Cornix trading bot review: 71% win rate, still lost money

I ran Cornix on a $600 Bybit sub-account for five weeks, following two Telegram signal channels, and closed 31 trades. Twenty-two of them hit at least one take-profit target. That's a 71% win rate, and the account still finished down roughly 4%.

When an account bleeds while the win rate looks great, everyone's first instinct is to blame the signal group. Mine was too. I spent an evening pulling up the channel's call history, ready to write the usual post about how signal providers cherry-pick their results. Then I went back through my own fills and found the actual culprit sitting in a dropdown I'd never touched, on the very first setup screen, left at its default.

Why does a 71% win rate lose money?

Because the win rate counts trades, and my P&L was decided by size. Those aren't the same thing, and Cornix's default configuration quietly makes them diverge.

Here's the question I should have asked before turning anything on: when do all of your entry orders actually fill?

Cornix splits a signal's position across multiple entry targets. Per Cornix's own entry ratios documentation, the default number of entry targets is 4, and the default strategy is "evenly divided," which puts 25% of your trade amount on each one. Sounds sensible. Scale in, get a better average price, standard DCA logic that every trading education post recommends.

Now answer the question. Target one fills when price touches the top of the entry zone. Targets two, three and four only fill if price keeps moving against your direction. So a signal that's immediately right gets you in with 25% of your intended size and then runs away without you. A signal that's wrong walks you down through all four levels until you're carrying the full position, right before it hits the stop.

Evenly divided entries hand you a quarter position on your best trades and a full position on your worst ones. That asymmetry isn't a bug and it isn't hidden. It's just the arithmetic of scaling into a zone, and it runs in the background of every trade the bot takes for you.

A 71% win rate on quarter-sized winners and full-sized losers is a losing system. Size decides your P&L, not the hit rate, and the entry ratio decides your size.

The trade that made it click was a SOL long in week three. The channel called it, price ticked down about 0.6% into my entry zone and filled all four targets within eleven minutes, then reversed and ran to the second take-profit. Great trade, full size, my biggest single win of the whole test. Two days later the same channel called an ETH long that was right almost instantly, and I got filled on target one only. Quarter size. Same signal quality, same bot, a quarter of the payoff, and nothing about the trade told me in advance which of the two it would be.

Is any of this Cornix's fault?

Mostly no, which is the annoying part of writing this review.

The documentation is plain about how it works. Cornix ships eight entry ratio strategies, including "one target" (100% on the first entry), "fifty on first target," and a decreasing exponential that puts 53.3% on entry one and 6.7% on entry four. The tools to fix the asymmetry are right there in the same dropdown that caused it. I just never opened it, because the default looked like the responsible choice and DCA has a good reputation.

So my honest, mildly unpopular take: on signal trades, turn the multi-entry DCA feature down or off. I moved to "fifty on first target" for the last two weeks and the shape of the results changed immediately, fewer trades reached full size and the average winner got bigger relative to the average loser. Two weeks is not evidence of anything, to be clear. It's a direction, not a result. But the feature everyone praises as risk management is the one that was quietly inverting my position sizing, and the fix was a single dropdown.

What does Cornix actually do well?

Execution, which is the boring thing you're paying for and the thing it genuinely nails.

It parses free-text Telegram calls into structured orders reliably. Across 31 trades I had zero missed entries and zero missed take-profits, including a 3am BTC call I slept straight through (I'd muted the notification channel four days earlier because the trade alerts kept waking me up, which turned out to be the single best risk control of the whole test). Stop-losses moved to breakeven when I told them to. Trailing take-profits behaved as configured. For a tool whose entire job is doing the thing at the moment you can't, that matters more than any feature list.

The custody model is also right. Cornix never holds your funds, it connects to your exchange through API keys, and their API key documentation explicitly tells you to leave withdrawal permission disabled. Read and trade only, plus futures permission if you need it. That's the correct answer, and it's stated in their own docs rather than buried in a support ticket, which is more than I can say for some platforms I've reviewed. I ran mine on a dedicated Bybit sub-account with its own keys, which is what I'd suggest to anyone testing a new bot. If you want the full version of that setup, it's in our self-hosted bot security checklist.

Is $32.99 a month worth it on a small account?

Depends entirely on your account size, and the answer for most people reading this is no.

Cornix's Advanced plan runs $32.99 a month (about a third less if you pay annually) and includes 2 API slots, 3 signal bots, 5 grid bots and 5 DCA bots. Premium is $52.99 and lifts the bot limits. There's a 14-day free trial with no card required, which is genuinely enough time to learn whether your signal source is worth automating.

Do the arithmetic on my test account though. On $600, that subscription is 5.5% a month you have to make before you've broken even on the tool, roughly 66% a year, and that's before exchange fees and funding. At $5,000 the same subscription costs you 0.66% a month, which is a rounding error. Cornix isn't expensive, it's just brutally size-dependent, and nobody selling signal automation says that out loud.

Can a good bot rescue a bad signal group?

No, and this is where I get a little cynical about the whole category. Most signal groups are a distribution business first and a trading operation second, and the calls exist to keep the channel active enough to sell the next tier of membership.

The CFTC's advisory on automated trading, titled AI Won't Turn Trading Bots into Money Machines, warns specifically about trade signal strategies promising outsized or guaranteed returns, and flags claimed "100 percent win rates" as a fraud marker. My test is a smaller, more mundane version of the same lesson: my channels weren't lying, their calls really did hit 71% of the time, and I lost money anyway. A published win rate with no position sizing attached is the most useless number in this industry.

Cornix executes whatever it's given, faithfully, at 3am, without hesitating. That's a multiplier on the quality of your source, in whichever direction that quality points.

The verdict

A 6.8 means the tool is better than my result was. Cornix did its job. I configured it badly, in the specific way its defaults invite, and five weeks of trades quietly paid for that lesson.

The limits here are real and worth stating: 31 trades across two channels in one market regime is a small sample, I didn't keep a per-trade ledger with exact fill prices, and a different pair of signal providers could easily have produced a different number. What I'm confident about isn't the 4% loss, it's the mechanism, because the entry ratio math holds regardless of which channel you follow. If you'd rather build the sizing logic yourself instead of inheriting someone's defaults, the build-in-public bot journey covers how I'm doing that, and the bot match quiz is a faster way to figure out whether signal automation is even the right category for you.

If you only do one thing before your first Cornix trade: open the entry ratio dropdown and change it off "evenly divided." Everything else you can tune later.