Most "AI trading bot" pages describe one clever model that picks winners. The reality of running an automated trading system is less glamorous and more useful to understand: the strategy is the small part. Most of the work is the control system around it, meaning the limits it cannot exceed, the evidence it must produce before it earns more trust, and the checks that catch it when it drifts.
This post describes those layers in general terms and where the honest limits are today. It does not publish strategy details, parameters or scores.
A strategy is not a system
A signal generator answers one question: is there a setup worth taking? A trading system has to answer several more:
- How much can be risked on one idea?
- How much can be open at once?
- What happens when data is stale or the exchange misbehaves?
- Who, or what, decides that a strategy has earned real capital?
- How do we notice, quickly, that something has stopped working?
Each of those needs its own control. Putting them all inside the strategy is how accounts get blown up, because the strategy is exactly the part that is most likely to be wrong.
Layer 1: risk limits enforced in code
Every position has a stop set at entry, and the size of any single trade is bounded relative to the account. Total exposure is bounded too. These are not guidelines: they are checks in the execution path, and a strategy that wants to break them is simply refused. The exact limits are not published.
The system also does not average down and does not use martingale sizing. Both turn a losing idea into a larger loss.
Layer 2: shadow validation before live capital
New strategies begin in shadow validation. They produce signals against live market data with no capital at risk, and their results are recorded as if they had been traded. Only an automated engine, not a person and not a model, can decide that a strategy has earned live capital, and it decides from accumulated evidence. A strategy that stops earning its place can be moved back.
See shadow mode for why this matters.
Layer 3: reconciliation and freshness checks
A bot's internal picture of its positions can drift from the exchange's: orders partially fill, connections drop, prices slip. The system compares its records against the exchange and stops new activity when they disagree. Stale market data is treated the same way: better no signal than a confident wrong one.
Layer 4: honest public reporting
The results page shows which result cohorts are published and why the others are not yet; losses count when they are. Every published figure carries its sample size, window and timestamp. When a number cannot be verified, the page says "unavailable" instead of showing an old or flattering one. The current operating mode is on the status page. Account balances, position sizes and symbol-level detail are never published.
Where the limits are
- Controls reduce risk; they do not remove it. A stop can be gapped through.
- Shadow results are not live results. Fills, slippage and timing differ.
- A small sample proves little. Evidence has to accumulate across regimes.
- Nothing here is a return forecast. No signal service can promise one.
How to use this
SuperKamouBot delivers signals; it does not trade your account. You subscribe, read each signal's entry, stop and +1R target, and place every order yourself on your own exchange account. It never takes custody of funds and never asks for your exchange credentials.
Pro is $59/mo and is the only plan on sale. See the pricing page, the how-it-works page and the results page, then judge for yourself.