▲ 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
#24 Grids with and without stops
Grid bots on 14 coins: what 30 months of prices say
New to grids? How a grid works
The numbers here come from our grid study: 14 coins, April 2024 to September 2026.
The rungs
A grid bot draws a range around the price — a bottom and a top — and splits it into evenly spaced price levels called rungs. Below the price, each rung holds an order to buy. When a buy fills, the bot places an order to sell that piece one rung higher.
An example: the price is 100. A grid 20% either side runs from 80 to 125, with 8 rungs:
80 · 85.3 · 90.9 · 96.9 · 103.2 · 110.0 · 117.3 · 125
The rungs under 100 (80, 85.3, 90.9, 96.9) wait to buy. Nothing is held yet: the grid only buys when the price comes down to it.
The gap
The distance from one rung to the next is the gap. Here each rung is 6.58% above the one under it. A wider range with the same 8 rungs has a wider gap:
| range either side | gap between rungs |
|---|---|
| 15% | 4.75% |
| 20% | 6.58% |
| 25% | 8.57% |
| 30% | 10.73% |
A round trip
The price dips to 96.9: the bot buys there. It rises to 103.2: the bot sells what it bought. That is one round trip, and it earns one gap.
With 1,000 USDT at 2x and 8 rungs, each rung trades about 250 USDT. A round trip at the 20% range earns about 6.58% of 250 — about 1.65% of the money put in. At 30% it is about 2.7%; at 15%, about 1.2%.
How often does that happen? In our study, a 20% grid on ETH made about 3.6 round trips a month. On calm BTC at 30% it was 0.6. On fast-moving memecoins at 15%, 8 to 15.
Fees
Every buy and every sell pays the exchange. A resting order (one that waits on a rung) pays 0.02%. So a round trip on 250 USDT pays about 0.10 USDT — about 0.01% of the money put in, against the 1.2–2.7% it earns. At these gaps the fees hardly matter. They matter for grids with very tight rungs, where each trip earns little.
The maths, step by step
Every number above comes from a few multiplications. Here they are, for the example: price 100, 20% either side, 8 rungs, 1,000 USDT at 2x.
- The range. The bottom is 100 × (1 − 0.20) = 80. The top is 100 ÷ (1 − 0.20) = 125.
- Dividing, not adding, makes the two sides match: a fall from 125 to 100 is 20%, the same as a fall from 100 to 80.
- The gap. The rungs are spaced by the same percentage, not the same price.
- From 80 to 125 is a rise of 125 ÷ 80 = 1.5625 times, made in 7 equal steps (8 rungs have 7 gaps between them).
- The step that, repeated 7 times, makes 1.5625 is its 7th root: 1.0658, a gap of 6.58%.
- So the rungs are 80, then 80 × 1.0658 = 85.3, then 85.3 × 1.0658 = 90.9, and so on up to 125.
- The size of each rung. At 2x, 1,000 USDT trades as 2,000. Split over 8 rungs, that is 2,000 ÷ 8 = 250 each.
- One round trip. A rung buys 250 and sells it one gap higher: 250 × 6.58% = 16.45 USDT. That is 16.45 ÷ 1,000 = 1.65% of the money put in.
- The fees. 0.02% on the buy and again on the sell: 250 × 0.0002 × 2 = 0.10 USDT, about 0.01% of the money.
- The stop. 7.1% under the bottom rung: 80 × (1 − 0.071) = 74.3.
- Each piece held loses 250 × (1 − 74.3 ÷ its buy price). The piece bought at 96.9 loses 58.3.
- The four pieces together lose 153.6, which is 15% of the 1,000.
Why a range breaks
A grid earns when the price goes up and down inside its range. It breaks when the price leaves:
- Above the top: everything is sold. The bot holds nothing and earns nothing until the price comes back down. In our study the price left a 20% range upward in about a quarter of months on ETH.
