Rigorous Validation. Every strategy earns its place.
A clean backtest takes an afternoon to fabricate. At Dalvox, every strategy completes eleven independent stress tests — across 14+ years of market history — before it is allowed to trade. Below is every single step, in order, with no omissions.
From raw idea to live account.
These are the exact steps, in the exact order, that every strategy completes before appearing in our catalogue. Click any step to see what it tests for and why it matters.
01
Strategy Generation & In Sample
60% of total data. Thousands of configurations tested. Only the statistically sound survive.
The process begins with a blank slate. We run thousands of algorithmic configurations — varying entry logic, exit conditions, risk rules, and parameters — against the first 60% of all available historical data. This is the In-Sample window: the controlled training environment where a strategy's logic is shaped.
At this stage, the selection is brutal. Strategies that only perform well because they happened to match a short patch of history are eliminated immediately. Only those that demonstrate statistically consistent, risk-adjusted returns — across multiple market regimes within the training window — pass the screen and move forward as candidates.
A shortlist of candidates built on statistically meaningful logic — not accidental pattern-fitting to a narrow window of history.
02
Quality Retest + Out-Of-Sample 1
Premium M1 data. First blind exposure to ~20% of history the optimizer never saw.
Surviving candidates are immediately re-run on higher-quality M1 data — which captures more granular price movement than the initial pass. This alone eliminates strategies that only looked good on lower-resolution data. Then, for the very first time, each strategy faces history it has never encountered: approximately 20% of total data, sealed off since day one of development.
This is the first blind test. There is no way to cheat it — the strategy either holds up on unseen data or it doesn't. One with a genuine edge continues to perform. One that was fitted to its training window collapses here, before any more time is invested in it.
That the initial results weren't a coincidence — the edge persists on unseen, higher-quality data.
Most algorithmic trading services consider this their entire process. We're at step 2 of 11.
03
Commissions & Slippage
Worst-case broker conditions forced on every fill.
Spreads are inflated, commissions are added, and worst-case slippage is injected on every single fill — simulating the kind of execution degradation you'd encounter with a bad broker during a high-volatility news event. The strategy must remain profitable under these handicaps, because real-world conditions are never as clean as a textbook backtest.
Profitability is genuine — not an artifact of unrealistically clean fill assumptions.
04
Multi-Timeframe Validation
Tested across candle intervals it wasn't built for.
The strategy is re-deployed on timeframes it was never designed for. A strategy overfitted to one specific candle interval fails here — because its edge only exists in the noise of that particular data resolution, not in the market itself. A robust strategy adapts and continues to perform because the logic captures something structurally real.
The edge is structural — not a coincidence of one data resolution.
05
Monte Carlo: Skip Trades
Trades randomly omitted across 1,000+ simulations.
A random percentage of trades is removed across thousands of independent simulations. If performance collapses when a handful of trades are missing, it reveals a dangerous truth: returns are concentrated in a few outlier wins, not distributed across a repeatable edge. The strategy must remain profitable even when its best individual trades are absent.
Profitability is broad and distributed — not riding on a handful of lucky trades.
06
Monte Carlo: Trade Order
Historical trade sequence randomised thousands of times.
The entire sequence of historical trades is shuffled and re-run thousands of times. This tests two things simultaneously: first, that the edge exists regardless of when trades occur in time — and second, that even in the absolute worst-case losing sequence (all losses landing back-to-back), the account drawdown remains within acceptable bounds. A real edge doesn't depend on lucky timing.
The edge is timing-independent, and drawdown is structurally bounded — not a product of sequencing luck.
07
Additional Markets
Deployed on instruments and indices it was never built for.
The strategy is run against entirely different markets — other indices, instruments, and session structures — that it was never optimised for. Overfitted strategies collapse immediately on foreign data because they were tuned to one instrument's noise, not to an underlying market dynamic. Robust ones continue to generate returns, confirming the logic captures something universal.
The logic exploits a real, transferable market inefficiency — not one instrument's specific quirks.
08
Monte Carlo: Parameters
Every configurable variable stress-shifted in all directions.
Every configurable parameter — stop distances, entry thresholds, timing windows — is systematically varied up, down, and sideways across thousands of combinations. A fragile strategy only works when parameters land on very specific, precise values: shift them slightly and it falls apart. A robust strategy performs across a range, because the underlying logic is what generates the edge — not any particular magic number.
No overfitting to specific parameter values. The logic drives performance — not a fragile configuration.
09
Monte Carlo: Price Data Perturbation
OHLC prices randomly shifted — entry and exit levels are never guaranteed in live trading.
The underlying OHLC price data itself is randomly modified across thousands of simulations — entry and exit prices are shifted to simulate execution uncertainty and real-world fill imprecision. In live trading, the exact prices from historical data never repeat. This test confirms the strategy does not depend on hitting specific historical price levels that might not exist in real conditions.
The strategy remains profitable even when prices deviate from the historical record — which is always the case in live execution.
Nine tests passed. Two remain — and they're the ones that matter most.
10
Out-Of-Sample 2 — The Final Blind Test
The most recent ~20% of all history. Sealed since day one. Opened only now.
The final approximately 20% of historical data — the most recent years, including current market regimes — has been completely sealed off since the first day of development. Not a single optimisation step has ever touched it. It represents the present: the market environment that is closest to where the strategy will actually trade.
Only after surviving all nine previous tests is the strategy run on this data for the first time. This is the closest thing to a live forward test that a backtest can offer. There is nothing left to adjust, no parameters left to tune. The strategy either performs or it doesn't.
The strategy is genuinely predictive. It didn't memorise history — it learned from it. Performance on the most recent data confirms the edge is current and real.
11
Tick-Precision Final Test — 100% of History
Every fill reconstructed from raw ticks. The definitive performance record.
The strategy is run one final time across the complete 14+ year dataset — from start to finish — using tick-level precision for every single entry and exit. This is the most computationally intensive simulation possible. Every OHLC candle is reconstructed from raw tick data, meaning fills are calculated at the exact price the market was trading at that microsecond — not an approximation of it.
The equity curves, drawdown charts, return figures, and risk metrics shown across the entire Dalvox platform come from this final run. It is as close as a backtest can get to real-world trading performance. No rounding, no estimation, no convenience.
The definitive, tick-accurate performance record. Every number you see on this site — every return figure, every drawdown, every equity curve — comes from here.
A strategy that reaches our catalogue has survived every single one of these tests. In the correct order. Without exception.
Most candidates don't make it through. The ones that do are the systems you'll find in our product tiers — and every performance figure you see on this site reflects the final tick-precision run described in Step 11.
All tests passed.
One decision left.
We are fully transparent — every trade across all our products is audited and available for review. Verify the numbers, then apply. We grant access to a limited number of accounts to protect the capacity of our systems.