Short answer. It sells, and then it runs out of things to sell. I opened a HYPE/USDT futures grid on Binance on 2026-08-26 — eight grids, geometric, 3x long, band trailing upward. Nine days later, five levels have closed. Every pending order on this bot is a buy. Eight buys, zero sells. The position I am still holding is about 0.045 times the investment — the bot has sold almost all of it on the way up. This is my second grid, and it is behaving like the opposite of my first one, which has closed three levels in a hundred and two days and is carrying about 3.4 times its investment.
| n | 1 run. One exchange, one symbol, one configuration, one continuous run. Nine days old. A single case, and a very short one. |
|---|---|
| Period | 2026-08-26 to now. Day 9 as of 2026-09-04. Still open. |
| Instrument | HYPE/USDT perpetual futures, grid, Long, cross margin |
| Exchange | Binance. This is not the exchange my gold grid runs on. |
| Leverage | 3x set on the bot. The position actually held is about 0.045× the investment as of day 9 — almost nothing. The setting is a ceiling, not a position. |
| Grids | 8, geometric (equal steps in percent, not in price) |
| Per level | 3.50%, as the bot states it. Dividing the band into eight geometric steps gives 3.54%, so the bot's figure and my own division agree to within a rounding. |
| Trailing | On, upward only. The downward continuation option is switched off. |
| Stop | None set. No take profit either. |
| Liquidation | 55.4% below the mark price, 52.7% below average entry, read 2026-09-04. |
| Chosen by | Me, using the grid simulator on this site and an AI alongside it. Not an optimiser handing me an answer, which is how bot one was set up. See below, including the conflict of interest that comes with it. |
| Fees | Not stated on this page yet. The gold page has its fee tier; this account's isn't up here. |
| Excluded | Nothing stated yet. |
This is the thing I did not expect to be looking at on day 9, and it is the clearest single fact I have from either bot.
Pending orders: eight buys, zero sells. Read 2026-09-04, they sit at 3.25%, 6.56%, 9.75%, 12.84%, 15.83%, 18.71%, 21.49% and 24.18% below the price at that moment — those distances move with the price, so they are a reading and not a setting. The whole book is underneath. There is nothing above.
Those eight distances also let you check the spacing without taking my word for it. Consecutive levels are 3.54% apart, all eight of them, to within two hundredths of a percentage point — which is the ladder itself, not my division of the band. The bot states 3.50% profit per grid; the gap between the two is small and I haven't chased it down.
Price rose. On a long grid, every level that fills going up is a sale — the bot sells a slice of what it holds and books the difference. It did that five times in nine days, and the five closes paid almost identical amounts.
Here is what the screen shows now, and I am keeping it to what the screen shows. Price is sitting 0.3% below the top of the band. The book is eight buys and no sells. The position is 0.045 times the investment. I did not watch the levels turn over, so I am not going to narrate how each sale became a buy order — that is how a grid is supposed to work, but it is a description of the mechanism and not something I observed. What I observed is the state it is in: at the top of its range, holding almost nothing, with its whole book underneath it.
It is not broken and it is not finished. It is out of inventory.
One thing I can't tell you about this: the band trails upward — that is switched on, and the downward continuation is switched off — but I don't know how far it has actually moved, because I didn't write down where it started. That is my omission, and it goes in the method box the day I can fix it.
The screen shows +4.66% of the investment after nine days. That figure is three things added together, and the exchange splits them out:
| component | day 9 |
|---|---|
| matched profit — the levels that closed | +4.46% |
| the open position, marked to price | +0.25% |
| funding paid | -0.05% |
| total shown on the screen | +4.66% |
Almost all of it is the grid. That is the opposite of what my other page has been documenting for three months, where the grid's contribution has been frozen for eleven days and the whole screen moves with the position. Here the position is a rounding error and the closed levels are the number.
I want to be careful about what that does and doesn't mean. It does not mean this configuration is better. It means price went up, in a direction this bot is built to sell into, for nine days. A long grid in a rising market prints matched profits and shrinks its position. That is mechanical. It would do the same thing on a bad configuration.
This is the line that is different between my two bots.
Bot one's settings came out of an optimiser. I asked an AI for the configuration that would have made the most money on past gold data, across the whole history, and I used what came back. I never asked how often it would trade. It has closed three levels in a hundred and two days, and the page for that run is me working out why.
These settings I worked out myself — on the grid simulator on this site, with an AI alongside me while I did it. The thing that tool reports, and that I never asked the optimiser for, is how often a configuration would have traded.
So let me say the obvious thing before someone else does. I am reporting on a bot whose settings I chose with a tool I built, on the site where I publish both the tool and the results. That is a conflict of interest and naming it does not remove it. What I can do is leave the numbers where you can check them and not quietly edit this page when they go the other way. Nine days is not evidence that the second method is better. It is nine days. If this run ends up where bot one is, that goes here too, in the same size type.
