1.25M Solana wallets, each pre-scored with win_rate, net_pnl_sol, roi, bundle_rate and a 0–1 bot_confidence label — a ready-made feature table for copy-trade backtests and model training.
Disclosure: This article contains affiliate links. If you sign up through them, MadeOnSol may earn a commission at no extra cost to you. This never affects our rankings, ratings, or reviews.
Building a product on Solana data?
Skip the Geyser pipeline — embed KOL flow, deployer reputation, and the all-DEX firehose over REST, WebSocket, or webhooks. The Business tier (€400/mo) is the self-serve embed license at 500k calls/day — Enterprise adds white-label & redistribution.
Read this before you build anything. A low bot_confidence does not mean "confirmed human." It means no bot signal was detected — and for a large share of wallets, that is simply because there wasn't enough behavior to evaluate. Several of the underlying signals (trade velocity, buy-size variance, bundle frequency) need multiple trades to compute. A wallet that placed one trade and never returned has no behavioral surface to flag, so it defaults to a low score by absence of evidence, not by being verified human.
The practical consequence: if you treat "low bot_confidence" as a positive human label, your training set is dominated by single-trade noise wallets, and your model learns to predict "inactive," not "skilled." Always combine bot_confidence with a sample-size floor — require a minimum number of closed positions before you trust any per-wallet statistic. This is the same trap behind the why 100% win rate is a red flag problem: a wallet with a single winning trade shows a 100% win rate and a clean bot score, and both are artifacts of having almost no data.
The inverse mistake is just as costly. A high bot_confidence is not a synonym for "bad wallet" — bots are often mechanically more profitable than the median human because their edge comes from infrastructure (Jito tips, MEV positioning) that no retail trader can copy. So high-scoring wallets can have excellent PnL and still be worthless to follow. For a training target that means bot_confidence and profitability are near-orthogonal, and for a copy-trade filter it means you want low bot_confidence AND real history, not raw PnL rank. Treat the score as a routing signal, then let win_rate, roi, and trade count do the ranking within the human bucket.
sol-smart-money is the wallet-scoring layer. It pairs naturally with the trade-level sol-kol-trades for backtesting and sits alongside the deployer and Robinhood Chain sets in the full datasets hub. If you want the live, continuously-refreshed version of these scores instead of a snapshot, the same classifier is served through the paid v1 API.
Roughly 1.25 million scored Solana wallets, one row each, with coverage beginning 2026-04-12. Each row carries win_rate, net_pnl_sol, roi, bundle_rate, bot_confidence, and last_seen.
It is a 0-to-1 likelihood that a wallet is a bot, from the same classifier that runs in the live API. Crucially, a low score means "no bot signal detected," not "confirmed human" — many low scores are just wallets with too few trades to evaluate.
Tools mentioned
Live health scores, average ratings, and direct links on MadeOnSol.
Crypto tax software that simplifies reporting for every blockchain
Real-time DEX charts and trading data across all chains
Multi-chain portfolio tracker with DeFi position tracking across Solana protocols
Enterprise crypto accounting and tax reporting for Solana
AI 'vibe trading' app with personalized Solana sell signals
Build with MadeOnSol API
KOL tracking, deployer intelligence, DEX streaming, and webhooks. Free API key — 200 requests/day.
New customers try Pro free for 5 days — card, cancel anytime.
import { MadeOnSol } from "madeonsol";
// Get a free API key at madeonsol.com/pricing
const client = new MadeOnSol({ apiKey: "msk_your_key" });
const { trades } = await client.kol.feed({ limit: 10 });
const { alerts } = await client.deployer.alerts({ limit: 5 });
const { tools } = await client.tools.search({ q: "trading" });Building a product on Solana data?
MadeOnSol isn't an RPC or another generic token API — it's Solana memecoin intelligence that's painful to build in-house, pulled from dual-region gRPC shred streams: sub-second from the on-chain event to your app.
Signals you'd otherwise build
KOL & smart-money flow from 1,000+ labeled wallets, deployer reputation, coordination clusters, linked-wallet entity resolution, and an all-DEX firehose.
Embed it in your product
The Business tier (€400/mo, self-serve) licenses you to display MadeOnSol data inside your own product — 500k calls/day, 10 WS + 5 firehose connections. Enterprise above adds white-label & redistribution rights and custom endpoints.
Evaluate first, commit later
Test everything on a free ULTRA key — no commitment. If it fits, we scope volume or white-label pricing.
