That percentage on the grid bot screen is one of the most misread numbers in crypto. It looks like the yield on a savings product, so people read it the way they read a savings product: "this strategy makes that much in a year." It does not mean that, and it is actually two different numbers.
They show up in two places: the estimated or backtested return you see while creating a strategy, and the annualized figure attached to a bot once it is running. The first is simulated, the second is converted. Neither is a forecast, but they fail in different ways. Below, each one in turn, plus the numbers actually worth watching instead. The mechanics are broadly the same across platforms that offer grid bots; Binance is just the venue used for examples here.
1. The running figure: pure extrapolation
Start with the one you stare at most. The conversion these bots generally use is:
Annualized = (realized grid profit ÷ capital committed) ÷ days running × 365
That is the whole thing. It predicts nothing. It takes what you have already earned and multiplies it out to a year at the current pace. Understand that one line and most of the confusion around this number dissolves.
The dangerous part is the denominator: days running. On a new bot it is tiny, so any amount of profit gets blown up. An illustrative set of numbers:
| Capital | Realized grid profit | Days running | Headline it produces |
|---|---|---|---|
| 1,000 | 4 | 2 days | 0.4% ÷ 2 × 365 ≈ 73% |
| 1,000 | 4 | 10 days | 0.4% ÷ 10 × 365 ≈ 14.6% |
| 1,000 | 4 | 60 days | 0.4% ÷ 60 × 365 ≈ 2.4% |
All three rows are the same four units of profit. Same strategy, same capital, same result — and purely because of elapsed time the label falls from 73% to 2.4%, a factor of thirty. So "my new grid shows an amazing annualized return" is neither luck nor a bug. It is what short-window extrapolation always does, and it will slide as you keep running.
2. Why it drifts down for no apparent reason
The other common puzzle: the bot has not lost anything, yet the annualized figure is lower every day. Same denominator. The numerator (realized profit) is frozen while the denominator (days) ticks up, so the quotient slides.
When does the numerator freeze for long stretches? When price is outside your range. The bot is posted but nothing fills, profit stalls, days accumulate anyway, and the figure keeps sinking. That decline is not a new loss — it is telling you the machine is idle. What to do about being out of range is a separate question, covered in price left your grid range.
3. The one thing to remember: it excludes your unrealized loss
If you take one sentence away, take this one: the annualized grid figure counts only the profit already booked from completed levels, not the unrealized P&L on the position you are still holding.
That explains the jarring pairing: a healthy double-digit annualized figure sitting right next to a total return that is negative. When price has broken below your range and you are holding coins bought at every level on the way down, that pairing is close to guaranteed. The bot really did fill levels on the way down and those profits are real — they just do not come close to covering the unrealized loss.
4. The creation-screen estimate: a different failure
When you are setting a strategy up, the interface usually offers an estimated or backtested annual return based on the range and level count you typed in. Its origin is completely different from the one above: it generally takes the volatility over some past window (7 days, 30 days, whatever the lookback is), estimates how many fills your settings would have triggered in that window, and annualizes the result.
In other words, it answers: "if the coming period behaves like the stretch of history I sampled, what would these settings have produced?" Its reliability therefore rests entirely on one assumption — that future volatility and fill frequency resemble that past window. The moment the market turns from chop to a one-way trend, fill frequency collapses and the assumption is void. Which is, inconveniently, exactly the market in which grids lose the most.
This is also why estimates seen just after a violent stretch look absurd. A triple-digit annualized label does not mean the strategy is strong. It means the window it sampled was volatile.
| The number | What it is really saying | What it cannot do |
|---|---|---|
| Creation-screen estimate / backtest | A simulation of a past volatility window applied to your settings | It is not a promise of future return, and it swings a lot depending on the lookback window |
| Running annualized figure | Realized grid profit extrapolated from days elapsed to 365 | Not a sustainable rate; excludes unrealized P&L; not comparable across bots with different runtimes |
| Total return / total P&L | Realized profit plus unrealized P&L | This is the number that tells you if you are up or down — though it still says nothing about the future |
5. What to watch instead
Three sturdier measures, in order of importance:
- Total return including unrealized P&L, in currency rather than percent. "How much has this machine made or lost me to date" is a question no denominator can distort. Percentages get stretched by time; amounts do not.
