QuantBloc ← Back to QuantBloc

Methodology

QuantBloc Methodology

Last updated: November 17, 2026

QuantBloc measures what Solana wallets actually did, then separates that from what a follower could realistically have reproduced. This page documents the definitions, coverage, assumptions and limitations behind the metrics shown on QuantBloc.

1. Data coverage

QuantBloc currently covers Solana trades on Pump.fun and PumpSwap, and no other venues. On PumpSwap, QuantBloc includes only tokens proven to have originated on Pump.fun.

Trades are tracked from 13 May 2026. Copy trader ("mirror") measures are available from 8 June 2026. Micro-buys are captured from 5 September 2026. Per-trade on-chain fees have been read directly from signatures since 24 September 2026; fees on older trades are estimated.

As of 16 November 2026, the Base contains 37.0 million wallets and approximately 2.05 billion trades.

The wallet table is rebuilt once a day in the morning, Paris time. Copy trader and backtest tables are rebuilt nightly. Recent-activity columns refresh every 10 minutes.

QuantBloc merges two independent real-time streams of Solana trades. A third stream is currently in shadow testing. Wallets are not pre-filtered: bots, routers and token creators can all appear in the Base, with flags available to exclude them.

Coverage limits

QuantBloc does not count trades routed through other DEXes. Measurements show that approximately 13% of the volume of migrated Pump.fun tokens trades on Meteora, which is outside the current coverage.

Timestamps record stream reception time, with a median delay of about +1 second relative to the on-chain block time. Ordering within a block does not rely on that timestamp: QuantBloc uses the Solana slot and transaction index.

Some trades pass through aggregators or routers that sign in their own name. Where the on-chain data makes it possible, QuantBloc re-attributes those trades to the real wallet.

SOL balance is read only for wallets active during the last 7 days. An empty balance therefore means not read, not zero.

History before 13 May 2026 is not part of the Base. A few legacy copy columns cover a curated pool of approximately 2,000 wallets over their full on-chain careers. Those columns therefore use a different observation period from the rest of the Base.

2. Trade, position and price

A trade is one buy or one sell by one wallet on one token.

A position is everything that wallet did on that token: all of its buys and sells.

Price is stored as the token's market cap in SOL at the time of the trade. Ratios such as multiples and potentials are therefore SOL-native, so the SOL/USD exchange rate cancels out.

USD values convert each trade using the SOL price for that trade's own hour. Wallet-page fees use the SOL price for the corresponding minute. Today's SOL price is not applied retroactively.

Very small micro-buys, such as buys of a few lamports or fewer than 100 tokens, are retained as evidence that the wallet touched the token, but excluded from stakes, multiples and gains because their implied price is not reliable.

Impossible price readings are removed before calculations. QuantBloc rejects readings contradicted by the surrounding trades, placeholder values, and levels at which the token never credibly traded. While a token is still on Pump.fun, entries also have a hard floor at the token's launch market cap, approximately 28 SOL.

3. Trader metrics

These metrics describe what the wallet itself did.

Gain (SOL / $) is SOL received from sells minus SOL spent on buys. Tokens still held remain counted as spent; no unrealised value is added.

Copyable gain uses the same calculation, but handles tokens the wallet received rather than bought. If a wallet sells more tokens than it purchased, for example after an airdrop or transfer, sell proceeds are capped pro rata to the quantity it actually bought. This keeps only the part a copy trader could have reproduced.

Multiple (median / mean) is calculated per closed position as sell value divided by buy value. Only closed positions are included. With fewer than three closed positions, the multiple carries very little information, so QuantBloc displays the number of observations next to it.

Success rate ("winrate") is the share of closed positions sold above cost. QuantBloc displays winrate but does not treat it as a measure of quality. Internal measurement found that ranking by winrate can select wallets that win very often while still losing money overall.

Entry delay is the number of seconds between token creation and the wallet's first buy.

Other trader metrics include holding time, sells per token, share of the best trade, and concentration. Concentration measures the share of stake placed on the wallet's biggest token, not the share of profit.

Wallets that created tokens are flagged. Their historical results can look exceptionally strong while being impossible for an external copy trader to reproduce.

Rug rate is the share of the wallet's buys during the last 60 days that occurred before a rug in that token. QuantBloc defines a rug as a price loss of at least 50% over the next two observed trades, followed by no recovery above half of the pre-drop level. Rug rate is considered meaningful only from 15 judged buys.

4. Copy trader metrics

Trader performance and copy trader performance are deliberately separated. A copy trader does not receive the wallet's execution price.

Block convention

QuantBloc prices a copy trader at the end of block N after the wallet's trade, where N is 0, +1 or +2.

  • Block 0 uses the last trade of the same block. It is theoretical and requires same-block infrastructure.
  • Block +1 uses the end of the following block.
  • Block +2 uses the end of the second block after the wallet's block: the time an alert takes to go out. In FULL COPY, the default is block 0, and block +1 or +2 can be chosen.

"End of block" means the last trade in that block by transaction order. It is always after the copied trade. QuantBloc publishes all three block settings so that the cost of latency is visible rather than assumed.

Mirror copy

The mirror copy trader reproduces the wallet's behaviour: it buys N blocks after each wallet buy and sells N blocks after each wallet sell, using end-of-block prices.

QuantBloc reports the mirror median multiple at block 0, +1 and +2, the mirror winning share, and the number of mirror positions. Results are fragile below approximately 10 mirror positions.

The difference between block 0 and block +2 measures latency sensitivity. A wallet that wins at block 0 but loses at block +2 can be a trap for copy traders.

Frozen entries % is the share of positions where nobody traded between the wallet's buy and block +2. In those cases, the displayed copy trader price is a stale quote rather than an observed executable price. This is typical of some snipers.