- Below the bottom: every waiting buy has filled. The bot holds all of it, bought at higher prices, and the loss grows as the price keeps falling. On ETH that happened in about 3 months in 10 at the 20% range.
That is why a long grid follows its coin. Over our 30 months, grids on coins that fell 80–98% lost money in 44 of 48 settings; without a stop, grids on the three coins that rose (BTC, XRP, ZEC) made money at every range. A grid smooths the ride; it does not change its direction.
What a stop does
A stop is the emergency exit: if the price falls a set distance under the range, the bot sells everything and stops. In our study the stop sat 7.1% under the bottom rung and sold once the price had stayed under it for 5 minutes.
In the example, the bot holds the 4 pieces bought under 100. If the price falls to the stop at 74.3, it sells them at a loss of about 15% of the money put in. That loss is now fixed: it can't get worse, and it can't come back.
What we measured:
- On slow falls the stop did its job. On BTC, ETH, SOL, ZEC and NEAR it sold within 9% of its level.
- In a crash it sold the bottom. On 10 October 2025 the whole market dropped for a few minutes around 21:20 UTC. On a 15% DOGE grid, the stop meant to sell 7.1% under the range sold 50% under its level. The month ended −89% with the stop and −25% without, because the price came back.
- On average the stop cost money. It helped the average month on at most 4 of 14 coins at any range.
Wider or narrower?
A narrow range trades often and earns a little each time, but the price leaves it more often. A wide range trades less, holds less at the bottom, and its worst months were smaller on 13 of our 14 coins (30% against 15%). In this period neither range gave a reliable profit: what decided the result was whether the coin itself went up or down.
Credits
- Grid trading is long-standing market practice with no known originator.
- The worked numbers, and the study they come from ("Grid bots on 14 coins: what 30 months of prices say"), are ours.
- Data: Binance's public futures prices; fees and order sizes are Bybit's.
Past prices; no real orders were placed. Every figure is a share of the money put in, or a count.
Last checked: 16 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?".
What we tested
A grid bot buys a little each time the price steps down and sells it again one step higher (see How a grid works). Grid trading is long-standing market practice with no known originator. We replayed one simple long grid on 14 coins, month by month, from April 2024 to September 2026 (30 months):
- 8 price levels ("rungs") spread over a range around the price at the start of each month: 15%, 20%, 25% or 30% either side;
- 1,000 USDT at 2x, so each rung buys about a quarter of the money's worth;
- started fresh at the start of every month and closed at the end of it;
- with no stop, or with a stop a little under the range (7.1% below the bottom rung) that sells everything once the price has stayed under it for 5 minutes;
- real exchange costs: 0.02% to place a resting order, 0.055% to sell at market, and the funding the exchange charged at the time.
The coins: the ten most-traded crypto perpetuals on Binance on 13 October 2026 (stablecoins, shares, metals and oil left out), plus the big memecoins: BTC, ETH, SOL, ZEC, NEAR, XRP, DOGE, GTC, STRK, RSR, PEPE, WIF, BONK, SHIB. HYPE was in the top ten but has only 16 months of history on Binance, so the next coin, RSR, took its place.
A hard period for long bets. 11 of the 14 coins ended lower than they started, 6 of them more than 80% lower. ZEC rose about 47-fold, XRP more than doubled and BTC rose 17%.
What we found
- A grid follows its coin. A long grid still owns the coin on the way down. Where the coin collapsed, the grid lost money in most settings, about 4–8% a month on STRK, GTC and WIF at the narrow ranges. Where it rose (ZEC, XRP), the grid made money. The grid softens the ride; it doesn't change its direction.
- Wider ranges lose less. At 30% either side the grid traded less (0.6 to 5 round trips a month) and its worst months were smaller on most coins. Its average month was positive on 9 of the 14 coins. But only 4 were positive in both halves of the period.
- The stop cost money on average. It helped the average month on at most 4 of 14 coins at any range.
