Research tools and data

Realized-PnL leaderboard

Browse the published ranking by timeframe, search an address, filter by closed round-trips or win rate, and switch between amounts and percentages. The ranking counts matched on-chain buys and sells only. It does not measure total wealth, unrealized gains, transferred inventory, or future performance.

Each scope/timeframe is capped at 1,000 ranked rows. Available windows and counts come from the current provider manifest. “All” in the category control means all advertised token categories, not all chain history. The older saved seven-day export contains 53,222 observed addresses and 14,245 qualifying addresses; its “all” timeframe duplicates those seven days. Neither overlapping windows nor scope rankings should be added together.

Trader cards

Compare realized results, matched volume, closed trades, and recorded PnL sparklines. The model band is a descriptive win-rate comparison. Trades may be dependent and traders were selected after results were observed, so it is not evidence of skill or a recommendation to copy a wallet.

Wallet connection uses Peapod's existing wallet panel. “View my research record” searches the connected address. Missing detail exports are shown as unavailable, never as evidence of no activity. Where available, the address view also shows recovered swap observations checked against Goldsky receipts and the data lake. These are not netted trades or matched closes, and have not been reconciled to published PnL. Token-transfer discovery can miss activity through routers or relayers.

Pool depth

Inspect the saved estimate of how much a pool could buy or sell before moving price by 1%. This walks initialized ticks; it is different from TVL and from valuing the whole band at one active-liquidity value. The flat-liquidity comparison can err in either direction.

The 66-pool study is not the entire chain. Depth is anchored to 13 September 2026 at 21:09:31 UTC; the swap tape ends on 14 September at 02:07:47 UTC. The snapshots are historical, not executable quotes now.

LP calculator

Choose a pool, range, holding period, and deposit size to see a distribution of historical fees minus impermanent loss versus holding. The dataset contains 5,650 sampled windows and pool-specific fee kernels. Historical swaps are replayed unchanged, assuming a marginal price-taking position. Results are not forecasts, and larger positions could have changed the market.

The upstream snapshot notes that its tape predates a planned gas-subsidy end of 29 September 2026. Treat that as an observation about this source snapshot, not a statement about current network pricing.

How sources are consolidated

Tokens are matched by Robinhood Chain ID (4663) and contract address. Symbols are labels, not identifiers. All 193 imported token-logo identities match the existing token registry. Pool IDs are separate identifiers and cannot be substituted for token addresses.

The historical wallet study uses different coverage and accounting from the matched-round-trip rankings. They remain separately labeled; their PnL and NAV figures are not added together. Existing Pods, wallet sessions, watchlists and XP remain on this staging origin after the Peapod rename.

Download the consolidation report or read the integration notes.

Download the recovery coverage and checksums. The view shows at most the latest 100 verified swap logs per recovered address.

Current trader research

The leaderboard and trader cards now load the provider's current build, available windows and RWA/Pons and ETH/USDG scopes. The build time appears above the ranking; each detail has its own observation window. Rankings and displayed trade lists are capped, so these are not complete wallet histories. Matched realized PnL is not account equity or unrealized PnL. Provider USD figures and our historical USDG accounting use different methods and must not be added.

The win-rate band uses an exact two-sided binomial test against a 50% baseline. Five or fewer closes cannot yield a significance finding. The test describes observed win frequency, not proven trading skill or future returns.

Pool depth, the LP calculator, recovered receipts and the wallet activity browser retain their explicitly dated saved datasets. If the current provider is unavailable, historical evidence remains available without being presented as current.

Wallet ranking metrics

Current trader rankings can be sorted by closed-trade PnL %, copy score, matched-flow coverage, realized PnL, win rate and sample count. These sorts reorder the provider's capped ranking; they do not establish a chain-wide ranking by the new metric.

Closed-trade PnL % is realized profit divided by matched purchase cost. The source's matched volume is purchases plus sales, so purchase cost is (matched volume − realized PnL) / 2. Open positions, unknown cost bases and gas remain excluded. This percentage is not a whole-wallet return.

Copy score v1 is an experimental screening heuristic, from 0 to 100. It combines 40% closed-trade return (capped at 20%), 35% the 95% Wilson lower bound on win frequency, and 25% matched volume as a share of total volume. The result is multiplied by min(closed round-trips / 30, 1) and min(matched volume / $1,000, 1). Nonprofitable records score zero; unavailable inputs remain unavailable. This score does not estimate the effects of copying latency, fees, liquidity, open positions or execution. It is not a validated prediction or a recommendation to copy a wallet.

The Sharpe & returns view uses the separate historical, transfer-adjusted 178-wallet study. It exposes annualized daily Sharpe, compounded portfolio return, drawdown, valid sample days and missing intervals. Fourteen valid days and complete return paths are selected by default. These are selected-portfolio estimates using inferred external flows and proxy prices; they are not live or verified whole-wallet returns. Sharpe's definition requires a return series; cumulative realized-PnL sparklines alone cannot supply it.

Observation periods

Current ranking pages display the actual UTC from_ts and to_ts of the ranking response and refresh every minute while visible. “Delayed data” means the observation window ends more than an hour ago or a cached response was served after a provider error. A fresh build timestamp does not make old observations current. Trader details show their own observation window and last trade separately; an inactive wallet's last trade can legitimately be older than the data window.

The wallet entry point opens current trader rankings. The former saved accounting browser remains at /wallet-activity/historical. Historical portfolio studies, depth and calculator snapshots retain their real dates. Chart range buttons are limited to the history actually available, and “All available” refers only to that observation window.