Levandor

slv1 Strategy Candidates — Post-Copy-Trading Decision Space

12 min read

slv1 Strategy Candidates — Post-Copy-Trading Decision Space

After slv1’s naive copy-trading returned a NO-GO (solv1-alpha-first-verdict-no-go-2026-07-04, 2026-07-04), this note opened as the menu of candidate strategies to reuse the assets already in hand. It has since become the scorecard of the full arc — 5 strategies explored, all resolved: four NO-GO and one validated only as a risk overlay. Each candidate below still names the assets/substrate it reuses and which of copy-trading’s killers it dodges, and now carries its verdict.

For Agents — scorecard hub, all five verdicts are in

The naive whole-book copy-trading thesis is dead, and so are the follow-on candidates. Do not restart any of them. Read the scorecard below before proposing anything new.

  • Resolved NO-GO ×4: copy-trading, launch-sniper, co-buy consensus, and (leaning) Solana arbitrage.
  • Validated as a RISK OVERLAY, not alpha: meme-beta timing — the +109.5% basket beta is real but bankable only as a risk-managed exposure overlay, not a stand-alone edge.
  • Still parked (unexplored): #4 Jito back-run MEV.
  • The reusable substrate remains: the offline backtest harness solv1-alpha, the shelved shred consumer solv1-rig, the GMGN browser-harvest path GMGN-ingester, and the passive arb-measure spread harness.

Scorecard — 5 strategies explored

#StrategyVerdictWhy
1Copy-trading (naive whole-book)NO-GOCohort +23.0% vs a random meme basket +109.5%; edge lived only in position-holds while scalps bled to entry decay; a mirror lands slot N+1 so latency can’t capture the leader’s entry.
2Data-filtered launch sniperNO-GOMost bot-saturated + rug-heavy niche; ShredStream speed is table-stakes not edge, and no quality filter cleared the adverse selection at a worthwhile margin.
3Co-buy consensus signalNO-GOK-of-N consensus dodged single-wallet survivorship but still couldn’t beat the basket beta; consensus events too sparse / late to pay.
4Meme-beta timing (risk-on/off)RISK OVERLAY, not alphaThe +109.5% basket beta is real and dominant — but capturable only as a risk-managed exposure overlay (sizing broad meme exposure up/down), not a stand-alone token-picking edge.
5Solana arbitrage (atomic DEX-DEX)leaning NO-GO (measurement confirming)First PRINCIPAL-SAFE strategy (atomic bundle reverts → lose fees not inventory) so it earned a look — but mainstream arb is a solved colocated machine, and the long-tail solo lane’s spreads persist because slippage eats them (realizable ~$0.33 median at trivial size). See solv1-arbitrage-feasibility-measurement-2026-07-05.

The meta-pattern across all five

Every “be fast / follow informed flow” edge turned out to be commoditized or structurally behind. The banked Jito ShredStream — the asset the whole project was built around — proved to be table-stakes in copy-trading, in the sniper, AND in arbitrage (three strategies in a row): necessary to not lose, never sufficient to win. The only openings the research found were thin niches where speed doesn’t matter and margins are near zero. The two things that ever produced value were orthogonal to speed:

  1. SELECTION / filtering — the adverse-selection filters added +14.8pp (filtered − raw) in the copy study. Choosing what to act on beat acting faster.
  2. BETA — a random 50-meme basket returned +109.5% and crushed every stock-picking cohort — but it is bankable only as a risk-managed overlay (strategy #4), not as alpha.

Net for slv1: latency is a cost of entry, not a source of edge; the durable levers are selection and risk-managed beta.

The pivot — why re-optimize around assets in hand

The first real verdict killed naive copy-trading on three counts: the cohort netted +23.0% (a survivorship ceiling, leaderboard-selected) while a random 50-meme basket did +109.5%, and the only surviving edge lived entirely in position/long-holds (+296.1%) — scalps bled −20.5% to entry decay. Rather than throw the work away, these candidates recombine the assets already banked into angles that structurally avoid those failure modes.

Assets in hand

The candidates are picked to make an unused or under-used asset load-bearing:

