These large on-chain accounts are not better or worse versions of each other — they are running completely different businesses. Two measurable behaviours tell them apart; a third number decides whether they are worth watching at all.
Sorting traders by how much they made is circular — you define the category by the outcome and then use the category to explain the outcome. We use two purely behavioural axes instead:
Performance is read as an outcome, never as an input to the classification.
| Type | n | Median MTM P&L | Earned per $1M traded | Days active | Coins | Fills/day |
|---|---|---|---|---|---|---|
| Low-frequency directional | 221 | $210K | $21K | 0.32 | 6 | 34 |
| Mid-frequency swing | 1,955 | $155K | $3K | 0.58 | 12 | 193 |
| High-frequency taker | 837 | $229K | $2K | 0.71 | 15 | 1,071 |
| High-frequency mixed | 552 | $314K | $1K | 0.89 | 31 | 1,280 |
| High-frequency maker | 227 | $353K | $443 | 0.85 | 20 | 1,577 |
Sample: addresses with peak exposure ≥ $1M and reliable exposure data. "Earned per $1M traded" = mark-to-market P&L ÷ volume in millions — the only measure comparable across frequencies.
| Type | Analogy | How they make money |
|---|---|---|
| Low-frequency directional | The hunter | Waits a long time, one shot one deer. In the market only 32.2% of days, but each shot is worth a lot |
| Mid-frequency swing | The seasoned retail trader | Trades the setups he recognises, in the market about half the time |
| High-frequency taker | A courier racing for orders | A little per trade, thousands of trades a day, winning on speed — and paying a toll every time |
| High-frequency maker | A toll booth | Cents per car, living on traffic. Quotes 20 coins, on duty 85.1% of days |
| High-frequency mixed | A courier who also runs a toll booth | Takes both sides, highest attendance of all |
| Low-frequency directional | $21K per $1M traded | |
| High-frequency maker | $443 per $1M traded | ← about 1/47 |
Median mark-to-market P&L runs the other way — the makers are highest. Which tells you high frequency earns on volume, not on rate: a tiny unit efficiency multiplied by an enormous amount of turnover.
Attendance is the clearest dividing line: 0.32 for low-frequency, 0.85 for makers, 0.89 for mixed. The last two are a job, not speculation.
It splits "this person makes money" into "here is what he makes money from." When a high-frequency maker is net short a coin, that is most likely market-making inventory — he passively absorbed someone else's buying, he is not bearish. Read that position as "whale is short" and you may have the direction exactly backwards.
That is why we show makers and directional traders separately rather than adding them into one "whale positioning" number.
These addresses were selected on whole-window mark-to-market profit, so every proportion here describes this profitable cohort, not the market as a whole. The thresholds (500 fills/day, taker ratio 0.8 / 0.3) are practical cuts through a distribution, not natural constants — change them and the group sizes move, but the direction of the efficiency-versus-frequency relationship does not.