Entries are market orders and every position is closed by the process itself within seconds at most. The run charges the real per-second spread but assumes the simulated fill price: requotes, partial fills and broker rejections are not part of the base case.
Past results do not guarantee future results.
How Unified validates the edge
What was measured, on what data, with what costs, and what remains unverified.
Validation pipeline
- 01Freeze↓
Code, model, thresholds and configuration are hashed with SHA-256 before anything runs. If a hash moves before the run ends, the run is void.
14/14 artefacts verified intact - 02Data coverage↓
The external tick history is walked instrument by instrument and classified as complete, partial or no sample. Instruments without data are declared, not dropped.
20 complete · 5 partial · 5 no sample - 03Out-of-sample window↓
Only the window after the training cut counts. The first year of the external history sits inside the training window and is reported separately, never inside the headline figure.
NEXORA HFT MODERNO: 2022-07-26 → 2026-07-26 · NEXORA HFT GUARDED: 2022-07-26 → 2026-07-26 · NEXORA AUTO LEGACY: 2023-02 → 2026-07 · NEXORA AUTO FUNDED: 2023-02 → 2026-07 - 04Tick replay with real spread↓
Every second of the history is replayed with the spread the broker actually quoted. Commission is what the account charges; slippage is zero in the base case and charged in the stress scenarios.
per-trade result series - 05Decomposition↓
The result is split by instrument, by month, by year and by direction. Negative periods and weak instruments are published as they are.
per-pair / per-month / per-year tables - 06Walk-forward↓
The window is cut into consecutive segments and each one is evaluated in order, so a result that only works in one regime shows up as a losing segment.
window table per engine - 07Monte Carlo, trade by trade↓
The trade sequence is resampled at random. With a positive expectancy and hundreds of thousands of trades this comes out positive almost arithmetically, which is exactly why it is not read alone.
100,000 simulations - 08Monte Carlo, by block↓
The same resampling in week-long blocks, which preserves losing streaks. This is the drawdown number that means something.
block distribution per engine - 09Stress↓
Wider spread, slippage, execution delay and partial loss of the result, one scenario per row, plus a scenario with everything at once.
stress table per engine - 10Dependency↓
The result is recomputed without the best trades, without the best month and without the best year, to see how much of it rests on a handful of events.
dependency block per engine - 11Overfitting probe (PBO)↓
Combinatorially symmetric cross-validation over the decision threshold. Published as measured, including when it is unfavourable.
PBO 67.06% - 12Independent feed↓
The frozen models are evaluated again, with no parameter retuned, on the tick history of a broker that took no part in building them. The two feeds are never mixed inside one figure.
MetaQuotes vs FTMO comparison - 13Gate list↓
The suite checks every certification condition and publishes the ones that fail. A single failure rejects the certification regardless of how strong the headline figures look.
23 conditions (NEXORA AUTO LEGACY certification suite) - 14Release freeze and rollback drill
The shipped build is hashed again, cross-checked against the validation freeze, and every artefact is restored from its backup to prove the rollback works.
20 artefacts · 36 restored · 0 failures
23 certification conditions
The suite checks every condition and publishes the ones that fail. A single failed condition rejects the certification even when the headline figures look strong. Source: NEXORA AUTO LEGACY certification suite. The HFT suites do not publish a numbered gate list of their own; their gates are the pipeline stages above, each with its own artefact.
| # | Condition | Required | Obtained | Result |
|---|---|---|---|---|
| 1 | PF global OOS | >= 3,50 | 3.553 | PASS |
| 2 | Drawdown global | <= 5% | 20.40% | FAIL |
| 3 | Recovery Factor | >= 3 | 145.05 | PASS |
| 4 | Esperanza neta positiva | > 0 | +2.66 | PASS |
| 5 | Compras rentables | PF > 1 | 4.069 | PASS |
| 6 | Ventas rentables | PF > 1 | 3.174 | PASS |
| 7 | Todos los anos positivos | 4/4 | 4 | PASS |
| 8 | Meses positivos | >= 70% | 100% | PASS |
| 9 | Walk-Forward trimestres | >= 75% | 100% | PASS |
| 10 | Monte Carlo positivas | >= 95% | 100.00% | PASS |
| 11 | Probabilidad de ruina | < 1% | 0.000% | PASS |
| 12 | Monte Carlo caida p95 | <= 15% | 3.46% | PASS |
| 13 | Monte Carlo caida p99 | <= 20% | 4.01% | PASS |
| 14 | Monte Carlo PF mediano | >= 1,50 | 3.361 | PASS |
| 15 | Estres x1,50 | PF >= 3,20 | 3.253 | PASS |
| 16 | Estres x2,00 positivo | > 0 | +269,180 | PASS |
| 17 | Sin dependencia de pocas ops | > 0 | +268,828 | PASS |
| 18 | Sin dependencia de pocos pares | > 0 | +64,964 | PASS |
| 19 | Concentracion del par mayor | <= 15% | 48.9% | FAIL |
| 20 | Pares positivos | 30/30 | 30 | PASS |
| 21 | Sin look-ahead ni repintado | 40/40 | 40 | PASS |
| 22 | Estabilidad de parametros | mayoria | 20/20 | PASS |
| 23 | Paridad bot vs laboratorio | exacta | 0,0e+00 | PASS |
Known limitations of this method
These are properties of the certification you are reading, stated in the present. A page that declares the limits of its own method is worth more than one that only shows results.
