The Lab

How we tested

• Bitcoin, Ether and Solana, every minute from October 2022 to September 2026: 48 months.
• A long grid set fresh each month. Its range runs from 15% under the starting price to the same distance over it, with a 30% range as a second check.
• 3, 5, 8, 12, 20, 30, 50, 80 and 120 rungs, evenly spaced in percent.
• Spot, at a 0.1% fee on every order, and futures at 1x, 2x, 3x, 5x, 7x and 10x, at a 0.02% fee and paying or receiving funding.
• Liquidation: a futures month that hits its liquidation price counts as −100%.
• Real order sizes: a 10,000 USDT account, with the exchange's smallest order and the coin's quantity step. A 1,000 USDT account is shown beside it.
• Fairness:
  • every setting is shown;
  • the first two years pick a winner, and the last two judge it;
  • every month is also run 20 times with its hours shuffled into random order: the same moves, without real prices' habits.
• The simulation follows our own bot's rules and was checked against its backtest. Over 12 random months the two agreed to within 0.05% of the money on average.

The maths, step by step

1. The gap between rungs. The top of the range is 1 ÷ 0.85² = 1.384 times the bottom. With n rungs there are n − 1 equal steps, each the (n − 1)th root of 1.384:

• 8 rungs: 4.75% apart;
• 20 rungs: 1.73%;
• 120 rungs: 0.27%.

2. What fees take from each round trip. A round trip earns one gap and pays the fee twice.

For example, at 120 rungs on spot: 0.2 ÷ 0.27 = 73% of what each trip earns goes in fees.

3. Real order sizes. 10,000 USDT over 120 rungs is 83.33 USDT a rung.

• With Bitcoin at 80,000, that is 0.00104 BTC. The exchange's step is 0.001, so each rung rounds down to 0.001 BTC, or 80 USDT: 96% of the money is used.
• With Bitcoin at 30,000, it is 0.00278 BTC, which rounds to 0.002: only 72%.
• With 1,000 USDT, a rung is 8.33 USDT, or 0.0001 BTC. That is under the smallest order, so the rung can't be placed at all.

4. Luck against shuffles. If real prices were no different from shuffled ones, the real month would be one of 21 in a fair draw. So it beats 19 or 20 of its 20 shuffles 2 times in 21 by luck. Of 324 settings, about 31 would do that by chance.

What we found

1. No rung count traded better than another. Each grid was compared with the same months shuffled. Every rung count came out about the same against its shuffles. Only 20 of 324 settings beat 19 or 20 of their shuffles, fewer than the 31 that luck alone would give.

2. Fewer rungs earned more, because they hold more coin. A 3-rung grid buys its middle rung near the start. On average 36% of its money sat in coins, against 16–24% for 120 rungs. Prices rose over these four years, so holding more paid.

Average month, 15% range, futures at 1x:

• Each step is within one or two standard errors of the next. The shuffled months, which end at the same prices, show the same order.
• Holding more coin also swings more: a 3-rung month's spread was about twice a 120-rung month's, and its worst month was worse (Ether: −18% against −12%).

3. Fees decide the tight end, and spot's fee decides it early.

• On futures, 120 rungs lost 14–15% of their takings to fees.
• On spot, about 70%. On the 15% range, spot lost money on average from 20 rungs on Ether and Solana, and from 50 on Bitcoin.

4. Leverage multiplies, it doesn't choose.

• The best rung count was 3 at every leverage but one.
• The share of months liquidated hardly changed with the rung count, but grew fast with leverage:

The 30% range was liquidated less (Ether at 10x: 4–8%; Bitcoin: never).

5. Picking on the first two years didn't carry. The first half picked 3 rungs almost everywhere. At 1x it earned only +0.2% to +0.9% a month in the second half. At 10x on the 15% range, the same pick lost money in the second half: Bitcoin −1.5%, Ether −3.9%, Solana −11.7% a month.

6. Real order sizes.

• At 10,000 USDT every rung could be placed on every coin and leverage, except Bitcoin at 1x. There, 120 rungs used a median 71% of the money, and none could be placed in the months Bitcoin cost most.
• At 1,000 USDT, Bitcoin at 1x could not place 20 rungs or more at all, and Ether at 1x not 80 or more.

7. Minute by minute is close enough. The test lets each rung fill at most once each way per minute. One week of Ether's trades was rebuilt second by second. At 50 to 120 rungs it found 4–10% more round trips, and the week's result moved by less than 0.05%.