Every tipster Discord can show you a 200% backtest. Most of them are curve-fit — tuned in hindsight to look great on past data and fall apart on tomorrow's. Here are the three flags you can spot without any technical background.
Flag 1 — No out-of-sample window
A legitimate backtest splits the data: tune the model on years 1–4, test it untouched on year 5. If the "results" only show one ROI number without "tested on 20XX–20YY data the model didn't see," assume curve-fit.
Flag 2 — Suspiciously narrow strategy definition
"Bet over 2.5 in matches where the away team has won 3 of the last 5 and the home team played a Tuesday game." That's not a strategy — that describes 11 historical matches. With enough parameters you can fit a bumblebee to your fridge magnet. Real strategies are defined before the data, in plain terms.
Flag 3 — No drawdown reported
Every real betting strategy has a worst losing streak. If the marketing shows you the equity curve without the maximum drawdown — what was the worst -10u stretch, when did it happen — they're hiding the period where you'd have unsubscribed.
How GSS handles it
Every Verified System publishes its 365-day out-of-sample ROI, its worst rolling-30-day drawdown, and the full bet ledger. The threshold (probability cutoff) is fixed before the test period — no in-hindsight tuning. If a system later degrades, the badge expires; no quiet rebrand.