This is a development ledger, not a product. Every trade on this page is the engine on paper accounts. On Aug 29 we stood the engine down: Sniper Gen 1 did not pass our standard. Our own backtesting promised results the live tape never delivered - we found four flaws in the simulator that had inflated every study, so the simulation numbers that used to sit on this page are gone, and they are not coming back until a number can be reproduced trade-for-trade on real bars. The trade record below stays exactly as it happened. We stepped back to a more advanced approach - built on the desk's market-structure method, tested day by day on the exchange's own bar replay, with a human correcting the machine - and it will earn its place here the same way: live, wins and losses alike. Stay tuned for the NEW SNIPER - codename EL SAFA7.
The engine's first campaign, on paper accounts: every signal published to this page in real time, as it fired. Wins, losses, the brutal days - nothing edited, nothing reset, nothing hidden. Only the machine's trades are counted, ever. The chapter is sealed and its record stands below, forever.
📕 Record finalThe second engine came out of the lab with a perfect report card. Its forward campaign on a fresh paper account lasted four days: the lab had graded its own exam wrong, and the tape said so quickly. We stopped it ourselves. Every one of its trades stands in the record below.
📕 Retired, 08-29Why this exists. When something goes wrong on this desk - a connection drops, a rule fails, something outside the engine touches its account - it is written here with the number, the cause, and the fix. Losses are never removed from the record; they are explained. Entries marked ⚠ in the full log jump here.
Most trading bots are sold on a backtest.
Ours failed its own - so we threw the backtest out, not the standard.
Until Aug 27 this section listed millions of simulated bars and thousands of simulated years. We found four flaws in that simulator - the last one credited profit targets the market had touched before the entry filled - and every study built on it was inflated. Those numbers are gone. Here is how the next engine is tested instead:
So Generation 2 is built machine-first - the rules live in the engine, and the engine refuses what the rules refuse. Nobody can talk it out of them.
The engine's account belongs to the engine. Any fill the machine didn't place is flattened immediately - no warnings, no co-existing. One trader per account: the machine.
Configuration changes are accepted pre-market only - mid-session the engine refuses its own config file, no matter who edits it. Frozen means frozen.
Fresh breaks are entered only after the bar closes beyond the level - the break must hold. Setups with stops too tight to trust are refused outright.
The safety laws above were right and stay. What goes is the way the strategy was proven: no more full-history simulations as evidence - the next engine is graded on the exchange's own bar replay, day by day, broker fills only.
Why this page exists. Every number here is generated automatically from the desk's trade database - the same engine that publishes each signal's full lifecycle to our private feed in real time. Losses are counted, displayed, and kept. No cherry-picking, no resets of the record, no screenshots-only-on-green. Chapter One ended and its record - the good days and the brutal ones - stays on this page forever. That's the deal.