|
no
経験
|
5
製品
|
48
デモバージョン
|
|
0
ジョブ
|
0
シグナル
|
0
購読者
|
I'm m a Forex en stocks trader en software developer
Florin Tudor
パブリッシュされたプロダクト
Queen of Gold Ai Ucb Agents は 適応型の MQL5 エキスパートアドバイザーで ルールベースのエントリーロジックと コンテキストに基づく UCB バンディット層を組み合わせたものです 目的は 取引機会を検出するだけでなく 特定の市場状況で最も効果的な行動を学習し その知識をセッション間で保持することです 動作方法 システムは決定論的なシグナルエンジンから開始します 基本トリガーは 連続するローソク足の動きに基づき 逆張りとして解釈されます 連続した弱気の動きは 買いシグナルになる可能性があります 連続した強気の動きは 売りシグナルになる可能性があります その後 このシグナルは より広い市場コンテキストで評価されます システムは次を分析します スプレッド 出来高の質 ティックフローと急激な動き 流動性までの距離と精密なエントリーの可能性 取引セッションの時間 収益環境と市場状態 口座の状態と低残高の制約 これらの特徴は コンテキストキーとしてエンコードされ UCB モデルで使用されます 意思決定層 各コンテキストに対して モデルは以下の行動を評価します
· 4
Florin Tudor
パブリッシュされたプロダクト
キング・オブ・ゴールド AI UCB エージェントは、適応学習と確率的選択を通じて取引判断を最適化するよう設計された高度な機械学習アルゴリズムです。 主な機能： コンテキストフィルタリングとシグナル最適化を備えた UCB デシジョンバンディット トリプルバンディットアーキテクチャ（意思決定層、実行層、プロファイル層） 適応リスク管理による動的ロットサイズ調整 コンテキストの永続的保存と取引履歴の初期化 学習テレメトリ統合およびパフォーマンス追跡 適応型マルチレジーム取引システム 自己学習実行と動的リスク制御 リスクに関する免責事項： 金融市場での取引にはリスクが伴い、すべての投資家に適しているとは限りません。過去の実績が将来の成果を保証するものではありません。 ユーザーはシステムを慎重に評価し、常に適切なリスク管理を適用することを推奨します。 構造的に取引する。 適応的に取引する。 機関的な論理で取引する。
· 4
Florin Tudor
パブリッシュされたプロダクト
マルチアームバンディットAIエージェント – 適応型学習＆意思決定エンジン マルチアームバンディットアルゴリズムによって駆動され、市場状況に応じて継続的に学習・適応する高度なトレーディングシステムです。このエージェントは、パフォーマンスフィードバックに基づき、リアルタイムで取引の実行、ポジションサイズ、および意思決定を最適化します。 意思決定ロジック UCB式を使用：exploit + explore、ただし exploration = C * sqrt(log(totalPulls + 1) / armPulls) コンテキストに基づく事前情報が、過去のパフォーマンスを参考に意思決定を導く STRONGアームは十分なコンテキスト検証後にのみ有効化 HOLDアクションは有利な条件下での取引実行を促すためにペナルティが付与される システムは市場ごとのコンテキストに応じて戦略を動的に進化させ、この最新バージョンでは安定したデータ駆動型の取引行動と向上した一貫性を提供します。
· 4
Florin Tudor
Maximum scalper AI 🔍 File Overview
UCB_STATE_V4 = a snapshot of your bandit algorithm’s state:
Which combinations it has tried
How many times it tried them
The profit/reward for each
Its internal scoring
Basically, it’s the memory of your AI.
🧠 Structure
1️⃣ Header
UCB_STATE_V4
1
→ Version + config info.
