Learning Quant Finance the Hard Way: Unpacking berlinguyinca-trading-strategies
How a public time capsule of Python scripts tracks one developer's battle against a bear market, and why the codebase was eventually shuttered to protect novice traders.
Please be aware they are supposed to show off some of the indicators and concepts you can use with freqtrade and might not be great for investing purposes. So please use them on your own risk.
- The repository serves as a historical ledger of a developer learning the Freqtrade API during the brutal 2018 cryptocurrency bear market.
- Early high-frequency trading scripts lost money, prompting a technical pivot toward paranoid, low-frequency execution strategies.
- The configuration flag sell_profit_only forced bots to bag-hold depreciating assets rather than realize losses, highlighting the desperate tactics used in volatile markets.
- The project was intentionally deprecated to prevent novice traders from losing capital on overfitted backtests.
The 2018 Time Capsule
Most articles about open-source trading bots focus on deployment strategies or hypothetical returns. This repository is different. It is an archived, deprecated time capsule from the 2018 crypto crash. The story here is not about discovering the perfect trading algorithm. It is about a developer learning quantitative finance in public during a hostile market.
We can trace the code's evolution from naive, high-frequency scripts that actively lost money, to paranoid, low-frequency scripts designed purely for survival. Ultimately, the repository concludes with an ethical decision to deprecate the code so novice traders would not risk their capital on experimental scripts.
The Simple Illusion
The Freqtrade framework relies on the IStrategy pattern. A developer defines indicators, buy logic, and sell logic. The earliest commits reveal a bot appropriately named Simple. It executed 85 trades in backtesting but yielded a net negative return. The standard pipeline ingests OHLCV data, populates technical indicators using TA-Lib, and triggers execution logic based on simple crossovers.
Coding for Survival
As the 2018 market crashed, the strategies mutated. The technical pivot is most visible in the configuration blocks. Developers began utilizing the experimental sell_profit_only flag. This controversial setting forces the bot to hold depreciating assets indefinitely rather than trigger a stop-loss. It is the algorithmic equivalent of refusing to admit defeat.
"experimental": {
"use_sell_signal": true,
"sell_profit_only": true
}
The Overfitting Trap
Comparing the early, chaotic bots to the later, highly restrictive bots reveals a common pitfall in quantitative finance. Developers often accidentally over-optimize their backtests to historical data. This creates bots that look perfect on paper but fail completely in live markets.
| Strategy Name | Total Trades | Average Profit % | Result |
|---|---|---|---|
| Simple | 85 | -0.33% | Loss |
| ReinforcedQuickie | 254 | 0.09% | Break-even |
| ClucMay72018 | 6 | 4.20% | High Conviction |
Archiving the Alpha
The repository was eventually deprecated. The author explicitly warned that these scripts were not suitable for actual investing. The community migrated to the official, governed freqtrade-strategies repo. This lifecycle highlights the value of failing fast in public, and knowing exactly when to turn the machine off.