フリーランスセクション

Title: Improve signal quality and frequency of a Python XAUUSD signal engine (realistic backtest required)

Python エキスパート 設計 ストラテジーの最適化 Python 製品デザイン

指定

PROJECT
GoldPulse is a live XAUUSD (gold) signal platform. Signals are generated by a
Python engine (FastAPI backend, MT5 data feed) and sent to subscribers on the
website and Telegram. Subscribers execute trades manually.

THE PROBLEM
- The engine produces few signals: about 0.7 setups per day, and sometimes
  nothing for several days.
- After realistic costs (spread + slippage), results are around breakeven
  or slightly negative (profit factor ~0.9).
- The engine has 12 analyzers chained as AND-gates. Many of them re-check the
  same ideas (swing structure, BOS/CHoCH, liquidity sweep, MTF alignment),
  so most candidates are rejected and the survivors are not selected by quality.
- A quality-grade layer already exists. A small subset of signals (with H4 trend,
  pullback entry, strong S/R level with 3+ touches, not from FVG, London/NY session)
  performs better, but it is too rare.
- Already tested and rejected with realistic costs: lowering the score threshold,
  relaxing counter-trend rules, and pure market entry (it fails because
  subscribers enter manually with a delay).

WHAT YOU WILL RECEIVE (signal files only)
- app/services/gold_analysis_service.py  (12 analyzers + evaluate)
- app/engines/  (S/R levels, structure, etc.)
- app/services/signal_service.py, signal_state_service.py, reversal_protection.py
  (signal manager: one active slot, duplicate protection, cooldown)
- app/services/outcome_service.py  (trade outcome measurement)
- app/services/signal_quality.py  (A/B quality grade)
- app/config.py  (sanitized: no secrets)
- XAUUSD candle data M15 / H1 / H4 / D1, June 2022 to September 2026 (UTC)
- A realistic measurement tool (parity-checked against the live engine)
  and previous test reports
No live server access, no database, no keys.

THE GOAL
At least 2 signals per day, while keeping the quality: profitable after
realistic costs. You may redesign the selection logic (not only tune thresholds).

ACCEPTANCE CRITERIA (measured with the provided tool, no exceptions)
1. Costs: spread 0.30$ + slippage 0.30$ on entry and on stop loss.
2. Management: close 50% at TP1 + move SL to breakeven; manager layer included.
3. Build period 2022-06 to 2024-12. Validation 2025-01 to 2026-09 must not be
   used for tuning; it is run once at the end.
4. Average R per trade > 0 after costs, and the 95% weekly confidence interval
   lower bound above zero.
5. Improvement in at least 75% of half-year periods.
6. Still positive after removing the 3 best trades.
7. Max drawdown not worse than the current engine by more than 20%.
8. At least 2 signals per day on average. Entries must suit manual traders
   (pending/limit orders, or a clear price range).
9. No look-ahead: closed candles only.
10. All changes behind .env flags, OFF by default, clean diff, no changes outside
    the signal files.

DELIVERABLES
- Modified files + a unified diff
- A short report with the full results table (build and validation separately)
- The test scripts used, so results can be reproduced

BUDGET
[ضع المبلغ هنا] USD

NDA required. The code and the data are confidential and may not be reused.
Please describe your experience with Python backtesting of trading systems
and with avoiding curve-fitting.

応答済み

1
開発者 1
評価
(287)
プロジェクト
359
29%
仲裁
37
27% / 65%
期限切れ
10
3%
取り込み中
パブリッシュした人: 10 codes
2
開発者 2
評価
(102)
プロジェクト
124
31%
仲裁
7
29% / 29%
期限切れ
5
4%
取り込み中
3
開発者 3
評価
(16)
プロジェクト
35
23%
仲裁
4
0% / 50%
期限切れ
2
6%
仕事中
4
開発者 4
評価
(1)
プロジェクト
1
0%
仲裁
0
期限切れ
0
仕事中
パブリッシュした人: 2 codes
5
開発者 5
評価
(2)
プロジェクト
2
100%
仲裁
0
期限切れ
0
暇
パブリッシュした人: 2 codes
6
開発者 6
評価
(19)
プロジェクト
22
18%
仲裁
9
33% / 44%
期限切れ
3
14%
仕事中
パブリッシュした人: 1 code
7
開発者 7
評価
(2)
プロジェクト
2
0%
仲裁
1
0% / 100%
期限切れ
0
暇
8
開発者 8
評価
(11)
プロジェクト
18
33%
仲裁
0
期限切れ
1
6%
仕事中
9
開発者 9
評価
プロジェクト
0
0%
仲裁
0
期限切れ
0
暇
10
開発者 10
評価
(54)
プロジェクト
67
42%
仲裁
15
27% / 60%
期限切れ
1
1%
仕事中
11
開発者 11
評価
(4)
プロジェクト
7
57%
仲裁
1
100% / 0%
期限切れ
0
暇
類似した注文
HELP WITH REMOVING A BUG/INFINITE LOOP FROM A CUSTOM INDICATOR 30 - 50 USD
Please may someone assist me with the indicator attached, it has a SERIOUS BUG issue in the sense that when I load it, and activate the template (attached as well) MT4 ALWAYS freezes and then eventually disconnects automatically, ALL THE TIME. It even gets worse to an extent that after disconnecting a number of times mt4 ends up not loading anymore, UNTIL I remove the indicator. Please I WILL NEED A DEMO (FOR A DAY)
Volume Integration for the Expert Advisor . Apply only if you have experience with optimizing core volume implementation . 30 - 50 USD
Most of the expert is already coded . Create a panel to show the Ratio for the Range . Trigger is based on ratio . Show Distribution / project starts after we clear Distribution

プロジェクト情報

予算
30 - 150 USD
締め切り
最低 1 最高 5 日

依頼者

(10)
出された注文14
裁定取引数0