OpenRiot v7.9.48 — The Machine Reads the Chart

“The crypto tracker told you to sell ZEC at 504. ZEC was at 490. It had been at 550 three months ago. It had been at 350 three months before that. The machine knew the full range — it had 180 days of closing prices in memory — but the sell limit was a fixed formula: entry times 1.20. As if the market cared what you paid. We gave the machine the range. We taught it to measure where the current price sits and set a target that makes sense — near the top if we’re already close, further out if we’re still climbing. Then we taught it something harder: which coin to buy. Not by RSI. Not by concentration filters. Not by some arbitrary threshold that made LTC look like a good idea when it was dead money. We taught it to read the chart. Position in the range. Upside potential. Thirty-day trend. Three multiplications. The machine now knows the difference between a coin that’s climbing and a coin that’s just expensive. The Turing Police would use a 20% trailing stop and call it ‘quantitative.’ We used six months of candles and called it common sense.” — The OpenRiot Crew, after the rotation targets finally made sense, v7.9.48

Release Overview

This release rewrites the crypto tracker’s brain. Two related problems: sell limits were a fixed multiple of entry price, and rotation targets were based on RSI — a momentum oscillator that couldn’t tell the difference between a coin that was climbing and a coin that was just cheap.

The sell limit now uses six months of price history. The rotation target now uses a range-based momentum score. The result: ZEC’s sell limit moved from $504 to $603. Rotation targets stopped suggesting LTC and started suggesting coins that are actually performing. The best performer now rotates to the second-best instead of uselessly saying “Hold.”

Also: feh can navigate images with arrow keys again, and .heic files are now supported.

📊 Crypto: Sell Limits That Read the Chart

The old formula was entry × 1.20. The new system finds the 6-month high and low from 180 days of OHLC data, calculates where the current price sits in that range, and sets the target:

  • Upper 25% (near highs): target = 6-month high × 0.98
  • Middle 50%: target = 6-month high × 0.90
  • Lower 25% (near lows): target = 6-month high × 0.75

Floor at current × 1.10, ceiling at current × 1.50. The multipliers cap below the high — selling above the high was asking the coin to break its own record. Real-world: ZEC $504 → $603, BTC $65,997 → $72,756.

🔄 Crypto: Rotation Targets That Read the Chart

The old rotation logic used RSI scoring. A coin with RSI 35 got a 50-point “oversold” bonus and beat a coin with RSI 65 that was actually climbing. LTC — a zombie with no catalyst — was the top rotation target for weeks.

The new system uses rangeScore:

score = position × upside × trend
  • Position: where current price sits in the 6-month range
  • Upside: 6-month high / current price — room to run
  • Trend: average of last 30 days / average of older data — is the coin actually moving up?

ZEC (climbing, high momentum) is the top target. SOL (near lows, upside) is second. LTC (flat, no trend) is dead last. The best performer rotates to the second-best instead of “Hold.”

🖼️ feh: Arrow Keys and .heic

feh.desktop had %f instead of %u and dropped --start-at. Opening an image from Thunar showed only that one image. Fixed: --start-at %u restores directory navigation. .heic/.heif MIME types added for iPhone photos.

A new config/applications/feh.desktop ships with the repo.

🧾 Files Changed

File Change
source/crypto/crypto.go NEW — rangeScore() and rangeRanked()
  for 6-month momentum scoring; rotation
  targets use range data instead of RSI;
  best performer rotates to second-best;
  sell limits use position-based multipliers
  (0.98/0.90/0.75); OHLC fetch 90→180 days
config/applications/feh.desktop NEW — --start-at %u, .heic/.heif
install/openriot UPDATED — Compiled binary

🗣️ Final Words

“The sell limit moved. The rotation target changed. LTC disappeared from every suggestion. The machine had stopped guessing. It was reading the chart — six months of candles, position, upside, trend. It knew the difference between a coin that was climbing and a coin that was just expensive. The best performer rotated to the second-best. The zombie coin got nothing. The Turing Police would have added a neural network. We added three multiplications and a 180-day fetch.” — The OpenRiot Crew, after the rotation targets finally made sense, v7.9.48