Keep reading
Analytics
On-chain datasets built for reproducible research: SHA-256 checksums, fixed temporalCoverage windows, identity-scrubbed pseudonymous rows, and schema.org/Dataset JSON-LD that makes every set citable and discoverable in Google Dataset Search.
Analytics
Four ready-to-analyze Solana datasets — 1.25M scored wallets, 1.5M priced KOL trades, 27k real deployers, plus genesis-complete Robinhood Chain — as scrubbed, checksummed CSVs with free 5,000-row samples.
Analytics
A €149 CSV of Robinhood Chain intelligence indexed from block zero — deployer reputation, smart-money scores, and token lifecycle since the 2026-04-30 genesis. No incumbent has genesis-complete RHC history, because the chain is too new to have one.
Analytics
Computing deployer reputation for ~27k real Pump.fun deployers means a validator gRPC feed, an indexing pipeline, terabytes of storage, and bond attribution — or a €99 file. Here's the honest cost breakdown, including where building your own still wins.
Analytics
The sol-kol-trades dataset stamps price and market cap at execution on every one of 1.5M+ rows — so you can compute realized copy-trade PnL with no price oracle and no look-ahead bias.
Analytics
As of mid-July 2026, only 23% of human alpha wallets on Robinhood Chain are mostly-memecoin — 66% mostly trade non-launchpad assets like stables, tokenized stocks and blue-chips. The chain built for stocks became a meme machine, but the money says otherwise.
Enjoyed this article?
Real-time KOL trades, Pump.fun deployer intel, and 1,200+ ranked Solana tools — free to explore.
Open the KOL TrackerThe sol-smart-money dataset is ~1.25 million scored Solana wallets (€149), one row each, pre-labelled with win_rate, net_pnl_sol, roi, bundle_rate, a bot_confidence score from 0 to 1, and last_seen. Coverage begins 2026-04-12, and it is the exact classifier output that powers our live API's alpha-flag endpoints — exported to a flat file you own. For anyone training a model or running research, that means the hard part (turning raw DEX swaps into per-wallet features and labels) is already done: you get a clean, wallet-keyed feature table, not a pile of transactions to reconstruct. Download it, with a free 5,000-row sample, at /datasets/sol-smart-money.
This is the buy-the-file companion to our on-site wallet intelligence. If you want the running methodology behind the score, the 952k alpha-wallets methodology post and how to read Solana smart money go deep. This post is about the dataset: what's in it, how to use it for ML, and the one caveat that will wreck your labels if you miss it.
| Column | Type | Meaning |
|---|---|---|
wallet | string | Scrubbed wallet key (pseudonymous, no name) |
win_rate | float | Share of closed positions with positive realized PnL |
net_pnl_sol | float | Cumulative realized PnL in SOL |
roi | float | Net PnL over total SOL invested |
bundle_rate | float | Share of trades landing in the pool-creation Jito bundle |
bot_confidence | float 0–1 | Likelihood the wallet is a bot |
last_seen | timestamp | Most recent activity |
Like every MadeOnSol dataset, it ships as CSV.gz, is identity-scrubbed and wallet-keyed, SHA-256 checksummed, and comes with a free 5,000-row sample so you can inspect the schema before paying.
Most on-chain "smart money" projects spend 80% of their time on feature engineering: pulling swaps, matching buys to sells, pricing them, aggregating per wallet, then labelling. This dataset hands you the output of that pipeline. A few concrete uses:
bot_confidencewin_rate, roi, and bundle_rate co-vary across the population to characterize what "smart money" actually looks like at scale.Yes. It is a pre-scored feature table, so you can use bot_confidence as a label and the other columns as features, or use the whole set for wallet selection and distribution research. Always apply a sample-size floor before trusting per-wallet stats.
This set is one row per wallet with aggregate scores; sol-kol-trades is one row per individual trade with price and market cap stamped at execution. Use this to pick wallets, use that to backtest their trades.
Yes. It is identity-scrubbed and wallet-keyed — addresses only, no names or personas. The data is pseudonymous on-chain information delivered as a checksummed CSV.gz.
sol-smart-money is 1.25M Solana wallets with the feature-engineering already done: win rate, PnL, ROI, bundle rate, and a production bot_confidence label, from 2026-04-12 on. It's the fastest way to start ML or copy-trade research on real on-chain behavior — as long as you remember that low bot_confidence means "unevaluated," not "human," and pair it with a sample-size floor. Grab the file and a free 5,000-row sample at /datasets/sol-smart-money.