You open the spot grid screen, tap [AI], and the upper limit, lower limit and grid count fill themselves in. Next to them sits a period picker: 7 days, 30 days, 180 days. Switch periods and every number moves. So which period do you pick, and can you hit Create on whatever it hands you?
No guessing needed here. Binance publishes the formula behind the spot grid AI range in its help center, and it is only a handful of lines; we checked each one against the official pages. The official wording is that the system generates parameters "based on the technical analysis of the symbol's price and the selected time period". In practice that means it looks back over a stretch of past prices, draws a box around them, and drops the box into your form.
How the upper and lower limits are worked out
The spot page describes a Bollinger band. Daily candles are taken over the period you chose — the past 7 of them for 7 days, 30 for 30 days, 180 for 180 days. From those it computes a moving average (MA) and a standard deviation, then:
- Upper band (UB) = MA + 3 × standard deviation
- Lower band (LB) = MA − 3 × standard deviation
The page puts it as "BBM is set to 3". Whether the average is built from closing prices or something else is not stated, and we won't fill that gap for them.
A sanity check comes next. Using 1-hour candles, the system works out an average price, (Open + Close + High + Low) ÷ 4, plus a drift figure, DR = (average price − MA) ÷ MA. Then it asks how far the upper band sits above that average price:
- If UB ÷ average price − 1 lands between the min range and the max range, the grid's upper limit = (1 + DR) × UB — the band shifted by however far the average price has drifted from the MA.
- If not, the upper limit becomes average price × (1 + default range), and the Bollinger number is thrown out.
The lower side runs the same check on 1 − LB ÷ average price. Inside the window, the lower limit = (1 + DR) × LB (note it is (1 + DR) here too, not (1 − DR)); outside it, the lower limit falls back to average price × (1 − default range).
All three ranges scale with the period: min range = 1% × n^(1/3), max range = 15% × n^(1/3), default range = 5% × n^(1/3), where n is 7, 30 or 180. Think of it as a fence around the band. A coin that has barely moved all week produces a box that is too tight; one fresh off a violent swing produces a box that is too loose. Either way the band gets replaced by a preset width tied to the period.
7, 30 or 180 days: how far apart they land
Plug n into the formulas and the fences come out as below. These percentages are our own calculation from the published formula, not figures printed on the official page; whatever the screen shows at the time takes precedence. Checked September 2026.
| Period | Data for the band | Min range | Max range | Default range |
|---|---|---|---|---|
| 7 days | Past 7 daily candles | ≈ 1.91% | ≈ 28.69% | ≈ 9.56% |
| 30 days | Past 30 daily candles | ≈ 3.11% | ≈ 46.61% | ≈ 15.54% |
| 180 days | Past 180 daily candles | ≈ 5.65% | ≈ 84.69% | ≈ 28.23% |
Cube roots grow slowly. Going from 7 days to 180 multiplies the period by more than 25, yet the fences only widen about 2.95 times. The band itself is what really spreads things out: if those 180 daily candles contain one big run up or down, the standard deviation balloons and the band opens with it. So the 180-day setting usually hands you a wider box, while 7 days hugs last week's prices.
Here is each path once, using made-up numbers purely to show the mechanics. None of this describes a real coin.
- The normal case. 30 days, MA = 100, standard deviation = 4, so UB = 112 and LB = 88. Average price 101, so DR = 1%. The upper band sits about 10.9% above the average price, inside the 3.11%–46.61% window, so the upper limit = 1.01 × 112 ≈ 113.1. On the lower side, 1 − 88 ÷ 101 ≈ 12.9%, also inside the window, so the lower limit = 1.01 × 88 ≈ 88.9.
- Too tight. 7 days on a coin that went sideways all week: MA = 100, standard deviation = 0.5, UB = 101.5. With the average price also at 100, the band is only 1.5% above it, short of the roughly 1.91% floor for 7 days. The band is dropped and the upper limit falls back to 100 × (1 + 9.56%) ≈ 109.6 — wider than the band would have been.
- Too loose. 180 days after a half-year with a big trend in it: MA = 100, standard deviation = 30, UB = 190. At an average price of 100 the band is 90% higher, over the roughly 84.69% ceiling, so again it falls back: 100 × (1 + 28.23%) ≈ 128.2.
