Hashprice is around $32/PH/day right now. An S19j Pro pulling 3.05 kW breaks even at roughly $0.044/kWh which means for a lot of fleets, only part of the day is worth running at all.

So I built a small agent that does two things:

1. Hourly curtailment. It reads your electricity price (fixed, ESIOS/REE for Spain, or any JSON endpoint), computes your actual break-even from your hashrate, power draw and the day's hashprice, and decides per hour: mine or pause. Fail-safe by design — if it can't get price data, it defaults to MINE. It never pauses blindly.

 2. Hashboard failure detection. This is the part I think is genuinely useful. Instead of comparing a board against a fixed threshold, it compares each board against its siblings in the same rack, using median/MAD instead of mean/std so an already-dying board can't poison the reference. That cancels out ambient temperature, grid voltage and difficulty variance — the noise that makes absolute thresholds useless. It also watches fan RPM normalized by temperature, which is the earliest signal there is: fans quietly working harder for the same temp usually means dust or a bearing going, weeks before anything else shows up.

Single Python file. Runs on a Pi or any spare box on your LAN. Reads miners over the standard CGMiner API on 4028. Nothing leaves your network — no cloud, no telemetry, no account. MIT licensed.

What I actually need from you:

I've validated it against simulated farms and I have no real fleet. The parser is written blind against firmware formats I've only read about, so it will probably break on contact with real hardware.

If you're willing, paste the output of:

    echo '{"command":"stats"}' | nc <your-miner-ip> 4028

plus your model and firmware version. Even one or two would help me a lot. If you'd rather not post it publicly, DM works.

**Full disclosure on the money,** because I'd rather say it up front than have someone find it later: the software is free and stays free, and the failure detection is free permanently with no catch. If you enable automated curtailment and it saves you money, my model is 25% of *net* savings — net meaning electricity avoided minus the mining revenue you lose while paused, and only counting pauses that were actually executed. The ledger lives on your machine, the formula is in the source, and advisory mode bills nothing ever. You can also just run the whole thing yourself and never pay me anything — it's MIT, that's a legitimate use.

Happy to answer anything about the statistics or the implementation.

software: hash guard.es , download on the page web.