Test a strategy on this market before it trades.
Axiom Finance turns your trading idea into a bot, tests it against years of history and on data it has never seen, then lets it practice on live prices like these with no real money. When the evidence holds, switch it on for real — with risk limits built in.
Live market feed
BTC/USDT
Waiting for a bot stream
Waiting
—
—
Funding and spreads across ~45 exchanges. Net.
Funding rates and price gaps across every venue we connect, priced after taker fees and the depth at the best price. The panel is the real screen, delayed. A free account opens the full terminal; live data, alerts and execution come with paid plans.
- Net means after taker fees on both legs and the depth at the best price.
- Your own API keys, trade-only. Keys that can withdraw are rejected at connect.
- P2P USDT routes in BRL and ARS: buy on one exchange's P2P desk, sell on another's, across seven exchanges.
- Delayed data stays free. No trial clock.
Funding, annualized
Loading the delayed screen…
Every venue we screen, and what it charges.
Listings, fees and feed health from the same snapshot as the screen above. These taker fees are the ones subtracted from every net figure.
Keys scoped to trading only — your funds stay in your exchange account.
The exchange list loads with the screen above.
Most ideas die here. That's the point.
Every strategy walks the same path, and each gate is allowed to end it. Nothing reaches live trading because you believed in it — it reaches live because it survived.
Illustrative diagram of the pipeline, not a performance result.
- 01
Backtest
Replay the rules over history with the fees and price slip you would actually have paid. Cheap, fast, and the easiest stage in the pipeline to fool yourself with.
What this stage checks
- Tested across calm, choppy, and trending markets
- Real costs included, not assumed away
- Worst losing stretch shown next to the gains
- 02
Monte Carlo
Run the same trades in thousands of different orders. If the results only hold up in one particular sequence, what you found was luck wearing a strategy's clothes.
What this stage checks
- A range of outcomes, not just one lucky run
- Shows how much the order of trades matters
- Your worst-case loss, surfaced up front
- 03
Walk-forward
Train it on one stretch of time, test it on the stretch that comes right after, move forward, repeat. This gate fails more ideas than the other four combined, and it is the only one that produces evidence live trading will accept.
What this stage checks
- Trained and tested on separate data
- Tested many times over, not just once
- You can see if performance fades over time
- 04
Paper
Trade it against real, live prices, in real time, with nothing at stake. Delays, partial fills, an exchange behaving differently than the data implied — these problems show up here or they show up with your money.
What this stage checks
- Orders placed against live prices
- Real trading costs included
- Every decision written to a log
- 05
Live
Real orders, on your exchange account, with your keys. Allocation starts small and the safety controls are armed before the first order goes out, not after the first bad day.
What this stage checks
- Starts small, scales up gradually
- Stop-loss and max-loss limits armed
- Continuous monitoring and an emergency stop
Schematic drawings of what each stage produces. Not results from a real run.
Then a written verdict
Every run ends in APPROVED, REJECTED, or NEEDS MORE DATA — stored against the exact settings that produced it. Change the settings and the evidence no longer counts.
3.5 million backtests. Most failed. Here is the ledger.
Our own research, failures included. Each row says what we tested, what came out, on how much data, and when it stops working.
| Tested | Result | Sample | Fails when |
|---|---|---|---|
| Everything we tried, against holding the coin | Nothing beat holding in a strong bull market | 3.5M backtests · 5 study rounds | Not a strategy: it is the baseline every row is measured against |
| Indicator signals as entries (RSI, MACD, moving averages, Ichimoku) | 0 of 400 passed | Final out-of-sample test | Always. We offer them as filters, never as signals. |
| Drawdown Shield (cash below a slow average, sized by volatility) | Max drawdown 74% → 46% | 18 of 18 coins · 86 of 90 periods | Strong rallies (gives back part of the gain) and sideways markets |
| Mean reversion with a volatility filter | Survived | 28 of 35 altcoins | Strong uptrends |
| Funding fade (against the crowded side, BTC and ETH) | Not refuted; re-audit pending | Paper only until the re-audit | Long trends with no reversal |
Past results from our own research; out-of-sample where stated. Not a forecast.
Run the same tests on your strategyResearch, not promises.
The tools you would expect from a research desk, in one place: build the strategy, sweep its parameters, read the paper log, and get a written verdict before anything reaches an exchange.
- Visual and code-level strategy building
- Reproducible backtests you can re-run
- Test many setting combinations at once, with the ranges recorded
- A full set of risk and performance checks, not just the profit number
- Paper and live execution from one place
- Auditable logs of every decision taken
| Parameter | Value | From | To | Step |
|---|---|---|---|---|
| RSI period | 14 | 7 | 28 | 1 |
| Moving average | 30 | 10 | 60 | 5 |
| Stop (ATR) | 2.0 | 1.0 | 3.0 | 0.5 |
| Take profit | 3.0 | 1.0 | 5.0 | 0.5 |
| Date / time | Pair | Side | Price | Result |
|---|---|---|---|---|
| 2026-05-12 09:41 | BTC/USDT | BUY | 64 812.40 | +0.62% |
| 2026-05-12 11:04 | BTC/USDT | SELL | 65 214.05 | +0.31% |
| 2026-05-13 08:22 | ETH/USDT | BUY | 3 118.72 | -0.18% |
| 2026-05-13 14:57 | ETH/USDT | SELL | 3 090.11 | -0.24% |
| 2026-05-14 10:35 | SOL/USDT | BUY | 171.88 | +1.04% |
- Robustness
- Solid
- Stability
- Consistent
- Execution
- Controlled
Illustration of the product interface. Figures shown are examples, not results.
The same live feed your bots trade on.
This is the same market data your bots would trade on, streamed straight to the page. Check the 24h move before you point a strategy at it.
Connecting to live market feed
Loading 24-hour prices. If the feed is slow, the panel retries on its own.
What the platform already ships
Workflows you can open in the product today. Not a roadmap, not a promise.
Validation stack
The part most bot platforms skip. Every strategy goes through the same battery before it can touch real money.
Exchange coverage
Trade on the venues you already use. Credentials stay encrypted and scoped to the permissions you grant.
Pair flexibility
Point a bot at whatever your exchange account supports, spot or perpetual, one pair or a basket.
Strategy toolkit
More than one generic trend bot — and each preset says plainly whether it cleared validation or not.
Operational controls
The guardrails that matter once real orders start going out, not just before.
Your keys. Your capital. Our discipline.
Two things decide whether a bad day stays a bad day: who can move your money, and what cuts your exposure when you are not watching.
Your keys never leave your control
You connect exchange API keys scoped to trading only. Funds stay in your exchange account — we never hold them, and a key with withdrawal permission is refused at connect time rather than quietly accepted.
- Credentials encrypted with AES-256 at rest
- Passwords hashed with Argon2id
- Withdrawal-scoped keys rejected on connect
- Per-tenant keys, rotated without downtime
Risk gets the last word
Limits on position size, total exposure, and how much you can lose in a stretch all run independently of the strategy. When a limit is crossed, exposure is cut whether or not the strategy agrees.
Schematic comparison, not a measured account.
Open the screen. Test before you trade.
A free account opens the arbitrage terminal on delayed data, and lets you backtest, walk-forward and paper-trade a bot, all before a cent is at risk.
Free account, no card. Delayed data and paper trading stay free.
Questions before you connect a key?
Write to the team. A person answers, not a bot.