Last Updated: April 2026
*Disclaimer: This article is for informational purposes only and is not financial advice. Crypto trading involves significant risk of loss. Never trade with money you cannot afford to lose. Always do your own research (DYOR).*
I have been telling myself for months that I was going to run a proper trading bot experiment. Not the lazy version where I throw $100 at a default preset and check back in three weeks. A real one, with a journal, with daily notes, with screenshots, with the embarrassing mistakes left in. So that is what this is. Week 1 of running automated crypto trading bots, fully documented, with the actual numbers and the actual feelings that came along with them.
I am writing this for two reasons. First, because most of the bot content online is either pure shilling ("I made 400% in a month with this magical bot!") or pure doom ("all bots are scams"). Reality is messier. Second, because I genuinely wanted to test whether the bots I have been recommending in articles actually deliver in a real account, in real conditions, with real fees deducted. Anybody can write theory. I wanted to see it run.
This first week covered the period from April 18 through April 24, 2026. The market was choppy. BTC ranged between roughly $61,400 and $66,800. ETH spent the week digesting a strong move. Funding rates were modestly positive. It was, honestly, a fairly clean test environment for grid bots and a moderately friendly one for funding strategies. Here is everything that happened.
The Setup: What I Was Actually Running
I deliberately kept the setup simple because I wanted to be able to attribute results clearly. Two bots, two exchanges, two strategies. No layering, no overlapping pairs, no manual interventions during the test period.
The first bot was a grid bot running on a built-in exchange platform, set on the BTC/USDT perpetual pair. I chose the perpetual rather than spot for capital efficiency, with leverage capped at 2x. The grid range was set from $58,000 to $70,000 with 80 levels, which gave me roughly $150 per grid step. Total capital allocated: $2,000. The bot had been live for about thirty hours before April 18, so I started counting from the first full day of operation.
The second bot was a 3Commas DCA bot connected to my Try Bybit account. I configured it to trade ETH/USDT with a base order of $40, safety orders of $80 each, up to seven safety orders, with price deviations spaced at 1.5% with a martingale-style scaling factor of 1.05. Take profit was set at 1.2% from average entry. Total budget allocated: $1,500. This kind of DCA bot is the kind of thing Try 3Commas has been refining for years, and I wanted to see whether the default-ish configuration held up in real conditions.
I want to be clear about one thing: neither of these setups is exotic. They are configurations a beginner could reasonably arrive at after watching three YouTube videos and reading two articles. That was on purpose. I did not want to test "what is the absolute best parameter set" because I cannot generalize from that. I wanted to test "what happens if a normal person sets this up reasonably and walks away," because that is what most readers will actually do.
Combined starting capital across both bots: $3,500. Both started fully funded. Both had tight maximum drawdown notifications wired through Telegram. I committed to not adjusting parameters mid-week, which was harder than it sounds.
Free: Crypto Trading Platform Cheat Sheet
Side-by-side fee comparison, ratings, and quick-pick recommendations for every major exchange and trading bot. Save hours of research.
No spam. Instant download on the next page.
Day-by-Day: What Happened on the Screen
Day 1 (April 18): Quiet day. BTC drifted between $63,200 and $63,900. The grid bot caught five round-trip cycles and netted about $14 after fees. The DCA bot opened one base position on ETH at $3,128 and closed it within four hours for $0.43 profit after fees. End-of-day combined P/L: +$14.43. Not exciting, but exactly the type of day grid bots love.
Day 2 (April 19): BTC pushed up to $65,100 before fading. Grid bot caught nine cycles, which felt good for a single day. Net contribution: +$23.10. The DCA bot opened and closed two ETH cycles for a combined $1.85 profit. The friction here is that ETH was barely moving in the deviation range I had configured, so the bot was essentially scalping tiny moves. Daily combined: +$24.95.
Day 3 (April 20): First problem. BTC dropped sharply to $61,400 in a few hours after a CPI-related macro headline I had not been watching. The grid bot kept buying all the way down, exactly as designed, but I felt that pit-of-stomach feeling I always feel when a strategy I am running drawdowns hard. Unrealized loss on the grid bot peaked at -$87. The DCA bot triggered three safety orders on ETH and was sitting at -$22 unrealized. Combined unrealized drawdown for the day: -$109.
Day 4 (April 21): BTC bounced to $63,800. The grid bot cleared a chunk of inventory it had bought yesterday. Net realized for the day: +$31.40 from grid, plus the DCA bot finally closing its big ETH position at break-even-plus for a $4.10 profit. Daily combined: +$35.50. By end of day, all unrealized losses from Day 3 had recovered.
Day 5 (April 22): Grindy upward day. Twelve grid cycles, +$28.70 from grid. Two clean DCA cycles on ETH, +$2.40 combined. Daily: +$31.10.