In the last two cases the range on your screen no longer has anything to do with the Bollinger band; it is just the average price plus or minus a fixed percentage. The official page doesn't say whether the interface tells you which path was taken. You can estimate it yourself: if the limits sit about the default range away from the current price for that period, the fallback probably kicked in.
Grid count: only the futures guide spells it out
The spot page covers the range and stops there. The grid-count formula is published in a different document, the futures grid AI parameters guide:
Grid Number = (1 + buffer) × (upper limit − lower limit) ÷ ATR, with the buffer currently set to 30%.
ATR, the average true range, measures how much a candle typically moves. That guide uses different candles from the range calculation: for 7 days, the past 168 thirty-minute candles; for 30 days, the past 360 one-hour candles; for 180 days, the past 1,080 two-hour candles. Read backwards, each grid step works out to about ATR ÷ 1.3, a bit narrower than one average candle's move. Longer periods use coarser candles for ATR, so the steps tend to get wider.
Carrying on with the 30-day example: a range of 88.9 to 113.1 is about 24.2 wide; with a hypothetical ATR of 1.5, the grid count ≈ 1.3 × 24.2 ÷ 1.5 ≈ 21.0. How that gets rounded isn't documented.
What the formula never looks at
Once you have read those few lines, the AI setting is easy to describe: it turns recent volatility into a price box, and on futures it slices that box by the average candle move. Several things never enter the calculation.
- Direction. The band opens symmetrically around the average and only slides by the current drift. Nothing in it estimates whether price goes up or down next.
- Whether you would hold the coin at the bottom. The range comes from the market, not from your balance or how long you plan to stay in.
- Fees. A wide range with many grids can leave each grid a thin margin, and the published method never compares that margin to a fee rate. Arithmetic spacing makes the top of a wide range the thinnest part; see arithmetic vs geometric grid spacing.
- What happens next. The box is built from the last 7, 30 or 180 days. Once the market leaves that history behind, price can walk right out of it — what to do when price leaves your range.
As a starting point it is fair: the range has a statistical basis, which beats typing a round number from gut feel. What it can't do is make the call for you. One practical check: tap both 7 days and 180 days and compare the two sets of limits. The further apart they are, the less last week resembles the last six months, and the shakier it is to copy either one as is.
After copying to manual: what to change
The AI tab also lets you copy the parameters over to manual creation, which carries the whole set into manual mode so you can edit each field. The official page also names two defaults: recommended parameters use Arithmetic Mode, and Sell All Base Coin at Stop is switched on. Go through them in this order:
- Could you live with the lower limit? If price slides all the way there, most buy orders have filled and you are mostly holding the coin. Only set a lower limit you'd be willing to hold at.
- Arithmetic or geometric. AI starts you on arithmetic. When the upper limit is far above the lower one, the top grids earn a visibly thinner percentage and geometric may suit better; the spacing guide shows how to check.
- Sell everything at stop, or not. It is on by default, which means stopping the bot sells the base coin you are holding. If you want to keep the coin after stopping, switch it off here.
- Grid count and amount per grid. Once the count is set, your total is split across the grids, and each share has to be big enough to place an order; see how much a grid bot needs to start. To try a few counts and watch the spacing change, run them through the grid spacing and grid count calculator.
One limit people forget: the official page says you can only set up 5 grid strategies per trading pair (whatever the setup page shows at the time takes precedence). Running a 7-day and a 180-day version side by side uses two of those five.
This piece covers only the algorithm behind the AI button. How range, grid count and per-grid profit trade off against each other, and which to settle first, is in how to set grid parameters.
Nothing here is investment advice. Crypto prices are volatile, grid strategies do not guarantee a profit, a drop below your range can leave you holding a losing position, and futures grids also carry liquidation risk. Judge your own situation and size accordingly. The MA, standard deviation, average price and ATR figures above are all hypothetical, used only to show how the formula works. How the percentages for the three periods were derived is explained above the table in section 2. Official pages to compare against: How to use auto parameters in spot grid trading and the futures grid AI parameters guide. As the official page itself puts it, grid trading is used at your discretion and at your own risk.