AI FX Bot Lab: Real Trading Experiments

Kimi | Japan FX Bot Lab

0.0 (0)
Tech News
Updated Daily

Can AI really trade forex? AI FX Bot Lab is a real-time experiment from Japan, where I build and test AI-assisted FX trading bots using MT5, Python, machine learning, and local LLM tools. I share live results, failures, risk lessons, and bot improvements from rule-based, AI-driven, and ML + LLM hybrid systems. Not financial advice. fxaibotlab.substack.com

15h ago

Why AI Trading Bots Need Brakes: The June 1–5 Weekly Review

In today’s episode, we review the weekly performance of our four MT5 automated trading bots from June 1 to June 5. The portfolio ended the week with a combined realized loss of -3,307 JPY. While it wasn’t a profitable week financially, it was arguably our most valuable week for system development. We break down each bot’s behavior to understand why giving AI complete autonomy is a risky game, and why the ultimate feature of a trading bot is a reliable “brake”: * GateGrid AI: The clear winner of the week, finishing at +707 JPY. Its multi-layered filtering system proved that a bot’s true power lies not only in finding entries, but in its ability to say “do nothing” and avoid bad trades. * BoundSniper: Finished at -868 JPY. As a pure execution bot, its losses confirmed that the MT5 execution layer is doing its job, but the upstream TradingView signal logic needs serious refinement and better filtering. * LLMBridgeTrader: Took a hard hit at -1,399 JPY. It clearly demonstrated that giving an AI full autonomy over position management (OPEN, HOLD, CLOSE, REVERSE) is dangerous without a strict “ML gate” to reject weak trading plans before they reach the market. * MLScore GF-T4: Ended at -1,747 JPY. It exposed a critical structural flaw: re-entering the market under the same unfavorable conditions immediately after a stop-loss. It highlighted the urgent need for re-entry logic, cooldown rules, and daily risk limits. The biggest takeaway from this week? AI can create brilliant trading plans, but the system still needs the final authority to hit the brakes. Join us as we discuss how we are using this week’s “valuable losses” as direct training and debugging data to build smarter, safer trading systems! #FX #MT5 #AITrading #AlgorithmicTrading #MachineLearning #SystemTrading This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit fxaibotlab.substack.com

2 min
1d ago

Turning Losses Into Training Data: The Evolution of AI Bots [June 5th Trade]

In today’s episode, we break down the June 5th parallel test of our four MT5 automated trading bots. The portfolio finished the session with an overall loss of -1,068 JPY. At first glance, it looks like a tough day. But for an AI-driven project, a red day with a clear diagnosis is far more valuable than a lucky green day. We dive into the distinct behaviors of each system and the major structural upgrades they inspired: * GateGrid AI (GBPUSD): The only profitable bot today, securing +199 JPY. Its conservative, multi-layered decision structure (combining CatBoost and Ollama) proved its worth by taking small profits and effectively staying out of trouble. * BoundSniper (USDJPY): Finished with a minor -100 JPY loss. As a pure execution bridge, its loss simply tells us that the upstream TradingView signal logic needs better exit controls, rather than indicating an execution failure. * LLMBridgeTrader (EURUSD): Took the hardest hit at -675 JPY. The AI’s immense freedom became a liability. In response, we discuss our massive upgrade: implementing a Machine Learning (ML) Gate powered by CatBoost to strictly filter the LLM’s “OPEN” and “REVERSE” trade plans before they reach MT5. * MLScore GF-T4 (GBPJPY): Ended at -492 JPY, but received the biggest structural overhaul. We’ve upgraded this bot to differentiate between “Breakout” (trend-following) and “Range” (mean-reversion) setups. With new historical backfill data, strategy-specific TP/SL settings, and strict daily safety limits, it’s evolving from a bot that simply guesses into a bot that learns from its logs. The ultimate lesson from today’s session is that our systems are shifting toward a new phase of development. Join us as we discuss how we are literally turning today’s financial losses into tomorrow’s training data! #FX #MT5 #AITrading #MachineLearning #AlgorithmicTrading #SystemTrading This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit fxaibotlab.substack.com

