 7 August 2026 carry an execution drag we found and fixed in our own audit; entries were filling 30 minutes to 18 hours after the signal candle rather than the backtested moments-after-close. Details and measured impact are in the research log. From 7 August 2026 fills land within about two minutes of the close, matching the tested construct.
All three books run fully automated on demo accounts (live since July 2026), each on its own capital, sized to a fixed volatility target. The backtest numbers above are real and honestly built, but the number that counts is the live, out-of-sample one, and that only comes with time. We'll publish it here as it accrues, wins and losses alike.
Systematic model output, not financial advice. Demo performance shown for transparency; it is not a promise of future results.
We wrote up the exact six-layer system behind these books (data, features, signal engines, backtesting, honest validation and a working broker-connected bot) as a complete handbook for beginners. Full Python code at every step, including the validation discipline most guides skip.
Educational guide, not financial advice. It teaches the engineering and the testing, and promises no profits.
Get the handbook →£29.99 one-off · instant PDF download · secure checkout via Whop · Sports edition also available