Day 6 (April 23): Choppy and frustrating. BTC traded inside a $400 range almost all day. The grid was firing, but tiny cycles meant fees ate more of the gross. Net: +$11.60 grid, +$0.90 DCA. Daily: +$12.50.
Day 7 (April 24): A late-day push to $66,800. Grid bot rotated nicely, +$22.40. ETH was sluggish, just $1.10 from one DCA cycle. Daily: +$23.50.
The Final Numbers (Net of Fees)
Now we get to the part that actually matters: the bottom line, with all fees, all funding payments, and all slippage included. Below is the comparison table I put together at the end of the week.
| Metric | Grid Bot (BTC) | DCA Bot (ETH on Bybit via 3Commas) | Combined |
|---|---|---|---|
| Starting capital | $2,000 | $1,500 | $3,500 |
| Trades executed | 47 round trips | 9 full cycles | 56 |
| Gross profit | $187.40 | $19.20 | $206.60 |
| Trading fees | -$42.10 | -$8.45 | -$50.55 |
| Funding payments | -$3.10 | -$0.55 | -$3.65 |
| Net profit (week) | +$142.20 | +$10.20 | +$152.40 |
| Net return (week) | +7.11% | +0.68% | +4.35% |
| Annualized (linear, theoretical) | ~370% | ~35% | ~227% |
| Max unrealized drawdown | -$87 | -$22 | -$109 |
| Sharpe-ish (gut feel) | Decent | Mediocre | Decent |
Two huge caveats on those annualized numbers. First, you cannot just multiply a one-week return by 52 and call it an annual return. Markets have regimes. The week I tested was a friendly grid bot week. A trending bear week would absolutely flip these numbers negative, possibly badly. Second, max drawdown is a one-week observation. The real max drawdown across a year is almost certainly worse, perhaps materially worse, than what I observed in seven days.
What Worked: An Honest Pros List
The grid bot did exactly what grid bots are supposed to do. In a range-bound week, it churned through cycles like a slow, patient money machine. I checked it maybe four times total during the week, each check lasting under two minutes. That is the entire point of automation: I was not glued to the chart, I was not making emotional decisions, I was not over-managing a mostly-fine system.
The fee structure on the perpetual was kinder than I expected. I had budgeted for fees to eat about 30-35% of gross. They actually ate around 22%. That delta is meaningful when you compound it over time.
The 3Commas-on-Bybit setup connected cleanly. API permissions were straightforward, the bot logged trades to the dashboard accurately, and the Telegram alerts were reliable. I have used cheaper or free alternatives in the past where the connection between platform and exchange was the actual point of failure. None of that here.
The risk controls held. I set hard cutoffs for unrealized drawdown that would have triggered manual intervention if hit. They were never triggered. The grid bot's worst moment was at -$87, well within the band I had budgeted for. That is not a brag, that is the result of conservative grid range and reasonable leverage. It is also the boring discipline that most posts skip over.
What Did Not Work: An Honest Cons List
The DCA bot's results were genuinely underwhelming. Net $10.20 on $1,500 of dedicated capital for a week is not zero, but it is barely above what I would have earned holding USDC at any reasonable yield platform. The configuration I used was conservative, which I will own, and ETH simply did not give me the volatility-around-a-trend that DCA bots feed on. If ETH had trended steadily down, my safety orders would have averaged in nicely and my take-profits would have triggered more often. If ETH had trended steadily up, my single base orders would have closed faster. Choppy-flat with no clear direction is the worst case for this style of bot, and that is what we got.
Capital efficiency on grid bots is a real cost most people undercount. I had $2,000 dedicated to a strategy that returned 7.11% net. Sounds great, until you remember that most of that capital was not "working" in the sense of being deployed in the market. Roughly half the capital sat as collateral or as not-yet-filled grid orders most of the time. That is normal for the structure, but it means my "real" yield on actively deployed capital was lower than the headline number.
Slippage on the perpetual was occasionally surprising. Two of the 47 grid trades had visible price gaps where my order filled noticeably worse than the grid level. Total estimated slippage cost: about $4.80, which is small in the context of the week but would compound to ~$250 over a year if it stayed proportional.
The mental tax was non-zero. I told myself I was running a hands-off test and would not look at the dashboard more than necessary. I broke that promise on Day 3 when BTC dumped. I was on my phone checking unrealized P/L like a worried parent. The bots did not need me. I needed me. That is information.
