Solana Build Night 4: Backtest Validation and Failure Testing

​Try to break your own backtest before the market does. Tonight is a student-led build night: a short talk, demos, a guided mission, then free build on your own project.

​We look at three ways a result can fool you. Leakage: when training data sits next to test data, the answer is already in the room, and an LLM agent can leak too if its model has seen the period. Luck: try enough strategies and one looks brilliant, then falls apart on fresh data. A broken feed: an oracle that stops updating can be caught in the data. Then we inject Solana failures into a simulated agent (RPC rate limits with HTTP 429, a stale oracle, an expired blockhash, a duplicate submission) and see what the risk check catches and what it misses.

​Mission: find the planted leak in the sample strategy, fix it with a purged split, and chart before and after. Then audit your own project, build a feed-health alert, or write a regression test for a failure you caused.

​Reads use public mainnet data; every write is devnet only; no real funds. The faults are simulated on top of real prices, and nothing attacks a live service. Results are simulated on recorded data and are not advice.

​Questions? Message @travisim on Telegram: https://t.me/travisim

​RSVP Information

​Capacity is limited. RSVPs are subject to approval and are strictly non-transferable. Please submit your RSVP to request access.

​By registering for this event, you acknowledge that the Solana University Ambassador Program and its partners will collect and use your personal information to manage your registration and attendance. Any dietary or accessibility requirements you choose to share will be used only to facilitate your attendance.

​🎁 Food is provided at this session. Claude credits and RPC credits are provided free upon request — just ask the hosts.

​💬 Join our Telegram group: https://t.me/+bdoXt3fyIiI1ZTBl