2 min
2d ago

Small Wins, Heavy Homework: Why Autonomous AI Needs Brakes [June 4th Trade]

In today’s episode, we break down the June 4th parallel test of our four MT5 automated trading bots. The portfolio ended the session with a total realized loss of -550 JPY, or -916 JPY when factoring in open positions. While the result wasn’t dramatic, the day provided a crystal-clear split between our systems: the strictly filtered bots won, while the autonomous AI bots struggled. We dive into the completely different behaviors of each bot to uncover why AI needs strict boundaries: * GateGrid AI (GBPUSD): Delivered the cleanest performance of the day. It secured two wins for +197 JPY and ended the session completely flat with no open exposure. It perfectly executed what a grid-style bot should do: get in, get out, and avoid unnecessary risks. * BoundSniper Bot (USDJPY): Finished in the green at +38 JPY. Acting as a simple executor for TradingView signals, it took an early hit but successfully recovered through a 75% win rate across four trades. * LLMBridgeTrader (EURUSD): Ended with a -281 JPY realized loss. As our most autonomous bot—capable of deciding whether to open, hold, close, or reverse—its flexibility became its downfall today. The AI’s decisions failed to produce a stable expectancy, proving that it desperately needs stricter filtering around confidence and stop-loss distances. * MLScore GF-T4 GB (GBPJPY): Took the heaviest hit, suffering a combined realized and floating loss of -828 JPY. The biggest issue wasn’t just the stop-loss; it was the fact that the bot immediately re-entered the market under the same difficult conditions. It highlighted the urgent need for a “cooldown rule” to prevent immediate re-entries after large losses. The ultimate takeaway from today’s session is simple but profound: Automation should not only decide when to enter. It must also know when not to continue. Giving AI freedom is powerful, but without structured risk filters and “brakes,” that freedom can quickly destroy your edge. Join us as we discuss the “heavy homework” ahead and how we plan to build these crucial safety nets for our autonomous bots! #FX #MT5 #AITrading #AlgorithmicTrading #RiskManagement #MachineLearning This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit fxaibotlab.substack.com

2 min
3d ago

Learning Starts With Losses: Turning a Red Day into Training Data [June 3rd Test]

In today’s episode, we break down the June 3rd parallel test of our four MT5 automated trading bots. The portfolio ended the session with a combined result of -999 JPY. At first glance, it looks like a simple losing day, but for our AI and machine-learning-based systems, these results provide invaluable training material. We dive into the completely different profiles of each bot to see what we learned: * GateGrid AI (GBPUSD): The cleanest and strongest performer of the day, securing +133 JPY. It perfectly demonstrated its selective design philosophy by taking exactly one trade, winning it, and leaving no open exposure. It proved that a bot’s real value often lies in deciding when not to enter. * BoundSniper (USDJPY): Finished at -584 JPY. Despite having a high win rate with 5 winning exits and 2 losing exits, the losses were simply too large. It serves as a stark reminder that a good win rate means nothing if your average loss isn’t strictly controlled. * LLMBridgeTrader (EURUSD): Ended slightly negative at -147 JPY. Because this bot relies on high AI autonomy (deciding to OPEN, HOLD, CLOSE, or REVERSE), today’s results showed that it still needs stricter guardrails and better confidence filtering around its stop-loss placement. * MLScore (GBPJPY): Closed at -401 JPY. It had two winning exits, but one oversized loss dominated the day. However, because MLScore accumulates learning data, this specific loss is crucial feedback that will help the bot avoid similar bad setups in the future. The biggest takeaway from today’s session is that for bots like GateGrid and MLScore, every trade is feedback. A losing day might be painful, but if the logs are used to refine the models, today’s losses will literally become tomorrow’s filters. Join us as we discuss how we turn a red day into smarter trading logic! #FX #MT5 #AITrading #AlgorithmicTrading #MachineLearning #SystemTrading This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit fxaibotlab.substack.com

2 min
4d ago

Valuable Losses: What a Red Day Teaches Us About AI Trading [June 2nd Trade]