Comparing What I Ran To Other Realistic Setups
I want to put my numbers in context against other configurations I considered, so the table above does not exist in a vacuum. If I had run a wider grid range ($55,000-$75,000 with 100 levels), I would have caught fewer cycles per week but with smaller drawdown risk if BTC broke out. If I had run a tighter range ($62,000-$66,000 with 60 levels), I would have caught more cycles but been completely stopped out if BTC had escaped that range. The chosen middle setup balanced expected hit rate against tail risk reasonably for a one-week window.
For the DCA bot, more aggressive deviation spacing (e.g., 0.8% steps) would have produced more cycles but risked deeper drawdown if ETH actually trended. A slower configuration with 2.5% steps would have produced almost no cycles in a quiet week. The 1.5% step I picked was a coin flip between those, and I think it was defensible, even though the result was modest.
If you are going to copy any of this: do not. Or rather, do not copy the parameters. Copy the methodology. Pick a configuration that matches your view of the market regime, allocate capital you can fully afford to be stuck in, write down your expected drawdown band before starting, and journal every day. The journal is the actual product. The dollars are a byproduct.
What I Am Changing For Week 2
I am leaving the grid bot exactly as is. It performed within expectations and the regime has not shifted enough for me to second-guess the parameters. If BTC breaks above $68,000 with conviction, I will pause the bot and reconfigure for a higher range. Until then, no changes.
I am adjusting the DCA bot. I am moving it from ETH to SOL/USDT, where I expect more daily volatility, and tightening the take-profit from 1.2% to 0.9%. I am also reducing the budget from $1,500 to $1,000 to free up capital for a third experiment I want to run: a small funding-rate carry trade on a major asset, with $500 of capital, using the same Bybit account. That experiment will be its own journal entry next week.
I am adding one boring discipline. Each day I will write three sentences in the journal: what happened, what I felt, what I did or did not do. The "what I felt" line is the one I am most curious about over time. I suspect a multi-month log will show me that my emotional state correlates poorly with bot performance and that my urge to intervene is almost always wrong. We will see.
FAQ
Q: Did the bots actually run unattended, or did you intervene?
I did not change parameters or close positions manually during the week. I did check the dashboards more than necessary, especially on Day 3, but I did not act on any of those checks. The bots ran as configured for the full seven days.
Q: Why didn't you use higher leverage to juice returns?
Because higher leverage on a grid bot mostly increases tail risk, not expected return. The grid still earns the same gross spread per cycle, but a single unfavorable trend move can wipe out weeks of gains. Two times leverage on a perpetual was already aggressive given the grid structure. I do not recommend higher leverage on automated strategies unless you have a deep understanding of liquidation mechanics.
Q: Are these returns sustainable?
Almost certainly not at this rate. The week was favorable for grid bots. A trending week, especially a trending-down week, would produce a flat or negative result, possibly significantly negative if BTC broke below the grid floor. A realistic expectation for a balanced grid bot strategy across mixed regimes is probably 1-3% net per month after fees, not 7% per week.
Q: Why 3Commas instead of just using the exchange's built-in bot?
Mostly because 3Commas lets me keep one configuration that works across multiple exchanges. If I want to add another exchange next month or switch the bot to a different exchange later, I do not have to relearn a new platform. The downside is the subscription cost, which is meaningful at small account sizes. For accounts under about $2,000 dedicated to bot capital, the built-in exchange tools are usually more cost-efficient.
Q: How much time did this actually take you?
Setup took roughly 90 minutes total, including configuring API keys, setting alert thresholds, and writing the journal template. Daily monitoring averaged about 6 minutes. End-of-week reconciliation, including pulling all the trade data and writing this article draft, took about 3 hours. So roughly 5-6 hours total for week 1, with most of that being the writeup. Ongoing weeks should be closer to 1 hour each.
Bottom Line for Week 1
Week 1 net return: +4.35% combined, or +$152.40 on $3,500 of capital. The grid bot carried the week. The DCA bot was a small contributor. The mental tax of watching live bots is real even when you tell yourself it should not be. The journal is, at this stage, more valuable than the dollar gains. Week 2 starts Monday with one parameter change and one new experiment. I will be back here next Sunday with the next entry.
If you are thinking about running a bot yourself, do me one favor: write down what you expect to happen, with numbers, before you start. Then compare what actually happens. The gap between those two things is where the real learning lives. The dollars are nice, but the gap is the lesson.
*Affiliate Disclosure: This article contains affiliate links to Bybit and 3Commas. If you sign up through these links, I may earn a commission at no extra cost to you. I only recommend platforms I actually use, and the trading results above were generated on the actual platforms I am recommending.*
*Disclaimer: This article is for informational purposes only and is not financial advice. Crypto trading involves significant risk of loss. Never trade with money you cannot afford to lose. Past performance, including the one-week journal results above, does not guarantee future results. Automated trading can produce losses faster than manual trading. Always do your own research (DYOR).*