2️⃣ Default/Fallback Strategy
-1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY
-1 probably = SELL
AGGRESSIVE + GREEDY = style
Trades = 0, Profit = 0 → never used
3️⃣ Important strategies that were actually used
1|1|1|1|1|0|2|0|0|NORMAL|MODERATE|EXEC
96 trades
Profit: 37.42
Score: 0.40
🔥 This is your main winning strategy
Chosen 96 times
Positive profit and positive score → bandit prefers this
-1|1|1|1|1|0|2|0|0|NORMAL|MODERATE|EXEC
22 trades
Profit: 13.32
Score: negative
Positive profit but negative score → probably high drawdown or risk
4️⃣ Other strategies
0 trades
0 profit
Not explored yet or never chosen
5️⃣ Symbol statistics
EURUSD
118 trades
Win rate ~55%
The bandit is learning, not random
6️⃣ Decision history
timestamp | chosen strategy | reward
Shows when a strategy was chosen and the outcome
❗ Key Problem (why it’s not working in Tester)
Strategy Tester in MT5 runs in a sandbox
DLL cannot write anywhere freely
It only writes to:
MQL5\Files\
Tester\Agent-xxx\MQL5\Files\
✅ Live accounts don’t have this restriction → that’s why it works there.
🔥 Recommended Testing Environments
Demo account – works like live, writes files normally
Forward testing on chart – watch bandit evolution
VPS – run continuously, bandit keeps learning
Conclusion:
Your bandit is working (96 trades + scores)
The algorithm is fine
The limitation is MT5 Strategy Tester, not the AI itself
UCB_STATE_V4
1
-1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY 0 1774229303 0 0.00000000 0.00000000 0 0.00000000 0.00000000 0 0.00000000 0.00000000 0 0.00000000 0.00000000
3
1|1|1|1|1|0|2|0|0|NORMAL|MODERATE|EXEC 96 1774229084 96 37.42000000 0.40015989 0 0.00000000 0.00000000
-1|1|1|1|1|0|2|0|0|NORMAL|MODERATE|EXEC 22 1774229084 22 13.32000000 -0.16765199 0 0.00000000 0.00000000
-1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY|EXEC 0 1774229303 0 0.00000000 0.00000000 0 0.00000000 0.00000000
6
0|1|0|0|2|0|0|AGGRESSIVE|GREEDY 0 1774229304 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|2|0|0|2|0|0|AGGRESSIVE|GREEDY 0 1774229363 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|2|0|0|2|0|0|AGGRESSIVE|CONSERVATIVE 0 1774229408 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|2|0|0|2|0|0|AGGRESSIVE|MODERATE 0 1774229415 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|1|2|0|1|2|0|AGGRESSIVE|CONSERVATIVE 0 1774596950 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|1|2|1|1|2|0|AGGRESSIVE|CONSERVATIVE 0 1774597077 0 0.00000000 0.00000000 0 0.00000000 0.00000000
1
EURUSD 1 118 82 0.55416158 0.12497833 50.00000000
2
EURUSD 1 1 1 1 1 0 96 66 37.42000000 -16.84000000
EURUSD 1 -1 1 1 1 0 22 16 13.32000000 25.44000000
0
2
1547155986 -1 -1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY 1 5.00000000 -1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY|EXEC 0 1774229304
1547159109 -1 -1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY 1 5.00000000 -1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY|EXEC 0 1774229361
UCB_STATE_V4 = a snapshot of your bandit algorithm’s state:
Which combinations it has tried
How many times it tried them
The profit/reward for each
Its internal scoring
Basically, it’s the memory of your AI.
🧠 Structure
1️⃣ Header
UCB_STATE_V4
1
→ Version + config info.
