BTC · ETH · BNB · SOL · XRP, live
See the crypto market clearly, without the hype.
- Buy, Sell or Wait on five coins, in plain words, as each candle closes
- The money and mood behind every move — flows, leverage, funding, news
- Every call graded in public including the wrong ones
Free during the test phase. New accounts are approved by hand.
- 35
- studies run
- 1
- of them shipped
- 6
- tests on unseen data
The research that failed is published too, not just the one thing that worked.
The last seven calls, unedited
Tested at 58% on 432 calls the model had never seen — though an audit in September 2026 found most of that is the market and not the model: it calls up on 79% of 1-hour candles, and about 53% of the candles it picked rose anyway. Below is what it has actually said most recently — consecutive, nothing removed.
Right more often than wrong, and still not enough. A trade has to gain 0.09% to cover fees both ways. In the September 2026 audit the average 1-hour call lost 0.15% after fees. That gap is why this site publishes a record instead of selling signals.
What you are reading is a statistical property of one market on one timeframe. It is not calibrated to your position, your stake or your risk, and it is not advice.
35 studies run, 1 shipped. The other 34 were ideas that failed testing — things kept off this page rather than sold to you.
Every test is hashed before it is scored. Ask any other signals service for a timestamped hash made before the call was published.
The rules this follows
Written down so they can be held against the record. Each one costs something, which is the only reason any of them is worth reading.
- Every call graded automatically
The moment a candle closes, the call is marked right or wrong. You can see the full record, by coin and timeframe.
- Real numbers, after fees
Every call carries what it came to after trading fees, not only whether the direction was right. The two rarely agree.
- Ideas tested before they reach you
New ideas are tested on data the model never saw. The most promising are locked in advance and tracked live, so results can't be tweaked after the fact.
- Calm, not scary
No flashing alerts or countdown pressure. Just what's happening and what it means, so you can decide for yourself.
Shipped lately
Every change and every research round, dated. Members see the full history.
- change
The market mood dial was drawn in green whatever it said, including when it said the background looked soft
At the top of the Market page a half-circle dial answers “how the market feels”.
- change
Four readings on the market page told you what happens next, which this site’s own research says they cannot
The market page scores roughly two dozen background readings as good, neutral or soft, and four of them ended with a prediction.
- change
The site now says what it measured rather than who measured it, and two figures gained the labels they needed
Fifty-four sentences across twenty-one pages said “we” — we tested, we checked, our research, how we know.
How to join
- 1
Create an account
Name and email. No card is asked for, because there is nothing to buy.
- 2
Wait to be approved
Every account is approved by hand, so it is not instant. Nothing opens until it is.
- 3
Choose how strict
From the top 1% of setups to every lean the model has. Each setting shows the accuracy it was tested at before you pick it.
It is free during the test phase. Not as an offer — nothing here has earned a price yet. If that changes, members will be told before it does, not after.
Questions
How accurate is it?
Tested at 58% at the setting published on 1-hour candles, over 432 calls the model had never seen. In September 2026 that figure was put through a control it had never faced, and it half failed. Against candles picked at random it holds up clearly — those score about 50%. Against simply buying on the same candles it does not: the model says up on 79% of 1-hour calls, about 53% of the candles it picked rose regardless, and what is left over cannot be told apart from luck. So 58% is real, and most of it is the market’s direction rather than the model’s skill. It has also never turned into profit: a trade needs 0.09% to cover fees, and in the audit the model’s own calls lost 0.15% each after fees — slightly worse than calls placed at random times, which lost 0.09%. On 15-minute candles, where the model leans neither long nor short, it passes the same control easily. The whole audit is on the updates page.
How do I know the losses aren't deleted?
Grading is automatic and happens when the candle closes, before anyone looks at it. The seven calls above are consecutive, with nothing taken out, and each row carries the price when the call was published and the price when its candle closed — enough to check any of them on your own chart.
Is this financial advice?
No. It is a statistical read of one market, produced automatically, and it can be wrong. It is not calibrated to your position or your risk. Only ever trade money you can afford to lose.
Which coins does it cover?
Bitcoin, Ethereum and BNB on 15-minute, 1-hour and 4-hour candles, plus Solana and XRP on 15-minute and 1-hour. Solana and XRP publish nothing on 4-hour candles. All fifteen combinations of coin and timeframe were re-measured in October 2026, and on 4-hour candles those two came back unresolved — over 210 and 150 calls, their accuracy could not be separated from how the candles they picked were moving anyway. Unresolved means not demonstrated, not ruled out.
Is it free?
Yes, during the test phase. Members will be told before that changes, not after. There is no paid tier to upgrade to today.
Is there a phone app?
Not yet — Android and iPhone are planned after the website. The site works on phones now.
Read the whole record first.
Every call ever published, graded and dated. The 35 studies behind them, including the 34 that failed. Decide afterwards.
Create a free account