We split trading into two parts from the start.
strategy, filters, entry point.
margin, averaging, stops, risk limits.
One part decides «where to enter», the other — «how to open and exit».
3 background — for reference points, 2 for entry point. AI is only an executor — it doesn't decide or adjust anything on its own.
30% of deposit always in reserve. Entry 0.35%, averaging ×4 and ×2. Max 5 positions. Hard drawdown cap — 30%.
Short targets: 0.3–5% by timeframe. Trailing stop quickly moves into profit — position almost cannot close in the red.
Many think trading means knowing what will happen in X minutes. That's not it. We always work with reference points, confirmations, and filters.
AI doesn't decide or update anything on its own. It simplifies work with algorithms, runs checks, delivers data. The strategy is set by us, not the model.
No «forecasts». First we find the reference point, then we wait for confirmation, then we check filters. Only then — entry.
We catch the reaction, not the «big move». Often after our take-profit the price goes another +10% — we take our small guaranteed piece and exit.
Before the bot even considers entering, it already has a fresh «reference point» for each coin. These reference points are prepared by three parallel simulations.
In case of restart or new machine: runs our entire strategy from the past year on all coins — up to the current moment. The output is an up-to-date reference point for each coin. After that, the simulation has done its job.
Continuously runs the parallel strategy on all our coins and updates the reference points. The reference is built from: Fibonacci retracements, candles, Fibonacci pivot points, divergences, local maxima/minima, etc.
The key difference: for the entry itself we look at many filters, but references must be updated without them . Here — volume, RSI value, Ichimoku and Bollinger. This way we always have the most accurate and undistorted reference point.
The reference exists — now we need to understand that we're right now in the entry zone and find the specific level.
The first simulation answers the question: «are we currently in a zone where we can open positions at all?»
The second simulation looks for which exact level to enter on. Our entries are always at levels, never «at market».
From the background simulation, an up-to-date reference point for the coin already exists.
Simulation A confirms: divergence, Fibo zone, BB — we're in a possible entry area.
CMD, BB, volume — everything confirms the reaction.
totalCap + BTC dominance + USDT dominance — the overall market gives the green light.
Simulation B holds the specific level. Price touched it — position opened.
We have our own margin and free-margin calculation. From the start, 30% of the deposit is set aside and doesn't participate in calculations — we work with 70% of the deposit.
For each entry: 0.5% of working margin. Given that working margin is 70% of the deposit, the actual entry volume = 0.35% of the client's total deposit.
If the market gave us a better point — we add to the position. That's the strategy's advantage: we average by level, not from panic.
The client has two independent limiters — by number of positions and by total drawdown. Both are calculated automatically by the bot.
The bot won't take more than five open trades at once — even if the strategy gives a signal.
If all open positions combined went to minus 30% of the deposit — all are automatically closed . The client physically can't go beyond 30% drawdown.
When a position's stop-loss has already been moved into profit (see next section) — the bot knows this position will close in the positive zone one way or another . So such positions are considered «already completed» for the limit.
Example: 5 of 5 open. In 3 of them, SL is already in profit. The bot considers that 3 more can be opened — totaling temporarily up to 8 positions, because the 3 «green» ones won't stay long.
Since our targets are short, we use a stepped stop-loss on positive movement. It quickly moves into profit and follows the price.
The values shown are an illustrative example. Real parameters are configured and may differ.
| Price moved | +0.4% | +0.5% | +0.6% | +0.7% | +0.8% | +0.9% |
| SL moves to | +0.2% | +0.3% | +0.4% | +0.5% | +0.6% | +0.7% |
Take-profit isn't «one-size-fits-all». It depends on the timeframe the position is opened on.
The analysis at the same time runs on even more timeframes — this gives multi-level confirmation.
On the shortest TFs.
On hourly and around.
On daily and weekly positions.
Five simulations that check each other. Until reference → zone → filters → macro → level converge, the bot doesn't enter.
Strict risk model: 30% of deposit always in reserve, max 3% of deposit per trade, max 5 positions total, total drawdown cut at 30%. Take-profits short and guaranteed, stop-loss quickly moves into profit — the position almost cannot close in the red.
This document describes only the general architecture and operating logic of the trading bot, not all strategies and details.