StratiumLab note
How Stratium’s agent timing works (lab note)
This lab note documents Stratium’s product timing constants — the same numbers wired into the desk for BTC, ETH, SOL, XRP. It is not a live performance track record. Use it to understand pacing before you watch paper fills or run a backtest.
Why does agent timing matter for AI trading demos?
Retail AI trading products often hide whether fills are throttled, batched, or instant. Stratium exposes timing so you can read a setup before a marker appears, and so backtest replay stays human-readable instead of flashing hundreds of bars per second. Without published pacing, a demo can look “smarter” simply because it fires trades faster than a human can audit.
Market data still comes from Coinbase Exchange public APIs (Exchange API docs). Timing constants govern when Stratium’s agent is allowed to act on that tape — they do not invent the mid price.
What are Stratium’s live-desk agent timings?
On the live paper desk, Stratium evaluates short-horizon momentum on Coinbase Exchange prices for BTC-USD, ETH-USD, SOL-USD, XRP-USD, then applies product-level pacing before a fill can land. The constants below are first-party product specs, not a third-party benchmark study.
| Signal | Value | Role |
|---|---|---|
| Trade cooldown | ~12,000 ms | Minimum spacing between fills so the tape can move and narratives stay readable |
| Decide → execute | ~1,600 ms | Delay after a setup locks so the UI can show intent before the fill marker |
| Think cadence | ~2,200 ms | Status updates continue even when price is quiet |
| Momentum windows | ~45 s and ~5 min | Short-horizon percent/dollar moves used for confidence |
| Live unlock | 20 USDC | Ledger minimum before Start live trading is available |
How fast is Stratium backtest replay?
Backtest mode loads historical Coinbase candles, runs the agent across the series, then advances a cursor so chart, fills, and stats update together. Replay milliseconds are a UI pacing layer — they are not the same as wall-clock live cooldown.
| Setting | Value | Notes |
|---|---|---|
| Base pace (1x) | ~140 ms / candle | Slow enough to watch price evolve |
| Fill dwell | ~720 ms | Extra pause when a new fill appears on the current bar |
| Speed presets | 1x / 2x / 4x | Transport controls on the backtest desk |
| Lookbacks | 1d / 3d / 1w / 1m | Pair with candle sizes such as 15m–4h |
How should you read these numbers as a lab note?
Treat this page like a product specification sheet. The values are intentional UX and risk-of-spam controls: a ~12 s cooldown prevents fill spam on quiet tapes; the ~1.6 s decide-then-execute window exists so narratives can render before the marker; backtest dwell (~720 ms) exists so humans can read a fill before the cursor moves on.
These constants are not evidence of alpha. A slower agent is not “safer,” and a faster replay is not “more accurate.” Accuracy of fills still depends on the decision logic and on whether you are in paper, backtest, or live USDC mode — see /guides/backtest-vs-paper-vs-live.
How can you verify these timings yourself?
- Open the live desk (/) and watch paper fills — when the agent is active, note the gap between consecutive entries. Gaps near ~12 seconds are consistent with the cooldown spec (quiet markets may wait longer).
- Compare Stratium’s printed mid to the same product on Coinbase Exchange within a few seconds — timing docs do not replace venue verification.
- Switch to backtest (or open /?mode=backtest), run a window, and use 1x replay to feel the ~140 ms bar advance and ~720 ms fill pauses.
- Check funding gates: live trading remains locked until the ledger reaches 20 USDC.
What should you not conclude from these timings?
These figures describe Stratium’s product behavior on Coinbase USD pairs (BTC-USD, ETH-USD, SOL-USD, XRP-USD). They are not proof of edge, not audited live P&L, and not a promise that live USDC trading will match paper or backtest paths. Crypto trading can result in loss of capital.
For a diligence checklist that covers venue naming, mode labels, and risk limits — not just pacing — read /guides/evaluate-ai-trading-bots. For how backtest, paper, and live differ on capital at risk, read /guides/backtest-vs-paper-vs-live.
FAQ
Frequently asked questions
Where do Stratium’s timing numbers come from?
They are first-party product constants used by the Stratium desk (cooldown, decide-then-execute delay, think cadence, and backtest replay pacing). They are documented here so visitors can cite them without reverse-engineering the UI.
Does the ~12 second cooldown apply in backtest?
Backtest applies the agent’s decision spirit on historical bars, then replays results with UI pacing (~140 ms/bar, ~720 ms fill dwell). Do not equate replay milliseconds with wall-clock live cooldown.
Is this a performance track record?
No. This lab note documents timing and UX pacing only. It does not publish win rates or returns as investment claims.
Where does Stratium get the prices these timings act on?
Prices come from Coinbase Exchange public REST and WebSocket APIs for BTC-USD, ETH-USD, SOL-USD, XRP-USD. Stratium’s timing layer paces decisions on that tape; it does not replace the venue mid.
How is live unlock related to agent timing?
Live trading unlocks at 20 USDC on the ledger. Timing constants still describe how the agent paces decisions; unlock only gates whether fills size from your deposited USDC.
Continue on Stratium
Use the desk to verify behavior on Coinbase prices — then fund only if it earns your trust.