Education note. GOGA Academy explains how the GOGA research and risk layer works so you can read its results critically — not as a promise. This article is the short version of the method: which data is used, how a backtest is run, how the readiness gate decides, and what the whole system does not claim.
1. Data: real published signals, not simulated numbers
Research starts from signals that were published in real time, not from synthetic examples. Each recorded signal keeps the symbol, the type (pump or dump alert), and the timestamp at which it was published. That history is paired with real exchange candles for the same symbols, so every claim can be re-checked against the raw record.
The archive is large and keeps growing. Where a number matters, it is published with its observation window and source, so it can be traced back to the data instead of being asserted from memory.
2. Backtest: bar-by-bar, no look-ahead
A strategy is simulated on historical candles the way it would have actually traded:
- Entry happens on the open of the candle after the signal — nothing that was not known at that moment is used.
- Stop-loss and take-profit are evaluated on every bar, not only at the close.
- コスト (fees and slippage) are modeled so results are not flattered by free-trade fantasy.
The point is not a pretty equity curve. It is to see how often a rule survives realistic conditions before anyone risks money on it.
3. The gate: deterministic, not an opinion
Readiness is decided by a fixed set of metric requirements — not by a mood, and not by a language model. To be marked for live, a strategy must clear a strict bar on a meaningful sample:
| 評価 | What it takes |
|---|---|
| LIVE_CANDIDATE | Clears the strict bar: a meaningful number of trades, a positive net result, controlled drawdown and low risk of ruin |
| PAPER_ONLY | Clears the paper bar but not the stricter live bar — paper first |
| NOT_READY | Fails even the paper bar: insufficient data, excessive risk, or no positive result on history |
Each strategy also gets a blended 0–100 readiness score from the same metrics so results can be compared — but the verdict is decided by the metric requirements above, not by score bands.
Two properties matter:
- It is deterministic. A rules-based formula decides readiness. Our AI analyst may explain a result, but it is never allowed to override the gate.
- It caps leverage. Leverage above 5× is blocked for live by a fixed rule — high leverage is not refused by taste but by arithmetic, and the strongest backtest cannot wave a high-leverage strategy straight to live.
4. What we do NOT promise
- 利益保証なし。 Historical performance, including any backtest shown here, is not a guarantee of future results.
- No custody. GOGA is non-custodial: you keep your exchange account and keys. We request trade-only permissions and never ask for withdraw scope.
- No investment advice. We publish analytics and education. The decision to trade is always yours.
- Paper ≠ live. Paper removes emotion and real-world slippage. A paper result is a starting point, not a prediction.
When this method does NOT work
Honesty about limits is part of the method:
- On short windows and after a market regime change, historical edges decay fast.
- On illiquid symbols, the modeled fill price may not be reachable in reality.
- A strategy tuned to death (overfit) can pass on history and fail forward.
- Leverage above the cap is never survivable in this system — by design.
5. Where results live
Methodology, curriculum and honest risk explanations live in the GOGA Academy; the paper floor lets you test strategies without live risk; the public live ledger shows real executed history with recomputable data.