BTCUSD Consensus EA MT5 selects Bitcoin trades on H1 through five embedded model ensembles and market-regime analysis. Leonid Arkhipov’s system limits itself to one position and three entries per UTC day, with equal ATR-based stop/target distances and a time exit. Its panel exposes model scores and the execution gates behind each trading decision.
Our assessment: this is a model-based Bitcoin EA with unusually detailed public documentation and downloadable tester reports. It suits traders who want to examine fixed entry logic and broker settings. The main considerations are default ×2 Recovery after losses and evidence drawn entirely from historical testing; the large simulated returns are not a monitored live record.
View BTCUSD Consensus EA MT5 features and pricing
We inspected the MQL5 listing, its 38-page English guide ZIP, release notes, comments and both original tester reports on October 5, 2026. We have not independently run the EA or audited the training pipeline. Original developer images appear below with their dates and evidence limits; the cover is separate promotional artwork.
What the five models actually do
The guide describes four base ensembles assessing long and short opportunities, plus a SELL META ensemble providing additional short validation. Together they contain 1,360 trained decision trees. Their inputs include closed H1 price structure, returns, volatility, EMA relationships and the four M15 quarters within an hour. The base feature set contains 61 inputs; SELL META adds the other model scores and regime flags.
Regime classification distinguishes upward, downward, flat, compression, expansion and transition conditions. A regime influences entry thresholds rather than forcing a trade in that direction. Insufficient model agreement or a failed execution gate leaves the system waiting. A large model count is a design detail, not independent proof of predictive accuracy.
These models are embedded and pretrained. According to the guide, they do not learn online from your account and do not require external Python, DLLs or cloud AI for decisions. Internal model thresholds and core stop/target rules are fixed in this build; user inputs primarily govern execution, risk, broker time and the interface.
One position, ATR exits and an informative dashboard
- Market
- BTCUSD H1 on one MT5 chart, with internal H1/M15 analysis and supported symbol suffixes.
- Trade constraints
- One Bitcoin position at a time; maximum three entries per UTC day. The guide describes no grid or averaging against a losing move.
- Initial exits
- Stop loss and take profit each use 1.0 signal-H1 ATR, giving approximately 1:1 planned distances.
- Time exit
- Close after 12 hours if neither stop nor target has been reached.
Actual trade outcomes need not finish at exactly +1R or −1R: costs, execution and the timed exit can change the realized result. The broker-validation gate can reject an entry when the intended stop or tick-size requirements cannot be met, rather than widening the validated stop to fit.
The panel shows regime, model evaluations, H1 cycle, consensus, risk/load, filters, open position and account information. Closed-trade cards help inspect individual outcomes. The original interface illustration uses historical candles; an updating panel described as a “live interface” is not evidence of live-money performance.
Default Recovery changes the meaning of 1% risk
Auto Risk uses equity and the monetary value of the intended stop; fixed-lot sizing is an alternative. Defaults are a 1% base Auto Risk setting, a 0.01 fixed lot and Maximum Margin Use of 85%. Broker minimum/maximum volume, lot steps, contract size and margin can alter the calculated position.
- After a loss
- Recovery, enabled by default, multiplies subsequent volume by ×2.
- After another loss
- Volume remains at ×2; it does not escalate to ×4, ×8 or further doubling.
- After a profit
- Return to the base sizing.
- Recovery disabled
- Remove this increased-volume step; signals and stop/target rules remain the same.
This is a bounded increase in sizing after losses, so the base 1% setting should not be advertised as a fixed maximum on every trade. A Recovery trade can approximately double planned stop risk before broker and margin adjustments. No-grid operation does not mean there is no recovery sizing.
The 85% margin limit controls volume using margin allocation; it is not a cash-loss or drawdown ceiling. One-position operation prevents a grid of simultaneous positions, but consecutive losing trades still accumulate account losses. Inspect actual cash risk and the Recovery state before evaluating a result as a constant-risk trade.
Broker time and spread behavior need careful setup
Default GMT Offset +2 and DST enabled correspond to winter UTC+2 and summer UTC+3. The models normalize broker timestamps to a common UTC grid. Incorrect settings shift H1/M15 candle boundaries and can change the features and signals. A broker using UTC+0 throughout the year instead needs offset 0 and DST disabled.
