Two bots, opposite relationships with inventory
Passivbot and Vesper both post limit orders on perpetual futures and both avoid predicting direction. Underneath they disagree about the one thing that matters: what a position on your books is for. That disagreement decides which of them suits your account, and it is not a matter of taste.
Passivbotself-hosted grid, public domain
Vesperhosted desk you connectWhat Passivbot is, in its own words
Passivbot describes itself on its repository as a bot that automatically creates and cancels limit buy and sell orders without trying to predict price movements, working as “a contrarian market maker, using price bands, EMA-derived context, and risk controls.” Its strategy is described as inspired by Martingale betting: small initial entries, then doubling down on losing positions in a grid. It is written in Python and Rust, released under the Unlicense (public domain), self-hosted, and lists Bybit, OKX, Bitget, Bitunix, GateIO, Binance, Kucoin, Hyperliquid and WEEX. The authors mark it “used at one's own risk.” (Checked on github.com/enarjord/passivbot, 5 September 2026.)
Vesper posts limit orders too, and also refuses to predict direction. The difference is what happens after a fill: we requote the other side and work the net position back toward flat, measured in seconds, because for us inventory is a liability we are paid to carry briefly.
- Both post limit orders and neither forecasts direction.
- Passivbot: inventory is the position, held and averaged into.
- Vesper: inventory is a by-product, cleared as fast as the book allows.
The disagreement that decides everything
Averaging into a losing position is a real strategy with a real edge: price spends most of its life going nowhere, and a bot that leans against small moves gets paid for that. The known cost is the shape of the loss: rare, large, and arriving exactly when the market stops chopping. Passivbot's own documentation is explicit about the risk and about the leverage involved.
Quoting both sides and flattening has the opposite shape: small, frequent, boring losses from adverse selection, and no single event that takes the account. We measured our own: 1% of cycles produced three quarters of the loss, and once we cut those the cost per million of turnover fell from 177 to 43.
Neither shape is better in the abstract. They are different bets about which failure you can live with, and you should pick the one you can hold through without intervening.
- Grid and averaging: many small wins, rare deep drawdown.
- Quoting and flattening: constant small bleed, no single fatal event.
- Pick the failure shape you can sit through, not the higher headline return.
Where Passivbot is the better answer
It is public domain. Not merely open source: it is released under the Unlicense, which is as close to “take it, it is yours” as software gets. You can read every line, change anything, and keep running it forever with no one's permission. No hosted product matches that.
It runs on nine exchanges including the largest centralised venues, which is a different order of coverage from our short list. It ships a backtester and an optimizer that iterates thousands of configurations. If you want to tune a strategy yourself against history, that machinery is already built and we do not offer it.
And it costs nothing but a server and your time.
- Public domain: read it, change it, keep it.
- Nine exchanges including the big centralised ones.
- Built-in backtester and optimizer for tuning your own configuration.
Where we are different
We quote a short list of perpetual venues and each one was measured before it was added. We record their order books continuously (over four million snapshots at a 5.5-second cadence and still counting), and the calibration comes from that recording rather than from a configuration you tune yourself.
That is also the honest limitation: you cannot tune it, read it, or take it with you. If having the strategy in your own hands matters, the public-domain option is right and we are not.
- Few venues, each measured before it was offered.
- Calibration from our own book recording, not from your parameter search.
- Closed. You cannot read or modify it, which is a real reason to choose otherwise.
| Passivbot | Vesper | |
|---|---|---|
| Stance on inventory | Held on purpose, averaged into | Cleared toward flat, measured in seconds |
| Strategy family | Contrarian grid, Martingale-inspired | Two-sided quoting, requote on fill |
| Failure shape | Rare and deep, when chop ends | Frequent and small, from adverse selection |
| Licence | Unlicense (public domain) | Closed |
| Where it runs | Self-hosted: clone, Python 3.12+, Rust | Hosted; your funds stay on the exchange |
| Exchanges | 9, including major centralised venues | A short list of perpetual venues, each measured first |
| Tuning | Backtester and optimizer you drive | Calibration derived from our order-book recording |
| Best for | Someone who wants to own and tune the strategy | Someone who wants a measured venue and no tuning |
How to choose without trying both
Ask which loss you would rather explain to yourself. If a rare drawdown that arrives all at once is acceptable because the quiet months paid for it, the grid family fits you and Passivbot is the mature, free, self-hosted way to run it.
If you would rather bleed a little continuously and never face a single event that ends the account, you want two-sided quoting, and then the question becomes whether the book you chose can pay for it at all. That is decided by the tick, not by the software.
- Rare deep loss acceptable, want to own the code → Passivbot.
- Prefer constant small bleed and no single fatal event → two-sided quoting.
- Either way: compute the tick value in basis points on your market first.
Pick the failure you can sit through
- A rare, deep drawdown is acceptable and you want to own the code: Passivbot. Mature, free, public domain, optimizer included.
- You would rather bleed a little every day and never face one fatal event: two-sided quoting, which is what Vesper does.
- Neither software changes the arithmetic: compute the tick value in basis points on your market before choosing either.
Questions
Is Passivbot a market maker?
Its own documentation calls it “a contrarian market maker”, and it does post two-sided limit orders. The difference from a classical market maker is what it does after a fill: it holds and averages into the position rather than working it back to flat.
Which one is safer?
Neither is safe; they fail differently. Averaging into losers produces rare, deep drawdowns. Two-sided quoting produces continuous small losses from adverse selection. The right question is which failure you can hold through without intervening.
Is Passivbot free?
Its repository states the Unlicense, which places it in the public domain. What you pay for is a server and your own time to run, tune and maintain it.
Can I run both?
On separate accounts, yes: they are different strategies. On the same account they would fight: one is trying to accumulate the position the other is trying to flatten.
What should I check before either one?
The value of one tick on your market in basis points: tick size ÷ price × 10,000. Below roughly 0.1 bp queue priority is cheap; above roughly 1 bp it eats most of the spread. No bot changes that arithmetic.