On the other bot, this took me until day 98, because I was reading the wrong screen. So I read it on this one on day 9.
Liquidation sits 55.4% below the mark price and 52.7% below my average entry. The plain division — 100 divided by 3x — would say 33%. It is nowhere near, for the same reason it was nowhere near on the gold bot: the leverage setting is a ceiling and the position is a fraction of it. At 0.045× the investment, there is very little here to liquidate. The margin ratio the exchange shows for this position is 0.04%.
There is no stop on this bot. On the gold one there is, and it is the nearer of the two edges — so on that bot the stop is what would end the run. Here there is nothing between the position and liquidation except distance. Right now that distance is large because the position is small. If the eight buys underneath all fill, both of those change at once.
I run two grids. They are on different exchanges, on different assets, with different settings, and they are nine and a hundred and two days old. I am not going to put their percentages side by side. One of them has had nine days to produce a number. Comparing those returns would be comparing a run to a rounding error and calling it evidence.
What is fair to compare is how they are built and what they are holding, because those are settings and positions rather than results:
| this run, day 9 | the gold run, day 102 | |
|---|---|---|
| grids | 8, geometric | 5, arithmetic |
| each level needs a move of | 3.50% | 4.16–5.25% |
| leverage set | 3x | 8x |
| levels closed | 5 | 3 |
| position held, as a multiple of the investment | 0.045× | 3.4× |
| pending orders | 8 buys, 0 sells | — |
| band trails | up (downward continuation off) | down — already moved one step |
The bottom half of that table is the part worth sitting with. Two things called the same thing, holding amounts that differ by a factor of about seventy-seven. One of them has sold its way empty into a rise. The other has bought its way heavy into a fall. Neither of those is a setting I chose; both are what the same kind of machine does when price moves in one direction for long enough.
And they are not two different fates. They are the same one at different points. If HYPE comes back down through those eight buy orders, this bot fills all eight, spends its cash, and ends up holding a leveraged long it cannot sell — which is precisely where the gold bot is today. The gold bot's band is set to trail down and has already moved a step. This one's is set to trail up. Same mechanism, opposite ends of it, and this time I get to watch it from the start.
The screen also shows an annualised return for this bot. I am not repeating the figure. It is nine days multiplied out to a year, in a week when the asset went up, and the number it produces is not about anything. Publishing it would be the single most misleading thing I could put on this site, and it would be the most impressive-looking. Those two facts are usually the same fact.
I also can't tell you yet who chose these settings or what this account's fee tier is. Both of those rows in the method box say so rather than being filled in with something plausible. They go up when I can state them.
read at the date shown — this run is open and the percentage moves intraday
| read at | day | total % | matched profit % | closed | status |
|---|---|---|---|---|---|
| 2026-09-04 | 9 | +4.66% | +4.46% | 5 | running |
One reading. There is no trend here and I am not going to draw one. The five closes happened on 2026-08-26, 08-27, 08-30, 09-01 and 09-04, and each one paid almost exactly the same amount. With equal percentage steps and equal order sizes I would expect that, but I haven't opened the individual fills to confirm that is the reason, so treat the explanation as mine and the amounts as the screen's.
This is one asset, one exchange, one run that is nine days old and still open. n = 1 run. It is not evidence about grid bots, it is not evidence about this asset, and it is very obviously not a result.
I don't publish USDT amounts anywhere on this site, and I don't publish price levels — the band, the individual grid prices, the entry, the liquidation price. The percentages above are ratios to a quantity I never publish, and that quantity is what you would need to get from a ratio to a size.
Two of the ratios on this page are weaker than the rest and I would rather say so than have someone point it out. The distances to my eight buy orders, and the distance to liquidation, are both quoted from the current price — and anyone can look up what this asset traded at on the day I read the screen. So those particular ratios sit closer to being levels than the others do. I'm publishing them anyway, for the same reason I publish the equivalent on the gold page: what the redaction protects is the size of the account, and size needs a quantity — how much of the asset I hold. That is the number I never publish, in any form, on any channel.
I also can't show you the fills themselves beyond their timestamps and what each one paid. The full export isn't done on either bot.
The gold grid has its own page, and it is the longer story: why is my grid bot barely filling? 5 grids, 102 days, 3 trades closed. Nothing on this page changes anything on that one.
Raw data isn't published yet. The CSV for this run, with a README repeating the method box above, goes here when it's up.
There is no signup link anywhere on this page and nothing on it pays me. I want to be exact about that rather than just quiet: I do have a referral code for this exchange too. It sits on the referral page with the other one, and I am deliberately not linking it from here. A record of a nine-day-old bot is the last place a signup link belongs, and if you want the code you can go and find it. If that ever changes, the disclosure goes at the top of this page, before anything else, not down here.
Every other record I've published →
educational only · no signals · no profit claims · not financial advice · dyor
bots ≤ 30% of my portfolio · core is index funds