Backtest evidence

How should a Renko brick height be chosen, and does the grid offset — which is not a setting — move the result?

Is Trading Renko Reversals Profitable? 1,080 Settings Tested

Finding

The grid offset, which no chart exposes as a setting, moved the result as much as the brick height did. Holding the brick at 20 pips and shifting the grid through ten positions moved the 2025 USDJPY result from +222 to +3,083 pips. The brick height itself had no fixed answer: hourly bars liked 15 pips in both years (+3,084.3 pips over 808 trades in 2025, +2,177.7 over 765 in 2024) while 15-minute bars wanted 30 pips in 2025 and 10 pips in 2024. A brick contains no time, but the timeframe survives inside it: the same 10-pip brick built on 2025 USDJPY produced 8,041 bricks from 15-minute bars, 5,270 from hourly and 3,298 from four-hour, a factor of 2.4. Recording fills at the brick close flatters the result because it absorbs the overshoot: on hourly 20-pip bricks that overshoot averaged 7.60 pips and favoured the trade in all 15 cells. Repricing the same trades at brick closes improved all 30 cells, turning +1,491.5 pips into +10,500. The look of the runs survives shuffling: across 15 cells the share of bricks reversing the previous one differed by at most 2.4 points, and the mean run length by −0.19 to +0.16 bricks.

Key results

Shifting the grid through ten positions, 20-pip bricks in 2025
the result moves from +222 to +3,083 pips
Bricks built from the same 10-pip height in 2025
8,041 from 15-minute bars, 5,270 hourly, 3,298 four-hour
Repricing fills at the brick close
improved all 30 cells; hourly 20-pip went from +1,491.5 to +10,500 pips

Scope

  • USDJPY. 2025 as the main window, 2024 out of sample, plus the two halves of 2025
  • 15-minute, hourly and four-hour bars
  • 6 brick heights (5 to 50 pips) x 2 reversal rules x readings (direction change, or N bricks for N = 2 to 5) = 720 runs, plus 10 grid offsets = 360, for 1,080 in total
  • 0.3 pip spread, 0.1 lots, filled at the candle close

Method

Measured with Formiq's backtester. Fills use the candle's close, not the brick's close. Two readings — a change of brick direction, and N bricks in the same direction — exiting on the opposite signal. As a control, the same year's closes are shuffled bar by bar and the same bricks rebuilt, so the run structure can be compared.

Limitations

  • USDJPY in 2024 and 2025 only.
  • Only ten grid offsets were measured. A 20-pip brick has twenty of them, so the measured spread is a lower bound.
  • The shuffled series discards each bar's shape, reconnecting closes only, so it can only be compared against rules that read closes.
  • Only two reversal constructions were tried. Some implementations require three bricks before a reversal.
  • Renko construction itself varies between tools: some build from closes and some from highs and lows. This test builds from closes.
  • Order-level slippage is excluded and the spread is fixed at 0.3 pips.