In today’s episode, we break down the June 2nd parallel test of our four MT5 automated trading bots. It was a tough session across the board, with the total realized loss hitting -1,429 JPY, and the overall equity impact reaching -1,567 JPY when including floating losses. However, from a system-evaluation perspective, it was an incredibly useful day. Today, the main question wasn’t about who won, but rather: “Which bot lost in the most controlled way?”. We dive deep into the completely different loss profiles of each bot to understand their structural weaknesses and strengths: * LLMBridgeTrader (EURUSD): The winner among the losing bots. It ended with the smallest realized loss of -185 JPY. It successfully demonstrated that its risk management can contain the damage when AI judgments or market conditions turn unfavorable. * GateGrid AI (GBPUSD): Finished at -206 JPY. While it showed resilience by securing small wins (+81 JPY and +19 JPY) earlier in the day, a single larger loss of -306 JPY pushed it into negative territory, highlighting the importance of preventing one bad trade from overpowering multiple wins. * BoundSniper Bot (USDJPY): Ended at -438 JPY. Since its job is purely to execute TradingView signals, today’s drawdown was not an execution failure, but a signal-quality issue. It serves as a reminder that upstream logic needs robust filters for choppy or reversing sessions. * MLScore GF-T4 GB (GBPJPY): The main source of today’s drawdown, closing with a -600 JPY realized loss and carrying a -138 JPY floating loss for a total impact of -738 JPY. We discuss why its risk-reward structure requires an urgent review, especially since the reward target is relatively tight compared to the stop range. Join us as we discuss why we aren’t stopping the test, but instead tightening our review loop. Because in automated trading, controlled losses are often far more valuable for improving systems than easy profits. #FX #MT5 #AITrading #AlgorithmicTrading #RiskManagement #TradingStrategy This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit fxaibotlab.substack.com

5 min
5d ago

The Power of Diversification: Why You Need Multiple Bots [June 1st Test]

In today’s episode, we break down the June 1st parallel test of our MT5 automated trading bots. The portfolio ended the day with a solid realized profit of +739 JPY (or +757 JPY including floating profit), winning an impressive 12 out of 13 closed trades. But the real story isn’t just about an easy winning day—it’s about how portfolio diversification protected our profits. We dive into the performance of each bot to see how their distinct architectures worked together: * GateGrid AI (GBPUSD): The strongest performer of the day. It secured +384 JPY across 4 flawless wins. Its complex design—combining model-based filtering, local AI judgment, and volatility checks—proved that its greatest strength is effectively avoiding low-quality entries. * MLScore GF-T4 (GBPJPY): Delivered the cleanest execution. It took one single trade and successfully hit its take-profit for +250 JPY, leaving no open positions or floating risks behind. * BoundSniper (USDJPY): Quiet and consistent. Acting as a disciplined rule-based executor, it closed 5 winning trades for +216 JPY. It proved once again that this bot’s true value lies in its strict discipline rather than complex intelligence. * LLMBridgeTrader (EURUSD): The only bot to struggle, ending with a -111 JPY realized loss. Despite winning two out of three trades, a single large stop-loss outweighed its combined gains, highlighting the ongoing challenge of risk asymmetry when an AI acts as a trading planner. The biggest lesson from today’s session is clear: a single AI bot can be fragile, but a diversified group of bots is resilient. Because we ran rule-based execution, AI planning, machine-learning scoring, and grid-style filtering simultaneously, the overall portfolio easily absorbed LLMBridgeTrader’s weak performance and remained comfortably positive. Join us as we discuss why a multi-bot structure makes individual weaknesses easier to see and manage, and why an imperfect day can still be a highly useful win. #FX #MT5 #AITrading #AlgorithmicTrading #Diversification #RiskManagement This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit fxaibotlab.substack.com

