QuintileRevT
- エキスパート
-
Vladimir Karputov
Create a personal work for me: https://www.mql5.com/en/job/new?prefered=barabashkakvn - バージョン: 1.12
- アクティベーション: 5
QuintileRev — a contrarian system for EURUSD H3
No machine learning. No curve fitting. The rule is published in full.
What it does
One idea, traded consistently: when price has just set an extreme within a 50-bar window
while sitting near a round number, the market reverts more often than it continues.
Fresh low → buy. Fresh high → sell. Hold exactly 50 H3 bars (about six days), one position
at a time, half closed at +300 points.
There are no hand-picked thresholds. The system compares the current feature value against
quantiles of the last 2000 bars, excluding the current one — it only ever looks at the
past, never at the future.
The numbers (MetaTrader 5 Strategy Tester, EURUSD H3, 2016.01.01 – 2026.08.02, 0.1 lot, FxPro-MT5, start balance $500)
| metric | value |
|---|---|
| net profit | +$3,510 |
| profit factor | 1.51 |
| maximal balance drawdown | $455 (25.2%) |
| recovery factor | 6.40 |
| Sharpe ratio | 0.52 |
| trades (closing deals, partials included) | 601 |
| worst single trade | −$388 |
This is a backtest, not a live account. I am not selling you proof of profit — I am
giving you the rule and the data so you can verify it yourself.
What you should know BEFORE buying (the parts most vendors leave out)
1. It trades rarely: ~30 entries per year — roughly one signal every 12 days. If you
want an active EA, this is not it.
2. The worst trade in history was −$388 on 0.1 lot. On a $500 account that is
three quarters of your money in one trade. **Recommended deposit: $2,500+ per 0.1 lot**,
or 0.02–0.03 lots on $500.
3. The edge is tied to the daily clock. Measurement showed that the same system on a
bar grid shifted by an hour or two loses up to half its quality. It works on H3 aligned
to terminal midnight — do not change the timeframe.
4. A ~6-day hold means swaps. On EURUSD they are moderate, but longs pay more per night
than shorts receive. This is already included in the figures above.
5. It does not tolerate spreads above ~20 points. Check your broker.
Why there is no free three-month trial
The system takes ~30 entries per year. In three months you would see seven trades —
not enough to prove or disprove anything; seven trades fit comfortably inside randomness.
Instead of a trial I give you more: ten years in the tester and the complete rule.
Load the demo into the MetaTrader 5 Strategy Tester, pick any period, and you will get the
same numbers within an hour. That is a more honest test than waiting a quarter.
What you get
- QuintileRev.ex5 with licensing handled by MQL5 Market;
- the full rule and every parameter documented — nothing hidden;
- the 2016-2026 tester report, which you can reproduce yourself;
- setup support.
Parameters
| parameter | default | meaning |
|---|---|---|
| InpLots | 0.1 | fixed lot |
| InpHoldBars | 50 | holding period, H3 bars |
| InpPartialTP | 300 | lose half at +N points (0 = off) |
| InpLowRegFrac | 0.5 | lot fraction in the "smooth" market regime (1.0 = off) |
| InpQLow / InpQHigh | 0.20 / 0.80 | entry quantiles |
| InpUseRdFilter | true | trade only near round levels |
The defaults are exactly the settings that produced the figures above. I do not recommend
re-optimising them on your own history: across ~300 trades any parameter grid will find a
"better" value that will not survive forward.
Honest note on risk
No system works forever. This one was positive in all four tested periods (2016-2017,
2018-2020, 2021-2023, 2024-2026), but the past is not a promise. Trade a size that lets you
survive a $455 drawdown per 0.1 lot — and twice that, too.
