▲ Work in progress. This site is built and researched in the open: pages, numbers and findings may change or be wrong, and the record has gaps. Not advice. More
The Lab
Ideas traders use, put to the record
← the Lab · grids · write-up
#30 Grid bots: does a small trend bet beside the grid smooth the bad months? Not shown
Past prices; no real orders were placed. Every figure is a share of the money, or a count.
Last checked: 15 October 2026.
Where the idea comes from
- Grid trading is long-standing market practice, with no known originator.
- A grid is a rebalancing machine: it buys after each step down and sells after each step up, which is the idea behind "Shannon's demon". See our theory page "Shannon's demon: making money from a price that goes nowhere".
- It earns only if prices come back to where they were. That is mean reversion, on our theory page "Mean reversion: does what drifts away come back?".
- The trend sleeve is time-series momentum, documented by Tobias Moskowitz, Yao Hua Ooi and Lasse Pedersen (2012). See our theory page "Momentum: do recent winners keep winning?".
The question
A grid earns when prices swing and suffers when they trend. A common suggestion is to keep a small trend sleeve beside it: a position that follows the trend (what Moskowitz, Ooi and Pedersen, 2012, call "time series momentum"), to make up for the grid's bad months. Does it smooth the worst months, and what does it cost in the good ones?
How we tested
- Bitcoin, Ether and Solana, October 2022 to September 2026: 48 months. The same grid as our earlier grid studies: 20 rungs, 15% either side of the month's first price, re-centred upward, on futures at 1x.
- The grid gets 80% or 60% of the money, and the sleeve the rest.
- The sleeve: a futures position as big as its money, long or short, set each day from yesterday's closes by one of three plain rules, not tuned:
- 20-day sign: long if the price is up over the last 20 days, short if down;
- 50-day sign: the same over 50 days;
- cross: long when the 20-day average is above the 50-day average, short when below.
- It pays the market fee on every change, and pays or receives funding.
- The test:
- smoothing: the pair against the grid alone, in the grid's worst quarter of months;
- cost: the same, in the grid's best quarter;
- in each two-year half.
- The control: the same sleeve with each month's days shuffled, 100 times. It is long and short on the same number of days, but at random. If trend-following is what helps, the real sleeve should beat the shuffled ones.
The maths, step by step
1. The 20-day sign. If the price was 100 twenty days ago and is 104 now, it's up, so the sleeve is long tomorrow.
2. A sleeve's month. Long for two days, as the price goes 100 → 110 → 99: ×1.10, then ×0.90, is ×0.99, a 1% loss before costs.
3. Smoothing, measured. Rank the grid's 48 months and take its worst 12. Say in one of them the grid lost 4% and the pair lost 2%: the pair did 2 points better that month. Average that over the 12 months. Do the same for the best 12 to get the cost.
4. Why any side bet "smooths". The grid's worst months were picked for being the worst. Anything that doesn't move with the grid lands closer to average in those months, so it lifts them. The shuffled sleeves show how much of the lift comes from that alone.
What we found
1. No rule passed. Against random-day sleeves of the same size, no real rule ranked above the middle on any coin:
| share of shuffled sleeves the real one beat, at 80/20 | 20-day sign | 50-day sign | cross |
|---|---|---|---|
| Bitcoin | 50% | 38% | 27% |
| Ether | 0% | 27% | 0% |
| Solana | 15% | 6% | 7% |
2. The worst months did look better, but not because of the trend.
- In the grid's worst quarter, the 80/20 pair did 0.6 to 3.0 points a month better than the grid alone.
- A sleeve long and short on random days did as well or better.
3. What it did to the months.
- More losing months: Bitcoin from 27% of months to 44–50%, Ether from 31% to 42–58%.
- The single worst month moved both ways:
- Bitcoin with the 20-day sign: −6% to −3%;
- Ether with the 50-day sign: −14% to −11%;
- Solana's worst month, −35% in November 2022, stayed;
- the cross made Bitcoin's worst month worse, −8%.
- The average month rose with the 50-day sign on Bitcoin (+0.88% to +1.02%) and Ether (+0.58% to +0.94%), and fell on Solana (+1.88% to +1.22%). Each change was within one standard error.
4. The sleeve on its own was a coin flip: from −2.7% to +3.3% a month depending on coin and rule, with worst months of −21% to −69%.
What it means
- A trend sleeve doesn't smooth the grid any better than a random side bet would.
- If the aim is smoother results, a separate book that doesn't move with the grid does the same without a timing rule.
- If the aim is a trend bet, it should be tested as its own strategy.
Limits
- Three coins, which move together; 48 months.
- The rules were not tuned, on purpose. A tuned rule would need its own test on data it hadn't seen.
- Each month starts the split again. A sleeve held for longer would behave differently.
Credits
- Tobias J. Moskowitz, Yao Hua Ooi and Lasse Heje Pedersen, "Time Series Momentum", Journal of Financial Economics, 2012.
- Moving-average crossovers are old trading practice, with no known originator.
- R. A. Fisher, The Design of Experiments, Oliver and Boyd, 1935: the permutation test behind the shuffled sleeves.
- The sleeve rules and the split are our own design; the grid follows our study "Grid bots on 14 coins: what 30 months of prices say".
- Data: Binance's public 1-minute futures prices; Bybit's public funding history and fees.
- PerpGremlin and Claude, Grid bots: does a small trend bet beside the grid smooth the bad months?, TerminalGremlin, 2026: the tests, the code and the words.
← Adaptive grids against fixedTwo opposing Bitcoin grids →
Educational research from public market data; any trading discussed is paper, never real. Nothing here is advice. The full notice.