- One night made most of the stops' worst months. On 10 October 2025, around 21:20 UTC, the whole market crashed for minutes. On 9 of the 14 coins (XRP, DOGE, GTC, STRK, RSR, PEPE, WIF, BONK, SHIB) it was the stop's worst month at every range: the 5-minute stop sold up to 69% under its level, and the prices were back by the month's end. On BTC, ETH, SOL, ZEC and NEAR the stops' worst months were other nights, and they sold within 9% of the level, as a stop should. Leave October 2025 out and the stop helped the worst month on 5–9 of 14 coins. A stop protects against slow falls and can hurt in a crash.
- Smaller coins can take everything. On RSR, STRK, PEPE, WIF and BONK, some settings lost the whole stake in a single month: the price fell through the range so fast that the position would have been liquidated.
- No sign of a built-in edge. We also ran each coin's months with the hours shuffled into random order: the same moves, but no trends and no tendency to come back. The real prices paid a grid more than most shuffles did on only 2–4 of the 14 coins, where luck alone gives about 4. In this period, crypto prices did not return to their range more than a random path would.
BTC's minimum order step. At 1,000 USDT and 8 rungs, each BTC rung is about 0.002–0.003 BTC, the smallest sizes the exchange allows. So the ladder held only about 84% of the money meant for it, and its rungs were uneven, in every month. More money or fewer rungs would fix that.
Average month by coin · 25% either side
What it shows: Each coin's average month over the 30, with the range 25% either side of the price: the bright bar without a stop, the dim bar with the stop, as a share of the money put in.
How to read it: Right of the line made money in the average month, left lost it. Where the dim bar sits left of the bright one, the stop cost money. Hover or tap a bar for its figure; every range is in the tables below.
The tables
Mean: the average month. Worst: the worst single month. Stop fired: the share of months the stop sold out. Left the range: the share of months the price went above the top, or under the bottom. Halves: the average month in the first 15 months, then the last 15.
Range 15% either side
| coin | price | no stop: mean / worst month | stop: mean / worst month | stop fired | round trips a month | left the range (up / down) | halves, no stop |
|---|---|---|---|---|---|---|---|
| BTC | +17% | +0.8% / -9% | -0.2% / -13% | 13% | 3.6 | 13% / 23% | +2.8% / -1.1% |
| ETH | -26% | -0.9% / -24% | -1.6% / -27% | 23% | 5.1 | 27% / 43% | -0.9% / -0.9% |
| SOL | -42% | +0.1% / -27% | -1.7% / -33% | 33% | 7.4 | 53% / 47% | +0.9% / -0.7% |
| ZEC | +4615% | +3.4% / -43% | -2.5% / -31% | 47% | 10.7 | 63% / 50% | +3.7% / +3.0% |
| NEAR | -27% | -1.8% / -30% | -6.9% / -34% | 50% | 11.2 | 67% / 57% | -2.9% / -0.5% |