Spread protection compares Ask−Bid with ATR, using a default maximum of 10% of ATR. The manual discloses an important historical difference: tester entries before 2026 intentionally bypass the rejection filter, although real Bid/Ask spread still affects execution. From 2026 onward the tester applies the enabled filter normally. The multi-year test therefore does not apply identical spread-rejection behavior throughout its history.
The developer recommends low-spread ECN/RAW-type conditions and testing on the intended broker’s real-tick history. His 500 USD deposit guidance assumes a 0.01 minimum lot and 1:500 leverage; different Bitcoin contracts and margin requirements can make that guidance unsuitable. Portability across broker symbols is different from identical fills or results. A stable VPS supports terminal continuity, but does not eliminate slippage or price-feed differences.
The long historical report differs from the promotional table
The original long tester report ZIP covers May 8, 2017–October 2, 2026 on BTCUSD H1 using 492,849,560 Dukascopy real ticks. It uses a USD10,000 initial deposit, 1:100 leverage, Auto Risk 1% and Recovery enabled at ×2. The report contains 1,532 trades, with 1,485 winners and 47 losers (96.93% wins).
- Gallery presentation
- Profit factor 64.01 and relative equity drawdown 5.14%.
- Downloaded report
- Profit factor 64.76 and relative equity drawdown 5.13%; relative balance drawdown is 3.05%.
- Report net profit
- USD45,491,981.17 from the simulated compounding run, not verified live income.
We retain the promotional image unchanged and disclose the differing report values rather than silently treating them as identical. The extraordinary compounded ending result depends on the simulation, sizing and historical model development; it should not become a target for a buyer’s account.
The developer describes rolling periods of approximately six months of training/optimization followed by six months of forward testing. We have not audited the raw training matrices or pipeline. That description does not establish that the entire displayed multi-year curve is unseen data, and historical price ticks do not independently validate future model performance.
The shorter test: claimed unseen data, still a simulation
The developer says training ended August 31, 2026 and September 1–October 2 was excluded from training and optimization. The original shorter report ZIP confirms that testing date range and its reported outcomes. It does not independently establish how the training split was enforced.
From a USD10,000 starting deposit at 1:100 leverage, the report shows USD3,503.49 net profit, 42 trades, 34 winners and eight losers (80.95%), with profit factor 4.54. Relative equity drawdown is 3.70% and relative balance drawdown 2.93%. Average winning and losing trades are USD132.14 and USD123.66 respectively. The settings again enable 1% base Auto Risk with ×2 Recovery.
This roughly one-month, 42-trade test is more focused than the whole-period curve, but it remains developer historical testing. Its 35.03% period gain is not a promised monthly rate, and the lower win rate illustrates how sharply a smaller sample can differ from the long aggregate.
Live evidence and the October VPS fix
No BTCUSD Consensus live signal was linked in the listing, comments, guide or seller page at this check. The seller’s Aurum signal belongs to a different EA; we do not present its growth or drawdown as Bitcoin evidence. There were also no public product reviews to capture or import as customer ratings.
A public October 3 support exchange concerned the EA’s large file failing to migrate to MQL5 VPS. The developer reduced embedded graphic resources, and the buyer reported that synchronization was resolved. The version 1.65 release notes confirm the resource optimization and state that models, signals and trading parameters were unchanged. This is feedback about one technical fix, not profitability evidence or a guarantee for every VPS.
Other products from Leonid Arkhipov
Compare Ghost XAU EA MT5 and its shared edition review, or Aurum AI EA MT5 and the Aurum guide. For Forex majors, see GBP Machine Pro with its GBP review, and EUR Machine Pro with its EUR review. These are separate systems with different market exposure and evidence.
Is BTCUSD Consensus worth considering?
The strongest practical features are the documented model ensemble, visible execution gates and controlled number of positions. The public guide and original reports allow a closer inspection than a promotional curve alone. A sensible assessment still needs correct broker time, actual contract risk and an explicit decision about default Recovery.
The current evidence supports studying the design and testing it under the intended broker’s conditions. It does not establish a live track record or justify expecting the historical win rate and compounded gains to continue.