  • Chainstack RPC — non-archival (getSignaturesForAddress retains only ~3h; an active wallet pages out at ~2.9h). This is a realtime + send-tx asset only — it is NOT a historical backfill source (that limitation is exactly what forced the GMGN harvest pivot). Good for live execution/submission, useless for lookback.
  • Jito ShredStream (banked) — pre-confirmation transaction visibility (see txs in shreds before they confirm). This is the rarest UNUSED asset in the stack: copy-trading never actually needed it (a mirror lands slot N+1 regardless of feed speed), so it has sat idle. Any strategy whose edge is acting on information before confirmation is where this asset finally earns its keep.
  • Jito bundles — atomic bundle submission with tips (ordering / front-or-back-run positioning, revert protection).
  • GMGN deep data via browser same-origin fetch() — the browser context’s Cloudflare cf_clearance beats the WAF (headless curl/reqwest = 403); yields wallet_activity + 1-second klines. Rate-limited to ~1.3k fetches/session.
  • solv1-rig (shelved) — the already-built dual-feed shred + Yellowstone/gRPC consumer and arrival-delta harness. It already knows how to ingest ShredStream — reusable substrate for any shred-driven strategy.
  • solv1-alpha backtest harness — the walk-forward go/no-go engine (score → adverse-selection filters → cohort → real-kline decay → paper-mirror → SOL-hold + random-basket benchmarks → 4-condition verdict). Reusable for any offline hypothesis on harvested data.

Meta-lessons that filter the candidates

Three lessons from the copy-trading study gate which of these are worth pursuing:

  1. Latency alone is commoditized — feed speed is table-stakes, not edge. A mirror lands slot N+1 no matter how fast you see the leader; raw speed buys you nothing durable on its own.
  2. SELECTION / filtering is the durable part — the adverse-selection filters (wash / copy-bait / age-funding) added +14.8pp (filtered − raw) in the copy study. The repeatable value was in choosing what to act on, not in acting faster.
  3. Beta dominates skill in a bull — a random 50-meme basket returned +109.5% and crushed the “skilled” cohort by ~86.5pp. In a rising tape, broad meme exposure swamps stock-picking; any candidate must either beat that basket or be that basket.

Candidate menu

#StrategyStatusShredStream useDodges copy-trading’s killers? (survivorship / entry-decay / beta)My read
1Co-buy consensus signalNO-GO — beaten by basket betaNo — offline, reuses harvestsurvivorship YES (K-of-N consensus, not one cursed wallet) · entry-decay PARTIAL (still a copy) · beta NO (still must beat the basket)Cheapest test in the whole menu — reuses the harness + co-buy detection + the 37-wallet data already in hand. Dodges the single-wallet survivorship that headlined the NO-GO. Best first probe.
2Data-filtered launch sniperNO-GO — filter can’t clear adverse selectionYES — detects new pump.fun launches pre-confirmationsurvivorship YES (no wallet universe) · follow-decay YES (you’re first, not following) · beta PARTIALThe one strategy that makes the banked latency asset load-bearing. No survivorship, no follow-decay. RISK: the most bot-saturated + rug-heavy niche on Solana — edge is entirely in the quality filter, not the speed → feasibility-first.
3Meme-beta timing (risk-on/off)RISK OVERLAY — beta real, not alphaNosurvivorship YES (no wallet picking) · follow-decay YES (no copying) · beta YES — it harvests the +109% instead of fighting itSimplest angle, no ShredStream. Times broad meme exposure rather than picking tokens. Parked — the hard part is a credible regime signal.
4MEV / back-run large swapsParked — unexplored (lowest priority)YES (+ bundles)survivorship YES · entry-decay n/a (different mechanic)Uses ShredStream + Jito bundles but competes head-on with professional MEV firms in a tip-auction — thin, adversarial, colo-sensitive (exactly the commoditized-latency arena the meta-lessons warn against). Lowest priority.
5Solana arbitrage (atomic DEX-DEX)leaning NO-GO — measurement confirmingtable-stakes (3rd time) — long-tail lane is latency-insensitivesurvivorship n/a · PRINCIPAL-SAFE (atomic bundle reverts → lose fees not inventory)First strategy that can’t lose inventory, so it earned a look — but mainstream arb is a solved colocated machine and the solo long-tail lane’s spreads persist because slippage eats them (realizable ~$0.33 median). See solv1-arbitrage-feasibility-measurement-2026-07-05.

1. Co-buy consensus signal — NO-GO

Verdict — NO-GO

K-of-N consensus successfully dodged the single-wallet survivorship curse, but consensus events were too sparse and arrived too late to beat the basket beta. Selection helped; it did not clear the +109.5% random-basket bar. Closed out.

Hypothesis: instead of mirroring one leaderboard wallet (survivorship-cursed, +23% is a ceiling), buy when ≥K quality wallets independently pile into the same token inside a short window. Consensus across the already-filtered cohort is a stronger, less overfit signal than any single wallet’s unrepeatable history.

Why it dodges the killers: the NO-GO’s headline failure was single-wallet survivorship. A K-of-N co-buy requires agreement across the cohort, which is far harder to fake and doesn’t lean on one wallet. Entry decay still applies (you’re still following), so pair it with the verdict’s position-holds-only refinement rather than scalping the signal.

Assets reused: solv1-alpha harness + co-buy detection + the existing 37 copyable wallets + already-harvested GMGN data. No new harvest needed → cheapest test in the menu.