| XRP | +136% | +3.1% / -15% | -0.6% / -61% | 23% | 7.6 | 37% / 40% | +6.1% / +0.1% |
| DOGE | -57% | -1.5% / -32% | -4.4% / -89% | 37% | 8.3 | 47% / 50% | -1.7% / -1.4% |
| GTC | -96% | -5.4% / -45% | -6.5% / -75% | 57% | 11.7 | 60% / 60% | -11.4% / +0.6% |
| STRK | -98% | -7.8% / -75% | -6.1% / -75% | 50% | 11.3 | 43% / 63% | -12.6% / -3.0% |
| RSR | -84% | -4.2% / -100% | -3.0% / -100% | 50% | 13.6 | 60% / 53% | +3.0% / -11.4% |
| 1000PEPE | -51% | -2.9% / -100% | -3.9% / -77% | 43% | 10.7 | 53% / 63% | +0.7% / -6.5% |
| WIF | -95% | -7.0% / -100% | -5.2% / -64% | 50% | 13.9 | 63% / 73% | -7.3% / -6.7% |
| 1000BONK | -86% | -3.3% / -100% | -1.9% / -58% | 47% | 15.0 | 50% / 77% | +3.1% / -9.7% |
| 1000SHIB | -81% | +1.0% / -30% | -2.9% / -58% | 37% | 9.4 | 43% / 53% | +0.8% / +1.1% |
Range 20% either side
| coin | price | no stop: mean / worst month | stop: mean / worst month | stop fired | round trips a month | left the range (up / down) | halves, no stop |
|---|---|---|---|---|---|---|---|
| BTC | +17% | +0.8% / -7% | +0.6% / -11% | 3% | 2.1 | 7% / 17% | +2.2% / -0.6% |
| ETH | -26% | +0.7% / -14% | -0.3% / -18% | 13% | 3.6 | 23% / 30% | +1.1% / +0.3% |
| SOL | -42% | +1.0% / -27% | +0.4% / -18% | 20% | 4.9 | 33% / 40% | +1.6% / +0.4% |
| ZEC | +4615% | +4.3% / -37% | -1.0% / -31% | 37% | 7.8 | 57% / 47% | +4.0% / +4.5% |
| NEAR | -27% | -0.4% / -25% | -4.0% / -25% | 27% | 7.6 | 47% / 50% | -0.8% / +0.1% |
| XRP | +136% | +2.1% / -17% | -1.7% / -44% | 20% | 4.4 | 27% / 30% | +4.3% / +0.0% |
| DOGE | -57% | -0.1% / -27% | -2.5% / -59% | 30% | 6.0 | 37% / 47% | -0.1% / -0.1% |
| GTC | -96% | -3.9% / -43% | -6.4% / -70% | 37% | 7.8 | 40% / 60% | -8.7% / +0.9% |
| STRK | -98% | -6.5% / -64% | -6.6% / -76% | 47% | 7.3 | 33% / 57% | -10.5% / -2.5% |
| RSR | -84% | -5.8% / -100% | -7.0% / -100% | 43% | 8.5 | 57% / 50% | +0.9% / -12.5% |
| 1000PEPE | -51% | +1.2% / -35% | -4.8% / -95% | 40% | 7.6 | 40% / 50% | +4.1% / -1.6% |
| WIF | -95% | -6.6% / -100% | -6.1% / -83% | 47% | 9.0 | 57% / 53% | -5.9% / -7.2% |
| 1000BONK | -86% | -2.9% / -100% | -3.9% / -88% | 40% | 9.1 | 43% / 63% | +3.5% / -9.2% |
| 1000SHIB | -81% | +0.1% / -17% | -2.7% / -44% | 23% | 5.2 | 30% / 40% | +0.8% / -0.6% |
Range 25% either side
| coin | price | no stop: mean / worst month | stop: mean / worst month | stop fired | round trips a month | left the range (up / down) | halves, no stop |
|---|---|---|---|---|---|---|---|
| BTC | +17% | +0.6% / -4% | +0.6% / -4% | 0% | 1.1 | 3% / 3% | +1.6% / -0.5% |
| ETH | -26% | +0.6% / -14% | -0.1% / -17% | 10% | 2.3 | 17% / 20% | +0.7% / +0.4% |
| SOL | -42% | +1.3% / -25% | +0.4% / -20% | 17% | 3.3 | 17% / 30% | +2.1% / +0.6% |
| ZEC | +4615% | +6.6% / -33% | +5.3% / -23% | 17% | 6.3 | 47% / 43% | +6.7% / +6.5% |
| NEAR | -27% | -0.7% / -25% | -2.5% / -27% | 17% | 5.0 | 30% / 43% | -1.6% / +0.1% |