- Compare against simply holding over the same period. A grid's competitor is not zero, it is the version of you who did nothing. What would holding the asset have produced over the same stretch? That is the only honest baseline for whether the tool was worth using in this market — and in a strong uptrend the comparison stings.
- Fill count and net profit per level. A grid eats by filling, so "how many levels have completed, and what is left of each one after a round trip of fees" is far closer to how the machine is actually doing than any annualized figure. Net profit per level depends heavily on the spacing method, covered in arithmetic vs geometric spacing; to run it on your real fee rate, use the grid profit simulator and the break-even calculator.
The annualized figure is not entirely useless. It has one legitimate use: comparing two bots that have been running for similar lengths of time. Two machines both thirty days in with comparable capital — the higher figure genuinely is filling more. Beyond that, treat it as an activity light, not an expectation.
6. FAQ
How is a grid bot's annualized return actually calculated?
The figure on a running bot is generally realized grid profit divided by the capital you committed, which gives the return so far, then divided by the number of days it has been running and multiplied by 365. It is pure arithmetic conversion with no forecasting in it, and it does not include the unrealized profit or loss on whatever you are still holding. The estimate on the creation screen is a different animal, derived from a simulation of past volatility and fill frequency, essentially replaying a stretch of history against your settings. The exact formulas vary between platforms, so check the current documentation for whichever one you use.
Why is the number so high on a bot I only just started?
Because the denominator is days elapsed, and on a new bot that denominator is tiny. Take an illustrative example: put in 1,000, earn 4 in grid profit over two days, and the return is 0.4 percent. Divide by 2 days, multiply by 365, and the headline reads about 73 percent. That exact same 4, earned slowly over 60 days, annualizes to about 2.4 percent. Same money, same strategy, a thirtyfold difference in the label, driven entirely by how long it has been running. So the flattering number in the first few days is neither luck nor an error; it is what short-window extrapolation always produces and it cannot be treated as a sustainable rate.
Does the annualized figure include unrealized profit and loss?
Usually not. It counts the profit that has actually been booked level by level, while the position you are still holding sits in a separate column as unrealized profit and loss. That is why you can see a jarring combination: a healthy-looking annualized figure next to a total return that is negative. When price has broken below your range and you are holding coins bought higher up, that combination is close to guaranteed. To judge whether a strategy is making money, read total return including unrealized profit and loss, not the annualized grid figure.
When is this number still useful?
It is genuinely useful for comparing two bots that have been running a similar length of time. Two strategies both thirty days in, with comparable capital, are fairly compared this way: the one with the higher figure really is filling more often and capturing more spread. It also works as an activity indicator, since it drifts down as the denominator grows during a stretch with no fills. There is exactly one thing it should not be used for: as an expectation of future return, or as something to line up next to a savings rate or a yield product. Those are different measures with different risk structures.
Can I trust the estimated return shown when I create the bot?
Treat it as a sanity check on your settings, not as expected income. It is generally derived from historical volatility multiplied by an expected fill frequency, which assumes the coming period behaves like the stretch of history it sampled. When the market turns into a one-way trend, fill frequency collapses and that assumption fails outright. Where it earns its keep is in comparison: change the level count or the range on the same pair and watch how the figure moves, which helps you rule out obviously bad parameter combinations. The absolute value, especially a triple-digit one, is mostly a reflection of how volatile the sampled window was.
Nothing here is investment advice. Crypto prices are volatile, grid trading does not guarantee a profit, and it carries the risk of being left holding a losing position and — on futures — of liquidation. Judge your own situation and size accordingly. The 1,000 of capital, 4 of profit and the 2 / 10 / 60 day runtimes are illustrative assumptions used to show how the conversion behaves, rounded for display; they are not a promise of returns or a market forecast. How each platform computes its annualized figure, what it calls it, and whether unrealized P&L is included all differ, so go by the current documentation and by what your own account screen shows. Official product page to check against: What is spot grid trading.