The Basics — No Jargon
What does StockClaw actually do?
It looks at cryptocurrency prices at regular intervals and guesses: "Will the price go up or down?"
If it thinks it'll go up, it buys. If it thinks it'll go down, it sells. If it's not sure, it waits.
That's it. Everything else is details about how it guesses and how well it guesses.
Why there are no performance numbers here
This page used to open with a win rate, a profit factor and a max drawdown. They came from a backtest that was wrong, so they have been removed rather than restated with softer wording.
Two faults, both reproduced since. The simulation settled trades at a price the live system cannot reach; give it the one cycle of delay the live system is actually subject to and most of the apparent edge disappears. Underneath that, the model was trained on indicators that were misaligned by one bar against each other, so part of the answer was already inside the question. With that corrected, the direction call is not distinguishable from a coin flip.
The live account is still in verification and its results are not published yet. When there are numbers on this page again they will be live ones, not backtest ones.
The Settings, Explained Simply
Every 4 Hours, About the Next 24
These are two different numbers and this page used to run them together. The cadence is 4 hours: the system wakes on each 4-hour candle, looks at prices, and re-decides. The horizon is longer: the model was trained to call the direction six bars out, so each call is about roughly the next 24 hours: not 1 minute, not 1 month, and not "the next cycle".
It only calls a direction when that move clears 0.5% either way. Smaller than that was labelled HOLD in training, so a quiet market is something it is built to sit out rather than guess at.
Calls 4 hours apart, each looking 24 hours ahead, overlap: a new decision arrives while the previous one still has most of its window left to run.
Take Profit (TP) & Stop Loss (SL)
TP = "sell when I've made enough"
If TP is set at +1.5%, and you bought BTC at $67,000, the system automatically sells when it hits $68,005.
SL = "sell when I've lost enough"
If SL is set at -1.0%, the system sells when it drops to $66,330. This limits how much you can lose per trade.
Without TP/SL, the system waits for the next cycle. With TP/SL, it can lock in profits early or cut losses early.
Why AI Instead of a Human Trader?
Human Trader
- • Gets scared when price drops → sells too early
- • Gets greedy when price rises → holds too long
- • Needs sleep, misses opportunities
- • Makes 3 wins, then 2 losses → panics, changes strategy
- • Can watch 1-2 coins at once
StockClaw AI
- • No emotions — follows the math every time
- • Runs 24/7, never sleeps
- • Makes 3 wins, then 2 losses → keeps going, no panic
- • Watches 6 coins simultaneously
- • Re-checks all 6 on the same 4-hour schedule, whatever the news
The Catch
Nobody has shown this makes money yet — including us. The backtest that suggested it did was broken, and the live account that would settle the question is still running. Markets change even when a result is real. This is a tool, not a money printer.
Never invest money you can't afford to lose.
Three Tiers — One Sentence Each
All three run the same model. Three settings change between them, not one: how big a move it needs, how confident it has to be, and how much of your balance it commits per trade. That last one is the part people miss.
Safe — "Only when it's really sure"
Trades the least, stakes the least per trade.
Like a sniper: it needs the largest move and the highest confidence before it fires, and it puts the smallest slice of your balance behind the trade when it does.
Standard — "The middle setting"
Trades more often, stakes half again as much.
Takes the smaller moves Safe skips, and needs less confidence to act on them.
Aggressive — "Acts on the smallest moves"
Trades the most, and stakes 2.5x what Safe does on each one.
Reacts to small moves and needs the least confidence to act. More chances to be right, and more of your balance at risk every time it is wrong. This is not a sensitivity dial — it is a different amount of money.
All tiers run on the same 4-hour cycle. Auto-trading is still in live verification and is not on sale; when it opens it is deposit-based, with a 20% fee on realized profit. The thing you can buy today is .
So which one should I pick?
- Safe — fewest trades, and the smallest share of your balance on the table at once.
- Standard — the middle setting on all three parameters.
- Aggressive — most trades, and 2.5x Safe's stake per trade. Bigger swings in both directions.
Same model in all three. What changes is the size of move it waits for, the confidence it demands, and how much it commits — so the tiers differ in risk taken, not only in how often they act. We can't tell you which performs best: that is exactly what live verification is for.
Backtest vs. live
A backtest is the strategy replayed over old prices. It is easy to make one look good by accident: ours settled trades at prices that were never actually available, and no amount of care in the write-up would have caught that. Only live trading can.
That's why we run a public Live Verification first: real capital, real trades, every result logged. It is still running, so there is nothing to report yet — and we would rather say that than publish the backtest.