- Local
- Web
- Server
- 24/7
gm.
your bot already traded.
your bot doesn't sleep. neither does the market.
Scalp with evidence, not hunches.Record the market tape. Replay it. Optimize for the market condition you actually trade in. Run the bot dry first, then decide with the numbers in front of you.
Dry-run = simulated, nothing at stakeLive = real USDC on Papertrade (despite the name)
Read the docsNo trading is open yet. Access starts with the waitlist, in dry-run.
How it works
Record, replay, optimize, run dry. Then decide.
Every step leaves a trail you can inspect. Nothing is promoted to real money because a backtest looked good.
Record
The bot writes the market tape (best bid and offer, trades, order book depth) to your disk, hour by hour, while it runs. The manager records it on your own computer. Nothing uploads that tape, and it is never offered for download.
Replay
Feed that tape back through the same decision engine the live bot uses. The same inputs give the same decisions, so a replay is an experiment, not a guess.
Optimize, per market condition
Pick the hours that look like the conditions you care about, such as trending, ranging or busy, and search for settings for each one separately. A set that only works in one regime says so.
Run dry
Launch the bot with simulated fills on the live feed. It trades nothing, but it makes every decision a live bot would. Compare how it behaves against the replay.
Then decide
Going live is a separate, deliberate switch with its own confirmation, and it spends real USDC on Papertrade. The optimizer's reliability tier is on screen when you make that call.
The optimizer
An optimizer that admits when it's guessing.
Searching thousands of settings over a short tape will always find something that looks great. That is overfitting, and most tools hide it. This one prints a reliability tier before it prints anything else.
- Walk-forward
- The tape is cut into consecutive folds. For each fold, the settings are searched on everything before it and scored on the fold itself, which the search never saw.
- Holdout
- With enough tape (two days or more), the most recent quarter is set aside completely. It is scored only when you explicitly ask to reveal it.
- Baselines
- Your starting settings and a pile of random valid settings are scored the same way. A pick has to beat the dice, not just beat zero.
- Latency stress
- The pick is re-run with slower fills. If the edge disappears when the exchange is a second slower, you will see that.
Reliability tiers, by evidence behind the pick
- noiseunder 10 rounds
Treat it as a coin flip. The report says so in its first line.
- anecdotal10 to 100 rounds
A story, not a result. Enough to be curious, not enough to rely on.
- indicative101 to 300 rounds · out-of-sample not negative
Worth a long dry-run. Still no promise.
- usableover 300 rounds · positive out-of-sample · survives slower fills · beats baselines
The bar for taking a configuration seriously, and it is deliberately hard to reach.
Most strategies lose. This tool tells you when it's guessing.
Market conditions
See the market you're tuning for.
Settings that suit a trending afternoon rarely suit a dead night. The manager shows recorded tape hour by hour, so you choose the conditions first and optimize second.
- drift up
- drift down
- no tape recorded
- bar height = traded volume
- trending
- ranging
- high volatility
- quiet
- busy
- like now
- Instrument
- ETH
- Time frame
- 13h to 17h, selected
- Reliability
- anecdotal
- State
- running
- Go live
- needs confirmation
A bot card shows its mode and the evidence tier, not a promise.
Presets such as trending, ranging, busy and quiet pick hours by percentile over the tape you have recorded, so they adapt to your data.
Safety model
Built to fail safe.
Dry-run is the default
Every new bot starts with simulated fills. Going live is a separate switch per bot, refused unless you type the bot's exact name.
A wallet-wide loss budget
One budget applies across every bot on a wallet, so ten bots cannot quietly lose ten times the limit. When it is spent, every running bot is paused.
Shared funds, first come first serve
Each bot has its own margin and leverage. Several bots can trade one wallet: a bot opens only if its margin is free at that moment, otherwise it skips the entry and says "Not enough funds". If the balance cannot be read, nothing opens.
Keys never leave your machine
The bots run locally. No key, key path or token is ever sent to us, and the manager only listens on your own computer.
The cloud copy is settings, not secrets
Optional sync stores your bot settings, saved time frames, optimizer summaries and round history. Bots pulled from the cloud arrive stopped and dry.
FAQ
Straight answers.
Is this financial advice?
No. stradefi is software that records, replays and tests strategies. It does not tell you what to trade, it does not know your situation, and it makes no claim about future results. Trading leveraged products can lose more than you expect, quickly.
Is it affiliated with Papertrade?
No. stradefi is an independent project. It is not affiliated with, endorsed by, or sponsored by Papertrade or Hyperliquid, and the names belong to their owners.
Does it trade with real money?
Only if you switch a bot to live. Dry-run is simulated: live prices, fake fills, nothing at stake. Live sends real orders with real USDC. Despite its name, Papertrade is not paper money (see the next question), so a live stradefi bot can lose real funds. Every bot starts dry, and live is a separate switch you confirm by typing the bot's name. Nobody has access yet: it opens through the waitlist, dry-run first.
What is Papertrade?
Papertrade is a real on-chain perpetuals exchange on HyperEVM (BTC and ETH only). You deposit USDC, and winning trades are paid from a USDC liquidity pool, with a payout queue when the pool is thin. There is no funding rate, opening a position is free, and winners pay a win fee. It is a third-party venue with its own risks: smart-contract risk, pool and queue risk, and outages. Papertrade allows automated trading, so bots are fine there. stradefi is not affiliated with it.
When can I use it?
There is no date yet. Join the waitlist and you will get one email when access opens. We will not send marketing mail.
What do you do with my email?
It is used to tell you when access opens. We also keep the time you joined, which version of the site you used (public or pre-release), the page language you signed up from, the request's IP address (deleted after 90 days) and its approximate location (country, region, city, timezone), to understand who is interested. See the privacy note for exactly what is stored and how to have it deleted.
Waitlist
Join the waitlist. Dry-run opens first.
One email when access opens. No newsletter, no tracking.