2️⃣ Default/Fallback Strategy
-1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY
-1 probably = SELL
AGGRESSIVE + GREEDY = style
Trades = 0, Profit = 0 → never used
3️⃣ Important strategies that were actually used
1|1|1|1|1|0|2|0|0|NORMAL|MODERATE|EXEC
96 trades
Profit: 37.42
Score: 0.40
🔥 This is your main winning strategy
Chosen 96 times
Positive profit and positive score → bandit prefers this
-1|1|1|1|1|0|2|0|0|NORMAL|MODERATE|EXEC
22 trades
Profit: 13.32
Score: negative
Positive profit but negative score → probably high drawdown or risk
4️⃣ Other strategies
0 trades
0 profit
Not explored yet or never chosen
5️⃣ Symbol statistics
EURUSD
118 trades
Win rate ~55%
The bandit is learning, not random
6️⃣ Decision history
timestamp | chosen strategy | reward
Shows when a strategy was chosen and the outcome
❗ Key Problem (why it’s not working in Tester)
Strategy Tester in MT5 runs in a sandbox
DLL cannot write anywhere freely
It only writes to:
MQL5\Files\
Tester\Agent-xxx\MQL5\Files\
✅ Live accounts don’t have this restriction → that’s why it works there.
🔥 Recommended Testing Environments
Demo account – works like live, writes files normally
Forward testing on chart – watch bandit evolution
VPS – run continuously, bandit keeps learning
Conclusion:
Your bandit is working (96 trades + scores)
The algorithm is fine
The limitation is MT5 Strategy Tester, not the AI itself
UCB_STATE_V4
1
-1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY 0 1774229303 0 0.00000000 0.00000000 0 0.00000000 0.00000000 0 0.00000000 0.00000000 0 0.00000000 0.00000000
3
1|1|1|1|1|0|2|0|0|NORMAL|MODERATE|EXEC 96 1774229084 96 37.42000000 0.40015989 0 0.00000000 0.00000000
-1|1|1|1|1|0|2|0|0|NORMAL|MODERATE|EXEC 22 1774229084 22 13.32000000 -0.16765199 0 0.00000000 0.00000000
-1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY|EXEC 0 1774229303 0 0.00000000 0.00000000 0 0.00000000 0.00000000
6
0|1|0|0|2|0|0|AGGRESSIVE|GREEDY 0 1774229304 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|2|0|0|2|0|0|AGGRESSIVE|GREEDY 0 1774229363 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|2|0|0|2|0|0|AGGRESSIVE|CONSERVATIVE 0 1774229408 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|2|0|0|2|0|0|AGGRESSIVE|MODERATE 0 1774229415 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|1|2|0|1|2|0|AGGRESSIVE|CONSERVATIVE 0 1774596950 0 0.00000000 0.00000000 0 0.00000000 0.00000000
0|1|2|1|1|2|0|AGGRESSIVE|CONSERVATIVE 0 1774597077 0 0.00000000 0.00000000 0 0.00000000 0.00000000
1
EURUSD 1 118 82 0.55416158 0.12497833 50.00000000
2
EURUSD 1 1 1 1 1 0 96 66 37.42000000 -16.84000000
EURUSD 1 -1 1 1 1 0 22 16 13.32000000 25.44000000
0
2
1547155986 -1 -1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY 1 5.00000000 -1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY|EXEC 0 1774229304
1547159109 -1 -1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY 1 5.00000000 -1|2|0|0|0|0|2|0|0|AGGRESSIVE|GREEDY|EXEC 0 1774229361
· 3
Florin Tudor
Wat is this ? MQL5 spam rewiew from clients?ho else get this from MQL5 WEB?is not normal to manipulate the clients
· 2
Florin Tudor
Aa wait for the deep but the agents will go whit the trend ,the gift 🎁 is verry advance so my clients will receive this multi arm bandit agents EA for Free!
· 2
Florin Tudor
My Agents are very bizzy:)) the market is very choppy:)) but no problem sir we must make 🤑💰 for master Tudor:)
· 3
Florin Tudor
inside the King of gold en Maximum Scalper the workers the Ai Agents are working day en night ,the file/data folder/mql5/Files is where you will see the agents keep the information en model also wat simbol they work on ,enjoy this is made by a trader for traders!!