First-block buys % measures buys made in the first block of a token's life. These bundle or launch-snipe entries are not reachable by a copy trader.

Copyable potential

Copyable potential is the highest credible level reached after the copy trader's block +2 entry divided by that copy trader entry.

The peak must occur after the copy trader entered. A token's all-time high is not used if it occurred before the entry and was therefore unreachable.

Counts of tokens reaching x10, x50 and x100 use the same rule.

Potential measures entry quality. It is a ceiling, not a realised gain.

Risk

Drawdown before peak (median %) is the lowest level reached between copy trader entry and the later peak. A value of -40% means the position had to withstand a 40% drawdown before reaching that peak.

Risk ratio is (potential - 1) / drawdown before peak. It becomes unbounded when drawdown approaches zero. QuantBloc displays values above 100 as >x100, while sorting on the underlying value.

Max loss (median %) is the lowest level reached during the token's entire observed life after entry. This metric is displayed for transparency even though it is not very useful for sorting, because most Pump.fun tokens eventually approach zero.

5. Fees

Wallet page

The wallet's own results are net of the fees paid by each trade where those fees are available on chain.

QuantBloc reads the network fee, priority fee, tip including Jito and similar services, trading-bot commission, account rent and platform fee from each trade.

Before 24 September 2026, fees are estimated rather than read directly from the trade.

Copy trader simulations

In FULL COPY, fees are configurable: bot commission, priority fee, tip and per-trade costs can be set separately and applied on buys and sells.

On the Pump.fun launchpad, slippage is calculated with the launchpad formula. Validation against real fills produced a 0.00% median error. After migration, on PumpSwap, prices are read from the pool reserves on chain, and the impact of the copy trader's own order on the pool is included, on buys and on sells.

6. FULL COPY

FULL COPY is a Premium wallet-page backtest designed to answer:

If I had copied every buy and every sell of this wallet over this window, with my stake and my bot's fees, what would I have now?

Entries and exits are simulated at the end of block 0, +1 or +2 after each copied trade.

Stake is set per buy in dollars. The user can choose one buy per token or copy all buys.

Optional exits include take-profit, stop-loss and time exit. A stop-loss must remain hit for 2 seconds and is filled at the worse price of the block where it was hit and the following block. By default, the simulation follows the wallet's sells.

Positions that were never sold are valued at their current price rather than removed from the result.

The engine is checked automatically every day on reference cases, with an alert if a result moves beyond the largest difference accepted during validation.

7. Top traders by bot

QuantBloc identifies 25 trading bots from their fee vaults on chain.

A wallet's gain "via a bot" is calculated as its position result multiplied by the share of SOL it spent through that bot, minus fees.

This feature is a watch list of wallets that traded well through each bot. It is explicitly not a promise that those wallets will be profitable to copy.

8. Tracker lists

Tracker lists are lists of Solana wallets to follow in a wallet tracker. There are five profiles, recalculated every Monday from the same data as the Base, and each list stays frozen until the next Monday.

  • Profitable wallets: clean, active wallets that win steadily without relying on one big hit. Ranked by average gain per trade, corrected for luck: a wallet with few trades or very uneven results moves down.
  • Gem hunters: wallets that repeatedly get into tokens that climb hard after they buy. Ranked by the number of tokens that did ×10 after their entry. This is what the token did, not what the wallet cashed.
  • Whales: profitable wallets whose usual stake is in the top 1% of the Base.
  • KOLs: wallets of known KOLs that traded over the last 30 days, ranked by 30-day gain. Tokens a KOL launched themselves stay in his gain, and a column shows who also creates tokens.
  • Dev wallets: wallets that launch tokens on Pump.fun and had tokens really migrate, over the previous 21 days (the last day is left out so tokens have time to migrate). Tokens quoted outside SOL and mayhem tokens are excluded. Their best token is the highest peak among their tokens, leaving out pools emptied right at migration, where a single buyer sets the price.

Every row is checked before publication. The rug share is checked again against its latest calculation, impossible values are looked for, and the peaks counted are compared with prices read from the pools on chain. A row that fails a check is removed, not corrected: a list slightly incomplete but right is preferred to a complete list with wrong figures. If a list would lose more than half of its candidates compared with the previous week, nothing is published and the previous week's list stays online.

The figures in these lists are the wallet's own. What a copy trader would have made is a different number, shown in each wallet's backtest (entry at block 0, +1 or +2, real fees). Lists can be copied as addresses or exported to paste into GMGN or Axiom, either the whole list or only the wallets ticked.

The selection bias described below applies here too: a list of the best wallets of the past weeks separates candidates, it does not forecast them.

9. Biases and limitations

QuantBloc publishes its limitations because backtests and historical wallet rankings can otherwise look more precise than they really are.

  1. Venue coverage. QuantBloc covers Pump.fun and PumpSwap only. Approximately 13% of the measured volume of migrated Pump.fun tokens trades on Meteora and is not included.
  2. Backtest fills are a ceiling. Backtests sell at the observed price. Real execution can be worse, especially during fast declines. On the real fills of QuantBloc's own trading bot (measured on Robinhood Chain, not yet on Solana), sells into rising prices achieved 99% of the target price, while sells during a crash achieved 33%.
  3. Timestamp delay. Stream timestamps are reception times and are approximately +1 second after on-chain block time. In-block ordering uses slot and transaction index instead.
  4. Survivorship. Wallets that stopped trading remain in the dataset. Active-wallet filters are available.
  5. Selection bias. Any list of "best wallets", including one built by sorting the Base, tends to look better in the past than it performs in the future. A sort separates candidates; it is not a forecast.
  6. Past performance. Past results do not predict future results.