| XRP | +136% | +3.0% / -15% | -0.3% / -41% | 13% | 3.3 | 23% / 20% | +5.0% / +1.1% |
| DOGE | -57% | -0.9% / -27% | -3.1% / -50% | 20% | 3.4 | 27% / 37% | -1.2% / -0.6% |
| GTC | -96% | -2.3% / -35% | -6.4% / -52% | 33% | 5.4 | 23% / 43% | -4.8% / +0.2% |
| STRK | -98% | -7.4% / -100% | -6.7% / -100% | 33% | 5.2 | 30% / 50% | -8.2% / -6.6% |
| RSR | -84% | -3.0% / -100% | -4.3% / -100% | 30% | 6.3 | 33% / 50% | +4.1% / -10.0% |
| 1000PEPE | -51% | +1.6% / -30% | -3.4% / -80% | 33% | 5.3 | 37% / 43% | +3.9% / -0.6% |
| WIF | -95% | -1.9% / -53% | -4.2% / -89% | 33% | 6.8 | 33% / 50% | -2.2% / -1.6% |
| 1000BONK | -86% | -2.5% / -100% | -5.0% / -100% | 33% | 6.4 | 30% / 47% | +4.1% / -9.1% |
| 1000SHIB | -81% | -0.1% / -20% | -3.2% / -36% | 17% | 3.3 | 17% / 30% | +0.9% / -1.2% |
Range 30% either side
| coin | price | no stop: mean / worst month | stop: mean / worst month | stop fired | round trips a month | left the range (up / down) | halves, no stop |
|---|---|---|---|---|---|---|---|
| BTC | +17% | +0.4% / -5% | +0.4% / -5% | 0% | 0.6 | 0% / 0% | +1.9% / -1.1% |
| ETH | -26% | +0.9% / -10% | +0.8% / -12% | 3% | 1.5 | 10% / 10% | +1.5% / +0.2% |
| SOL | -42% | +0.6% / -20% | -0.3% / -26% | 10% | 2.1 | 13% / 20% | +1.0% / +0.3% |
| ZEC | +4615% | +6.9% / -18% | +4.4% / -27% | 17% | 4.8 | 40% / 23% | +6.1% / +7.7% |
| NEAR | -27% | +0.6% / -20% | -2.4% / -28% | 13% | 3.5 | 20% / 27% | +2.6% / -1.3% |
| XRP | +136% | +2.3% / -15% | +0.5% / -31% | 7% | 2.2 | 17% / 17% | +3.9% / +0.8% |
| DOGE | -57% | -0.5% / -21% | -1.9% / -38% | 13% | 2.4 | 20% / 20% | -0.6% / -0.3% |
| GTC | -96% | -2.9% / -37% | -4.8% / -53% | 13% | 3.6 | 23% / 37% | -3.5% / -2.3% |
| STRK | -98% | -3.1% / -49% | -4.7% / -73% | 23% | 4.0 | 17% / 37% | -6.4% / +0.2% |
| RSR | -84% | -0.6% / -49% | -4.4% / -82% | 23% | 3.7 | 23% / 33% | +3.0% / -4.2% |
| 1000PEPE | -51% | +3.0% / -24% | -0.9% / -74% | 20% | 4.4 | 27% / 33% | +7.0% / -0.9% |
| WIF | -95% | -0.3% / -41% | -3.5% / -87% | 30% | 5.1 | 27% / 33% | -0.8% / +0.2% |
| 1000BONK | -86% | +2.3% / -25% | -3.3% / -88% | 23% | 5.0 | 23% / 33% | +6.3% / -1.7% |
| 1000SHIB | -81% | +0.5% / -17% | -1.2% / -35% | 10% | 2.4 | 10% / 20% | +2.2% / -1.2% |
Across the coins (counts)
| range | stop helped the average | stop helped the worst month | the same without October 2025 | positive in both halves (no stop / stop) | real beat 15 of 20 shuffles (no stop) |
|---|---|---|---|---|---|
| 15% | 4 of 14 | 5 of 14 | 4 and 8 of 14 | 3 / 0 of 14 | 2 of 14 |
| 20% | 1 of 14 | 5 of 14 | 1 and 9 of 14 | 3 / 0 of 14 | 3 of 14 |
| 25% | 1 of 14 | 2 of 14 | 2 and 8 of 14 | 4 / 1 of 14 | 4 of 14 |
| 30% | 0 of 14 | 0 of 14 | 1 and 5 of 14 | 4 / 2 of 14 | 2 of 14 |
The maths, step by step
The arithmetic of one grid (the rungs, the gap, a round trip, the stop) is worked through on How a grid works. This page adds four ways of reading the results.