· 3
Florin Tudor
🎁 Maximum Scalper In 1 hours ,One day free Download for all clients to be able to tested out on Demo or Cent account!after 2 days the price will come back on ! Test the Agents en if you find something Rong just message me !
· 3
Florin Tudor
🎁 Maximum Scalper In 2 hours One day free Download for all clients to be able to tested out on Demo or Cent account!
I will make some set.files in comments!!
I will make some set.files in comments!!
· 3
Florin Tudor
Maximum Scalper AI has evolved from a simple rule-based reversal scalper into a locally adaptive MQL5 Expert Advisor with persistent AI-driven decision logic.
This upgrade introduces a self-learning multi-layer bandit system that adapts directly from real trade outcomes.the agent is build in so it will make his files for every simbol is trading in your mql5 files data folder
This upgrade introduces a self-learning multi-layer bandit system that adapts directly from real trade outcomes.the agent is build in so it will make his files for every simbol is trading in your mql5 files data folder
· 2
Florin Tudor
Maximum Scalper AI has been upgraded from a rule-based reversal scalper into a locally adaptive MQL5 expert advisor with persistent AI-style decision memory.
The update adds a multi-layer local bandit framework that learns directly from trade outcomes without any DLL dependency. The EA now uses:
a UCB decision bandit for HOLD / SMALL / NORMAL / STRONG
an execution bandit for MARKET / LIMIT
an auto-close bandit for DIRECT_CLOSE / TRAIL
Its learning context is no longer simplistic. It now evaluates signals using:
consecutive bullish/bearish bar structure
volume regime and volume ratio
volume profile bias
spread conditions
session timing
adaptive performance mode
greed/conservative signal state
equity vs balance condition
floating drawdown stress
safe mode state
The EA also gained persistent on-disk learning memory, so it can retain what it has learned across restarts. It stores:
UCB trade-decision memory
execution memory
auto-close memory
symbol/timeframe performance memory
setup-pattern statistics
active and pending contextual link
The update adds a multi-layer local bandit framework that learns directly from trade outcomes without any DLL dependency. The EA now uses:
a UCB decision bandit for HOLD / SMALL / NORMAL / STRONG
an execution bandit for MARKET / LIMIT
an auto-close bandit for DIRECT_CLOSE / TRAIL
Its learning context is no longer simplistic. It now evaluates signals using:
consecutive bullish/bearish bar structure
volume regime and volume ratio
volume profile bias
spread conditions
session timing
adaptive performance mode
greed/conservative signal state
equity vs balance condition
floating drawdown stress
safe mode state
The EA also gained persistent on-disk learning memory, so it can retain what it has learned across restarts. It stores:
UCB trade-decision memory
execution memory
auto-close memory
symbol/timeframe performance memory
setup-pattern statistics
active and pending contextual link
· 3
Florin Tudor
The King of Gold.mq5 file (version 9.08) has been integrated with UCB (Upper Confidence Bound) Bandit AI - an advanced machine learning algorithm that optimizes trading decisions.
Core Features
UCB Decision Bandit with contextual filtering and signal boosting
Triple-Bandit Architecture (Decision, Execution, Profile bandits)
Dynamic lot multipliers and smart risk management
Persistent context savings and trade history warm-up
Complete learning telemetry
Core Features
UCB Decision Bandit with contextual filtering and signal boosting
Triple-Bandit Architecture (Decision, Execution, Profile bandits)
Dynamic lot multipliers and smart risk management
Persistent context savings and trade history warm-up
Complete learning telemetry
· 3
Florin Tudor
On BTC the new version of King of Gold AI performing verry well in first day of testing
· 2
Florin Tudor
The king of Gold MQL5 file will also be updated whit AI decision engine to fit MQL5 market requirements, after testing on cent account will be uploaded
· 2
Florin Tudor
First Day in testing ,the mistake he made will be remember en the good outcome also for every simbol is trading separately
· 3