Why we deleted Bitcoin from our trading bot

The reasoning that got us there

The bot was built for gold. Gold is volatile, trends cleanly, and respects structure well enough that order blocks and liquidity sweeps mean something. Bitcoin is also volatile. It also trends. It trades around the clock, which looked like an advantage over gold's session limits.

So we enabled it. The logic was that a strategy reading structure and momentum should not care what the underlying asset is — the patterns are the patterns. That reasoning is extremely common in this industry, and it is wrong.

What actually broke

“Volatile” is not one property. Gold and bitcoin are volatile in structurally different ways, and every parameter in the bot encodes an assumption about which kind.

Stop distances were calibrated to gold's noise floor. A stop that sits safely outside routine movement on XAUUSD sits directly inside routine movement on BTC. The bot was not being stopped out by being wrong about direction — it was being stopped out by noise, then watching price continue to its target without it. A 28% win rate with the bot frequently right about direction is what that looks like from the outside.

The spread guard was tuned for a metal. Our ceiling for gold is one dollar. On BTC that threshold is meaningless — it either blocks everything or nothing depending on the hour, so a filter designed to prevent bad entries was effectively switched off.

Around-the-clock trading was a liability, not a feature. Gold has sessions, and sessions create structure — opening ranges, session highs and lows, predictable liquidity. Bitcoin's weekend tape is thin and moves violently on low volume. Strategies built on session structure had nothing to hold onto.

Why we deleted it rather than fixed it

The obvious move was a second parameter set — BTC-specific stops, its own spread ceiling, weekend handling. We could have done that. We would then have been maintaining two tuning surfaces, two sets of regime thresholds, and two backtest suites, in exchange for a symbol with no demonstrated edge.

The alternative was to spend that same effort on gold, where the bot already had a measurable edge. Seven gold-specific strategies came out of that decision. None of them would make sense on bitcoin, which is rather the point.

We also did not want to leave it in as a toggle. A disabled feature buried in settings is a trap: eventually a customer switches it on, gets our 28% win rate, and reasonably concludes the bot does not work. Better that the product does one thing than that it offers a second thing badly.

What 117 trades does and does not prove

Being straight about the limits: 117 trades is a real sample but not an enormous one. It is enough to say this configuration did not work on this instrument over this period. It is not enough to say bitcoin is unsuitable for automated trading in general — plenty of people trade it profitably with systems built for it.

The honest conclusion is narrower and more useful: parameters do not transfer between instruments, and a bot advertising a long list of supported symbols is usually telling you it has one parameter set and a lot of optimism. Ask which symbols it was actually tuned and measured on. Usually the answer is one.

The same rule killed 24 strategies

Bitcoin was not a special case. Of 34 strategies built for this bot, 24 are disabled in the shipped configuration — some because they lost real money with a track record attached, others because they never fired at all. Four trade a default install.

Over roughly three months of live trading on the developer's account, the strategies now switched off accounted for 929 trades and about $330 of loss. The ones still enabled accounted for 76 trades and a small profit. Removing things is most of what made the configuration work. The full list is here, including every cut and the reason.