The base case charges the real per-second spread, zero commission (what this account charges) and zero slippage; one point of slippage is charged in the stress scenario. There is no live dataset, so real slippage and rejection rates remain unmeasured.
Nothing on this page is a live result. Every figure comes from historical tick data. A LIVE VERIFIED figure would carry that label and there is none today.
5 instruments returned zero valid instants and 5 returned a partial history. That is missing data, not a failure. Those instruments stay in the universe and appear in every table as NOT CERTIFIED - INSUFFICIENT SAMPLE.
On the two metals the MetaQuotes feed declares a point value 8.79x below what its own contract implies, while the external feed is consistent with the contract. Profit Factor is a ratio and survives, but any euro figure for metals measured on MetaQuotes is understated - profit and risk alike. The fix is documented and deliberately not applied: touching it inside a frozen validation would void the run.
Professionally managed strategies typically run a profit factor between 1.2 and 2.0. The figures here are measured over historical ticks at the minimum lot with a small per-trade expectancy. They are not a claim that such a ratio survives live execution, size or capacity. Read them next to the drawdown, the trade count and the stress table, never alone.
Combinatorially symmetric cross-validation returns a probability of backtest overfitting of 67.06% on the decision threshold: the configuration that wins in one slice of time tends to under-perform in the next. Published as measured. Its scope is the threshold only - model and geometry were frozen before this run.
A large share of the net result comes from a single instrument. The per-instrument table publishes the exact share so the concentration stays visible instead of being averaged away.
Artefact freeze
Code, model and configuration were hashed before the validation started and re-hashed after it finished. If a single hash had moved, the run would be void.
FROZEN AT26 Jul 2026, 14:08
VERIFIED AT26 Jul 2026, 16:02
INTACT14 / 14VALID — no artefact moved during the run
| Engine group | Artefact | SHA-256 (32) | State |
|---|---|---|---|
| HFT MODERNO/GUARDED | modelo_certificado.pkl | 8c429e4ee243d72b81b9a5d789ae132f | INTACT |
| HFT MODERNO/GUARDED | umbral_par.pkl | 69cc1b6c80ca10af9b34f7c88b1bbb7c | INTACT |
| HFT MODERNO/GUARDED | scalper_hft.py | 5f67439147047dd0656c7f72359955f0 | INTACT |
| HFT MODERNO/GUARDED | cert_30pares.json | 6a6b2e7a7b4196655043316871f8efc7 | INTACT |
| HFT MODERNO/GUARDED | certificacion.json | bbcb5d0c6ab32484c24a3423775176ba | INTACT |
| AUTO LEGACY | auto_gbt.pkl | 3da7987164f9a05855d55934ee88bc2e | INTACT |
| AUTO LEGACY | auto_selector.py | 52e4cd7841bd0cf722e0559ec821e4c7 | INTACT |
| AUTO LEGACY | auto_mtf.py | d52dffdd650afda9c25bc14493821df2 | INTACT |
| AUTO LEGACY | auto_bot.py | dd7206f0cb092437db184a5f217c0aea | INTACT |
| AUTO LEGACY | auto_legacy_m5_autotrader.json | 0cd45c4d6e5d37d0a8cd4303b37baf4a | INTACT |
| AUTO FUNDED | auto_gbt_ftmo.pkl | 7781977fb62d308a015c70ab7cfef7ff | INTACT |
| AUTO FUNDED | auto_fondeo.py | 8861e345de3ddc455c3a8a8b67e05f15 | INTACT |
| AUTO FUNDED | auto_funded_institutional.json | 17ffff6d409a45fc87c31f9330a05ffa | INTACT |
| AUTO FUNDED | FTMO_160K_FREETRIAL_2STEP.json | f5615cfa53f2cfa95419cd44b9cb9b1f | INTACT |
Every engine page carries the same evidence for its own certificate: NEXORA HFT MODERNO · NEXORA HFT GUARDED · NEXORA AUTO LEGACY · NEXORA AUTO FUNDED