- The mean month is the average of the 30 monthly results. BTC at 20% with no stop averaged +0.8% a month. Its worst month was −7%.
- The halves: the same average over the first 15 months and over the last 15. A result that is real should show in both. BTC at 20% was +2.2% in the first half, then −0.6% in the second.
- A shuffle: take a month's 720 hours and put them in random order. The month keeps every one of its moves, but it loses any trend and any habit of coming back. Each month was shuffled 20 times.
- Beat 15 of 20 shuffles: the real month paid the grid more than at least 15 of its 20 shuffles. If real prices had nothing special, the real month would be just one more shuffle among 21. It would land in the top 6 of the 21 by luck alone, 6 times in 21 (29%). Across 14 coins, that is about 4 coins by luck. The 2–4 found is no more than luck.
Cautions
- 30 months is short, and most of these coins fell over it. A long grid on falling coins is measured in hard conditions.
- Each month starts fresh and ends closed. A grid left running for many months behaves differently, and could lose more than any one month here.
- The prices are Binance's, including the minute-by-minute "mark" price the stop watches; the fees, funding and order sizes are Bybit's. On the night of 10 October 2025 each exchange's prices moved differently; another exchange's stops could have sold at other prices.
- A loss is counted as at most the whole stake (a liquidation), never more.
- The stop's month-without-the-crash check was added after we saw that one night drove the result. It is labelled as such, not hidden.
How it was measured
Prices: Binance USDT-perpetual hourly candles and one-minute mark prices, each monthly file checked against its published checksum. Fees, funding and order steps: Bybit's (maker 0.02%, taker 0.055%). The grid's orders: the same order planner as our paper bots, replayed hour by hour; a resting order fills only if the price trades through it. The control: 20 shuffles of each month's hours, each hour keeping its own minutes.
The mark price is the price a futures exchange uses to value positions and to decide liquidations. It is built from several exchanges' spot prices, so one wild trade can't liquidate anyone. For example, a sell that knocks the traded price down 3% for a second may barely move the mark.
Update, 16 October 2026: a caution
Added after this page was published. The text above is left as it was.
- The coins were picked after the window. They were the most traded on 13 October 2026, after the 30 months ended. That favours coins that survived and grew, ZEC most of all, and leaves out coins that faded or were delisted. Our page "Survivorship: how the coins were picked, and what it does" shows how much that can matter. Our DCA, QFL and grid follow-up pages use the same list.
Update, 21 October 2026: plain words
Added after this page was published. The text above is left as it was, except the explanations this lists.
- Now explained in plain words, with a small example, where it first appears: mark price.
Credits
- Grid trading: long-standing practice, with no known originator.
- R. A. Fisher, The Design of Experiments, Oliver and Boyd, 1935: the permutation test behind the shuffled months.
- The monthly replay, the stop test and the controls are our own design.
- Data: Binance's public prices; Bybit's public fees, funding history and order steps.
- PerpGremlin and Claude, Grid bots on 14 coins: what 30 months of prices say, TerminalGremlin, 2026: the tests, the code and the words.
research, not a recommendation to trade · generated 2026-10-13. Prices: Binance Vision (data.binance.vision), used under CC BY-NC-SA 4.0; what is drawn from it here is shared under the same licence.
← Grid bots, followed upMore rungs or fewer? →
Educational research from public market data; any trading discussed is paper, never real. Nothing here is advice. The full notice.