Transparent
Intelligence.
Kvantix computes one score per coin from live market microstructure — orderbook imbalance, funding, momentum. Each score is written to a hash-chained ledger that cannot be rewritten. Outcomes are measured against public Binance prices you can recompute yourself. No hidden win-rates. No claim you can’t check — including the ones we’ve had to retract.
An audit trail, not a promise
The signature of Kvantix: a tamper-evident ledger of every prediction.
Live signals
All eight coins scored and logged in the open. We no longer split them into “product” and “control” — the data did not support that division, and we say so below.
Market wire
Latest headlines from the market. Source-attributed, uninterpreted — the news is context, not advice.
What we measure — and what we don’t
Honesty is the product. We measure one thing and claim nothing more.
We measure predictiveness
For every signal we record the coin’s actual return 4, 12 and 24 hours later, and we split the results by market regime — trending up, trending down, ranging. A signal has to work in all of them. The 24-hour horizon failed that test and has been closed. The question we answer in public: does a higher KAS precede higher returns? Nothing else.
Claims are locked before they’re tested
When we find a pattern, we write down exactly what we expect — which assets, which thresholds — and hash-lock it with a timestamp. It is then measured only against data that did not exist when the claim was made. We can’t quietly move the goalposts. One such test is open right now and resolves in mid-September.
An edge has to survive fees
A 0.04% gap per trade is not an edge if a round trip costs 0.06%. We measure net of realistic trading costs and we publish the answer when it’s negative — which, so far, it usually is. We also don’t simulate a strategy with entry, stop and take-profit: that’s exactly where track records get flattered.
History can’t be rewritten
Each prediction is hash-chained to the last. A retroactive edit breaks the chain visibly. Outcomes live in public price data — you don’t have to trust us, you can check.
The record so far
Built from live operation. We show the numbers as they are — including the run where our own engine was wrong.
The number that looked best was the most misleading
For a period, our v1 engine’s 24-hour results looked outstanding: a correlation of −0.39, with the lowest-conviction bucket finishing positive 72% of the time. Read at face value, that is a powerful contrarian signal, and it would have been easy to build a product on it.
It wasn’t a signal. It was a short bias that happened to coincide with a falling market. We found it by splitting the results into up-days and down-days: the gap was positive on down-days and negative on up-days. A signal that only works while the market falls isn’t an edge — it’s a directional bet in disguise, and it breaks the moment conditions change.
That test is now standard in everything we publish. It is also why the 24-hour horizon no longer exists in our engine.
This is a live test, and right now it is a failed one. The numbers above come from our first engine (v1), computed in real time from the tamper-evident log below. Across more than 10,000 observations spanning up-trends, down-trends and ranging markets, v1’s high-conviction bucket underperformed its low-conviction bucket in every single regime. It doesn’t rank the market weakly — it ranks it the wrong way round. Two claims previously made on this page do not survive the data and are retracted: that the engine works on liquid majors (BTC is among the weakest), and that the 24-hour horizon was strongest (it was the first one we closed). A rebuilt engine now runs in shadow — logged, measured, and published to no one, including us, until it passes a test it could have failed. Paid access stays closed. That is the whole point of doing this in the open.
Access
Tiers are set. Nothing is for sale until a signal passes a test it could have failed. One such test is currently open and resolves in mid-September.
Observer
- Delayed signals
- Full, verifiable track record
- Independent hash verification
Pro
- Real-time signals · all eight coins
- Telegram push
- API access with per-user key
Quant
- Full API + raw sensor data
- Orderbook metrics · liquidation feed (beta)
- Webhooks to your own executor