Cost / risk: lowest cost (fully offline, data in hand). Residual risk: still exposed to beta-swamps-skill — it must clear the same random-basket bar the naive cohort failed.

2. Data-filtered launch sniper — NO-GO

Verdict — NO-GO

The most bot-saturated and rug-heavy niche on Solana; ShredStream speed is table-stakes and no quality filter cleared the adverse selection at a margin worth the risk. The banked latency asset was necessary-not-sufficient here too. Closed out.

Hypothesis: ShredStream sees new pump.fun launches pre-confirmation → snipe only launches that pass a quality filter (deployer history, initial liquidity, socials, holder distribution, …), submitting via Jito bundles.

Why it dodges the killers: there is no wallet universe, so no survivorship; you are the first buyer, so there is no follow/entry decay (the mechanic that killed scalps). This is the single strategy where the banked ShredStream latency asset is genuinely load-bearing instead of decorative.

Assets reused: ShredStream (banked) + the shelved solv1-rig shred consumer (already ingests shreds) + Jito bundles + Chainstack send-tx (realtime submission) + GMGN/quality data for the filter.

Cost / risk: this is the most bot-saturated and rug-heavy niche on Solana. Per the meta-lessons, speed is table-stakes here — the edge is entirely in the quality filter, not the latency. Hence feasibility-first before any build; research is running → ~/coding/slv1/docs/sniper-feasibility.md.

3. Meme-beta timing (risk-on/off) — RISK OVERLAY, not alpha

Verdict — validated as a RISK OVERLAY, not alpha

The +109.5% basket beta is real and dominant — but bankable only as a risk-managed exposure overlay (when to size broad meme exposure up/down), not as a stand-alone token-picking edge. The value is in sizing/timing exposure, not selection. Kept as an overlay concept; rejected as alpha.

Hypothesis: don’t pick tokens — time the whole meme complex. The copy study’s most brutal fact was that a random 50-meme basket did +109.5%; beta dominated. So harvest the beta: go risk-on/off on broad meme exposure driven by a regime signal.

Why it dodges the killers: no wallet selection (no survivorship), no copying (no decay), and it captures rather than fights the beta that swamped the copy study.

Assets reused: mostly market-level data (can reuse the GMGN harvest path). No ShredStream, no rig — the simplest stack.

Cost / risk: parked because the entire difficulty is a credible regime/timing signal — without one, “time the meme complex” is just leverage on noise.

4. MEV / back-run large swaps — parked (lowest priority)

Hypothesis: use ShredStream to spot large swaps pre-confirmation and back-run them via Jito bundles.

Why lowest priority: it does use the banked assets (ShredStream + bundles), but the counterparties are professional MEV firms in a tip-auction — margins are thin, the game is adversarial and colo-sensitive, and it lives squarely in the commoditized-latency arena the meta-lessons flag as not a durable edge. Parked, lowest priority.

5. Solana arbitrage (atomic DEX-DEX) — leaning NO-GO

Verdict — leaning NO-GO (measurement confirming)

The first PRINCIPAL-SAFE strategy in the whole arc — an atomic Jito bundle reverts if the spread is gone, so you lose fees not inventory (flash loans push at-risk capital → 0). That safety is why it earned a look after the copy/sniper/co-buy angles had already died. But desk research found mainstream arb is a solved colocated machine and the only solo lane is latency-insensitive long-tail; the passive arb-measure harness then showed the persistent spreads persist because slippage eats them — realizable profit ~$0.33 median at trivial size, negative above it. A decision-grade 180-min run is confirming. Full write-up: solv1-arbitrage-feasibility-measurement-2026-07-05.

Hypothesis: capture cross-venue price divergences on long-tail pump.fun→PumpSwap × Meteora-DLMM pairs the colocated giants under-index, executed as an atomic Jito bundle (revert-protected) with flash-loaned capital.

Why it dodged the killers (and why that wasn’t enough): no wallet universe (no survivorship), no following (no entry decay), and — uniquely — no inventory risk (atomic revert). It failed on a new axis: the spreads are not profitably capturable — on $25–100k pools your own trade’s slippage eats the 2–3% gap, so the mid-price “opportunity” is an artifact of un-capturability, not standing edge.

Assets reused: DexScreener discovery + Chainstack WSS accountSubscribe (realtime pool state) + on-chain reserve / active-bin decoders + Jito bundles (the execution path, left un-exercised — this stayed passive). ShredStream would be table-stakes for the 3rd time if it were ever built.

Cost / risk: the measurement harness is zero-capital and already run — the finding kills the economics before any build. The one positive-persistent event was a commoditized latency race (Meteora bin price lagging PumpSwap ~3 slots), and the profit model is an optimistic upper bound (DLMM depth ~6× thinner than modelled → truth is worse).

Graph View

Backlinks