QuantLabsNet.com Public Quant Analytics Group

I’ve been running a pretty large universe of automated trading bots — over 200 of them are currently flagged as profitable. Originally I wanted to surface only the absolute cream‑of‑the‑crop using a strict filter. When that turned out to be crushing, I relaxed the gate logic to something bare‑bones, and I honestly expected a lot more bots to pass.

The ones that pass all filters is gold, copper, and natural gas. 

Here’s what I was working with.

Strict filter criteria (applied to every bot):

• Minimum trades: ≥ 20

• Win rate: ≥ 50%

• Max drawdown: ≤ 50%

• Grade: Must be A+, A, B+, or B

• Profit factor: ≥ 100.0

• Last 3 months profitable: at least 2 of the 3

• Current month profitable: must be profitable

• 
  

Under these rules, hardly anything survived. So I decided to drastically relax the entry gate, just to see how many bots could at least show some basic consistency.

Relaxed gate logic (what I actually used for today’s report):

• profitable == True

• positive_months >= 1 (at least one profitable month — any month, ever)

That’s it. No win‑rate, no drawdown cap, no grade, no profit factor, no recent‑month streak. Just “it’s green overall and it’s managed to be profitable in at least one single month.”

Out of the 200+ profitable bots as of today, only 10 strategies made it through this relaxed filter.

And here’s the kicker: all 10 were trading gold, copper, and natural gas. Not a single equity, forex, or crypto bot passed. Only those three commodities.

I’m not sure whether to be impressed by how selective even the relaxed filter is, or worried that the rest of the profitable bots are just noise. Either way, it’s a stark reminder that “profitable” can be a very loose label, and the moment you require even one month of proven profitability, the list shrinks to a tiny handful of commodity‑focused strategies.

Curious if anyone else has run similar scans across a large bot pool. Did you see similar clustering in hard commodities?