Your Bot Was Never Going to Keep Working
You build a bot. First week: +8% returns. By week four: +3%. By week eight: -2%. You didn't change the code. The market changed.
This is model decay. It kills 87% of retail trading bots before day 60.
Professional traders know this. That's why they retrain, recalibrate, and rebuild constantly. Amateur bots become deadwood because nobody tells you this part when they teach you to code.
What Is Model Decay?
Model decay happens when the patterns a bot learned stop matching real market conditions. Your EA trained on 6 months of data in a low-volatility, range-bound market. Now we're in a trend. Or a crash. Or a central bank meeting shifted everything overnight.
The bot doesn't know that happened. It keeps executing the same logic on a market that no longer fits the pattern.
In machine learning terms, this is concept drift. The relationship between input (price, volume, RSI) and output (profitable entry) drifts as market regime changes. Your model was right. The market moved. Now your model is wrong.
Most traders think bots are "set it and forget it." They're not. Professional bots are "set it, monitor it daily, retrain weekly."
Why 60 Days Is the Cliff
Research from quantitative trading firms shows that most algorithmic edge degrades 40-60% within 8 weeks of live deployment. Not always immediately. Sometimes slower. But the cliff exists.
Here's why: your training data is historical. Historical is not current. The longer you trade without retraining, the larger the gap between "what worked" and "what works now."
Week 1-2: Your bot caught the regime it was trained on. Win rate holds. Week 3-4: Market starts to shift. Win rate drops 10-15%. Week 5-6: Regime fully different. Your bot loses on setups it used to win. Week 7-8: Decay accelerates. Bot underperforms by 30%+ vs. live market conditions.
By day 60, the bot isn't just unprofitable. It's a liability. It keeps trading, keeps losing, keeps eating spreads.
The Signs Your Bot Is Decaying
You don't need to wait 60 days to know it's happening. Watch for these:
- Win rate drops but signal count stays the same. The bot's still finding setups, just not winning on them. Classic decay.
- Profitable pairs stop working. Your EURUSD bot made 12% last month. This month it's flat on EURUSD but still triggers 40+ trades. The pattern broke.
- Drawdown increases while average win decreases. The bot takes bigger losses and smaller wins. That's regime shift killing edge.
- Peak equity flatlines then reverses. Equity was climbing. Now it oscillates or drifts down. The momentum stopped.
If you see any of these, the model decayed. You need to retrain.
Why Professionals Retrain Constantly
Professional quants don't build one EA and deploy it. They build a framework that retrains automatically.
The framework works like this:
- Live monitor. Track in-sample vs. out-of-sample performance daily.
- Trigger retraining. When live performance drops 15% below backtest, retrain on the last 90 days of data.
- Parameter adjust. Shift stop-loss, take-profit, or signal thresholds based on current volatility and trend direction.
- Deploy new version. Put the updated EA live. Keep the old version running on a separate account to compare.
- Loop weekly. Repeat every 7 days. The best traders retrain daily.
This is why professional EA development includes monitoring dashboards and retraining protocols. It's not a feature. It's the entire game.
The Hidden Cost of Decay
A decayed bot doesn't just stop working. It actively hurts you.
Say your bot trades $10 per pip. Over 60 days it places 400 trades. If decay costs you 0.5 pips per trade on average (from losing setups that used to win), that's $2,000 in slippage and losses you didn't see coming.
Most traders blame themselves. "Maybe I was wrong about the strategy." No. The strategy was right. The market regime moved.
The cost of inaction: another 60 days of decay on another bot while you figure out why this one failed. That's another $2,000+ lost while you're learning the lesson the hard way.
What Continuous Learning Looks Like
Here's the thing: you can't prevent model decay. It's built into live trading. But you can manage it.
The traders who stay profitable across market regimes do three things:
One: they track regime changes daily. Is the market trending or ranging? Volatile or calm? ECB meeting this week? Each regime favors different parameters.
Two: they've built multiple bots for multiple regimes. One EA for trending markets. One for consolidation. One for volatility expansion. Professionals don't run a single bot. They run a portfolio.
Three: they retrain before the bot breaks. They don't wait for 60 days of degradation. They retrain every week. New data in. Old backtests out. Parameters updated. New version deployed.
This is why custom EA development with ongoing monitoring from professionals matters. A $300 expert advisor that includes a weekly retraining protocol is cheaper than a $50 bot you have to rebuild every month.
The Framework Professional Traders Use
If you're going to build or deploy a bot, use this framework to stay ahead of decay:
- Days 1-7: Monitor obsessively. Track every trade. Compare live performance to backtest. Watch for degradation.
- Days 8-14: First retraining. Pull the last 30 days of live data. Retrain the model. Does it improve? If not, adjust parameters manually.
- Days 15-30: Weekly retrains. Make it a routine. Every Sunday, retrain on the last 60 days. Deploy Monday morning.
- Days 31-60: Switch to multi-bot strategy. If this bot decayed, you should have built a second one for the new market regime. Swap allocation from decayed bot to new bot.
- Beyond 60 days: deprecate. Archive the old bot. Build a new one from current market data. Use the old bot as a reference only.
This is the difference between a bot that lasts 60 days and a bot that lasts years.
Key Takeaways
- Model decay is inevitable. Your bot's edge degrades 40-60% by day 60 because market regime shifts. It's not your fault. It's physics.
- The cliff comes between weeks 5-8, when performance drops 30%+ and losses accelerate. That's when most traders give up.
- Professionals don't "set and forget." They retrain weekly. They monitor daily. They run multiple bots for multiple regimes.
- The cost of one decayed bot is $2,000+ in lost pips over 60 days. The cost of preventing decay is one retraining protocol.
- If you're building bots yourself, set up a weekly retraining schedule now. If you're buying them, make sure monitoring and retraining are included.
Your Next Move
You have two choices: build a bot and watch it decay every 60 days, or build one with a retraining framework baked in.
If you're building from scratch and want to skip the three-month learning curve on decay management, we can build you a custom MT5 EA with a weekly monitoring dashboard and automated retraining protocol. From $300. Working demo in 45 minutes. Full backtest report included.
Most importantly: we build it so it stays profitable as market regimes shift. Not because we're magic. Because continuous learning is built in from day one.
Tell us your strategy. We'll build the bot. Then we'll show you how to keep it working.