4 min
6d ago

The Power of Skipping: Introducing ML_ScoreAnalyst for GBPJPY

In today’s episode, we introduce the newest addition to our automated trading lineup: ML_ScoreAnalyst. Running on a real account, this bot targets the highly volatile GBPJPY 15-minute chart (M15). However, the core focus of this episode isn’t about how often it trades, but rather why it rejects most of the setups it finds. We dive into the architecture of this new “score-based” trading bot and what makes it fundamentally different from our previous systems: * Candidate vs. Decision (The Scoring System): ML_ScoreAnalyst doesn’t just blindly fire orders when a condition is met. First, it identifies a breakout candidate based on price action and ATR. Then, it passes that data to a CatBoost machine learning model, which calculates an entry probability score from 0 to 100. If the score doesn’t beat our strict threshold, the bot logs the decision as a “SKIP” and stays out of the market. * Built for Continuous Evolution: This is not a “set and forget” system. Every decision and its underlying market features are saved to a CSV log. This creates a continuous feedback loop where we can back-test different Stop Loss (SL) and Take Profit (TP) combinations, label the outcomes, and retrain the CatBoost model with verified live data. * Multi-Layered Safety Checks: Operating on a live account requires extreme caution. We discuss the strict, multi-layered safety protocols built into the bot, including preventing duplicate orders on the same candle, checking broker filling modes, and strictly isolating Dry Run, Demo, and Real account environments to prevent catastrophic execution errors. The ultimate value of a trading bot isn’t just in its entry logic, but in its ability to selectively stay out of unfavorable conditions. Join us as we explore the mechanics of ML_ScoreAnalyst and why an intelligent “SKIP” is often the most profitable decision a bot can make. #FX #MT5 #MachineLearning #AlgorithmicTrading #CatBoost #SystemTrading #GBPJPY This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit fxaibotlab.substack.com

4 min
May 31

Simple Rules Beat Advanced AI: The May 25–29 Weekly Bot Review

Simple Rules Beat Advanced AI: The May 25–29 Weekly Bot Review In today’s episode, we review the weekly performance of our three MT5 auto-trading bots from May 25th to May 29th. The portfolio ended the week in the red at -1,560 JPY, but the contrasting results gave us a fascinating and unexpected insight: sometimes, simple rules completely outperform advanced AI. We break down the distinct behaviors of each bot to understand what went right and what went wrong: * BoundSniper Bot: The only profitable bot this week, securing +393 JPY. Acting purely as a rule-based executor with no AI, it stuck faithfully to TradingView signals without hesitation. It proved that simple execution can avoid large losses and provide ultimate stability, though we also learned the importance of monitoring overnight swap costs that can silently eat into small profits. * GateGrid AI: Finished the week at -913 JPY. Despite a strong defensive showing early in the week, a single hard hit of -765 JPY wiped out all of its accumulated gains. It perfectly demonstrated the classic weakness of grid trading: when one single loss exceeds your average win, the total ends up negative. * LLMBridgeTrader: Our most experimental, fully AI-driven planner took the hardest hit, ending at -1,040 JPY. Because the AI is given full discretion over position management (OPEN, HOLD, CLOSE, REVERSE), a run of bad calls caused the drawdown to balloon rapidly. However, it showed encouraging resilience by clawing back some profit at the week’s end, highlighting the urgent need to build tighter “guardrails” and maximum loss caps. The ultimate lesson from this week? A good automated trading bot isn’t just judged by the size of its profits, but by how it loses and where it manages to stop the damage. Join us as we discuss why our simplest bot won the week, and outline our specific plans to completely rework the exit strategies for our AI models moving forward. #FX #MT5 #AITrading #AlgorithmicTrading #SystemTrading #TradingStrategy This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit fxaibotlab.substack.com

4 min

Can AI really trade forex? AI FX Bot Lab is a real-time experiment from Japan, where I build and test AI-assisted FX trading bots using MT5, Python, machine learning, and local LLM tools. I share live results, failures, risk lessons, and bot improvements from rule-based, AI-driven, and ML + LLM hybrid systems. Not financial advice. fxaibotlab.substack.com

Creator

Kimi | Japan FX Bot Lab
Years Active

2K
Episodes

22
Rating

Clean
Copyright

© Kimi | Japan FX Bot Lab
Show Website

AI FX Bot Lab: Real Trading Experiments